&EPA
            United States      Industrial Environmental Research
            Environmental Protection Laboratory
            Agency        Research Triangle Park NC 27711
                        EPA-600/9-84-025b
                        November 1984
            Research and Development
Fourth
Symposium on the
Transfer and
Utilization of
Particulate Control
Technology:

Volume II.
Electrostatic
Precipitation

-------

-------
                                DISCLAIMER


     This document has been reviewed in accordance with U.S.
Environmental Protection Agency policy and approved for publication.
Mention of trade names or commercial products does not constitute
endorsement or recommendation for use.
                                   ii

-------
                             ABSTRACT
     The papers  in  these three volumes of Proceedings were presented
at the Fourth Symposium on the Transfer and Utilization of Paticulate
Control Technology held in Houston, Texas during  11 October through 14
October 1982, sponsored by the Particulate Technology Branch of the
Industrial Environmental Research Laboratory of the Environmental
Protection Agency and coordinated by the  Denver Research Institute of
the University of Denver.

     The purpose of the symposium was to bring  together researchers,
manufacturers, users, government  agencies, educators  and  students to
discuss new technology and to provide  an effective means for the
transfer of this technology out of the laboratories and into the  hands
of the users.

     The  three major categories of  control technologies  -
electrostatic precipitators, scrubbers, and fabric filters  - were the
major concern of the symposium.  These technologies were discussed
from the perspectives  of  economics; new technical advancements in
science and engineering; and applications.  Several papers  dealt with
combinations of devices  and technologies, leading to a concept of
using a systems approach  to particulate control rather than device
control.   Additional topic areas  included novel control devices,  high
temperature/high  pressure  applications,  fugitive  emissions,
measurement techniques,  and economics and cost analysis.

     Each  volume  of these proceedings contains a  set  of related
session topics to provide  easy access to a unified technology area.

     Since the  spirit  and style of the panel  discussion are not
reproducible  in  print, the initial remarks presented by the panelists
have been  included  in  the volume to which their input to the panel
pertained, in  the interest  of providing unified  technological
organization.
                               111

-------
                               CONTENTS
VOLUME II  - CONTENTS	     V
VOLUME I   - CONTENTS	    ix
VOLUME III - CONTENTS	   xii
          Section A - Industrial Applications
MODELING OF WET BOTTOM AGITATOR SYSTEMS FOR ELECTROSTATIC
PRECIPITATORS ON RECOVERY BOILERS  	     1
  M.A. Sandell, R.R. Crynack

DESIGN AND PERFORMANCE OF ELECTROSTATIC PRECIPITATORS
UTILIZING A NEW RIGID DISCHARGE ELECTRODE DESIGN 	    17
  G.R. Gawreluk, R.L. Bump

DEVELOPMENT AND EVALUATION OF NEW PRECIPITATOR EMITTER
ELECTRODE	    35
  R.L. Adams/ P. Gelfand,

INDUSTRIAL APPLICATIONS OF TWO-STAGE TUBULAR ELECTROSTATIC
PRECIPITATORS  	    51
  H. Surati, M.R. Beltran
          Section B - Advanced Technology


PILOT DEMONSTRATION TWO STAGE ESP TEST RESULTS	    65
  P.  Vann Bush, D.H. Pontius

EVALUATION OF PRECHARGERS FOR TWO-STAGE ELECTROSTATIC
PRECIPITATORS   	    84
  G.  Rinard, D. Rugg, M. Durham

INITIAL EXPERIMENTS WITH AN ELECTRON BEAM PRECIPITATOR TEST
SYSTEM	    96
  W.C.  Finney,  R.H. Davis, J.S. Clements, E.G. Trexler,
  J.S.  Halow, 0. Tokunaga

EXPERIMENTS WITH WIDE DUCTS IN ELECTROSTATIC PRECIPITATORS . .  .   Ill
  E.  Weber

A RECONCILIATION:  WIDE VERSUS NARROW SPACED COLLECTING
PLATES  FOR PRECIPITATORS  	   126
  D.G.  Puttick

PULSE CORONA AS ION SOURCE AND ITS BEHAVIORS IN MONOPOLAR
CURRENT EMISSION	   139
  S.  Masuda, Y. Shishikui

-------
          Section C - Fundamentals

A NEW CORRECTION METHOD OF MIGRATION VELOCITY IN DEUTSCH
EFFICIENCY EQUATION FOR CONVERSION OF ELECTROSTATIC PRECIPITATOR
SIZING FROM A PILOT-SCALE TO FULL-SCALE	   154
  F. Isahaya

DISTORTION OF PULSE VOLTAGE WAVE FORM ON CORONA WIRES DUE TO
CORONA DISCHARGE   	   169
  S. Masuda, H. Nakatani

ELECTROSTATIC PRECIPITATOR ANALYSIS AND SYNTHESIS  	   184
  T. Chiang, T.W. Lugar

COMPUTER MODEL USE FOR PRECIPITATOR SIZING	   194
  G.W. Driggers, A.A. Arstikaitis, L.A. Hawkins

IMPROVEMENTS IN THE EPA/SRI ESP PERFORMANCE MODEL	   204
  M.G. Faulkner, R.B.Mosley, J.R. McDonald, L.E. Sparks

NUMERICAL SIMQLATION OF THE EFFECTS OF VELOCITY FLUCTUATIONS
ON THE ELECTROSTATIC PRECIPITATOR PERFORMANCE  	   218
  E.A. Samuel

CORONA - INDUCED TURBULENCE	   230
  M. Mitchner, G.L. Leonard, S.A. Self

VELOCITY AND TURBULENCE FIELDS IN NEGATIVE CORONA
WIRE-PLATE PRECIPITATOR   	   243
  H.P. Thomsen, P.S. Larsen, E.M. Christensen,
  J.V. Christiansen

THE EFFECT OF TURBULENCE ON ELECTROSTATIC PRECIPITATOR
PERFORMANCE	   261
  D.E. Stock

FACTORS  LEADING TO ELECTRICAL BREAKDOWN OF RESISTIVE DUST
LAYERS AND SUSTAINED BACK CORONA	   271
  P.A. Lawless, L.E. Sparks

ELECTRICAL BREAKDOWN OF PARTICULATE LAYERS  	   288
  G.B. Moslehi, S.A. Self

E13CTROMECHANICS OF PARTICULATE LAYERS	   306
  G.B. Moslehi, S.A. Self

LATERAL  PROPAGATION OF BACK CORONA IN TWIN-ELECTRODE TYPE
PRECIPITATORS	   322
  S. Masuda, T. Itagaki

FIRST MEASUREMENTS OF AEROSOL PARTICLE CHARGING
BY FREE  ELECTRONS	   337
  J.L. DuBard, M.G. Faulkner, L.E. Sparks
                                 vi

-------
          Section D - Operation & Maintenance
GAS FLOW DISTRIBUTION MODEL TESTING	   349
  D.R. Cook, J.M. Ebrey, D. Nbvogoratz

AIR FLOW MODEL STUDIES	   369
  L.H. Bradley

COLLECTING ELECTRODE RAPPING DESIGNED FOR HIGH EFFICIENCY
ELECTRIC UTILITY BOILER ELECTROSTATIC PRECIPITATORS  	   384
  A. Russell-Jones, A.P. Baylis

ELECTROSTATIC PRECIPITATOR AND FABRIC FILTER OPERATING AND
MAINTENANCE EXPERIENCE 	   401
  P.R. Goldbrunner, W. Piulle
          Section E - Conditioning
ECONOMICAL FLY ASH COLLECTION BY FLUE GAS CONDITIONING	   416
  E.L. Coe, Jr.

EXPERIENCES AT DETROIT EDISON COMPANY WITH DECLINING
PERFORMANCE OF SULFUR TRIOXIDE FLUE GAS CONDITIONING
EQUIPMENT	   430
  L.A. Kasik, W.A. Rugenstein, J.L. Gibbs

ESP CONDITIONING WITH AMMONIA AT THE MONROE POWER PLANT OF
DETROIT EDISON COMPANY 	   444
  E.B. Dismukes, J.P. Gooch, G.H. Marchant, Jr.

FLY ASH CHEMISTRY INDICES FOR RESISTIVITY AND EFFECTS ON
ELECTROSTATIC PRECIPITATOR DESIGN AND PERFORMANCE  	   459
  H.J. Hall
          Section F - Control Systems
A NEW ENERGIZATION METHOD FOR ELECTROSTATIC PRECIPITATORS
MITSUBISHI INTERMITTENT ENERGIZATION SYSTEM	   474
  T. Ando, N. Tachibana, Y. Matsumoto

SOME MEASURED CHARACTERISTICS OF AN ELECTROSTATIC
PRECIPITATOR OBTAINED USING A MICROCOMPUTER CONTROLLER	   489
  M.J. Duffy, T.S. Ng, Z. Herceg, K.L. McLean

ELECTROSTATIC PRECIPITATOR ENERGIZATION AND CONTROL SYSTEMS  . -   499
  K.M. Bradburn, K. Darby
                                VI1

-------
APPLYING MODULAR MICROCOMPUTER CONTROL ELEMENTS IN A
PRECIPITATOR CONTROL SYSTEM  . . .	   521
  I.M. Wexler
          Section G - Plenary Session
THE CURRENT STATUS, FUTURE DIRECTIONS, AND ECONOMIC
CONDITIONS IN THE APPLICATION OF ESP'S	   534
  S. Oglesby

AUTHOR INDEX 	   539
                                viii

-------
                               VOLUME I

                          FABRIC FILTRATION

          Section A - Fabric Filters:  Fundamentals
THEORY OF THE TEMPORAL DEVELOPMENT OF PRESSURE DROP
ACROSS A FABRIC FILTER DURING CAKE INITIATION  . . .
  E.A. Samuel
PULSE JET FILTRATION THEORY - A STATE-OF-THE-ART ASSESSMENT. . .    22
  R. Dennis, L.S. Hovis

LABORATORY TECHNIQUES FOR DEVELOPING PULSE JET COLLECTORS. ...    37
  R.R. Banks, J.T. Foster

OFF-LINE PULSE-JET CLEANING SYSTEM 	    48
  T.C. Sunter
          Section B - Fabric Filters:  Measurement Techniques


FIELD EVALUATION OF THE DRAG OF INDIVIDUAL FILTER BAGS	    62
  W.T. Grubb, R.R. Banks

A DUAL-DETECTOR BETA-PARTICLE BACKSCATTER GAUGE FOR MEASURING
DUST CAKE THICKNESS ON OPERATING BAG FILTER AND ESP UNITS. ...    77
  R.P. Gardner, R.P. Donovan, L.S. Hovis

MIT FLEX ENDURANCE TESTS AT ELEVATED TEMPERATURE 	    91
  J.T. Foster, W.T. Grubb

THE ONE-POINT IN-SITU CALIBRATION METHOD FOR USING A BETA-
PARTICLE BACKSCATTER GAUGE FOR CONTINUOUSLY MEASURING DUST
CAKE THICKNESS ON OPERATING BAG FILTER AND ESP UNITS	   107
  R.P. Gardner, R.P. Donovan, L.S. Hovis


          Section C - Fabric Filters:  Coal Fired Boilers
PULSE-JET FABRIC FILTER EXPERIENCE USING NON-GLASS
MEDIA AT AIR TO CLOTH RATIOS OF 5 TO 1 ON A PULVERIZED
COAL FIRED BOILER	   121
  G. Pearson, D.D. Capps

START-UP AND OPERATION OF A FABRIC FILTER CONTROLLING
PARTICULATE EMISSIONS FROM A 250 MW PULVERIZED COAL-FIRED
BOILER	   132
  C.B. Barranger, N. Spence, J. Saibini
                                 ix

-------
VOLUME I CONTENTS  (Cont.)

PERFORMANCE OF A 10 MW FABRIC FILTER PILOT PLANT AND
COMPARISON TO FULL-SCALE UNITS  	 ,  	   148
  W.B. Smith, K.M. Gushing, R.C. Carr

THE DESIGN, INSTALLATION, AND INITIAL OPERATION OF  THE W.H.
SAMMIS PLANT, UNIT 3 FABRIC FILTER	   164
  D.R. Ross, J.R.  Howard, R.M. Golightley

RESULTS FROM THE FABRIC FILTER EVALUATION PROGRAM AT
COYOTE UNIT #1	   179
  H.J. Peters, A.A. Reisinger, W.T. Grubb, M. Lewis

BAGHOUSE PERFORMANCE AND ASH CHARACTERIZATION AT THE
ARAPAHOE POWER STATION	   192
  R.S. Dahlin, D.R. Sears, G.P. Green

AN EVALUATION OF FULL-SCALE FABRIC  FILTERS ON UTILITY
BOILERS	   210
  J.W. Richardson, J.D. McKenna, J.C. Mycock

STATUS OF SPS INVESTIGATION OF HARRINGTON STATION UNIT 2
FABRIC FILTER SYSTEM  	   226
  R. Chambers, D,  Harmon

UPDATE OF SPS PILOT BAGHOUSE OPERATION  	   239
  R. Chambers,  S.  Kunka, D. Harmon

THE USE OF SONIC AIR HORNS AS AN ASSIST TO REVERSE  AIR
CLEANING OF A FABRIC FILTER DUST COLLECTOR	   255
  A. Menard, R.M.  Richards
           Section D - Fabric Filters;  Electrostatic Enhancement
 ELECTROSTATIC STIMULATION OF REVERSE-AIR-CLEANED
 FABRIC FILTERS	    287
   D.A.  Furlong,  G.P.  Greiner, D.W. VanOsdell, L.S. Hov}s

 ELECTRICAL STIMULATION OF FABRIC FILTRATION: ENHANCEMENT BY
 PARTICLE PRECHARGING	.........    303
   G.E.R.  Lamb, R.  Jones,  W.  Lee

 ESFF AS A FIELD EFFECT	    316
   L.S.  Hovis, G.H.  Ramsey, R.P. Donovan

 ELECTRICAL ENHANCEMENT OF FABRIC FILTRATION:  PREGHARGING
 VS.  BAG ELECTRODES	  .  .  „    327
   R.P.  Donovan,  L.S.  Hovis,  G.H. Ramsey

 PERMEABILITY OF DUST  CAKES COLLECTED UNDER THE  INFLUENCE OF
 AN ELECTRIC FIELD	 «  •  •    342
   D.W.  VanOsdell,  R.P. Donovan, D.A.  Furlong, L.S. Hovis


                                  .x

-------
VOLUME I CONTENTS  (Cont.)

          Section E - Fabric Filters:  Practical Considerations
HIGH VELOCITY FABRIC FILTRATION FOR INDUSTRIAL COAL-FIRED
BOILERS	   357
  G.P. Greiner, S. Delaney, L.S. Hovis

OPTIMIZING THE LOCATION OF ANTI-COLLAPSE RINGS IN FABRIC
BAGS    	   382
  J. Musgrove

PULSE JET ON-LINE CLEANING FILTER FOR FLY ASH	   420
  W.G. Wellan

TOP INLET VERSUS BOTTOM INLET BAGHOUSE DESIGN  	   431
  R.M. Jensen

UPGRADE OF FLY ASH COLLECTION CAPABILITY AT THE CROMBY
STATION	   446
  T.J. Ingram, R.J. Biese, R.O. Jacob

HIGH SULFUR FUEL, FABRIC FILTER STARTUP EXPERIENCE 	   460
  P. Hanson, L. Adair, R.N. Roop, R.B. Moyer

FUNDAMENTAL STRATEGIES FOR CLEANING REVERSE AIR BAGHOUSES. ...   482
  M. Ketchuck, M.A. Walsh, O.F. Fortune,
  M.L. Miller, M. Whittlesey,
          Section F - Dry Scrubbers
DESIGN CONSIDERATIONS FOR BAGHOUSE - DRY S02 SCRUBBER
SYSTEMS	   494
  O.F. Fortune, R.L. Miller

RESULTS OF BAGHOUSE AND FABRIC TESTING AT RIVERSIDE  	   506
  H.W. Spencer III, Y.J. Chen, M.T. Quach

REACTIVITY OF FLY ASHES IN A SPRAY DRYER/FABRIC FILTER FGD
PILOT PLANT  . . ;	   521
  W.T. Davis, R.E. Pudelek, G.D. Reed
          Section G - Plenary Session
FABRIC FILTRATION - AS IT WAS, HAS BEEN, IS NOW
AND SHALL BE	   536
  E.R. Frederick

AUTHOR INDEX 	   551
                                 XI

-------
                              VOLUME III

                               CONTENTS

          Keynote address

PARTIOJLATE CONTROL TECHNOLOGY AND WHERE IT IS GOING	   XV
  K.E. Yeager


          Section A - Economic Comparisons
A COMPARISON OF A BAGHOUSE VS. ESP'S WITH AND WITHOUT GAS
CONDITIONING FOR LOW SULFUR COAL APPLICATIONS  .	     1
  W.H. Cole

APPLICATION OF THE BUBBLE CONCEPT TO FUEL BURNING SOURCES AT
A NAVAL INDUSTRIAL COMPLEX	,	    12
  C.S Thompson
          Section B - Mechanical Collectors
CYCLONE PERFORMANCE:  A COMPARISON OF THEORY WITH
EXPERIMENTS	    26
  J.A. Dirgo, D. Leith

HIGH FLOW CYCLONE DEVELOPMENT	    41
  W.B. Giles

CYCLONE SCALING EXPERIMENTS  	    53
  W.B. Giles

TEST METHODS AND EVALUATION OF MIST ELIMINATOR CARRYOVER ....    66
  V. Boscak, A. Demian
          Section C - Coal Characterization
FILTRATION CHARACTERISTICS OF FLY ASHES FROM VARIOUS COAL
PRODUCING REGIONS	    81
  J.A. Dirgo, M.A. Grant, R. Dennis, L.S. Hovis

FLY ASH FROM TEXAS LIGNITE AND WESTERN SUBBITUMINOUS COAL;
A COMPARATIVE CHARACTERIZATION	,	    97
  D.R. Sears, S.A. Benson, D.P. McCollor, S.J. Miller

USE OF FUEL DATABANKS FOR THE EFFECTIVE DESIGN OF STEAM
GENERATORS AND AQC EQUIPMENT 	 ......
  N.W. Frisch, T.P. Dorchak
                                xii

-------
VOLUME III  CONTENTS (Cont.)

          Section D - Inhalable Particulate Matter
 DEVELOPMENT OF INHALABLE PARTICULATE (IP)  EMISSION FACTORS ...    131
  D.L. Harmon

 INHALABLE PARTICULATE MATTER RESEARCH COMPLETED BY
 GCA/TECHNOLOGY DIVISION	    141
  S. Gronberg

 RESULTS OF  TESTING FOR INHALABLE PARTICULATE MATTER AT
 MIDWEST RESEARCH  INSTITUTE  	    154
  K. Wilcox, F. Bergman,  J. Kinsey,  T. Cuscino

 INHALABLE PARTICULATE EMISSION FACTORS TEST PROGRAMS  	    166
  J.W. Davison,

 CHARACTERIZATION  OF PARTICULATE EMISSION FACTORS FOR
 INDUSTRIAL  PAVED  AND UNPAVED ROADS  	    183
  C. Cowherd,  Jr., J.P. Reider, P.J. Englehart

 CONDENSIBLE EMISSIONS MEASUREMENTS  IN THE  INHALABLE
 PARTICULATE PROGRAM  	    198
  A.D. Williamson, J.D. McCain
          Section E - Advanced Energy Applications
GAS CLEANING AND ENERGY RECOVERY FOR PRESSURIZED FLUIDIZED
BED COMBUSTION	   211
  A. Brintanann, P.M. Kutemeyer

DEMONSTRATION OF THE FEASIBILITY OF A MAGNETICALLY
STABILIZED BED FOR THE REMOVAL OF PARTICULATE AND ALKALI ....   226
  L.P. Golan, J.L. Goodwin, E.S. Matulevicius

TEST RESULTS OF A HIGH TEMPERATURE, HIGH PRESSURE
ELECTROSTATIC PRECIPITATOR 	   241
  D. Rugg, G. Rinard, J. Armstrong, T. Yamamoto, M. Durham

COAL-ASH DEPOSITION IN A HIGH TEMPERATURE CYCLONE  	   256
  K.C. Tsao, A. Rehmat, D.M. Mason

DUST FILTRATION USING CERAMIC FIBER FILTER MEDIA — A STATE-
OF-THE-ART SUMMARY —	   271
  R. Chang, J. Sawyer, W. Kuby, M. Shackleton,
  O.J. Tassicker, S. Drenker

HIGH TEMPERATURE AND PRESSURE PARTICULATE FILTERS FOR FLUID
BED COMBUSTION	   282
  D.F. Ciliberti, T.E. Lippert, O.J. Tassicker, S. Drenker
                                xiii

-------
VOLUME III CONTENTS  (Oont.)

MOVING BED-CERAMIC FILTER FOR HIGH EFFICIENCY PARTICULATE
AND ALKALI VAPOR REMOVAL AT HIGH TEMPERATURE AND PRESSURE   ...   300
  D. Stelman, A.L. Kohl, C.A. Trilling

TESTING AND VERIFICATION OF GRANULAR BED FILTERS FOR REMOVAL
OF PARTICULATES AND ALKALIS	   318
  T.E. Lippert, D.F. Ciliberti, R. O'Rourke

BAGHOUSE OPERATION IN GEORGETOWN UNIVERSITY COAL-FIRED,
FLUIDIZED-BED BOILER PLANT, WASHINGTON, D.C.	   335
  V. Buck, D. Suhre
          Section F - Novel Devices
PARTICLE CAPTURE MECHANISMS ON SINGLE FIBERS  IN THE PRESENCE
OF ELECTROSTATIC FIELDS	    347
   M.A. Ranade, F.L. Chen, D.S. Ensor, L.S. Hovis

PILOT DEMONSTRATION OF PARTICULATE REMOVAL USING A CHARGED
FILTER BED	    362
   P.H. Sorenson

PILOT DEMONSTRATION OF MAGNETIC FILTRATION WITH CONTINUOUS
MEDIA REGENERATION	    370
   C.E. Ball,  D.W. Coy
          Section G - Plenary Session
NOVEL PARTICULATE CONTROL TECHNOLOGY  	    386
   S. Masuda

AUTHOR INDEX	    406
                                xiv

-------
                 MODELING OF WET BOTTOM AGITATOR SYSTEMS FOR
               ELECTROSTATIC PRECIPITATORS ON RECOVERY BOILERS

                by:  Michael A. Sandell
                     Robert R. Crynack
                     Air Pollution Control Division
                     Wheelabrator-Frye Inc.
                     5100 Casteel Drive
                     Coraopolis, Pennsylvania  15108
                                   ABSTRACT
     Wet bottom precipitator designs are becoming more common on black liquor
recovery boiler applications.  A black liquor-filled pan under the precipi-
tator mixes the collected salt cake as it is circulated with rotating
agitators.  Undesirable buildup can result due to improper mixing and poor
flow patterns.  In order to verify that the design of a complex wet bottom
agitator system is adequate, three dimensional scale modeling should be
employed.  The paper presents the results of several such model studies.
The results discussed include a) visualization of flow and mixing patterns,
b) identification of poor mixing areas, c) specification of location and
size of baffles to improve areas of poor mixing, d) determination of agita-
tor speed and direction of rotation, e) evaluation of paddle design and
orientation, and f) location of feed and drain pipes.  The theory of mod-
eling, which considers the temperature and viscosity of the black liquor and
which establishes the scaling factors and model fluid parameters, is also
discussed.  This modeling technique provides a valuable tool for users,
consultants, and manufacturers to minimize on-line problems.


                                BACKGROUND
     Precipitators have been used in the pulp and paper industry on recovery
boiler applications since the late 1930's.  Initially dust removal was
accomplished with pyramidal hoppers.  Because of the hydroscopic nature of
salt cake, extensive heat tracing and insulation were necessary for satis-
factory dust removal.  To overcome this, drag conveyor dust removal systems
were developed.  With drag conveyors, the salt cake is collected in a dry pan
beneath the precipitator collection zone and then removed from the pan by a
scraping mechanism.  This type of system had high maintenance requirements
because of the many moving mechanical components.  To overcome some of the

-------
problems with drag conveyors, the wet bottom, design evolved and came, into use
on direct contact evaporator (DCE) recovery boiler precipitators around
1948.  (1)

     During the '60's and early '70's when environmental considerations
increased, the recovery process was scrutinized for ways of reducing malodor-
ous emissions.  The principal source of these emissions was the contact of
the black liquor and flue gas in the DCE.   One technology response was the
non-contact evaporator (NCE) and dry bottom precipitator (ESP).  The NCE
boiler and dry bottom ESP began commercial operation in the USA in 1969.
Many of the first NCE recovery units and their dry bottom precipitators had
problems requiring rebuilding and had long down times because of the precipi-
tator.

     Because of these early dry bottom problems, the first NCE recovery wet
bottom precipitator came on line in 1972.   In 1974 the second NCE wet bottom
unit came on line.  Limited testing has indicated that odorous emissions from
these wet bottoms are negligible.  By 1978 an industry survey showed that
wet bottom ESP's comprised 25% of new NCE's installed.

     The design of a wet bottom generally calls for a pan filled with black
liquor under the entire collecting surface of the ESP.  The pan, with rotary
agitators, mixes the collected dust and the black liquor.  This wet bottom
concept is easily integrated into the black liquor recovery process. (2)
The black liquor which has been circulated through the pan is fired in the
recovery boiler.  The salt cake collected by the ESP which has been mixed
with the liquor in the wet bottom is eventually returned to the boiler.

     The high solids concentration of the black liquor and inadequate mixing
of the salt cake and black liquor have caused salt cake build-ups in several
wet bottom agitator pans.  These buildups can reach to the collecting area
of the precipitator and short out the electrical fields.  The shorting of
several fields causes emission levels to increase.  This can create a mainte-
nance headache for the operating company who has to enter the precipitator
and remove the salt cake buildup.  The purpose of this paper is to discuss
model studies performed to increase the degree of confidence that the full-
scale agitator systems will operate satisfactorily with little or no build-
up problems.

                                  THEORY
     For complete similarity between a model and the full-scale system to
exist, it is necessary to have geometric, kinematic and dynamic similarity.
Geometric similarity exists when the ratios of all corresponding dimensions
of the model and the full-scale system are equal.  Kinematic similarity
exists when the streamlines are geometrically similar.  Dynamic similar-
ity exists when the ratios of all forces are the same.  There are five
dimensionless parameters which define the fluid dynamics of a system.
These are the Reynolds, Froude, Weber, Euler and Mach numbers.  For dynamic
similitude to be maintained, these five dimensionless parameters for the

-------
model must equal the five dimensionless parameters  of  the  full  scale  system.
In actuality, it is not practical to maintain the equality of all five
parameters.  Therefore, the model must at least  consider those  dimensionless
parameters which are most important to the system being modeled.

     For modeling an agitator type system, the most important parameters
are those associated with the Reynolds number and the  Froude number.  The
Reynolds number is important because it defines  the relationship between size,
viscosity and velocity.  The Froude number defines  the relationship between
velocity and gravitational acceleration.   The Mach  number  which defines com-
pressibility, the Weber number which defines surface tension and the  Euler
number which defines pressure were all judged to be of minor significance
in these model studies.

     The Reynolds number is the ratio of the inertial  forces to the frictional
forces usually in terms of convenient flow and geometrical parameters.  It
is given by the equation

                       Re -
                              \i      V

where p - density, L - length, V = velocity, y = absolute  viscosity and
v « Kinematic  viscosity.  The Froude number is  the ratio  of the inertial
forces to the force of gravity.  Given a free surface, the form of the
waves will be directly affected by the force of  gravity and so  the Froude
number is significant.  It is given by the equation

                             j.£'
                                 LG

where V = velocity, L - length, and G = gravitational  acceleration.   By
equating the Reynolds and Froude numbers of the  model  and  the full scale
system, scaling factors between the two systems  can be determined.  The
length scaling factor and velocity scaling factor is determined to be
               L2     \ *2J             V2

As can be seen both the length and velocity scaling factors  are dependent
upon the ratio of the kinematic viscosities of the fluids used  in the model
and full scale systems.

                                 MODEL DESIGN
     The objective in modeling a wet bottom ESP is to provide information
which can be used to increase the degree of confidence that the agitator
system for the low odor recovery boiler will function satisfactorily with
little or no particulate buildups.  The parameters of the full scale system
which are important in the model program are dependent upon the black liquor
properties and the physical design of the wet-bottom.  Those black liquor

-------
properties which are important  are  1)  viscosity of the black liquor,
2) solids concentration of the  Hack liquor, 3) liquor temperature, 4) liquor
feed rate.  The elements of the wet bottom design which are important are
1) pan size, 2) number of agitators and 3) agitator size.  The black liquor
characteristics should be given both for normal operating conditions and
worst-case conditions.

     The parameters of the full scale system set the limits and variables
for the wet bottom model.  The  length scaling factor should be restricted  to
the range of 6:1 to 12:1.  It is  felt that any model scaled butside this
range would either be too small or  too large to be easily constructed or
modified after construction.  The exact length scaling factor is  dependent
upon the ratio of the kinematic viscosity of the black liquor in  the full
scale system to the kinematic viscosity of the fluid used in the  model.  The
kinematic viscosity of black liquor is dependent upon many factors, with
temperature and concentration of  dissolved solids being two of the most
important.  (Figure 1) (3)
                          1000
                        LU
                           0.2

                           0.14
                              20  30  40  50 60 708090 ilC 'JO

                                  TEUPERATiJRE, °C
                               * Where DS is dissolved solids
                          FIGURE 1 KINEMATIC VISCOSITY OF BLACK LIQUOR

-------
     It has been our experience that the slow flow areas in the full scale
system have the greatest effect on the operation of the wet bottom system.
Slow flow areas allow the salt cake collected from the precipitator to drop
out of suspension and buildups to occur.  These salt cake buildups can
occur to the extent the precipitator fields are shorted and efficiency
of the precipitator reduced.  While it is not practical to simulate the
actual operation of the precipitator above the model, the relative motion
of the fluid can be observed.  Thus, the thrust of the model study is to
minimize the fluid residence time in the slow flow areas.

     The variables of the model which can be examined are 1) direction
of agitator rotation, 2) agitator rotational speed, 3) paddle porosity,
4) paddle angle from agitator arm, 5) synchronized vs non-synchronized
agitator rotation, 6) slow dispersion area baffling, 7) fluid viscosity,
8) liquor feed locations, 9) liquor feed rates, 10) normal and emergency
overflow locations and 11) liquor level.  For each variable specific items
were noted.  These included 1) identification of slow flow areas,  2) overall
flow patterns, 3) fluid swirling and eddying, and 4) fluid splashing.

                        MODEL EQUIPMENT AND PROCEDURES


     The results of three different model studies are presented in this paper.
The models varied in the number of agitator arms and black liquor viscosities.
The systems examined had 2, 3, and 6 agitators  The viscosities examined
ranged from 20 centistokes to 2,000 centistokes.  This corresponds to a dis-
solved solids concentration in the black liquor of 55% to 70% respectively.

     The agitator arms have been generally made from carbon steel rods on
which the metal paddles were attached.  The metal paddles were attached to
the agitator arms using screws and nuts.  This allowed the paddle angle to be
adjusted quickly and easily.

     The agitators were driven from above the liquid rather than below as in
the full scale system.  This eliminates the need for sealed bearings and
allowed maximum flexibility in the model.  The speed of agitator rotation
was adjustable from 1 RPM to the maximum.  The direction of rotation could
be reversed either by the drive mechanism or by a loop in the drive chains.
Synchronized vs non-synchronized rotation was examined using both duplicate
drive mechanisms and by offsetting the agitator arms 90  from one another.

     Several fluids were used in the models examined.  These included water,
ethylene glycol, and a glycerin and water mixture.  The selection of the
fluid was based upon its physical properties.  The kinematic viscosity of the
modeling fluid is the critical factor in the scale factor determination.
Other factors which were considered were availability, ease of cleaning, low
odor and low toxicity.  The added advantage to the selection of the glycerin
and water mixture was the ability to change the proportion of glycerin to
water and thus, change the kinematic viscosity.

-------
     Because it was not practical to simulate the actual operation of a
precipitator above the model, only the relative motion of the fluid could be
observed.  This was accomplished by dropping a given amount of dye into the
model fluid and observing the relative motion and speed with which it dis-
persed.  In order to compare the dispersion rate of the dye in one area with
respect to another, the rate of dispersion was timed.  The sides of the model
pan were labeled A, B, C, and D with the respective .corners being identified
a corner AB, corner BC, etc. (Figure 2)  Dispersion was defined as 100%
of the dye being removed from the immediate area.  Slow flow areas and over-
all flow patterns were examined in the greatest detail.

                                RESULTS

DIRECTION OF AGITATOR ROTATION
     The direction of agitator rotation is dependent upon the design of the
 full scale system.  System complexity varies with the number of agitator
 arms, the size of the arms, the size of the pan, and the viscosity of the
 fluid.  Because of this variation in complexity, each system must be ev-
 aluated individually.  When all the agitators of a given wet bottom
 rotate in the same direction, a fluid flow which rotates around the perimeter
 of the tank develops.  The fluid flow in the center of the pan, however,
 lacks the force of the main flow.  The result of this lack of force produces
 stagnant areas of swirling liquor between any two agitators.  (Figure 2)
     When two adjacent agitators rotate in different directions, the liquor
 is moved through the center of the two agitators.  However, a stagnant area
 now develops where the two agitators rotate toward each other.  (Figure 3)
 The direction of the rotation for each system must be selected  on the basis
 of the location of slow flow areas and the overall flow pattern of the
 black liquor for that system.

-------
     xxxx-yy  *  f    ////*/ / /  /    .'///*////
             *S  f  r r     f/fr"fff7     f 77  7  ~ / /  /  /
   ^ ^  ^ 7  *y  7 7  7^

                    \
     ////-////     f /  /
    r  7  7  7  ~7  7 7  7    ~7^
  FIGURE Z AGITATORS ROTATING IN THE SAME DIRECTION
FIGURE 3. AGITATORS ROTATING IN OPPOSITE DIRECTIONS

-------
AGITATOR ROTATIONAL SPEED

     While the rotational speed of the agitator arms is discussed, it is, in
reality, the agitator tip speed not the revolutions per minut'e which is the
major factor necessary to properly mix the collected salt cake with the black
liquor.  Results indicate that the tip speed of the agitator arms has a sig-
nificant effect on the dye dispersion rates and the overall mixing of the
model fluid.  Increased tip speed reduces the dispersion time measured in
each slow flow area.  Table 1 shows the effect of increased rotational
speed, i.e., tip speed, on the dye dispersion rates. It is assumed that the
increased arm speed imparts increased momentum to the viscous fluid, provid-
ing the force necessary to drive the fluid into the corners and slow flow
areas, thus, decreasing the dispersion times.

 TABLE 1.  COMPARISON OF DISPERSION TIMES FOR VARIOUS AGITATOR ARM SPEEDS
 Agitator Speed                           Dispersion time (sec)
      (RPM)                    AB        BC        CD        AD      Center
7
9
11
180+
40
30
180+
50
35
180+
60
50
180+
60
50
75
70
40
     It was also demonstrated in the model studies that the agitator tip
speed has an optimum speed.  At tip speeds greater than this optimum speed,
the agitator arm can cause excessive wave action and splashing.  This higher
speed may cause a salt cake buildup on the pan walls above the fluid level
or on the anti-sneakage baffles in the precipitator.  Proper sizing of the
agitator drive systems can considerably reduce equipment costs as well as
energy consumption.

PADDLE POROSITY

     The results of the model study indicate that the outermost paddle is the
most important with respect to salt cake buildup.  The interior paddles
should be porous to allow the fluid to pass through them.  This allows for
a more complete mixing of the salt cake and black liquor.  The outermost
paddle, however, should be less porous.  This less porous paddle has the
beneficial effect of driving the fluid outward from the agitator arm into
the slow flow areas.  The less porous paddle imparts a greater momentum to
the fluid flow and thus, decreases the size and dispersion times of the slow
flow areas.

PADDLE ANGLE

     Four different paddle angles were examined, 0  , 15 , 30 , 45° from the
center of the agitator arm, angled such that the liquor is pushed outward as
the agitator rotates.  The 30° angle of the paddle on the agitator arm gen-
erally provided lower dispersion rates in the corners than did a 0°, 15 , or

                                     8

-------
45  paddle angle.  Table 2 Illustrates these results.

      TABLE 2.  COMPARISON OF DISPERSION TIMES FOR VARIOUS PADDLE ANGLES
 Agitator Speed
   (RPM)
Corner
  Dispersion Time (sec)
0°     15°     30°     45°
9
9
9
9
9
11
11
11
11
11
AB
BC
CD
AD
Center
AB
BC
CD
AD
Center
60+
60+
60+
55
*
60+
60+
60+
55
35
60+
60+
60+
60+
36
56
60+
60+
60+
47
40
50
60
60
75
30
35
50
50
40
55
60+
60
180
30
180
180
180
180
20
*  No measurement taken

     Also examined was the effect of a paddle angled such that the liquor was
drawn inward toward the center of the agitator as it rotates.  Paddles angled
in this manner had higher residence times in the slow dispersion areas than
paddles angled outward.

SYNCHRONIZED VS NON SYNCHRONIZED AGITATOR ROTATION

     Investigations performed in the model studies indicated that the differ-
ences found between the synchronized and non-synchronized operations were
small.  The non-synchronized rotation, however, generally produced slightly
less dispersion times.  Normally in full scale systems the agitators are
operated independently of one another.  It is difficult, therefore, to syn-
chronize or non-synchronize the agitator operation at all times.  In fact,
the agitators generally are non-synchronized because of their independent
operation.  It was apparent, however that the optimum configuration in either
mode of operation (synchronized or non-synchronized) provides sufficient
fluid movement and mixing patterns.

SLOW DISPERSION AREA BAFFLING

     In all the model studies performed, with no baffles in position, areas
of slow flow and circular motion develop in all the pan corners. (Figure 4)
The dispersion times measured with no baffles in position were exceedingly
high in all cases.  It was clearly indicated that stationary baffles are nec-
essary in all corners and other areas of slow dye dispersion.  The effect of
stationary baffles on dispersion times is presented in Table 3.  A graphic
presentation of the effect of stationary baffles is presented in Figure 5.

-------
     y I  i i .  t I  II


                                             \
                                                   /I  t  X -. /  /  I	I
  FIGURE 4. FLOW PATTERNS W!TH NO BAFFLES
   / *  f f _  f



FIGURE 5.  FLOW PATTERNS WITH BAFFLES
                                   10

-------
       TABLE 3.   COMPARISON OF DISPERSION TIMES WITH AND WITHOUT BAFFLES

Corner
Baffles
No
No
No
Yes
Yes
Yes
Agitator Speed
(RPM)
7
9
11
7
9
11
Dispersion Time (sec)
AB
120
120
120
50
30
25
BC
120
120
120
60
35
45
CD
180+
180+
150
45
25
20
AD
180+
180+
150
45
25
20
     Examination of the location of stationary baffles indicates that baffles
should be placed so that no additional corners are created.   The high vis-
cosity of the black liquor prevents the fluid from sweeping  into and out of
the corners without the help of baffles to direct the fluid  direction.
Therefore, baffles should be placed away from the pan walls  and positioned
in such a way as to direct the black liquor to the slow flow area.

LIQUOR FEED LOCATIONS

     The placement of the inlet feed was investigated.  Three locations were
examined.  A feed located in the center of one of the short  sides and a feed
in either corner of the short side were investigated. (Figure 6)  In one
corner the feed entered from either the bottom of the pan or through the side
wall.  The feeds were positioned such that the inlet feed flow enhanced the
mixing and flow pattern of the fluid.

     At the lower viscosities, at each corner location, the  mixing and flow
enhancement caused by the inlet feed flow eliminated the need in that corner
for baffling.  At the higher viscosities (1,000 cs) a backflow area developed
upstream of the inlet feed.  A baffle was positioned in this corner to reduce
this backflow area and enhance the fluid pattern into and out of the corner.
In all feed placements the inlet feed flow had no effect on  the overall flow
and mixing patterns except in the location where the inlet jet stream
occurred.

     The inlet feed should be located under the inlet field  of the precipi-
tator and positioned so that the liquor flows across the inlet field.  This
placement allows the fresh black liquor to contact the greatest amount of
collected salt cake and allows a longer mixing time prior to firing.
                                    11

-------
      FIGURE 6. INLET FEED LOCATIONS

INLET FEED RATES

     Examination of  varying liquor feed rates  indicates that as the feed rate
increases, the possibility of a  slow flow area developing at that location
lessens.  The critical factor in feed rates is the lowest probable black
liquor flow.   The lowest feed rate expected may not provide the fluid flow
enhancement necessary to eliminate baffling in that location.   •

NORMAL AND EMERGENCY OVERFLOW LOCATIONS

     The location of the normal  and emergency  overflow pipes was examined
to determine whether they should be centered in the corner or offset either
upstream or downstream of the corner.  The overflow pipe was placed in
various positions to determine the optimum location.  Due to the close prox-
imity of the agitator arms, the  overflow pipe with its required baffling
could not be positioned upstream of the corner without interfering with the
agitator arm.  Table .4 presents  a comparison of the dispersion times measured
between the centered and the offset overflow pipes.  As can be seen, the off-
set and centered pipes had approximately the same  dispersion times with no
baffles in position.  Once the baffles were positioned, however, the offset
overflow pipe dispersion times were approximately 5 seconds less than the
centered pipe.
                                     12

-------
             TABLE 4.  COMPARISON OF DISPERSION TIMES (SEC)
                  FOR CENTERED VS OFFSET OVERFLOWS
        Position
Baffles
Dispersion Times (sec.)
Centered
Offset Downstream
Centered
Offset Downstream
No
No
Yes
Yes
27
28
24
19
     The normal and emergency overflow pipes should also be positioned
under the outlet fields of the precipitator.  This placement along with the
location of the liquor feed under the inlet fields permits the maximum cross-
circulation and mixing time between the black liquor and the collected salt
cake.

FLUID VISCOSITY

     Different model fluid viscosities were used in the model studies.  These
viscosities ranged from 20 cs to 2,000 cs in the full scale system.  The flow
patterns which developed with the more viscous fluid was the same pattern as
was present with the less viscous fluid.  The difference between two viscos-
ities is that there is less fluid eddying as the arm rotates with the higher
viscosity fluid.  It is also more difficult for the viscous fluid to change
direction.  This causes waves slapping against the side of the pan rather than
flowing into and out of a corner.  Because of the high viscosity the fluid
also did not mix as readily.  The higher viscosity fluid for the above reasons
produced longer dispersion times in the corners and other slow flow areas
even with baffling in place.

FLUID LEVEL

     The optimum fluid level appears to be at the top of the agitator arm.
If the fluid level is above the top of the arm, surface material is not
moved and dissolved.  If the fluid level is too low, there are no delet-
erious effects other than less than full utilization of the agitator
system.

     The fluid level must also be considered with respect to controlling gas
sneakage around the active collecting area of the precipitator.  Anti-
sneakage baffles have generally been installed in full scale systems such
that they extend down into the black liquor.  This gives an excellent gas
seal to eliminate sneakage below the precipitator.  However, this procedure
creates serious agitation problems.  When the fluid level is above the agita-
tor arms and paddles, effective mixing of the salt cake which falls to the
surface of the black liquor is not achieved.  By extending the baffles into
the liquid, the surface is divided into individual segments.  The salt cake
falling onto the surface between any two baffles remains in that area until
                                     13

-------
it is dissolved and mixed with the fluid below the baffle level.   For this
reason it is recommended that the anti-sneakage baffles remain above the
liquor level.  The sneakage area should be kept to a minimum.   This generates
a trade-off between the amount of gas sneakage with its corresponding loss
of efficiency and the improvement in agitation.

                                 CONCLUSIONS
     In order to examine the flow patterns of a wet-bottom ESP, considera-
tion was given to the performance of a model study.  At the time there was
concern that a laboratory scale model might not adequately simulate a full
scale system especially the use of a model fluid other than black liquor.
Investigation of the philosophy and theory of fluid flow modeling provided
confidence that such a model study could be performed and adequately
simulate the full scale system.  The models studied and their full-scale
counterparts have verified this confidence.

     Although it was not practical to reproduce particulate buildup conditions
in the model, an adequate simulation of the flow patterns was obtained.  Flow
patterns were observed in the model under a number of prototype parameters
and model variables.  Many of the flow patterns and dispersion times were as
expected, but several flow patterns were surprisingly different than would be
intuitively expected.

     The conclusions which can be reached upon examination of the model
studies performed are 1) the agitator arm tip speed rather a specific RPM is
of critical importance.  The optimum tip speed for each installation is de-
pendent upon a combination of factors such as liquor viscosity, pan size and
agitator size.  A tip speed in excess of the optimum tip speed can signifi-
cantly affect the component costs of the full scale system.  This excessive
tip speed can also cause undesirable wave action and splashing.  2) The outer-
most paddles on each agitator arm should be less porous than the inner pad-
dles.  The less porous paddle is necessary to impart the momentum necessary
to force the fluid into the corners of the mixing chamber. 3) The inner
paddles should be porous to allow greater mixing as the arm rotates and lesser
torque requirements.  4) The paddles should be angled approximately 30  from
the agitator arm such that the liquor is pushed outward as the agita-
tor rotates.  5) Stationary baffles are necessary to direct the fluid flow
into corners and other slow flow areas.  Baffles should not be positioned to
create additional corners in the mixing chamber.  6) The black liquor feed
should be located in a corner of the wet bottom.  The feed may enter from
either beneath the wet bottom or through the end wall of the casing.  The feed,
however, should be directed in such a way as to enhance the black liquor fluid
flow.  The feed should also be located under the inlet fields of the precipi-
tator.  7) The overflow pipes should be offset in a corner of the wet bottom
rather than centered in the corner.  The overflow pipes should be located
under the outlet fields of the precipitator.  This combined with the place-
ment of the feed under the inlet fields provides the maximum cross-circulation
of the black liquor.  8) The liquor.level should not extend above the top of
the agitator arms nor should anti-sneakage baffles extend down into the black


                                     14

-------
liquor.

     Wet bottom agitator systems cannot be perfectly modeled in the labora-
tory.  Therefore, several basic assumptions must be made in order to practi-
cally model the system.  Parameters such as black liquor viscosity were
approximated but in reality can vary significantly.  It should be noted that
the investigation and subsequent recommendations of a model study may not
eliminate mixing or buildup problems which can occur in the wet bottom.  It
is our opinion, however, that the model studies discussed in this paper ade-
quately simulate their full-scale counterparts and conclusions can be drawn
with a high degree of confidence.  With the possibility that larger evapora-
tors and recovery boilers will be employed by the pulp and paper industry,
it is our belief that complex wet bottom agitator systems should be modeled.

     The work described in this paper was not funded by the U.S. Environ-
mental Protection Agency and therefore the contents do not necessarily
reflect the views of the Agency and no official endorsement should be
inferred.
                                     15

-------
                                 REFERENCES
1.  Freyaldenhoven, Ray J. ,  Peaces, Michale M., and Henderson, J. S.
    The recovery boiler precipitator (A Special TAPPI Task Force Report)
    The NCE recovery precipitator wet bottom design and operating exper-
    ience.  Paper presented at the 1981 TAPPI Engineering Conference,
    Atlanta, Georgia.

2.  Balasic, Paul J. and Peaces, Michael M.  A closer look at electro-
    static precipitator wet bottoms.  TAPPI  Vol. 63, No. 6.  June 1980

3.  Environmental Pollution Control - Pulp and Paper Industry, Part 1 -
    Air. EPA-625/7-76-001.   U.S. Environmental Protection Agency,
    October 1980, Pg. 10-47.
                                    16

-------
    DESIGN AND PERFORMANCE OF ELECTROSTATIC PRECIPITATORS
       UTILIZING A NEW RIGID DISCHARGE ELECTRODE DESIGN

           by:  Gary R. Gawreluk
                Robert L. Bump
                Research-Cottrell, Inc.
                Somerville, New Jersey 08876
                          ABSTRACT

     A survey of recent electrostatic precipitator buying
practices indicates that rigid electrodes,as compared to
weighted wires and rigid frames,are becoming more and more
the electrode of choice for new equipment purchases.  This
paper presents some of the reasons behind this trend and
discusses the rigid electrode design of one manufacturer.
Additionally, the design and performance of three instal-
lations that are utilizing this rigid electrode are examined.
                             17

-------
                        INTRODUCTION

     Most people in our industry are familiar with the trend
that emerged in the United States in the early 1970's towards
more reliable discharge electrode systems for electrostatic
precipitators.  The trend towards heavier and mechanically
stronger discharge electrodes was triggered in part by the
increased use of lower sulfur fuels and in part by increased
regulatory pressures which dictated the need for more
reliable equipment.

     During this period, the Europeans were able to pene-
trate the U.S. market with a relatively well-established
product that was perceived as offering an improved measure
of reliability.  The product they offered was the rigid
frame electrode and the product displaced was the weighted
wire.

     What most may not realize is that now, in the early
1980's, a new trend is evident with regard" to discharge
electrodes that promises to tip the scale back once again
in favor of American technology.  This development is the
rigid electrode.

                         BACKGROUND


     Generically, and briefly, most precipitator discharge
electrodes may be classified in three categories (Figure 1):

     o  Weighted Wires consist of thin elements approxi-
        mately 0.1" in diameter which hang the full length
        of the collecting electrode, stabilized and plumbed
        with a heavy weight at the bottom.  Wire lengths
        generally do not exceed 36 feet.

     o  Rigid Frames also contain thin emitting elements,
        but limit the vertical span to roughly 6 feet and
        are tensioned in a structural framework.  The two
        most common rigid frame designs are the bedspring
        type and the mast type.

     o  Rigid Electrodes, on the other hand, combine
        mechanical integrity and stability with corona
        generating properties in the same member.  This
        results in an electrode that is easily fabricated,
        installed, and virtually impervious to electrical
        arc erosion or mechanical abuse.

     The ongoing trend that was previously mentioned is
illustrated in Figure 2.  This graph illustrates the per-
centage of power industry contract awards by electrode type

                             18

-------
over the past 5 years.  What is readily apparent from this
graph is that while the percentage of weighted wire contracts
has continually diminished since 1978, the rigid electrode
has emerged as a major factor in the marketplace.  From a
paltry 8% share of the contracts in 1978 the rigid electrodes
have succeeded in capturing 50% of all contracts for the
past two years running.

     A study of precipitator buying practices in other
market segments further supports this trend.  Figure 3
depicts the trend towards rigid electrodes that has been
observed in the pulp and paper industry, which represents
the second largest market for precipitators, while Figure 4
shows the trend for all miscellaneous industries combined,
including pulp and paper, petrochemical, cement and metal-
lurgical.

     Why is that rigid electrodes have been so successful in
such a short period of time?  The answer lies not only in
the inherent features of the electrode but in the design of
the precipitator systems of which they are a part.  As a
designer and supplier of rigid electrode precipitators,
Research-Cottrell has observed that the following design
features and performance benefits have contributed in large
part to its dramatic success in the marketplace:

     o  No More Electrode Breakage - Rigid electrodes,
        fabricated in many cases of heavier gauge steel
        than the collecting surfaces, have put an end
        to grounded out bus sections due to electrode
        burn-off.

     o  High Electrical Sectionalization Capability -
        The benefits of a high degree of electrical
        sectionalization in direction of gas flow has been
        well documented.  Rigid electrodes permit this to
        be done more easily and economically than with
        most rigid frames.  Some currently operating units,
        for example have up to a dozen electrical fields,
        each less than five feet in length.

     o  Rappers Out of Gas Stream - Since all rappers are
        roof mounted, they are not exposed to the corrosive
        and erosive effects of flue gas and particulate
        like the tumbling hammer, mechanical-type systems.
        Additionally, this location permits them to be
        inspected, and serviced if need be, while the
        precipitator is on line.

     o  Adjustable Rapping Intensity Capability -
        Magnetic impulse, gravity impact type rappers
        permit the adjustment of rapping intensity,

                              19

-------
        either individually or by group,  from a central
        control panel.   This permits on-line rapping
        optimization  for both low and high resistivity
        cases, providing full range adjustment of col-
        lecting surface accelerations from 0 to well over
        100  g's at the  least responsive points.

     o  Ease of Installation - Many rigid electrode precipi-
        tator designs incorporate a high degree of shop
        fabrication which minimizes field erection cost
        and  duration.  Among the items shop fabricated
        are  the rigid electrodes, collecting plates^ upper
        and  lower electrode support/alignment frames and
        rappers.  Additionally, in the Research-Cottrell
        design, electrode alignment is simplified since
        each electrode  automatically hangs plumb from a
        single bolt connection.  Moreover, an entire bus
        section can be  assembled outside the precipitator
        and  lifted into place.

            RIGID DISCHARGE ELECTRODE DEVELOPMENT

     Before  considering a few case histories, it might be
appropriate  to take a look at the basic precipitator design
as well as some of the  history behind its development.
                                              I'M
     Research-Cottrell  commercialized the Hi-R   rigid elec-
trode precipitator in 1978.  This culminated 6 years of
research and development on this design which included not
only the development  of a new rigid discharge electrode,
trade-named  the Dura-Trode  , but also a much improved col-
lecting plate.

     The main objectives of the Dura-Trode development pro-
gram were to develop  an electrode that was inherently rigid,
unbreakable, easy to  fabricate and handle, and would distri-
bute corona  evenly along its entire length.  The basic ele-
ments of the program  designed to meet these objectives
included:

     o  Corona field  studies in both a laboratory precipi-
        tator as well as in a large electrolytic tank
        capable of reproducing typical precipitator oper-
        ating voltages.

     o  Tooling and production studies at our electrode
        fabricating facility.

     o  Construction  feasibility studies performed by
        individuals responsible for erecting scores of
        precipitators.
                             20

-------
     o  Mechanical strength, electrical stability, and
        accelerated life testing performed in a full scale
        test tower.

     The last phase of study before going commercial involved
investigation of the actual electrode in service.  This was
accomplished in both pilot precipitator work as well as
through backfitting hundreds of the electrodes in full scale
operating units.

     The final result is a roll-formed, structural member
fabricated from heavy gauge steel  (Figure 5).  Corona sets
up at discrete points on the tips of the scallops, while the
flat portion of the electrode body serves to enhance the
field strength at the collecting surface.  Actual measure-
ments of field strength show that this additional mass pro-
vides approximately double the field strength of wire type
electrodes at equivalent distances from the electrode with
equivalent gas passage spacing and energization levels
(Figure 61,

     As expensive and time consuming as it is, a well-
developed, successful R&D effort is a necessary pre-requisite
for introducing a new product into the marketplace, particu-
larly one as traditionally conservative as the power gener-
ating industry.  Successful demonstration of that product in
full-scale commercial installations, however, is a necessity
for continued participation in that market.  The previous
discussion illustrates how this first objective was met.
The following case histories serve to support the second.

                       CASE HISTORIES
CASE HISTORY #1:  COMBINATION BOILER

     The first commercial Hi-R rigid electrode precipitator
went on line in late 1980 at a major pulp and paper complex
in the Southeast.  The application required the collection
of fly ash from a 500,000 Ib/hr power boiler designed to fire
pulverized coal from Virginia, West Virginia, and Eastern
Kentucky, or a combination of bark and coal  (Figure 7).
Design sulfur was 0.8%.  This application presented the po-
tential for both high and very low resistivity operation;
high with the coal-only case and low with the combination
bark firing.  Our previous experience with more than a half
dozen similar combination boiler applications dictated a
precipitator incorporating a low treatment velocity and
variable intensity rapping in order to maximize collection
and minimize rapping re-entrainment losses.
                             21

-------
     Guaranteed removal efficiency was 99.5% for the coal-
only case and 99.0% for the bark/coal combination.  Design
gas volume, at 380°F, was 257,000 ACFM for coal, but approx-
imately 20% higher for the combination.

     Performance tests run by an independent testing labora-
tory at full load conditions with the unit burning coal only
yielded collection efficiencies from 99.75% to 99.82%.  The
clear stack was indicative of an opacity level well below
the 20% guarantee level.  At present this unit continues to
perform with no visible emissions.

CASE HISTORY #2:  RECOVERY BOILER

     Another southeast paper mill was the site of a second
Hi-R rigid electrode precipitator which also went on line
in 1980.  That unit was designed to collect salt cake
 (.sodium sulfate) from a 158,000 Ib/hr low odor recovery
boiler  CPigure 8).  Salt cake is a relatively moderate re-
sistivity ash but, because of its sticky,tenacious nature,
requires a high level of rapping acceleration to dislodge
it from both discharge and collecting electrodes.  The rap-
ping system supplied can deliver in excess of 50 G's to all
points in the collecting surfaces.

     The precipitator designed for this application incor-
porated a heat jacket around the entire casing in order to
minimize corrosion potential, and a wet bottom ash collection
system with floors and sides fabricated of type 316 stain-
less steel.

     This unit, which was started up only 6 months after the
beginning of erection, was designed for an outlet loading of
0.0255 gr/scfd.  Design gas volume was 450,000 ACFM at 420 F.
Performance tests, showed that the Hi-R precipitator limited
emissions to 0.01 gr/scfd which was significantly better
than the guarantee level.  This was equivalent to a col-
'lection efficiency of roughly 99.7%.  The precipitator also
produced a visibly clear stack.

CASE HISTORY #3:  UTILITY BOILER

     Introduction of Research-Cottrell's Dura-Trode rigid
electrode into the power industry came in 1979 with an award
for four Hi-R precipitators to serve 1150 MW of generating
capacity at New England Electric*s Brayton Point Station.
These precipitators are a part of the largest single coal
conversion project in the U.S. to date.

     The application involved collection of fly ash from
three pulverized coal-fired boilers burning Eastern bitum-
inous coal with sulfur content ranging from 0.8 to 1.5%

                              22

-------
(Figure 9).   Design gas volumes from these two 250MW units
and a 650MW unit ranged from 856,000 ACFM for each of the
smaller units to 2,210,000 ACFM for the larger unit.  The
new precipitators were located in series behind existing
precipitators built in the early 1960's that had treatment
velocities of 6 to 8 feet per second.

     Research-Cottrell's scope of supply on this project
included not only the precipitators but also the flues from
air heater to stack, heat insulation, field wiring, control
houses, access facilities, and model study, as well as
installation of all material supplied.

     The first unit was brought on line in March, 1981, the
second in June, and the third in January, 1982.  Although
outlet emissions for state compliance were 0.08 lb/10  BTU,
contract levels were set at 0.019 gr/acf, approximately
equal to 0.06 lb/106 BTU, and had to be met with one entire
field across the precipitator de-energized.

     Performance testing on the first two units has proven
that the precipitators are operating well within guarantee
with average emissions less than 0.02 lb/10  BTU and test
opacities at 0%.  Although test results on Unit #3 are not
yet available, the clear stack gives every indication that
compliance on that unit will also be achieved.

                         CONCLUSION
     The previous case histories illustrate only a few of
the many diverse applications for which contracts have thus
far been received for Dura-Trode rigid electrodes (Figure
10}.  Other applications include cyclone fired as well as
pulverized coal fired power boilers, burning coal with
sulfur contents ranging from 0.5 to 4%, as well as appli-
cations in the petroleum, cement and other miscellaneous
industries.  All told more than 32 precipitators with Dura-
Trodes have been sold to date.  In the power industry alone,
this represented by more than 3500 megawatts of generating
capacity.

     More than a decade ago, United States industry was
offered what was perceived to be a more reliable alternative
to the weighted wire design.  The ensuing years found rigid
frame type precipitators being more and more widely accepted.
But the advent of the American design rigid electrodes has
ended their corner on the market.  If recent successes in
both sales and demonstrated performance are any indication,
these rigid electrodes will more and more become the design
of choice in future precipitator purchases.
                              23

-------
     The work described in this paper was not funded by
the U.S. Environmental Protection Agency and therefore the
contents do not necessarily reflect the views of the Agency
and no official endorsement should be inferred.
                              24

-------
                                     Figure 1
                       Discharge  Electrode Types
to
01
                       Rigid frame
                       (bedspring)
 Rigid frame
(strung mast)
Rigid electrode
(Dura-Trode™)
  Weighted wire
   (shrouded)

-------
      100
      80-
to
5  60-
    
-------
                                    Figure 3
                         Major ESP Awards (%)
                               By Electrode Type
                              (U.S. Pulp & Paper)
     100
                                       Rigid electrode
      80-
ro
60-
    0)
    Q.
      40-
       Rigid frame
       (bedspring)
                    Weighted wire
      20-
        1978
                  1979
   1980
Year (Nov. — Oct.)
1981
1982

-------
to
00
      100-
      80-
60-
      40-
      20-
       1978
                                     Figure 4
                          Major ESP Awards (%)
                                By Electrode Type
                              (U.S. General Industry)
                                         Rigid electrode
                                              Rigid frame
                                        (bedspring & strung mast)
                        Weighted wire
                 1979
     1980
Year  (Nov. — Oct.)
1981
1982

-------
                  Figure 5
      Rigid Discharge Electrode
Bolted
Connection
                       Main Structural Member

                       Corona Generating Vanes
                     29

-------
              Figure 6

Electrostatic Field Plots

           .109 Wires
          Dura-Trode
         Rigid Discharge Electrode
  Note: Numbers designate percent field strength.
                 30

-------
U>
I-1
                            Figure 7
                      Case History #1
                   Combination Boiler
           APPLICATION:  P-C fired boiler burning
                        E. bituminous coal (0.8 % sulfur)
                        or coal/bark combination

Gas volume (ACFM)
Gas temperature (°F)
Inlet loading (Gr/ACF)
Outlet loading (Gr/ACF)
Efficiency (%)
DESIGN*
257,000
380
2.0
0.01
99.5
ACTUAL*
295,455
405
2.7
0.006
99.78
   *Coal only

-------
                 Figure 8
          Case History #2
          Recovery  Boiler
APPLICATION: Salt cake collection from a low odor,
            black liquor recovery boiler

Gas volume (ACFM)
Gas temperature (°F)
Inlet loading (Gr/SCFD)
Outlet loading (Gr/SCFD)
DESIGN
450,000
420
7.5
0.0255
ACTUAL
423,885
380
2.7
0.01

-------
                               Figure 9
                        Case History #3
                         Power Boilers
            APPLICATION:
Two 250 MW and one 650 MW
P-C fired boilers burning
E. bituminous coal (0.8 — 1.5% S)
U)
u>
          DESIGN
ACTUAL
Gas volume (ACFM)
Unit #1
Unit #2
Gas temperature (°F)
Unit #1
Unit #2
Outlet loading (lb./106 BTU)
Unit #1
Unit #2
Opacity (%)
Unit #1
Unit #2
856,000
856,000
260
260
0.06
0.06
20
20
1,031,000
1,007,000
295
275
0.019
0.017
0
0

-------
                Figure 10
    Hi-R™ Precipitators with
Dura-Trode™ Rigid Electrodes
APPLICATIONS:
       Pulverized coal fired boiler
       Cyclone fired boiler
       Oil fired boiler
       Combination coal/bark/oil fired boiler
       Low odor recovery boiler
       Conventional recovery boiler
       Fluidized catalytic cracking unit
       Cement kiln
       Alumina calciner
       Lead glass furnace
Total precipitators        —   32

Total megawatt capacity   — 3,790

-------
                        DEVELOPMENT AND EVALUATION
                 OF A NEW PRECIPITATOR EMITTER ELECTRODE

              by:   R. Adams and P. Gelfand
                   Air Correction Divison, UOP Inc.
                   Norwalk, Connecticut 06856
                                   ABSTRACT
     This paper  reviews  the laboratory and field development, and evaluation asso-
ciated with  the  development,  of  a new  rigid  emitter.   The paper  discusses the
experimental procedures and establishes  the comparative criteria for the evaluation  of
different  electrodes.  Laboratory  studies were  conducted to measure the  current
density patterns  formed  by various electrode designs.   Final laboratory data are
presented comparing three different types of emitters, two rigid and one wire.  The
data result in a comparison of emitter performance on a theoretical basis.  Final field
data on a full-scale precipitator are presented which verify the laboratory results.

                                 INTRODUCTION

     As part of the development of a new electrostatic precipitator, "ADVANCOR,"
the Air Correction Division of UOP  embarked on a program of intensive investigation  of
precipitator electrodes.  The goals of the "ADVANCOR"  design  were to provide  an
increased  performance, highly reliable  precipitator.  It  was recognized that a key
element necessary to achieve this goal was the development  of a  new electrode that
had superior  electrical corona characteristics and  could be made virtually unbreakable.
The development  program involved  both laboratory and field tests and retrofitting of a
full-scale  precipitator with  new emitting electrode design.   These  efforts and the
results obtained will be discussed.

                    HISTORY OF PREOPrTATOR ELECTRODES

     There are fundamentally three generic types  of precipitator electrodes (Figure  1).
A brief description of each follows.
Weighted Wire

     The  weighted wire electrode is a free-hanging electrode suspended from the top
of a precipitator  and tensioned with sufficiently heavy weights at the bottom of each
electrode  to  prevent vibration caused by the electrical field or the gas flow.  Usually
these weights are guided by a grid framework held by insulators to the hopper or shell
to prevent swaying. In general,  the geometry of the electrode can take a variety  of
shapes—round, square, strip-like, star, barbed wire.


                                        35

-------
Rigid Frame

     The rigid frame electrode consists of a structural frame usually made with pipes
containing rows  of  emitting electrodes similar to those  used with  weighted  wire.
Usually, the structural  frame takes on  the form  of  a box  that  is cross-braced  for
stability on all sides. To prevent electrical  or mechanical vibration of the electrodes,
they are  made in  short  lengths (two  to three  meters) stretched, or  tensioned,  and
mounted between pipe cross members that are installed parallel to gas flow.

Rigid Electrode

     The rigid electrode is  a free-hanging, mechanically rigid assembly consisting of
flat or rolled  plates,  pipes, or masts, with discharge  points or projections mounted
thereon.  Rigid electrodes are suspended from the top of the precipitator and guided at
the bottom with  a structural framework.  Depending on the design, stabilizing  bracing
insulators may not be required.

                           ELECTRODE REQUIREMENTS

     The essential requirements of a precipitator discharge electrode are:

      o    Effective corona generation:   high  electric  fields for charging and col-
           lecting,  and  uniform  current  density at  the collecting electrode with low
           expenditures of energy

      o    Cleanability:   a mechanical  shape  that permits  efficient use  of  rapping
           energy to keep the electrode free of  dust build-up  that could reduce corona
           generation

      o    Good gas mixing  characteristics to help move the dust toward the collecting
           plates

      o    Relatively low-cost fabrication

      o    Ease  of installation and maintenance

      o    High  reliability:  should be virtually unbreakable

      There are many claims as to advantages and disadvantages of  the  great variety of
discharge electrodes in present use.

      The weighted-wire system, while offering good corona discharge,  is prone to wire
breakage due to electrical erosion (sparking) and  hopper overfilling.

      The rigid frame system, while it  prevents the shortcomings of the weighted-wire,
is frequently subject to misalignment due to thermal  distortion of the structural  frame.

      The  rigid electrodes are  free  to expand and resist breakage while at the same
time providing good corona discharge and lower power consumption.

      Air Correction, therefore, concentrated its design of a new electrode on the rigid
type.


                                         36

-------
                              LABORATORY TESTING

     Over a several year period, virtually all electrode shapes and designs were tested
and evaluated in our laboratory.  Special apparatus was  built to measure the current
distribution produced  by an  emitter (Figures 2  and 3).  This equipment permitted
mapping of the current density over the full surface of the plate (Figure 4).

     In addition to the  current density tests, air load tests were conducted to obtain
corona  discharge curves for  each  emitter.   Discharge curves were measured with  a
simulated high-resistivity dust layer.  Finally, the candidate electrodes were tested in
our laboratory under a simulated hot gas and dust condition (Figure 5).  This was done
using heated air and re-entrained flyash.

     The dust deposition pattern obtained for the mast electrode is shown in Figure 6.


     The deposition pattern resembles those reported by S. Self, et al, Ref.  Electro-
mechanics & Re-entrainment of Precipitated Ash. Dust densities appeared to vary over
the surface.  Coarse particles were near the  point discharges; fine  particles were
opposite the central mast. The fundamental criteria used  in evaluating these tests were
average electrical field,  specific current  density,  and good  unformity of  current
distribution on plate surface. Two "ADVANCOR" electrode candidates were selected, a
barbed plate design and a mast electrode (Figure 7).

     The large  diameter central mast (pipe)  has  the  advantage  from a gas flow
perspective of creating a turbulent wake just downstream  of the mast and in  the center
of the gas passage.  Wake helps drive dust particles toward the collecting electrodes,
thus adding collection of fines participate. The central mast construction also serves as
a conduit for mechanical energy helping to keep the emitting points clean and free of
build-ups.

     Laboratory tests with air  or hot gas and re-entrained flyash  are  excellent  to
preliminarily select or evaluate  emitting electrodes but must always be followed  by
field tests, in  situ, to confirm  the  effectiveness of an electrode  design  under real
operating conditions.

Criteria for Electrode Evaluation during Fieid Tests


     The next step in  this development program was to test the candidate  electrodes in
a full-scale operating precipitator.  The plan was to install a limited number of test
electrodes and  compare electrical  operation.   An evaluation criterion has  to  be
developed that would permit comparisons to be made with only electrical data available
at the  electrode  terminals.   To  develop  this  criterion, the  following  fundamental
considerations were employed.
                                         37

-------
                        FUNDAMENTAL CONSIDERATIONS

Force on a charged particle (F):

              F    =   qE

where:        q    =   charge
              E    =   electrical field strength

But from field charging theory:
           ••stat
where:
              k    =   relative dielectric constant
              a    =   radius of particle
             EQ    =   charging electric field
             e     =   8.84 x 10~12 permitivity of free space

              E    =   E_ the precipitating electric field

From Stokes law and Electric Force Balance:

              W   =
                                                                                 (1)
                                                                                 (2)
where:
                                                                                 (3)
              n
                         iran

                   =   viscosity of the field

                                 1  ,
              q   =   qstat
where:
and
              t    =   time of exposure
              T    =   time constant of charging

Let
Then

J
t
W2a
t
           •*stat
so that   W
           2a
                                  (sec)
                  =   current density
                      qstatx  z-EL
                              6 iran
                                       k-1
E0ED
  n
                                                                                 (5)



                                                                                 (6)

                                                                                 (7)



                                                                                 (8)


                                                                                 (9)
so that the precipitation rate W2& for a particle size is a function of E  E .
                                                                  o~p'
                                         38

-------
But, on an electrode terminal basis:

              E   oc Voltage applied/distrance between emitter and ground

but ED is related to current density by:
              E   =  /   20YL)   (i_(*f]   +   f  V0   }2                  (10)
                         4TreQk                      r In  b
                                                         a

in cylindrical geometry and similarly in parallel ducts for similar duct geometries.


             E    cc / j                                                       (11)

           W2a   =  Cj  / j EQ                                                 (12)


This relationship becomes the basis for a first order evaluation of emitter performance,
so that by measuring the terminal voltages and currents, comparison on the basis of
theoretical improvement can be made.


                                 FIELD TESTING

Site Selection Criteria

     To ensure that the candidate  electrodes would be tested  under representative
conditions, the following criteria were used:

     o     Existing ACD Precipitators

     o     Cooperative Client

     o     Convenient Location

     o     Flexible Operation

     o     Pulverized Coal Fired Boiler

     o     High Resistivity Flyash Application

     Mead Paper  Corporation, one of Air Correction's customers, had a fire occur in
one of the three  weighted-wire  precipitators built  by Air Correction in Lynchburg,
Virginia. Mead asked Air Correction  to rebuild this unit.  Air Correction embarked on a
two-phase program.  Phase I - place as much of the damaged precipitator in service as
was possible, then install, at full-scale, the Air Correction test electrode stations inside
a portion of the precipitator, run  the  tests, verify the results, select the electrode, and
then enter Phase n - rebuild the complete unit with the final electrode selection for full
characterization and testing (Figure 8).
                                        39

-------
Description of Mead Installation Boilers

      o    Three B<5cW "Sterling" p.c. fired boilers

      o    Boiler rating 90,000 Ibs. st./hr.

      o    Fuel:  Eastern bituminous coal, 12,500 Btu/lb, 0.8% sulfur

Mead Precipitators

      o    Three Design 24(666)26-1-3 (22,464 ft2 of plates)

      o    90,000 acfm at 450°F (each) SCA 249.6

      o    94.43% collection efficiency

      o    Follow existing cyclone collectors (98% to 99% overall efficiency)

      o    3.5 Ibs/MM Btu dust loading

      o    0.16 Ibs/MM Btu emission limit

                          EVALUATION TEST PROGRAM
Phase I

      The two candidate electrodes and a weighted-wire control were installed at Mead.
Figure 7 shows these emitters.

      o    Mast
      o    Barbed Plate
      o    Weighted- Wire (Control Electrode)

      Figure 9 shows  the  precipitator modifications and locations of our test stations.
Measurements were taken with these three stations and comparisons  made in accor-
dance  with the performance evaluation  criteria.  Results  are shown in Figure 10.  An
analysis of these data indicate that the barbed emitter failed to perform over a period
of time and that  it demonstrated a sensitivity to the  dust characteristics at full-scale
but that the mast emitter demonstrated operation well into the acceptable region of
performance.   Therefore,  the  mast  electrode  was selected  for Phase  II in  the
retrofitting of the precipitator.

Phase n
     Air Correction  replaced all damaged collecting electrodes and  retrofitted with
the new mast-type emitters.  Also  installed was a new emitter and  collecting plate
rapper system with electronic controls.  Then tests were conducted  to fully charac-
terize the performance of the rebuilt precipitator.  These measurements were made to
determine:   mass  emissions versus  power, resistivity, particle  size  dependent  effi-
ciencies, and mass performance of the rebuilt  precipitator,  as  well  as an  identical
adjacent weighted-wire precipitator (Unit #2).
                                        40

-------
                                PHASE H RESULTS

Operating Conditions

      Resistivity of the ash was  10   ohm cm at a gas temperature of 460°F. A typical
laboratory resistivity is shown in Figure 11. A good agreement with in-situ resistivity
measurement was obtained.  The particle inlet loading from  the mechanical collector
was approximately 0.4 gr/acf with a mean micron  diameter of 6  and a geometric
standard diameter  of 4.  SO3  measurement  taken using ASTM  Method D 3226-737
averaged  3.56 ppm of SO,;  typically,  1.15% sulfur coal was  being  burned during the
testing.  Boiler operation was held  at full load during most tests, except for one group
that was conducted at a low load.   The results of the mass emissions versus electrical
power density are shown  in Figure  12.  Mass performance tests were on Unit  //I with
the mast ("ADVANCOR") electrode  and Unit #2 with weighted wire.  The results shown
the performance  enhancement that was obtained with  the "ADVANCOR" design.  An
enhancement of 21% based on the increased W^ (modified Deutsch  omega) was obtained
when comparing  Unit //I  with Unit //2.  Similarly, terminal  electrical conditions for
Unit //I and Unit //2 as measured during the test, theoretically predicted an enhance-
ment of 22% based on the evaluation criteria discussed earlier.

      Interestingly, independent  of the field test, Mead flyash was introduced into the
hot gas laboratory apparatus equipped with the Mast electrode. The improvement was
measured at 27%.  All these  tests  strongly suggest that  the  "ADVANCOR" electrode
will provide increased collection efficiency.  More data on other installations  will be
needed to absolutely confirm this.

                                   SUMMARY

The objectives for ACD's new emitter have been achieved.

      1.    Effective  corona generation:  high electric fields for charging and collecting
           and uniform current  density at the collecting electrode  with low expendi-
           tures of energy.

      2.    Cleanability:   a mechanical  shape that permits  efficient  use  of  rapping
           energy to  keep the electrode free of dust build-up that  could reduce corona
           generation.

      3.    Good gas mixing characteristics to help move the dust toward the collecting
           plates.

      4.    Relatively low-cost fabrication.

      5.    Ease of installation and maintenance.

      6.    High reliability: should be virtually unbreakable.

     The  work described in this paper was not funded  by the U. S. Environmental
Protection Agency and therefore the contents  do not necessarily  reflect the views of
the Agency and no official endorsement should be inferred.
                                        41

-------
to
                        CLASSIFICATION OF ESP'S
                    BY DISCHARGE ELECTRODE DESIGN









.
.




.
























i
i
i
i
i
i
» i
*
i
|
i
i
i
i
i
i
i
i
i
i i
i
i i
i i
i
< <
i i
i i
i i
i i
1 i
i i
1 i
1 i
1 i
i i
' i
i ,
i (
1 ,
1 |
1 1
' 1
1 |
1 |
1 1
1 1
1 1
| 1
| 1
1 1
1 1
1 1
1 <
| 1
1 '
1 !
1 i
1 i
1 i
i i
i i
i i
i i
i i
i i
i i
i i
i i
i i
i i
i *
i *
i i
i '
i <
i *
i '
i '
i '
i '
^ RIGID RIGID
-^ FRAME ELECTRODE
              WEIGHTED
               WIRE
                        FIGURE 1

-------
10
              LABORATORY APPARATUS FOR
           MEASUREMENT OF CURRENT DENSITY

                    FIGURE 2
LABORATORY APPARATUS WTTH TEST ELECTRODE
             FIGURE 3

-------
CURRENT DENSITY PATTERN
  WEIGHTED WIRE DESIGN
        FIGURE 4
   ELECTRODE TEST APPARATUS

        FIGURE 5
            44

-------
DUST DEPOSITION PATTERN MAST ELECTRODE
             FIGURE 6
   FULL SCALE TEST PRECIPITATORS
            FIGURE 8

               45

-------
             BARBED PLATE
                  MAST

TWO PROTOTYPE EMITTERS THAT MET
     PRELIMINARY OBJECTIVES
             FIGURE 7
               46

-------
BARBED
 PLATE
WEIGHTED
  WIRE
                           SUPPORT
                           INSULATOR
                            (TYP.)
                            PRIOR TO
                         MODIFICATIONS
                          TRANSFORMER
                           RECTIFIER
                             (TYP.)
                             AFTER
                         MODIFICATIONS
                         SUPPORT INSULATORS
                         FOR TEST ELECTRODES
    PHASE I - EMITTER  EVALUATION
         TEST SET-UP AT MEAD

               FIGURE 9
                  47

-------
CO
z
UJ
Q


Z
UJ
oc
QC

o

I

z
g
I-
o
      COMPARATIVE BASIS OF

EMITTER PERFORMANCE EVALUATION


              START TEST (BARBED PLATE)


              ^  START TEST

                   (MAST)
                      ACCEPTABLE

                        REGION
 MECHANICAL

  CLEANING  «l
(BARBED PLATE)  |


UNACCEPTABLEJ\i

   REGION   j y


            11
END TEST

 (MAST)
    WEIGHTED WIRE

     CONTROL LINE)
             END TEST (BARBED PLATE)
     FUNCTION - VOLTAGE GRADIENT
              FIGURE 10



                 48

-------
*>.
vo
          O

          2
          X
          o
             10"



              4


              2
     4


     2


    10'



     4


     2
          >   4

          '    O
          CO

          CO
          UJ

          ec
1000
                     200
                      i
                              300
                                                 TEMPERATURE,°F
400
                     500      600     700

                    I  I  '  '  '  ' I '  ' '  ' I  '
                                         800   900
                   1  I  I
                   I   I   I   I
       VC+273  80  90  100
                            2§6
                             2.4
              i  i  i  i  i i  i  i i  1 1  1 1 1  1 1 1 MM linn
i   i  i   i  t  i  i  i  i  i  i  i i  i i i  1 1  1 1
l^^™^i^^"^M^HB^B^H^H^^^H^H|^B^ll^"^"|

     2.2   I     2.0    I     1.8 I       J
  2.2   I     2.0



TEMPERATURE, °C


    FIGURE 11
                                      li
                                      TT

                                      1.4>
         '16    '    1.41

300           400      500

-------
1JOO
                PERFORMANCE CURVE
            EMISSIONS VS. POWER DENSITY
                                   UNIT No. 2
                                    DOMAIN
                                     UNIT No. 1
0.01
               5.0    7.0    9.0    11    13    15
                POWER DENSITY (WATTS/FT2 )

                      FIGURE 12
                       50

-------
                           INDUSTRIAL APPLICATIONS
                                OF TWO STAGE
                     TUBULAR ELECTROSTATIC PRECIPITATORS

                 by:  Harish S. Surati and Michael R. Beltran
                          Beltran Associates, Inc.
                             Syosset, NY  11791
                                  ABSTRACT

     Two stage Tubular precipitators incorporate best features of both the
single stage and two stage type designs.  Extremely high collection effi-
ciency in sub-micron region makes these precipitators ideal choice for pro-
cesses having a very high concentration of sub-micron organic mist like
Retort Oil Shale or Coal/Wood gasification.  Design parameters, field test
data and operating data from installations on these processes are discussed
in the paper.

     Tubular precipitators have also found wide acceptance in ferrous and
non-ferrous metallurgical applications involving sub-micron emissions, most
notably from smelting, sintering, and scarfing operations.  Installations
on Electric Arc and Rotary Hearth Furnace exhausts are described.

     Several installation details from chemical processes like Sulfuric Acid
Manufacturing, Sulfonic Acid Production, Ammonia Scrubbing of S02> etc. are
covered in the paper.  Comparisons with single stage type precipitators are
made wherever applicable.
                                INTRODUCTION

     Although the two stage electrostatic precipitator was first conceived
and patented in 1910, until recently most of its use was confined to in-plant
air cleaning.  During the last two decades, plate type designs have been
modified and made suitable for industrial applications involving organic
emissions,, most notably in asphalt saturating, plastic curing, food proces-
sing, printing, textile finishing, and heat treating industries.  A detailed
discussion of plate type two stage precipitators in various industrial appli-
cations can be found in authors' other publications (1-3).  However, plate


                                     51

-------
type designs are inadequate where very high particulate loadings and/or ex-
tremely corrosive components are present in the exhaust stream.  Two stage
Tubular precipitators were developed for such applications.

     Two stage Tubular precipitators incorporate best features of both the
single stage and two stage type designs.  Unlike single stage precipitators,
two stage Tubular precipitators have a passive electric field (no corona dis-
charge) in the collecting section.  Thus, very high electric field strengths
can be maintained and very high collection efficiencies can be achieved even
in the submicron range.  Unlike conventional plate type two stage designs,
Tubular precipitators have wider spacings between discharge and collecting
electrodes, thus allowing their use in applications involving very high par-
ticulate loadings.

     Two stage Tubular precipitators consist of a short ionizing section
followed by a comparatively very long collection section.  The discharge
electrode is in the form of a rod or tube with a series of sharp discharge
points at the end, and is centered in the collecting tube.  Various collec-
ting tube geometries have been utilized over the years, the most common being
the round and hexagonal.  The square configuration shown in Figure 1 is a
slight variation of the hexagonal shape and was chosen because of ease in
manufacturing.  The square tube geometry is much more space efficient than
the round shape.  Comprehensive treatment of theoretical and design para-
meters affecting two stage Tubular precipitators can be found in authors*
other publication (4),
                           INDUSTRIAL APPLICATIONS

     Two stage Tubular precipitators have been installed on a variety of
industrial emissions.  These can be broadly divided into five catagories.

     (1)  Alternate Energy Sources
     (2)  S02 Scrubbing Processes
     (3)  Primary and Secondary Non-Ferrous Metals Industry
     (4)  Steel Industry
     (5)  Chemical Industry

     All of these applications involve heavy concentration of submicron par-
ticles.  Submicron particles are formed by condensation phenomena or by gas
phase reaction, where the product of reaction has very low vapor pressure at
the reaction temperatures.

     Generally particles larger than 1 micron in diameter scatter light by
true reflection.  Thus the loss of light is proportional to the projected
surface area.  For submicron particles, the relationship is more complex,
since the particle diameter is comparable to or less than the wavelength of
visible light spectrum (.4 to .7 micron range).  Total number of particles
increases inversely as the cube of the diameter.  The total projected sur-
face area increases as the reciprocal of diameter.  Thus for the same con-
centration of particulates, viewed through the same linear distance, plume
density increases with decreasing particle size.


                                      52

-------
     In many applications described in this paper, serious corona quenching
(current suppression) situations are encountered.  When particle density
approaches ion density, corona quenching occurs.  When the particle concen-
tration 'is mainly in the submicron range, such conditions arise even with
particulate loading of a few tenths of a grain/CF.  The particles are charged
to the same polarity as the unipolar corona ions.  Thus, the electric field
gradient near the discharge electrode decreases, reducing the flow and mobil-
ity of ions.  Since for submicron particles, diffusion charging is the prin-
cipal means of particle charging, reduction in ion mobility affects particle
charging.  In two stage precipitators, higher ion densities are obtained than
is possible in conventional single stage units.  Corona current is a measure-
ment of rate of deposition of corona ions on the grounded electrode.  Since
the charged particle mobility is order of magnitude lower than the corona
ions, contribution of the charged particles depositing on grounded electrode
to the corona current is minimum.  This explains, in part, why even in heavy
current suppression situations where current is suppressed by a factor of 40
or more, very high precipitator efficiencies are possible.  To combat corona
suppression type situations, high intensity ionizers, positive corona, and
several electrical fields in series are generally required.

     The electrostatic force exerted on the particle depend on the electro-
static field, particle diameter, and dielectric constant.  The residence time
and the vertical distance needed for the particle to migrate from the dis-
charge electrode to the grounded electrodes increases with decrease in par-
ticle diameter and with lower dielectric constant (good electric insulators).
For these reasons condensed hydrocarbons (dielectric constant of 2-5 and sub-
micron size) have very low migration velocities.  Water droplets (dielectric
constant of 78) are thus very easy to collect compared to organic mist.

     In summary, heavy concentration of submicron particles 1.  create corona
suppression, 2.  decrease migration velocity, and 3.  create higher plume
densities.

ALTERNATE ENERGY SOURCES

Retort Oil Shale

     Oil shale is a fine grained, compact, sedimentary rock containing an
organic material called kerogen.  Heating the oil shale to about 900°F decom-
poses this material to produce oil shale.  Commercially, in situ combustion
and surface retorting methods have been tried.  Exhaust gases from oil shale
retort contain a considerable amount of water (40 to 50 grain), and hydro-
carbons (3-4 grain/SCF).  Some dust as shale fines, and carbon is also pre-
sent.  The carrier gas is combustible and has high heating value.  Very high
tar mist removal efficiencies are required.  Since hydrocarbon mist is formed
by condensation, the majority of particles are below one micron in size.
Low dielectric constant combined with very high loading of submicron parti-
cles create very tough conditions for efficient precipitation.  Two Tubular
precipitators in series operating 250 fpm throughput velocity have 99% plus
collection efficiency.  Since the carrier gas is combustible, every flange
connection is sealed tightly to .prevent inleakage of oxygen.  The method of
insulator purging is of paramount importance with inert gas and cleaned

                                     53

-------
process gas being used successfully.  Advanced designs incorporate insulating
oil type seal to eliminate purging requirement altogether.

Coal/Wood Gasification

     Coal gasification has received much attention in recent years.  Similar
processes have also been developed for wood gasification and municipal refuse
gasification.  Precipitators used in these processes are for gas clean up
rather than for pollution control.  Coal tar, fine unburned carbon, char ash,
and voltalized heavy hydrocarbons are present in the gasifier exhaust.  The
gas cleaning train generally consists of a cyclone or some mechanical pre-
filter after the gasifier, then a heat exchanger to recover heat, a medium or
low energy scrubber to cool the gas and remove heavy tar particles, and a wet
Tubular ESP to remove fine hydrocarbons.  The particulate loading after the
gasifier can be as high as 9-10 grain/CF of char (in case of wood gasifica-
tion) and almost as much of hydrocarbons.  The particle size distribution is
generally bimodal with the char and carbon particles :j.n 2 to 15 micron range
and the condensed hydrocarbons in the submicron range.  Outlet loading from
the precipitator of 0.003 grain/CF or less are required.  Presence of H2S,
ammonia and water creates fairly active corrosive conditions.  Generally
304 stainless steel construction is used.

S02 SCRUBBING PROCESSES

     Many processes generate S02 in concentrations that are too low to be
handled effectively in acid plants but are high enough to violate air quality
regulations.  Scrubbing of S02 gases is generally accomplished using one of
the following processes.

     1.  Lime/Limestone Slurry Systems
     2.  Soda Ash/Caustic Scrubbing
     3.  Double Alkli Processes
     4.  Ammonia Scrubbing

     Major industrial sources where these desulfurization techniques have
been utilized are:

     a)  Utility plants
     b)  Recovery boiler off gases in pulp and paper industry
     c)  Cogeneration using petroleum coke
     d)  Tail gases from some metal smelting operations
     e)  Tail gases from single absorption type acid plants.

     FGD systems for utility plants is a topic for a seperate paper by itself.
We will cover remaining four processes in this paper.

Paper Industry

     In paper making, large quantities of steam are required at three stages.
First for cooking the wood chips, then to seperate the individual cellulose
fibers from the binding material lignin, and finally for evaporation of water
in which cellulose fibers are dispersed for paper making.  Economic consid-

                                      54

-------
erations have led to development of some special type of boilers using pulp
mill by products.  These include sand impregnated salt water borne bark and
hogged wood fired boilers and recovery boilers using black kraft liquor and
spent sulfite liquor.

     In ammonia based acid sulfite pulping processes, recovery boiler off
gases are passed through an ammonia absorber.  In the absorber, Ammonium sul-
fite and Ammonium bisulfite are produced by the following reactions.

     2NH3 + H20 + S02 ^ (NH4>2SC>3  Ammonium Sulfite

     NH3 + H20 + S02 ^ Nlty HS03  Ammonium Bisulfite

     These reactions take place in the liquid phase and the ratio of the sul-
fite to bisulfite produced depends on the pH of the solution.  To maximize
absorption of S02 and to minimize ammonia partial pressures, the absorber is
operated at minimum practical temperatures.  Figure 2 and Figure 3 (5) show
the number size distribution and the mass size distribution of the particu-
lates in the Ammonia absorber off gases.  Very dense plume is observed at the
absorber stack.  The particle size is extremely fine, between 0.1 and 0.5
microns in diameter.  The particles are not completely soluble in water.  In
the past, FKP construction Brink type fiber bed filters have been used to
control these emissions.  However, pluggage and gradually increasing pres-
sure drop through the fiber bed plague these systems.  Moreover, when pH con-
trol is not very accurate, free ammonia is generated, causing corrosion of
the glass fiber.

     Tubular precipitators used on these applications are constructed of
fiberglass.  A specially conductive fiberglass resin was developed for this
application.  This eliminates the need for, and also the problems associated
with, maintaining water film on the collecting electrode.  The high voltage
discharge electrode is made of graphite and high moly stainless steel.  Car-
penter-20 discharge discs have also been used.  The particulate concentration
is generally in 0.4 to 0.6 grain/CF range for properly operated units.  In-
crease in absorber temperature or in pH of the scrubbing liquor can signifi-
cantly increase particulate concentration.  Two pass Tubular precipitators
have collection efficiencies in excess of 99 percent under entire range of
process conditions.  The collection efficiencies were measured using modi-
fied EPA method 5 and a forward light scattering photometer.  Excellent
agreement was found between these two measurements.  Heavy current suppres-
sion was encountered.  This is believed to have been caused by heavy concen-
tration of extremely fine particles and enormous moisture loading (saturated
stream at 160°F).

Cogeneration

     Cogeneration using petroleum coke feedstock has been investigated.  The
exhaust from tagentially fired pulverized coke boiler is first cleaned by a
Dry ESP.  Ammonia scrubbing or Double alkali processes are used for S02
scrubbing.   The presence of heavy metal impurities in the fuel act as a cat-
alyst to promote further oxidation of S02 to 803.  Tubular precipitators are
used to collect fine acid mist and fine sulfite-bisulfite emissions.  The S02

                                     55

-------
concentration has to be reduced to 5-10 ppm level to comply with opacity reg-
ulations.

Metal Smelting

     Primary and secondary metal production usually involves smelting of the
ore in a reduction furnace.  Sulfur oxides generated during this operation
are generally cleaned and taken to the acid plant if the concentration is
above 5 percent.  Weak gases are generally scrubbed using lime/limestone
slurry processes and emissions are collected using a baghouse.  However, in
some applications where caustic/soda ash or Ammonia scrubbing is used and/or
where appreciable quantity of acid mist is present, Tubular precipitators are
used to control emissions.  These precipitators are operated wet.  Particu-
late loadings of about 0.3 to 0.4 grain/CF are common and heavy moisture load
is usually present.  Corrosive conditions dictate use of plastic or special
alloy construction.

Acid Plants

     In older acid plants and in single absorption plants, 98% S02 conversion
efficiency can be achieved.  Tail gases from these plants require S02 scrub-
bing systems to comply with Federal regulations on allowable SC>2 emissions.
Ammonia scrubbing is widely used.  Tubular precipitators are used to control
acid mist and .sulfite-bisulfite emissions.

NON-FERROUS METALS INDUSTRY

     Two stage Tubular precipitators have been used in following Non-Ferrous
Metal operations.

     1.  Zirconium Calcining
     2.  Silver/Gold Refining
     3.  Molybdenum Roasting
     4.  Nickle Recovery using Electric Arc Furnace

Zirconium Calcining

     Zirconium and Hafnium are used in the nuclear industry for the fuel rod
casings.  These metals are immune to corrosion attack from most of the chem-
icals and can withstand very high temperatures.  Neutrons pass through Zir-
conium, whereas they are absorbed in Hafnium.  Thus the nuclear reaction can
be controlled by use of Zirconium and Hafnium tubes.  Zirconium and Hafnium
are mined as Zircon sand.  The ore is chlorinated, selectively precipitated,
passed through seperation operation, chlorinated again, and then reduced.
The exhaust from the calciner contains Zirconium oxide, Hafnium oxide, a
trace amount of elemental sulfur, some chlorides and sulfuric acid mist and
sulfur dioxide.  The exhaust is first treated in a caustic packed bed scrub-
ber.  The two pass Tubular precipitator made of FRP is used to remove fine
particulates and acid mist.  High moisture loading and high concentration
(.4 to ..5 grain/CF) of submicron particulates cause severe current suppres-
sion.  Two pass in series are effective in combating the suppression effects.
Collection efficiencies in 97 to 99 percent range are observed.

                                     56

-------
Silver/Gold Refining

     In the bisulfate slime fusion process, filtered slime obtained from the
tank house electrolytic refining operations is fused in two rotary batch
kilns.  The typical charge for the rotary kiln consists of about 46% by
weight slime, 41% by weight sulfuric acid and remaining sodium sulfate.  The
slime consists of appreciable quantities of Copper, Silver, and Selenium.
Silica, Lead, Tellurium, Arsenic, Gold, and other organic materials are also
present.  The fusion of slime in the rotary kiln is a batch process and re-
quires approximately six hours.  The fume laden gases pass through a primary
spray quench type scrubber followed by a venturi scrubber and finally through
a Tubular wet ESP.  The size of particles entering precipitator is thus in
submicron range.  Selenium in the crystalline form sticks tenaciously to
solid surfaces, so collecting tubes are continuously flushed.  Mild current
suppression conditions are present.  Generally the particulate loading is in
0.1 to .2 grain/CF but at times can go as high as .4 to .5 grain/CF.  Since
the exhaust volumes are small from these processes, a low throughput velocity
(high SCA) single pass unit is used for this application.  Collection effici-
encies of 98-99% are obtained on this application.

     Tubular precipitators have also been used on photographic film incinera-
tor exhausts for recovering silver halides.  Carbon and other finely divided
impurities are also present in the exhaust gases.  The gases are first passed
through a quencher/scrubber and then through the Tubular precipitator.  FRP
construction is used to prevent chloride corrosion attack.  Two passes are
used to obtain extremely high collection efficiency.

Molybdenum Roaster

     Molybdenum disulfide is oxidized in the multilevel hearth furnace to
Molybdenum trioxide and sulfur dioxide.  The exhaust is first passed through
a baghouse and then through a lead lined quench scrubber.  The exhaust stream
entering the precipitator contains organic compounds and sulfuric acid mist
with some oxides of Molybdenum, Selenium, Rhenium, and Mercury.  Chlorides
and trace amounts of fluorides are also present in the air stream.  Fiber-
glass reinforced plastic with synthetic veil on the inside surface is used
as a material of construction.  Particulate and acid mist loadings of .6 to
.7 grain/CF have been measured.  High SCA, two pass Tubular precipitator
system has collection efficiency in excess of 99 percent.  The cleaned gases
are then taken to an acid plant.

Nickle Recovery

     Variety of wastes are generated during steel making process.  These
wastes are generally contaminated with slag, oil, and water.  Wastes from
specialty steel making processes contain appreciable amounts of Nickle and
Chromium.  Baghouse dust, mill scale, and grinding swarf are blended with
crushed coal, powdered limestone, and high alloy grit.  The mixture is pel-
letized, dried, and then reduced in a rotary hearth furnace.  Nickle and Iron
oxides are completely reduced.  The hot reduced pellets are mixed with addi-
tional fluxes and coarse metallics for adjustment of slag and metal chemis-
try.  This is then fed into electric arc furnace where Chromium oxide is

                                     57

-------
reduced.  The exhaust from electric arc furnace is taken to a high pressure
drop (45" w.c.) venturi scrubber.  The submieron particulate emission  from
Venturi was still very high (0.1 to 0.2 grain/CF).  Apart from causing opa-
city problems, the particulate carryover was causing maintenance problems for
the high pressure fan.  Two pass Tubular precipitators installed on  this pro-
cess completely eliminated this problem and plume opacity is reduced to
almost  zero.  The unit is operated continuously wet to prevent accumulation
of lead and zinc on the collector plates.  Throughput velocity of 7  fps was
used for this application.  Current suppression was of order of two  to three-
fold.

STEEL INDUSTRY

     Typical applications where Tubular precipitators are used in the  steel
industry are:

     1.  Scarfing
     2.  Sintering
     3.  Coke Oven Exhaust

Scarfing

     Very fine Iron oxide particulates are created during scarfing opera-
tions.  Particulate loadings of 1 grain/CF and over are commonly encountered.
The particle size is mainly in submieron to 2 micron range.  The exhaust
stream  coming from scarfer is generally completely saturated.  Wet Tubular
precipitators are used in this application.  Collection efficiency of  99% and
higher  are  required to meet with opacity regulations.  ASTM 304 L stainless
steel construction is used.  The collected particulates are very easy  to wash
off.  Tubular precipitators used on this application have very high  collec-
tion efficiencies even at high throughput velocity, as can be seen from Fig-
ure 4.

Sintering

     Sintering is generally used to beneficiate ores by a high temperature
agglomeration process.  Sintering process transforms raw ore into a  product
which is uniform in size, has not many fines, is convenient to handle, and
has better  chemistry.  Particulate loading from these processes range  in 0.1
to 0.5  grain/CF.  Very high amount of condensable organic matter is  also pre-
sent (0.05  to 0.3 grain/CF).  Sintering machines using recycle draft and
strand  cooling have lower emissions and lower exhaust volumes.  Tubular units
using stainless steel construction have been used in this application.

Coke Oven Exhaust

     Exhausts from coke oven batteries are cooled in a quencher and  then
cleaned in  a Tar mist type Tubular precipitator.  The exhaust contains Tar,
fine unburned carbon, ash, etc.  The coke oven gas after being cleaned can
be used as  a fuel source.  Part of the cleaned gas is further cleaned  in a
Fuel gas precipitator to be used to fire coke oven batteries.  Since the
carrier gas is combustible, all the precautions required for oil shale pre-

                                     58

-------
cipitators are also applicable here.  Insulator purging is done using cleaned
gas from the Fuel gas precipitator.

CHEMICAL INDUSTRY

     The Tubular precipitators are used as acid mist precipitators in:

     1.  Sulfuric Acid Plants
     2.  Sulfonation Plants

Sulfuric Acid Plants

     Tubular precipitators have been used in metallurgical type acid plants
to clean up smelter gases before they can be taken to the acid plant.  Tubu-
lar precipitators are also used in sulfuric acid plants using H2S or spent
sulfuric acid as a feed material.  Lead and FRP construction have been used
in this service.  In some applications Hastealloy has been used.  Conductive
FRP construction, square tube two stage Tubular precipitators have several
advantages over conventional lead tube type units.  Both sides of the tubes
are used for collection so height of the unit is reduced by a factor of two.
The FRP housing covers entire unit, thus this type of precipitator can be
designed for more than 20 inches w.c. negative pressure (lead tubes are gen-
erally not designed for more than -20 inches w.c.).  The units are shop
assembled, thus significantly reducing installation costs.  Two stage, two
pass Tubular precipitators used on Copper smelter off gases and on Gold and
Arsenic roaster off gases have achieved 99.5% plus efficiency.  The normal
criteria used in the acid industry, to check the precipitator efficiency, is
to have have a 10 meter run of ductwork after the precipitator.  This length
of duct should be optically clear if the precipitator is performing satis-
factorily.

Sulfonation Plants

     Surfactants are organic compounds that have both a water soluble (hydro-
philic) and a water insoluble (hydrophobic) group.  The hydrophilic group for
the most commercially available anionic surfactants is either a sulfonate or
a sulfate.  The hydrophobic portion is generally a hydrocarbon (Cg-Cig) in a
straight or slightly branched chain.  Oleum is most frequently used for sul-
fonation reaction.  The reaction can be given by:

     RH + H2S04 "ZZ RS020H + H20

     Very dense white plum is generated during transfer of oleum to the stor-
age tank.  Fiber bed filters with absorption spray type devices are used for
control of these emissions.  The emissions from the sulfonation reactor have
also been handled using filters.  However, pluggage problems are encountered
during manufacture of some detergents.  Two stage wet type Tubular precipi-
tators have been used in this application quite successfully.  The acid mist
loadings are generally 0.25 to 0.3 grain/CF.  Very fine size distribution
(0.1 to 0.3 micron) is encountered.  Low throughput velocities are required
to achieve high collection efficiencies (99.5% plus).
                                     59

-------
     Tubular precipitators have also been used on the detergent spray tower
emissions.   Detergent slurry is sprayed from the top of the detergent spray
tower.   Hot gases are drawn countercurrently upwards.  The organic oils are
evaporated creating dense plume.   The exhaust also contains carryover deter-
gent fines and a high moisture load.  Wet two stage Tubular precipitators are
used to control opacity of the exhaust stream.
                                 CONCLUSION

     Two stage Tubular precipitators are ideally suited in applications in-
volving high concentrations of fine particulates or for control of organic
and acid mist.  Increased use of these units in difficult and corrosive appli-
cations in envisaged.
                                 REFERENCES

1.   Surati, H.S., "Two Stage Precipitator for Hydrocarbon Emission Control,"
     IEEE-IAS Annual Meeting, Chicago 1976, pp 340-345.

2.   Surati, H.S., and Beltran,  M.R., "Heat Recovery on Organic Electrostatic
     Precipitators," Annual Industrial Air Pollution Control Seminar, 1976
     pp 5-1 - 5-10.

3.   Beltran, M.R., "Smoke Abatement for the Carpet Industry," Carpet and Rug
     Industry, May 1973, pp 30-34.

4.   Surati, H.S., Beltran, M.R., and Raigorodsky,  I., "Tubular Electrostatic
     Precipitators of Two Stage Design," Iron and Steel Engineer, December
     1980, pp 32-36.

5.   Environmental Protection Technology Series,  "Ammonia Absorption/Ammonium
     Bisulfate Regeneration Pilot Plant for Flue Gas Desulfurization"
     EPA-600/2-77-149, August 1977,  pp 27-28.
                                     60

-------
   'A-A'
  /B-B/
SINGLE STAGE
                             r
                             B
                       COLLECTION
                       SECTION
                      IONIZATION
                      SECTION      \
TWO STAGE
    Figure 1.  Round and Square Electrode Geometry


                       61

-------
o:
LU
CD
  LU
^ N
5™ r i

CJ CO
LU LU
peri-
sh
I3O
Z •-•
  Q
     10?
0=1-
0
•z.
o
CJ
106
       0.003
   i	L
 0.01                           0.1


INDICATED  PARTICLE DIAMETER,/*
                                                              0.13
  Figure 2.  Number-size distribution of particulate in absorber plume.
                                    62

-------
                    PARTICLE DIAMETER,
Figure 3.  Mass-size distribution of particulate  in absorber plume.
                              63

-------
     100 -
u- o
LU Z
LU UJ
 I  -•
CD O
O «
rH U.
    _
   tu  95
 LU U
 D_ LU
      90
1
I
I
I
I
                100     200      300      400      500      600


                 PRECIPITATOR  THROUGHPUT VELOCITY.,  FPM
                                                   700
          Figure 4.   Precipitator Efficiency as a function of Throughput
                     Velocity for Scarfing fume
                                    64

-------
                PILOT DEMONSTRATION TWO-STAGE ESP TEST RESULTS

                    by:   P.  Vann  Bush and Duane H. Pontius
                          Southern Research Institute
                          Birmingham,  AL 35255-5305
                                    ABSTRACT

     Results of the  evaluation  of  the  performance of a large pilot scale two-
stage ESP under a  range  of  opeating conditions are presented in this paper.  A
three electrode precharger,  followed by  four collector sections with a total
SCA of 286 ft2/kacfm (56 m2/m3/s)  and  a  gas flow capacity of 30,000 acfm (850
mvmin)i was tested  under two ash  resistivity conditions.  Background informa-
tion on system operation since  start-up  in early 1981  as well as results from
recent tests are presented.

                                   DISCLAIMER

     The research  described in  this article has been funded wholly or in part
by the U.S. EPA under  Contract  68-02-2683  to Southern Research Institute, and
it has been subject  to the  Agency's required peer review and policy review.

                                 INTRODUCTION

     The two-stage concept  for  electrostatic precipitation of high resistivity
dust has been under  investigation  at Southern Research Institute since 1974
with the funding and general guidance  of the Particulate Technology Branch of
EPA's Industrial Environmental  Research  Laboratory (Research Triangle Park,
NC).  The majority of  effort has been  expended to develop and demonstrate a
charging stage.  The three-electrode charger has been the principal focus of
development since  early  1977 (1,2).  Less  attention has been given to the de-
sign of the collector  stage of  the two-stage system.

     Tests of a large-pilot demonstration  scale system incorporating a three-
electrode charger  and  a  four-section collector stage have been underway since
early 1981.  The system  has  a gas  volume capacity of 30,000 acfm and an aver-
age, as-tested specific  collection area  (SCA)  of 286 ft2/kacfm.  It is
installed at the TVA Bull Run Steam Plant.   Figure 1 is a photograph of the
pilot demonstration  facility.

     The test program  has three objectives:   1)  to demonstrate reliable, high
collection efficiency  in the two-stage system;  2) to demonstrate the charging
effectiveness of the precharger stage;   and 3)  to evaluate the feasibility of
retrofitting a conventional  ESP with a precharger stage.  All of the objectives
are to be achieved with  high resistivity dust.   The first two objectives have
been addressed.  Most  of the remainder of  the  test program will address the
third objective and  will be completed  by March 1983.

                                      65

-------
                                                                   4100-283
Figure 1.  Photograph of the pilot demonstration facility.
                          66

-------
                                  BACKGROUND

     The design and installation of the two-stage  system has been described
previously (3).  Without repeating the details,  Figure 2 shows the internal
arrangement of electrodes in the system.   The  four collector-stage electrical
sections are fitted with 3/8-in. diameter  rods for discharge electrodes.  This
design was selected in order to achieve the high electric fields at low current
densities which are desirable  for collector performance in two-stage systems.
The design has proven to be a  good choice  for  the  two-stage system.

     There were only minor problems associated with the start-up of the pilot
demonstration system.  A wiring problem in the precharger power supply and
modification of the ash disposal system caused some delay in achieving steady
operation.  After 1 month of operation it  was  determined that additonal reac-
tance was needed in the precharger power supplies  in order to achieve optimum
performance.  The power sets for the precharger  stage were of special design
not really adequate for field  service.  Nevertheless,  the addition of linear
reactors in the precharger stage power sets allowed operation over a broader
range of voltages.

INITIAL PERFORMANCE TEST

     The first thorough characterization of the  performance of the two-stage
system was conducted in June and July  1981.  The fly ash resistivity was meas-
ured to be 1-3 x 1011 ohm cm at the flue gas temperature of 250-260°F (120-
127°C).  Throughout the test period the collector  stages were operated at peak
voltage levels corresponding to 10-15 nA/cm2 current densities and 4 kV/cm
electric fields (50 kV applied voltage).   This voltage-current operating level
is evidence of the success of  the 3/8-in.  diameter rods in meeting the design
criteria for the collector stage.  The precharger  was operated with 30-40
nA/cm2 current density and 3.0-3.5 kV/cm electric  field.   The inlet mass load-
ing for the test was a high 6.12 gr/scf (14 g/m3).  Both Method 17 mass loading
measurements and cascade impactor measurements indicated an overall system
collection efficiency of 99.7%.  This is equivalent to 0.35 lb/106 Btu particu-
late emissions.

     There was no apparent performance benefit due to the precharger during
these tests.  This was in spite of particle charging effectiveness, as deter-
mined by charge/mass (Q/m) measurements, at levels expected from theoretical
predictions for the given particle size distribution and electrical conditions.
The mathematical model of ESP  performance  (4)  was  used to determine what the
two-stage system efficiency should be under the  operating circumstances with
precharger off and precharger  on.  Using the typical values for the contribu-
tion to the emissions due to nonideal effects  (gas velocity standard deviation
= 0.25 and gas sneakage = 10%), the model  predicts a collection efficiency of
99.656% without the precharger, and an efficiency  of 99.731% with the precharg-
er using an SCA of 266 ft2/kacfm.  This difference of 22% in penetration,
though substantial, was less than the scatter  in the data.
                                      67

-------
                                       FIELD 1
                                                                   FIELDS 2 &3
                                                                                              FIELD 4
oo
                    PRECHARGER
                    FIELD
      GAS FLOW
         , DISCHARGE RAPPING
          ARRANGEMENT
                                                               COLLECTOR GUIDES
                                                               & BAFFLES
BOTTOM
DISCHARGE
FRAME
                                                                                                                 COLLECTOR
                                                                                                                 ASSEMBLY
                                                                                                                     620-249
                                          Figure 2. Internal arrangement of the two-stage system.

-------
     The high efficiency  that  was  measured,  99.7% for 2 x 1011 ohm cm resistiv-
ity ash with an  SCA of  286  ft2/kacfm,  indicates excellent performance on the
part of the downstream  collector.   The benefit of using large diameter dis-
charge wires in  high resistivity ash collection needs further study.

     Information on particle size  distribution and system electrical character-
istics derived from the test data  and  electrical data from conventional ESP
tests was used in the mathematical  model  to  predict the efficiencies for the
precharger-collector system and a  conventional ESP at four values of SCA.  The
pilot system dimensions were used  for  all the conditions, and gas volume flow-
rate was varied  to change the  SCA.  The two-stage system model and the conven-
tional ESP model are plotted in Figure 3. The data points for the measured
efficiencies for the pilot  demonstration  ESP system and the full-scale Bull Run
Steam Plant ESP  are shown.  The data points  indicate the model predictions may
be used to approximate  the  difference  in  penetration between the two-stage
system and conventional ESP's  at a  given  SCA or to estimate the difference in
SCA required to  achieve a given emission  level.

SYSTEM AND OPERATING MODIFICATIONS

     Measurements made  during  the  initial test indicated that 50% of the outlet
emissions were due to rapping  reentrainment  losses.  The rapping cycle in the
outlet field was decreased  from 12  raps per  hour to 1  rap per hour.   Optical
particle counter measurements  showed that a  substantial reduction in outlet
emissions was achieved  as a result  of  this change.  These data are summarized
in Table 1.

                 TABLE 1.  RAPPING  CONTRIBUTIONS TO EMISSIONS
     Particle Diameter Range  (pm)
Percent Due to Rapping

0.464
1.285
2.631
4.990
10.009
>17.003

- 1.285
- 2.631
- 4.990
- 10.009
- 17.003

1 2 raps/hr
1.778
2.773
32.98
72.69
82.34
92.66
1 rap/hr
21.86
22.38
31.15
56.29
67.97
20.39

     After the July 1981 test period the pilot  demonstration system was operat-
ed continuously through October with no performance  degradation.   During Novem-
ber a problem in the Bull Run Steam plant heat  exchangers  caused  an approxi-
mately 50°F drop in the flue gas temperature  and  concomitant reduction in re-
sistivity.  At this condition the outlet emission level  was  reduced to 2.4 x
10—3 gr/scf (5.5 x 10—3 g/m3), less than half the level  measured  at 260°F.  The
                                      69

-------
          10-1
         10-2
       o
       III
       z
       UJ
       Q.
         10-3
         10-4
                                        PILOT DEMO ESP

                                        TEST
                                           BULL RUN ESP'
  TWO-STAGE

  MODEL

I	I      J
                                           CONVENTIONAL.
                                           ESP MODEL
I       I
                                 90.00
                                 99.00
                                 99.90
                                       O
                                       \L
                                       u.
                                       UJ
                                       I-

                                       UJ
                                       u
                                       tc
                                       UJ
                   100   200   300   400   500

                               SCA, ft2/kacfm
                        600    700
                                                       4100-17J
                                 99.99
Figure 3.  Model predictions for penetration as a function of specific collection area.
         The stars represent actual field data.
                               70

-------
plant remained in this condition  until  March 1982 at which time a major mainte-
nance outage occurred.

     The ESP mathematical model confirms  the desirability of another system
modification.  Figure 4  shows  a theoretical  prediction of the precharger bene-
fit to system performance as a function of fly ash resistivity.  Three curves
are shown, corresponding to ideal  collection and two degrees of nonidealities.
Variations in ash resistivity  are  reflected  in the model input data as varia-
tions in operating voltages and currents  (based on experimental data).  In
order to more clearly demonstrate  the precharger effect on system performance
it is desirable to operate at  higher ash  resistivity than is available at
260°F.  The resistivity of the ash at Bull Run Steam Plant is shown as a func-
tion of temperature  in Figure  5.   The peak resistivity occurs at about 325°F
(163°C).  In order to boost the flue gas  temperature in the pilot demonstra-
tion system, some ductwork was added to the  system during the plant outage
which permits mixing hot gas from upstream of the plant's heat exchangers with
the main slipstream  from the cold-side  duct.  This addition provides an operat-
ing range of 250 to  350 °F.  The measured  peak in resistivity is 4 x 10^ ohm cm
at 325°F.

     One other modification was made to the  pilot ESP system during the plant
outage.  The power supply system  for the  precharger stage was replaced with
commercial-grade transformer/rectifier  sets  and controllers.  This was the last
component in the system to be  developed to the industrial standard.

                                  RECENT TESTS

     After completing the system modifications and allowing a period of stabi-
lization it was decided to repeat  the tests  at 260°F before testing at the
maximum resistivity condition.  A  new measurement tool was used to set up oper-
ating conditions for these tests.  A continuous charge probe derived from the
design used by EPA personnel in in-house  experiments was used to determine the
optimum charging condition for the precharger with 260°F flue gas.  The current
due to charged particles impacting the  probe sensor (shown in Figure 6) was
measured with an electrometer.  A  maximum reading was measured when the pre-
charger was in the following operating  condition:  corona voltage = 17-20 kV,
corona current = 20-25 mA, grid voltage = 2-4 kV, and grid current = 40 mA.  At
this condition, there was steady sparking in the precharger.  Reducing the grid
voltage until sparking diminished  produced the next highest probe reading.
Essentially, the only change in V-I conditions was a change in grid current to
0 mA and less than 1 kV measurable decrease  in grid voltage.  Q/m measurements
were made at these two conditions.  The average Q/m measured for the first
condition was -1.45 UC/g with  a maximum of -2.01 yC/g.  The average Q/m value
at the latter condition was -1.05  yC/g  with  a maximum of -1.06 yC/g.  The
charge probe does indicate charging performance in this application and is
therefore a valuable aid in optimizing  the precharger.

EVALUATION OF THE PILOT DEMONSTRATION SYSTEM AT 260°F

     The test program to characterize the pilot demonstration precharger-col-
lector system was conducted in July 1982.  The objectives of the test wo to
                                      71

-------
        1.0
o
cc
LU
O
cc.
<
X
u
HI
cc
S:



O
01
    O
    cc
    01
    o
    cc
    <

    o
    o
   01
    01
    CL
        0.5
                     IDEAL

                     10% SNEAKAGE, ag = 0.25

                     20% SNEAKAGE, ag = 0.50
         1010
                                        p, Si-cm
                                                   1012
  1013

4100-236
Figure 4, Theoretical predictions of precharger benefit as a function of fly ash resistivity.
                                        72

-------
          1013
          1012
        o
       tn
       UJ
          1010
           1Q9
1000/T(°K) —-  3.0
       °C
                    I     I     I     I     I     I     I     I
      I
           I
      I
           I
      I
     2.8
60   84
141  183
2.6  2.4
112  144
233  291
2.2  2.0
182  227
359  441
1.8   1.6
283  352
541   666
                             TEMPERATURE
1.4   1.2
441  560
826  1041
  4100-234
 Figure 5.  Ash resistivity versus temperature at Bull Run Steam Plant.
                                 73

-------
                                                           4100-101
Figure 6. Continuous charge probe sensor.
                   74

-------
1) determine the system collection efficiency  and 2)  determine the percent of
penetration decrease attributable to  the  precharger stage.

     The precharger stage was optimized for the  test  using  the continuous
charge probe.  An example trace of the probe reading  is shown in Figure 7.  As
shown/ the action of the grid is very important  in achieving high charging
effectiveness with these test conditions.  The average values of voltages and
currents for the precharger stage are shown in Table  2.  The average Q/m ratio
measured at these settings was -1.75  yC/g.  This compares to a theoretical

                 TABLE 2.  PRECHARGER OPERATING  CONDITIONS
          Corona discharge electrode voltage  =   15  -  17  kV
          Corona discharge electrode current  =   25  -  30  mA
          Charging current density            =  100  - 120  uA/ft2
          Charging electric field strength    =      2 kV/cm
          Grid electrode voltage              =    2  -   3  kV
          Grid electrode current              =   45  -  50  mA
prediction of -1.81 MC/g for the given charging  conditions  and particle size
distribution, and thus indicates essentially no  degradation in charging with
the optimum precharger operating condition.

     The collector stage was operated at maximum stable  applied voltage in each
section.  These operating values and the corresponding currents and current
densities are shown in Table 3.

                  TABLE 3.  COLLECTOR OPERATING  CONDITIONS
                              Electric                      Current
     Section   Voltage(kV)  Field(kV/cm)  Current(mA)  Density(yA/ft2)
1
2
3
4
45.5-46.5
45
50-51
50-51
3.6
3.5
4
4
14.5
25
14-16
26
6
16
9
11

     Mass loading measurements which were made during  the  July test show a
"precharger on" penetration of 0.0033  (99.67% collection efficiency)  and a
"precharger off" penetration of 0.0052  (99.48% collection  efficiency).   This
36.5% difference in penetration due to the precharger  is a substantial  enhance-
ment in performance for this fly ash resistivity  (measured to  be 2-5  x 1011 ohm
cm at this test condition).
                                      75

-------
                                                    CORONA POWER
                                                    TURNED OFF
                                                                  4100-205
Figure 7.  Example trace of the continuous charge probe output.

                          76

-------
     The particle size distribution  and  fractional  efficiencies for particles
with diameters in the range  0.5 to  10.0  ym were measured with University of
Washington Mark V cascade impactors.  The  precharger on and precharger off
penetration-efficiency curves obtained from the inlet vs. outlet comparisons
are shown in Figure 8.  The  overall  mass collection efficiencies as measured by
the impactors are 99.52% with precharger on and 99.36% with precharger off.
This is a 25% reduction in penetration due to  the precharger stage.

     Electrical aerosol analyzer  (EAA) and optical  particle counter systems
were also used to determine  the percent  of collection efficiency due to the
precharger for particles in  the diameter range of 0.04 to 1.7 pm.  These data
showed a consistent difference in particle concentration throughout the measur-
ed size range of approximately 20%.

     The test data all point up an enhancement in performance attributable to
the precharger varying from  20 to 36% depending on  the measurement technique.
The major difference between this test and the test of July 1981 was the pre-
charger operating voltages and currents.   A fourfold increase in current densi-
ty was achieved for the recent test.  Several  factors may be contributing to
this difference, including the newly installed commercial-grade power supplies
for the precharger, and a reduced inlet  mass loading (measured to be 4.3 gr/scf
(9.8 g/m3) instead of 6.1 gr/scf as  in the 1981  test).

     Another difference between the  1982 and 1981 tests was the decrease in
precharger off collection efficiency.  The possible explanation for this dif-
ference is a reduction in inlet MMD  from 32 to 13 ym.  This reduction correlat-
es with the lower inlet loading.  It is  not certain what factors produced this
reduction.

TEST AT MAXIMUM RESISTIVITY

     A test program was developed to characterize the performance of the pilot
demonstration precharger-collector system  when subjected to the maximum fly ash
resistivity available.  In order to  achieve this condition,  a sample flue gas
stream from the boiler exhaust upstream  of the heat exchangers was mixed with
the cold-side slipstream.  This produced a maximum  ash resistivity of approxi-
mately 4 x 1012 ohm cm at 325°F.

     The configuration of the precharger-collector  system was unchanged from
the test at 260°F gas temperature.   The  3/8-in.  diameter rods were left in the
collector stage for discharge electrodes.   The voltage-current relationships
for the four collector sections at the high resistivity ash loading condition
are shown in Figure 9.  The voltage  before sparkover (and concurrent sharp
change in the curve slope) was less  by an  average of 8 kV than when the ash
resistivity was 2-5 x 1011 ohm cm in previous  tests.  This V-I behavior is a
typical result of back corona.  The  collector  sections were operated at the
maximum applied voltage point on the V-I curve.  Typical operating values for
collector section voltages and currents  are shown in Table 4.
                                      77

-------
    99.9
    99.8
    99.5
      99
      98
      95
      90
             I   I  I  I I III!     I   I  I I I INI     I   I  I  I I III
                                          >	•
             PRECHARGER OFF    __
       	PRECHARGER ON
         ~   III! Mill     I   I  I I I Mil     I   I  1  I INT
                                     0.1


                                     0.2
                                     05
                                        *
                                     1  S
                                                                ui
       10-1
10°
PARTICLE DIAMETER,
    10
   102
4ioo-sia
Figure 8. Precharger on and precharger off efficiency curves from tests at 260PF.
                                  78

-------
    140
    120
    100
1
DC
CC

O
     80
     60
     40
     20
                    10
                                                       COLLECTOR SECTIONS
                                                       O T/R 1
                                                       D T/R 2
                                                       A T/R 3
                                                       O T/R 4
20            30
  VOLTAGE, kV
40
     50

4100-211
                 Figure 9.  V-l behavior of collector sections at 325°F.

                                     79

-------
                  TABLE 4.  COLLECTOR STAGE OPERATING DATA
                             Electric                       Current
     Section  Voltage(kV)  Field(kV/cm)   Current(mA)     Density(nA/cm2)
1
2
3
4
40-42
40-42
42-44
42-44
3.2
3.2
3.4
3.4
5-10
5-10
8-12
10-20
2-4
3-7
5-8
4-9

     During the early portion of the test program the  precharger stage was
operated at voltages and currents comparable  to  those  maintained during the
test at 260°F.  After determining that the  operating range of the precharger
was limited by sparking on the grid electrode, several days of testing were
completed with the precharger off.  The  system was then taken off-line, and
the precharger stage was inspected to determine  the cause of the spark-limit.
Three grid electrodes were found to have suffered severe physical damage.  It
was concluded that operation of the pilot demonstration system for prolonged
periods at temperatures near and below the  acid  dewpoint had resulted in
deterioration of the grid material and had  produced irregular surface edges
from which corona and electrical breakdown  could occur.  The industrial power
supply for the precharger stage allowed  system operation to be maintained
under heavy sparking conditions.  During these heavy sparking episodes the
grid suffered the damage.

     Two of the three damaged grids were replaced with available spare grids.
The third damaged grid was replaced with a  grid  from the lane nearest the
casing wall, leaving that 1 lane out of  18  with  a grid electrode on one side
only.  Though this left the precharger stage  in  a nonideal condition, testing
resumed.

     The effects of replacing the damaged electrodes in the precharger stage
were slightly altered voltage operating  levels and more restricted operation
so as to avoid heavy sparking*  The precharger voltages and currents are sum-
marized in Table 5.
                                      80

-------
                   TABLE 5.  PRECHARGER STAGE  OPERATING DATA


Before repair
After repair
Corona
Voltage
(kV)
- 17-20
- 15-17
Grid
Voltage
(kV)
2-4
4-6
Charging
Field
(kV/cm)
~2.0
-1.5
Charging
Current
(mA)
20-25
15-20
Current
Density
(nA/cm2)
87-109
65- 87

     The mass loading measurements taken since the  damage  to  the  precharger has
been known show no statistically significant difference  in system collection
efficiency with or without the precharger.  The precharger-collector system
efficiency has been 98.87%.  The collector  alone  has  had an efficiency of
98.75%.  At this level of efficiency an SCA of 370  ft2/kacfm  (72.8 m2/m3/s)
would be required for 99.7% collection efficiency.

     The charging effectiveness of the precharger was measured  with Q/m sam-
ples.  The average of these measurements is -1.4  yC/g.   This  is compared to a
theoretical Q/m value under these electrical and  ash  size  distribution condi-
tions of -2.4 yC/g.  This indicates some performance  deterioration.

     A strip chart record of continuous mass monitor  readings indicated an
increase in rapping reentrainment emissions at the  325°F gas  temperature.  This
change in nonideal conditions, along with the precharger mechanical problems,
would cause a decrease in the performance benefit of  the precharger.  This is
evident in Figure 4.  Any increase in nonideal effects would  cause a decrease
in expected precharger performance benefit.

     The mechanical problem in the precharger stage may  be seriously altering
its electrical behavior and charging efficiency.  The lane with damaged and
missing grid electrodes constitutes 5.5% of the active area of  the precharger.
That much gas sneakage would be detrimental to performance by itself.  However,
if that lane is in severe back corona (and  it most  probably is),  a large frac-
tion of the precharger current may be confined to that lane.  This would dras-
tically vitiate the charging effectiveness  of the precharger.

                                 FUTURE PLANS

     The first test planned for the remainder of  the  pilot demonstration pro-
gram is a repeat of the test at maximum ash resistivity.  This  will occur as
soon as damaged precharger components can be replaced.   An effort to reduce, or
quantify, the contribution other nonidealities present in  the system are making
to the emissions will be made prior to this test.

     The final test program will be an evaluation of  the precharger applicabil-
ity as a retrofit to conventional ESP's.  For this  test  the discharge elec-
trodes downstream collector sections will be replaced with 1/8-in. diameter
wires to conform to standard ESP practice.  From  these results  of this test a
                                      81

-------
cost analysis for precharger retrofit  versus alternate techniques to achieve
emission limits will be produced.

                            SUMMARY  AND  CONCLUSIONS

     The evaluation of the pilot demonstration system has been underway for 20
months.  Only minor modifications have been required to achieve a rugged com-
mercial-grade two-stage system.

     The performance of the two-stage  ESP  system has been evaluated under two
                                     11                   19
ash resistivity conditions:  2-5 x  10•"• •"•  ohm cm and 4 x 10" ohm cm.  One obvi-
ous conclusion from the tests is the excellent performance of the downstream
collector stage.  The 3/8-in. diameter rods used for discharge electrodes have
proven very effective in providing high  electric field strengths.  The collec-
tor efficiency was ~99.5% for the 2-5  x  1011 ohm cm ash resistivity.  Even at
4 x 10 12 ohm cm ash resistivity the  collector achieved 98.75% efficiency in
spite of increased rapping reentrainment emissions.  This is very good given
the SCA of 286 ft2/kacfm.  (The full scale precipitator at Bull Run has an SCA
of 570 ft2/kacfm  (112 m2/m3/s)«)  Questions remain unanswered relating to
larger diameter discharge wires.

     The precharger-collector system was demonstrated to operate at high effi-
ciency, ~99.7%, in collecting 2-5 x  1013- ohm cm resistivity ash.  The precharg-
er reduced mass emissions by more than 20% under these conditions.  Theory
would predict an  even greater benefit  at higher ash resistivity.

     The ash resistivity in the pilot  system was raised by increasing the flue
gas temperature.  The precharger-collector system was tested at 325°F where the
resistivity was measured to be 4 x  1012  ohm cm.  Some mechanical problems de-
veloped in the precharger due to extended  operation below acid dewpoint.  These
problems, together with increased rapping  reentrainment,  cloud test results at
the high resistivity condition.  The precharger has not been shown to contri-
bute to enhanced  collection efficiency.  Tests with the precharger restored to
good mechanical condition will be made as  soon as possible.
                                      82

-------
                                  REFERENCES

1.  Pontius,  D. H., and L. E. Sparks.  A Novel Device for Charging  High
    Resistivity Dust.  APCA Journal, 27(7):698-700,  1978.

2.  Pontius,  D. H., P. V. Bush, and L. E. Sparks.  Field Evaluation of  a Two-
    Stage ESP for High Resistivity Dusts.  Staub, 40(11):   473-477, 1980.

3.  Bush, P.  V., and D. H. Pontius.  Pilot Demonstration of the  Precharger-
    Collector System.  In:  Third Symposium on the Transfer and  Utilization of
    Particulate Control Technology:  Volume I. Control of Emissions from Coal
    Fired Boilers, EPA-600/9-82-005a, July 1982.

4.  Mosley,  R. B., H. H. Anderson, and J. R. McDonald.  A Mathematical  Model of
    Electrostatic Precipitation (Revision 2).  EPA-600/7-80-034  (NTIS No.
    PB80-190994), February 1980.

                               ACKNOWLEDGEMENTS

    The authors express special thanks to Todd Snyder of Southern Research
Institute for his management and organization of all on-site  activities at the
pilot demonstration facilities.

    We also  express our gratitude to Leslie E. Sparks of EPA  for his guidance,
insight, and general supervision as Project Officer of the pilot demonstration
program.
                                      83

-------
                        EVALUATION OF PRECHAR6ERS
                                  FOR
                  TWO-STAGE ELECTROSTATIC PRECIPITATORS

                                  by

                             George Rinard
                              Donald Rugg
                            Michael Durham
                        Denver Research Institute
                         Denver, Colorado 80208

                               ABSTRACT

     Two-stage precipitators are being considered  for  high resistivity dust
collection.  Indications  are that the size of a two-stage precipitator for
this  application may be considerably smaller than a conventional  wire plate
design.  DRI is presently evaluating two-stage designs using a  7.08 m^/s
(15000 ACFM) pilot plant and slipstream of the Valmont  Power Plant in
Boulder,  Colorado.   Results of the evaluation of two precharger designs are
given.

       This  paper  has been reviewed  in  accordance with the U.S.
       Environmental Protection  Agency's  peer  and  administrative
       review  policies and  approved  for  presentation and
       publication.

                             INTRODUCTION

     The  Valmont Test Facility, designed and constructed by the  Denver
Research Institute and sponsored by EPA is presently operational.  This
facility  was constructed to provide a versatile  tool  for evaluating novel
electrostatic  precipitator concepts  on an operational  source of  high
resistivity dust.   Evaluation has begun on  the tri-electrode precharger
developed by Southern Research Institute  (SoRI).  Preliminary results of
this  work are  presented.   In  addition,  voltage-current characteristics of
the cooled-pipe charger/collector have been determined, and the results are
given and compared with those  obtained  in the laboratory.

                          VALMONT TEST FACILITY

     The  experimental facility  is  set  up on the grounds  of the Valmont
Station of Public Service of Colorado (PSCo)  in  Boulder,  Colorado.  The
facility  is shown  in Figure 1.  The test arrangement consists of the ESP,  a
gas  delivery  and return  system,  particulate sampling  station,  the
instrumentation and  sampling trailers, storage facilities, and support
equipment made  accessible by a road.
                                  84

-------
     Because  of a conditioning system  located just  downstream  of the air
preheater on the No. 5 unit, the slipstream is taken just  upstream on the
hot side of the  preheater.  The gas is  returned to the main duct  downstream
of the air preheater and just upstream of the I.D. fans.  A bypass duct is
installed between the supply  and  return  duct  complete with a  240 kW elec-
tric heater to  preheat the system.  Preheating is important to avoid sudden
changes in  temperature  and  to prevent  condensation.   The supply  and return
ducts are preheated by opening a  valve located  in a second by-pass loop.
This valve  allows circulation  of  flue  gas  due to the pressure differential
between the two legs of the system.

     An electric-motor-driven  gas  circulating fan  is  provided.  The fan is
complete with  an opposed-blade outlet control  damper to ensure  constant
flow through  the ESP.  A  venturi  flow  meter located  downstream  of the ESP
is used to monitor the  flow through the  system.   For  cold-side operation  a
Luhr-Interel design gas/air heat exchanger  is used to cool the gas.   The
heat exchanger is equipped with  a self-cleaning mechanism to  keep the
outside (gas-side) surface of the  tubes clean.

     The particulate sampling station is located between the  supply and
return  ducts.   It consists of  a prefabricated building, 3.0  x  5.5 m (10 x
18 ft) in size  and  is installed at an  elevation of approximately  3.7 m (12
ft) above ground.

     The two  trailers are for instrumentation and gas sampling.   The trai-
lers are 2.4 m (8 ft) wide and 12.2 m (40 ft) long and contain a computer,
controls,  and  gas sampling  and analysis  equipment.   They are equipped  with
heaters, air conditioners, water storage, and  waste disposal holding tanks.

     The test  facility  has  presently  been  operated  for  approximately  1100
hours during  which  200  hours  of testing has been  completed.   The descrip-
tion of the  plant,   its operating  conditions,and the  average values  of the
measured flue gas conditions are shown in Table  1.

     It is extremely important  in  pilot scale  work  to have good flow condi-
tions and eliminate   sneakage  as much  as  possible.   The  flow  distribution
for the test precipitator was modeled in the laboratory to ensure  good
quality flow.  Baffles were installed at  the top and bottom of each collec-
tor section,  in the  hoppers, and along the side of each  outside  plate.

     During system shakedown  the  flow  distribution  was  measured by taking
gas velocity traverses at  the inlet,  center, and outlet  sections of the
precipitator.    Figure 2  shows a histogram of the  flow  characteristics  at
the inlet.  This is  characteristic of the flow throughout the  precipitator.
The measurements indicate that  the  design  goal  to meet or exceed  the
Industrial Gas  Cleaning Institute  (IGCI) standards was accomplished.
                                   85

-------
                                  TABLE 1

                            OPERATING DATA FOR
                     VALMONT STATION OF PUBLIC SERVICE
                                OF COLORADO

                             PLANT DESCRIPTION

UNIT:          No. 5

RATING:        180 MW

BOILER:        Combustion Engineering Tangential  Fired

COAL:          50% Energy Coal  - Colorado (0.6 - 0.8% S)
               50% Colorado-Wyoming Coal  Co.
ON LINE:       1963

                           OPERATING CONDITIONS

                       TIME                   BOILER LOAD
                   2400 to 0700                70- 90 MW
                   0700 to 1700               100-160 MW
                   1700 to 2400                80-150 MW

      At Peak Conditions     MAX           MIN            AVG
        Boiler Load         162 MW         92 MW         126 MW
      Hot-Side Temp      390°C  (735°F)  310°C (590°F)  343°C (649°F)


                            FLUE GAS CONDITIONS

GAS CONSTITUENTS:

     HoO (Method 4)                             8%
     S&3 (Controlled Condensation System)     < 0.1 ppm
     S02 (Pulsed Fluorescence)                 200 ppm
     C02 (Gas Chromotography)                  12%
     02 (Gas Chromotography)                    8%

FLY ASH:

     Mass Loading      4.95 g/SCM (2.16 gr/SCF)   2.7 g/SCM (1.18 gr/ACF)

     Size Distribution   14 micrometers Mass Median Diameter
                         4.5 Geometric Standard Deviation

     Resistivity (In Situ)   @ 150°C (300°F)      2.0 x 1012 ohm-cm
                             @ 120°C (250°F)      7.5 x 1011 ohm-cm


                                    86

-------
                              TEST PROGRAM

     Besides  the  overall  performance of the ESP, the performance and  char-
acteristics of  the  individual charging  and collecting  components  are  to  be
analyzed  tot further  the understanding of  the precipitation process.
Cumulative charge/mass, charge on single particles, mass  loading, and
particle size distribution can be measured at ports located just downstream
of each precharging section.  Load cells mounted on  each  collector  section
will be used  to determine:  the  rate at  which the dust is  collected  in each
section, dust thickness  prior to  rapping, the effectiveness of  the  rapping
for removing the dust,  and the effectiveness  of downstream charging and
collecting sections  in precipitating  reentrained  dust from  upstream
sections.

     The parametric tests have been designed to  characterize the precharger
sections,  the collector  sections,  and  various  combinations of  precharger/
collector  configurations.  The initial stages of each parametric evaluation
of a novel concept will   involve an  optimization of the  rapping  frequency
and intensity.  The collector load cell  readings,  voltage-current char-
acteristics,  and  outlet  opacity will also be used to optimize the rapping.
In addition, the location  of upper and lower ash level  detectors can  be
changed so that  it will be  possible to optimize the amount of ash  in a
hopper to reduce  boil up.

     Besides  operating the  prechargers  and collectors at  the highest  vol-
tages possible, tests  will be run to determine  the  effects of reducing the
operating levels.   Tests will  also be run  in which one or more of the
prechargers will  be deenergized  to determine the  proper ratio  of precharger
to collector  sections.   This  can  be determined  by  analysis of  the  overall
performance and by  running charge/mass tests  after  each collector.  If the
particles exiting  a  collector  section are  not  charged, then another
precharger section  is  be required.

     After the parametric tests are run to determine the operating char-
acteristics  of the novel ESP devices,  the  unit will be operated under
optimum conditions at  two different  flow rates to determine operating
performance as  a  function of  precipitator  size.   This information  will  then
be used in an economic evaluation of the device.

     The first  phase of  the test  program  involves an  evaluation  to deter-
mine the optimum  two-stage  precipitator system.  This  involves  testing two
different  prechargers interfaced  with flat-plate high-voltage  electrodes  in
the collector sections.   The  first  precharger that  is being tested is the
tri-electrode precharger  developed  by SoRI.   Preliminary  results for  this
device are given  below.

     The first phase of  the planned test  program will be directed toward
new precipitator  designs  the  second  phase will  be directed  toward  improving
existing  precipitators  operating  on  high  resistivity dust.  The plates  in
all four collector sections will be replaced  with standard  wire-plate
configurations.  Potential  improvements will  be in the form of advanced
controllers,  pulsed excitation,  addition  of a precharger,  and lowering the
operating temperature of the precipitator.  The third phase of the test
program will  involve operating the ESP  at hot side-conditions.


                                   87

-------
             PRELIMINARY RESULTS FOR TRI-ELECTRODE  PRECHARGER

     Provisions are made in the test  precipitator for  as many  as  five
precharger  sections.  In the testing of  the  tri-electrode precharger, three
prechargers are  being used.  Each precharger is separated  by two collector
sections  with 11.3  sec/m  (57.6  ft2/kacfm)  of collector  area. The final  pre-
charger is  for testing of McLean's (1982) findings  that the precharger
should be  a good collector.

     The  experimental set  up for  measuring  V-I  characteristics of the tri-
electrode  precharger is shown in Figure  3.  The resistor R^  was  added to
increase  the range  of  operation of  the  grid voltage.   Without R^  the grid
floats up  to sparking potential with very  low corona current.

     Figure 4 gives the measured clean V-I characteristics  for  the  tri-
electrode  precharger.   It should  be pointed out  that these curves  are
plotted on voltage vs  current (the reverse  of standard practice for VI
curve).  The  reason for this is that by  varying R^ the voltage of  the  grid
or corona to grid voltage is a function of the corona current.  As can be
seen from  these  curves, the operating grid voltage  is reduced for increased
Rb.  From this figure  it  can  be  seen that  the corona current  vs  gas  gap
voltage (Vc-Vg) is not  a function  of grid  voltage.   Since this is the case,
the remaining V-I  data  will  be plotted  as Ic vs (Vc-Vg).

     Figure 5 gives a comparison  of the V-I data  obtained at  Valmont  with
that measured by DRI and McLean in  the  pilot precipitator  in EPA/IERL-RTP's
Particulate Technology Branch laboratory.   The current is normalized by
plotting  corona  current in microamps per meter length  of corona wire.  This
allows comparison  of the  data  for the different sized  precipitators.  This
curve shows good argument  for  these cases.

     Figure 6 shows V-I data after stable  operating conditions  had  been
obtained  under flue gas conditions.   From this figure it  can  be seen  that
the corona onset voltage is considerably higher than for the clean case.
Also as attempts are made  to increase the gas gap voltage  above 18 kV,  back
corona is  initiated.

     A charge probe consisting  of a fine  mesh  screen approximately  15 x 30
cm  (6 x 12 in.) was used to measure  negative charging  vs corona  current.
The probe  current was measured  with a Keighly 616  picoammeter.  The  results
are given in Figure 7.  As can be seen from this figure charging  does not
vary with  grid voltage.

     The second fact evident from Figure 7 is that, as the corona current
is increased above  about 3 mA, charging  decreases  because of the initiation
of  back corona.

     Figure  8 presents the same data  as Figure  7 except this time the
corona current is  held  constant and corona  voltage  varied.  As can be seen
from this  figure,  charging is  not a function of grid voltage.

     The  next phase of testing  of  the tri-electrode  precharger will  involve
charge-to-mass measurements at electrical  operating conditions that  produce
optimum charging.   That will be followed by  mass efficiency tests.


                                    88

-------
               V-I RESULTS FOR COOLED-PIPE  CHARGER/COLLECTOR

     The second charging  device to be tested  by  DRI  is the cooled pipe
charger/collector (CPCC) (Rinard  et al., 1982).   CPCC's  are presently
installed  in  the  second  and  fourth  charger section  of  the precipitator  at
Valmont.  This  was done to allow clean and dirty V-I data to  be collected.

     Figure 9 shows  the  clean and dirty V-I characteristics for the CPCC
that was  obtained in  the laboratory (Rinard et al., 1982).  This figure
shows that the back  corona that is  present when the pipes are not cooled
vanishes for  sufficiently cooled pipe surfaces.   In  this case the  dirty  V-I
curves for the cooled pipes are essentially the same as the clean char-
acteristics.

     Clean and  dirty curves for the  CPCC installed at Valmont are  given  in
Figure  10.  These results are  very similar  to those  obtained in the  labora-
tory.  The dirty  curves  with cooled pipes are shifted to the right of the
clean characteristics.   This  is most likely due  to space charge effects  and
the fact that the dirty curves were obtained with actual flue gas.  The
clean curves  were for  air-load  conditions.

     A final  comment is  in order for the corona onset voltage of  Figure  10
when compared to  Figure  6.  As can be seen  the corona onset voltage  did  not
increase for  the  CPCC  for  dirty  conditions as it did for the tri-electrode
precharger.   This is  most  likely due to the difference  in corona  electrode
design  (wires for the  CPCC and  barbs  for the tri-electrode)  and the higher
corona  current  density obtained for  the CPCC.

                               REFERENCES
McLean,  Kenneth  J.  Analysis of the Electrical and Charging Characteristics
of a Three-Electrode  Precharger.  In:   Third  Symposium on  the  Transfer  and
Utilization of Particulate Control  Technology, Volume  II, EPA-600/9-82-005b
(NTIS PB83-149591), pp. 304-313, July  1982.

Rinard, G., M. Durham, and D. Rugg.  Development of a Charging Device  for
High-Resistivity Dust Using Heated  and  Cooled Electrodes.   In: Third
Symposium  on the  Transfer  and  Utilization  of  Particulate  Control
Technology, Volume  II,  EPA-600/9-82-005b (NTIS PB 83-149591),  pp.  283-294,
July 1982.
                                   89

-------
FIGURE 1.   Valmont ESP Test Facility,

-------
10
c
o
0
S_
«4-
>-
O
Ul
:D
cy
U_
2^
o
i— i
I—
CQ
I-H
C£.
h™
CO
5




0

0

0

0

0
0

0
0



.7

.6

.5.

.4 -

.3
.2-

.1 •
.0
OPEN LOOP
24 READINGS TAKEN

-

-

"

-
-

•

r\ Cfl










1












i


— ^-V = 2I5
»..,!,













~L, ,
ion ten
\J        %^\^        I VJT N^        i*^s^
          AIR VELOCITY, cm/s

FIGURE 2.  Velocity distribution  histogram from the Valmont
           TEP No. 1 charger  section
                                         •CORONA WIRE
FIGURE 3.  Experimental  setup for measuring V-I characteristics
           of the  tri-electrode precharger.
                       91

-------
40
35
                                    8       10       12
                                   CORONA CURRENT,  mA
                                              14
.16
    FIGURE 4.
VI characteristics of tri-electrode  precharger No. 1 at
clean conditions.
                                    92

-------
               500
             f

             "3 300
                      . — Me Leon EPA LAB

                      « — mi EPA LAB

                      o — ORI VALMONT PC* \


                                          10    12

                                        (VC-V,I *V
FIGURE  5.   Comparison of DRI  and McLean  data for  clean tri-electrode
            prechargers at  (150 C).
             600
             500-
             400-
             300-
             200
              100
                                         10    12
FIGURE 6.   Dirty VI  curves for tri-electrode precharger No.  1  at
            150°C (300°F).
                                       93

-------
                                    «€-»
FIGURE 7.  Charge probe measurements at 150°C(300°F)  as a function of
           corona current.
                                            I, = 5.*
                                            !,=•»•»
                                    3     4

                                     V|IV
FIGURE 8.  Charge probe measurements  at 150°C(300°F) as a function of grid
           voltage.
                                    94

-------
            1000


            900


            800
O PIPE WATER TEMP 380C|IOO°F)

D NO WATER IN PIPES
                            AVERAGE FIELD STRENGTH, kV/»
 FIGURE 9.  Voltage-current characteristics of  6.0 cm (2,375  in.)  cooled
            pipes  with 3.2 mm  (1/8 in.) electrodes.
                                        + Dirty T  =  150 DC (Pipes with 45 oc
                                                 y            Water)

                                        o Dirty T  =  150 °C


                                        • Clean Tg =  150 °C
FIGURE 10.   Clean and dirty V-Icurves  of  the cooled electrode precharger
             at Valmont.
                                      95

-------
                        INITIAL EXPERIMENTS WITH AN
                   ELECTRON BEAM PRECIPITATOR TEST SYSTEM

               by:  W. C. Finney, R. H. Davis, and J. S. Clements
                    Department of Physics
                    Florida State University
                    Tallahassee, Florida 32306

                    E. C. Trexler
                    U.S. Department of Energy
                    Germantown, MD 20545

                    J. S. Halow
                    Morgantown Energy Technology Center
                    Morgantown, WV 26505

                    0. Z. Tokunaga
                    Japan Atomic Energy Research Institute
                    Takasaki 370-12
                    JAPAN
                                  ABSTRACT


     As part of the Department of Energy's Advanced Environmental Control
Technology Program, a laboratory-scale Electron Beam Precipitator test
system (EBP) has been designed and constructed at the Florida State Univer-
sity to investigate particle charging and collection under a wide variety of
experimental conditions.  The system consists of a rectangular, closed-
circuit wind tunnel composed of a number of interchangeable modules includ-
ing an electron beam (e-beam) precharger followed by a collecting section.
A model aerosol enters the e-beam precharger, is charged in an ionization
zone, and then passes downstream to the collector.  At a gas velocity of
10 ft/sec through the precharger the velocity in the collector is 2.5 ft/
sec, giving the collector an SCA of 200.

     Experimental results are presented of a preliminary investigation of
e-beam precharger ionization for a wide range of electron beam currents and
precharger electric fields.   Experiments planned with the EBP include the
determination of the charging and collection efficiency of the system using
resuspended particles with properties similar to high resistivity fly ash.
                                     96

-------
                                INTRODUCTION
     The Department of Energy  (DOE) in pursuit of its mission to assist in
meeting the nation's energy needs is concerned that the use of the large U.S.
coal resource is neither economically nor environmentally inhibited.  Approx-
imately 70% of our domestic coal, however, is characterized by fly ash
having high resistivity which causes problems with existing electrostatic
precipitators (ESP's) when they are employed to meet the existing new source
performance standards.  Future emphasis on controlling the small size
respirable particles may find existing ESP systems inadequate.  The research
and development target to meet future particle control needs should be to
develop systems which in addition to being small, economical, and reliable,
will effectively remove high resistivity ash (> 2 x lo10 fi -cm) and small
respirable particles  (0.01 - 2 ym).

     Progress has been made recently in several methods for enhancing the
performance and reducing the cost of ESP's.  Some of the interesting evalua-
tions of novel control devices which were sponsored by other groups include
the EPRI/APS High Intensity Ionizer (1), the EPA/SRI Tri-electrode Precharger
 (2), the EPA/DRI Temperature Controlled Electrode ESP (3), and the University
of Tokyo, Japan, Boxer Charger (4).  The Electron Beam Precipitator project
at the Florida State University  (F.S.U.) is being sponsored by DOE because
there is the potential for significant improvement  over existing control
devices and because this work is a companion effort to several DOE funded
SOX/NOX removal projects.  There is the possibility that there might evolve
from this work an integrated SOx/NOx/Particle removal system which would
offer many functional and economic advantages over separate systems.

     Electron beam ionization  (EBI)  for particle charging in electrostatic
precipitators has been the subject of a number of studies at F.S.U. (5,6,7).
EBI in air was examined in depth using a parallel electrode apparatus within
which the ion species produced by the beam could be separated by an electric
field.  This work characterized three ionization regimes (primary, satura-
tion, and secondary ionization) and provided the foundation crucial to under-
standing the behavior of an electron beam in an actual precipitator.  En-
hancement of the total ionization by a high electric field (secondary ioniza-
tion) has been shown to be a beneficial although inefficient means of in-
creasing the ion current available for particle charging (8).

     The experimental program examining the factors influencing EBI resulted
in parameters for the design of a laboratory-scale electron beam-energized
electrostatic precipitator.  The Electron Beam Precipitator (EBP) test system
at F.S.U.  is unique in this country.  It is singularly capable of treating
particles with an electron beam-energized ESP under a wide variety of
experimental conditions.   A realistic comparison with other pilot ESP's was


                                     97

-------
one goal of this program; therefore, the EBP is relatively large for a labor-
atory scale system.  A thorough investigation of the process of electron beam
precipitation is planned including the measurement of particle charging and
collection efficiencies using resuspended test aerosols having properties
similar to high resistivity fly ash.

     A description of the basic electron beam ionization process will precede
a review of the recently commissioned EBP, which was detailed in a recent
paper (9).  An explanation of the designed operation of the electron beam
precharger and the results of some initial ionization experiments using the
new precharger will be presented.

                          ELECTRON BEAM IONIZATION
     An electron beam is a concentrated, focused stream of accelerated elec-
trons produced by a continuously emitting electron beam source.  At the
Florida State University the e-beam source is a Van de Graaff electron accel-
erator which is rated at a maximum beam energy of 3 million volts  (MeV)
and a maximum current of 1 mA.  A wide choice of electron beam energy and
current is one advantage of this machine, which is continuously variable in
these parameters.  Beam geometry or positioning can be altered to suit par-
ticular experimental conditions as can the scanning frequency of the beam,
which is usually set to zero.

     The F.S.U. 3 MeV accelerator resides in an underground vault and is per-
manently installed and therefore non-transportable.  The electron beam
energy, current, and current distribution are remotely adjusted and measured
at the accelerator control console which is separated from the target room
vault.  Small, industrially reliable electron beams are presently in use in
such processes as plastics treatment, paint curing, and electron beam weld-
ing.  Electron beam ionization has retrofit possibilities on existing pre-
cipitator installations, and field testing of electron beam ionization will
require such a portable commercial electron beam.

     Electrons created in the accelerator are passed down a flight tube at
high vacuum and out into air through a thin foil "window" (Figure 1).  The
high energy, narrow beam spreads and scatters upon colliding with air mole-
cules, creating a bipolar plasma of positive and negative ions and additional
electrons.  Positive ions are formed by electron stripping while negative
ions result from electron attachment.  In air, the energy of ionization is
32 eV per ion pair, so a 1 MeV energy beam produces about 30,000 ion pairs
per electron.  The total number of ions produced depends upon the beam
current.  When an electric field is applied to the electron beam-produced
plasma, a separation of the ion species and electrons occurs; positive ions
are attracted to the cathode and negative ions and electrons migrate to the
anode.  With no electric field the ions recombine or diffuse out of the
plasma.  Either positive or negative ions can be selected to charge par-
ticles, but the nonworking monopolar ion species must be removed from the
ionization zone.
                                     98

-------
     Corona wire ionization and electron beam ionization differ in several
respects.  Corona emission-producing ionization has a finite onset voltage
while EBI begins as the beam voltage is raised from zero.  Monopolar ions
of either sign are created by a corona but EBI produces a bipolar plasma.
EBI has many orders of magnitude more working ions than corona; therefore
copious ionization is available for particle charging.

                 ELECTRON BEAM PRECIPITATOR AND SUBSYSTEMS
     The Electron Beam Precipitator is basically an instrumented wind tunnel
composed of a number of specialized modules separated by ductwork sections
(Figure 2).  A "racetrack" or closed-circuit arrangement is used to maximize
energy conservation and to control humidity.  Different ductwork configura-
tions can be assembled as desired including various e-beam entry locations
or a single pass arrangement.  To resist corrosion and for mechanical
strength the wind tunnel is entirely constructed of stainless steel.  Room
temperature operation is planned initially but high temperature operation
is not precluded by the design of the racetrack.

     The specifications of the wind tunnel ductwork are as follows:
Outside Dimensions = 20 ft long by 8 ft wide; Single Pass Length = 45 ft;
Ductwork Size = 12 in by 12 in (1 ft2);  Construction = #304 12 gauge stain-
less steel; Gas Velocity Range = 3 to 30 ft/sec; Maximum Gas Volume =
2000 ACFM; System Pressure Drop = 5 in water gauge.  Specialized modules of
the racetrack include the fan, an aerosol entrainment module, two monitoring
modules, the precharger and collector composing the precipitator, and an
absolute filter.  The monitoring, aerosol entrainment, and precipitator
modules each have removable plexiglass tops and opening ports on three
sides for observing precipitator conditions  and mounting of monitoring
systems or aerosol injectors.  Perforated, removable baffles of 50% open
area smooth gas distribution at several locations in the wind tunnel.  EBP
subsystems which are associated with the wind tunnel are several high voltage
power supplies for the precipitator, gas condition, particle, and charge
monitors, a microcomputer with a data conditioning system, three aerosol
generators, a closed-circuit TV system,  and the humidity control device.

     Circulating the gas and test aerosol through the wind tunnel is a
stainless steel centrifugal fan rated at 2000 ACFM at 7" water gauge pressure
drop.  A variable opening outlet damper downstream from the blower controls
the air velocity.  To ensure a particle free gas flow for proper experi-
mental control, a H.E.P.A.-type absolute filter is positioned upstream of
the fan intake.  Aerosols larger than 0.3 ym are collected with a 99.97%
efficiency by the high volume stainless steel frame filter.  For experiments
with monodisperse aerosols, the filter can be removed.  On the back or return
leg of the wind tunnel are two long ductwork sections with removable plexi-
glass tops.  One section just downstream from the fan outlet damper houses
a humidity control device composed of a set of refrigerated coils and a
compressor/condenser unit.  The relative humidity within the racetrack can
be reduced from ambient (50-80%)  to 10% or less and can be controlled within
these limits.
                                     99

-------
     Aerosol entrainment into the gas stream is accomplished in the module
just upstream from the 90° elbow turn preceding the first monitoring module.
Particle injection is through a stainless steel tube inserted into the duct.
The gas stream is smoothed by perforated baffles up- and downstream from the
sampling tube, which can be aimed against or with the flow.  Aerosols of two
basic types are resuspended in the electron beam precipitator.  The basic
properties of electron beam precipitation (charging and collection efficien-
cy, maximum field and current) should be investigated using a monodisperse
aerosol at low to medium loadings.  High loadings of mono- or polydisperse
bulk powders are needed to test the EBP under more realistic conditions.
Polystyrene latex beads (PSL) in 1.0 and 3.0 ym diameter sizes and aluminum
hydroxide (Hydral), which is a cheap bulk powder of nominally 1.0 ym diam-
eter, serve as monodisperse, basic study test aerosols at low and medium
loadings, respectively.  Polydisperse fly ash at high loadings will also be
tested after the EBP is completely characterized with PSL and Hydral.  A
Collision nebulizer-based fluid atomization aerosol generator entrains PSL
and a BGI Wright Dust Feeder, a rotating drum air injection bulk powder
feeder, resuspends Hydral or fly ash.  For high loading applications using
bulk powders, a fluidized bed with a screw feeder will be employed.

     Characterizing the operating conditions of the EBP requires a compre-
hensive group of monitoring systems.  These monitors must have remote con-
trols and readouts located in the e-beam control room since the EBP is in
the accelerator target vault.  Monitoring modules are located upstream and
downstream from the precipitator for the purpose of before and after measure-
ment of a number of parameters.  The determination of the operating charac-
teristics of the gas stream and the particle collection efficiency of the
precipitator is the primary function of the two monitoring stations.  Ex-
traction tubes and monitor sensors within the duct are mounted in 3" sampling
ports and on the plexiglass module tops.

     Several monitor systems measure gas stream conditions in the wind tun-
nel.  A hot wire-type anemometer including a solid state temperature sensor
is fastened to a side port of each of the monitoring modules.  Velocity and
temperature signals are wired to digital readouts in the control room.
Electronic ion exchange relative humidity/solid state platinum thermometer
probes are fastened to plexiglass tops at two points:  on the return leg of
the EBP downstream from the humidity control device and on the monitoring
module adjacent to the collector.  Signals are sent to a multichannel
analyzer which provides a display and hard copy of the data.

     Two types of particle samplers monitor the EBP:  real-time and time-
integrated.  Optical particle counters give a. real time indication of the
number and size of particles present in a continuously sampled gas stream.
Remote sensing units located upstream and downstream from the precipitator
send signals to a multichannel analyzer which can then be compared.  A
multistage cascade impactor and an absolute filter holder sample the gas
flow downstream from the collector in a time integrated fashion.  Particle
size and mass measurements are obtained for discrete intervals of time
depending upon the aerosol concentration in the duct.
                                     100

-------
     A microprocessor-based data acquisition and analysis system was devel-
oped to coordinate most of the data outputs of the electron beam precipitator
(Figure 3).  It will be used to accumulate, condition, process, and store
all signals from the energized sections of the precharger and collector and
from the EBP electronic monitors.  The computer is composed of a Z80 micro-
processor, two 8" floppy disc drives, a CRT terminal with keyboard, and a
printer/plotter.  Interfaced to the computer are voltage and ion current
signals from the precipitator through a yMac data conditioning system, the
velocity and temperature sensors, and the optical particle counters and
humidity/temperature probes through the multichannel analyzer.  Data analysis
subsequent to acquisition is performed using Basic and Fortran languages.

     A closed-circuit TV system is necessary to remotely determine pre-
charger and collector ion currents on the high voltage side of the circuit.
It can also be used to monitor various precipitator conditions visually.
The camera, lens, pan-tilt head, and tripod are in the accelerator target
room while the video monitor, remote control panel, and video cassette
recorder are contained in a rack in the control room.

     The precharging configuration that is being tested initially in the EBP
is the "separated" precharger (Figure 4).  It operates at duct velocity and
is upstream from the collector.  The cross-shaped precharging module is the
point of electron beam entry into the EBP.  A central 1 ft^ precharging
region through which all of the system gas passes is designed to accept
several precharger electrode configurations.  Particle charging occurs within
the e-beam ionized interelectrode volume.  To insulate the interior of the
stainless steel precharging module, thin plastic sheets cover the walls,
floor, and any exposed metal framing.  A plexiglass cupola is fitted to the
plexiglass module top for insulation of the high voltage connections.  Elec-
tron beam current is measured immediately after the beam enters air by a
movable probe located in the precharging module.  Scattering foils and a
collimation baffle respectively diffuse or restrict the electron flux which
enters the precharging region as needed.  A thick beam dump prevents the
e-beam from exiting the precharging module.  The precharging of particles
is discussed in another section of this paper.

     Since the precharging region is 1 ft2 cross-sectional area and the col-
lector is 4 ft2 area, a 1:4 expansion section connects the two modules.  A
removable 35% open area perforated baffle is mounted just upstream from the
collector for gas distribution.  Charge-to-mass (Q/M) monitors are inserted
in a port downstream from the precharger within the 1 ft2 section of the
transition.  Q/M measurement devices are used to quantify the effectiveness
of the e-beam precharger prior to particle collection.  A total Q/M sampler
based on a Faraday Cage encased absolute filter (10) will measure total
sample mass and total particle charge.  It is a time integrated instrument
and uses a cascade impactor extraction sampling train.  Several types of
single particle Q/M systems exist including a modified Millikan oil drop
apparatus (11)  and a crossed laser beam device (12) which measure the charge
and mass of individual particles.  After assessing the applicability of
these systems one or more will be chosen to obtain single particle Q/M
data.
                                     101

-------
     Gas entering the collector is split into two channels each of which has
three sections for a total of six typical wire-plate precipitator fields
(Figure 2).   The collector module frame is constructed of T and L angle
stainless steel to which exterior walls with long opening access ports are
welded.  Plexiglass top and bottom covers seal the module.  Twelve stainless
steel collector plates are mounted to standoff insulation and then to the
collector frame.  The six corona wire assembles are each composed of six
stainless steel corona wires suspended between aluminum buss rods which are
insulated from the collector frame by standoff supports.  A set of centrally
located plates can replace the corona wire assemblies for investigation of
a plate-plate collector, and a single 12" wide channel can replace the two
6" wide channels.

     The specifications of the collection module and its internal components
are as follows:

        Collector Module Dimensions  =  7 ft 6 in long by 4 ft 3 in high
                                        by 1 ft 3 in wide.
     -  Corona Wire Height  =  48 in (122 cm).
     -  Corona Wire Diameter  =  0.109 in (2.8 mm).
     -  Wire-to-Wire Spacing  =  4.5 in (11.4 cm).
     -  Wire-to-Plate Spacing  =  3 in (7.72 cm).
        Plate-Plate Spacing  =  6 in (15.2 cm) .
        Plate Dimensions  =  4 ft high by 2 ft 6 in wide.
     -  Single Plate Area  =  10 ft2.
     -  Total Plate Area  =  120 ft2.
        Collector Sparkover Voltage  =  55-65 kV.

     For:  Precharger Velocity  =  10 ft/sec, Collector Velocity  =  2.5 ft/
                                   sec, Gas Flow  =  600 ACFM:
     -  Collector SCA  (ft2) per 1000 ACFM Gas Flow  =  200.

     Three high voltage, high current DC power supplies are used to energize
the precharger and collector.  Two 130 kV, 30 mA units provide voltage to
the precharger while a 100 kV 100 mA supply powers the collector wires.  The
collector fields can be sectionalized if necessary.  Manual regulation of
the precipitator voltages is accomplished remotely from the accelerator con-
trol room.  Power supply voltages and currents are monitored at the control
racks; the T-R sets are adjacent to the EBP in the vault.

                          ELECTRON BEAM PRECHARGING
     Single stage precipitators operate by charging and collecting particles
in the same location in the unit.  Separating the two functions into a two
stage system can offer advantages over a conventional configuration.  The
overall removal rate can be improved by optimizing precharging and collec-
tion in individual sections of the precipitator since the electrical and
physical requirements of each are different.  Much of the recent work on
ESP's has focused on precharging as a more efficient means of driving up
the charge  on particles to the highest attainable limit.  Electron beam
precharging makes use of high ion current densities and electric fields


                                    102

-------
to accomplish this goal.

     The electron beam precharger is an ion current separation and collection
device for charging particles passing through the precharging module of the
EBP.  It provides a high electric field for particle charging but allows an
unimpeded flow of particle-laden gas.  High ion current density is necessary
for maximum charging rate and space charge enhancement of the electric
field.  An electrode configuration which approximates the electrical proper-
ties of a plane without restricting flow was the desired goal.

     The precharger electrode apparatus is composed of two racks of parallel
rod electrodes at the upstream and downstream boundary of the electron beam
ionization region in the precharger module (Figure 3).  Precharger rack
rods are suspended between pairs of buss rod/holder combinations and the two
racks are fastened together and to the precharging module sides.  The elec-
trode array is positioned in the precharging module through which the test
aerosol flows at duct velocity.  E-beam delivery ^.s perpendicular to the gas
flow direction and to the rack electric field.  Rod-to-rod and rack-to-rack
spacing is variable, but the design being tested initially has the follow-
ing specifications:

        Two Identical Racks of 9 Rods Each.
     -  Rod Height  =  11 in (27.9 cm).
        Rod Diameter  =  0.125 in (0.32 cm).
     -  Rod-to-Rod Spacing  ==  1.25 in (3.2 cm).
     -  Rack-to-Rack Spacing  =  6 in  (15.2 cm).
        Rods 1, 2, 8, 9 (Cathode and Anode)  =  Non-metered (guard).
     -  Rods 3, 4, 5, 6, 7 (Cathode and Anode)  =  Metered.
     -  Metered Rack Plane Area  =  67.2 in2  (433.5 cm2) for Each Rack.
        Ionization Volume  =  645 in^  (10,570 cm^).

     The electron beam is delivered into a region bounded by the two parallel
racks of rod electrodes.  Electron beam ionization forms a plasma of posi-
tive and negative ions and free electrons between the racks.  The charge
density is separated by an electric field which is imposed by the precharger
electrodes.  The upstream cathode rack (-) attracts positive ions while the
downstream anode rack (+)  attracts negative ions and electrons.  As each
electrode is approached from the center,  the average charge becomes increas-
ingly monopolar.  Particles in the gas stream traversing the charging zone
are first charged positively near the cathode, pass through the central
"grounded" plasma, and then charged negatively near the anode boundary
region.  Charged particles pass through the anode rack because their mobil-
ity is at least several orders of magnitude lower than that of ions and
electrons, which migrate rapidly to electrodes of opposite polarity.  Ions
and electrons not attached to particles ("waste" ions)  are attracted to the
racks and removed from the flow going to the collector.

     One effect of electron beam precharging which may occur is that of
particle impaction on the anode leading edge, causing a highly resistive
buildup or cake which may reduce the maximum electric field within the
charging region.  Solutions are envisioned which include mechanical elec-
trode cleaning, addition of guard electrodes, and resistivity modification


                                    103

-------
of the dust cake.  Since energetic electrons ionize solid particles as well
as the gas in which they are entrained (13) , the anode leading edge caking
problem may be solved by the direct irradiation of the dust layer, leading
to increased conductivity and electric field.  Electron beam irradiation
may also reduce the resistivity of entrained particles flowing through the
precharger, but this effect must be investigated further to determine its
magnitude.

     Another possible problem in successful precharging is that of space
charge expansion of the charged aerosol after it leaves the precharger and
before it reaches the collector.  Several solutions are being studied should
the effect be large.  A configuration where an e-beam precharger could be
placed just upstream from the collector with little or no transition section
would minimize space charge problems.

                           EXPERIMENTAL RESULTS
     Experiments have determined the maximum electric field and ion current
density achievable in the two rack precharger described earlier.  The elec-
trode system was positioned in the precharging module along the electron
beam axis in an arrangement similar to that for the parallel plate electrode
geometry used in the earlier characterization of ion current density regimes.
Experimental conditions for the preliminary investigation of the rack pre-
charger are summarized as follows:  Two Rack Precharger specifications as
described above; Electron Beam Energy  =  1.2 MeV; Beam Current  =  10 to
10,000 nA; E-Beam Tube End to Precharger Rack Distance  =  50  cm.  Ion
currents vs. average electric field  (total voltage/rack separation distance)
curves were obtained for five electron beam currents and the results are
shown in Table 1.

   TABLE 1.  MAXIMUM ION CURRENT DENSITY AND SPARKOVER VOLTAGE FOR FIVE
                ELECTRON BEAM CURRENTS USING THE PARALLEL RACK
                             ELECTRODE SYSTEM


Beam
Current
(nA)
0
10
100
1,000
10,000
Maximum
Electric
Field
(kV/cm)
6.3
6.2
6.0
5.5
5.2

Sparkover
Voltage
(kV+)
48
47
46
42
40
Maximum
Ion
Current
(yA)
0
24
98
565
2350
Maximum
Ion Current
Density*
(mA/m2)
0
0.54
2.26
13.05
54.27
 *43.3  yA  ion current  =   1 mA/m2  ion current density.
                                     104

-------
     The experimental results show that as electron beam current was in-
creased, the maximum electric field and sparkover voltage decreased gradually.
Higher beam current caused more ionization so the interelectrode volume
became more conductive, resulting in a lower sparkover voltage.  Sparking was
between cathode and anode rods at a variety of locations.  The maximum aver-
age electric field obtained was above 6 kV/cm which is high, but it is not
the highest attainable.  Electric fields for particle charging of 8-10 kV/cm
are anticipated with improved precharger design and construction techniques.
Collected ion current increased with beam current because more ions were
formed by the intensified beam.  At the maximum achievable electric field
for each beam current, very high ion current densities were obtained.  For
the described rack geometry the ion current density of 54 mA/m2 at 10,000 nA
(10 viA) beam current is at least 250 times as high as that found in a conven-
tional corona wire electrostatic precipitator with similar dimensions (14).
Measured ion current densities using the two rack configuration compare
favorably with results achieved using a parallel plate electrode system.

      Two plots  of ion current and  ion  current density vs. precharger elec-
 tric  field  strength  are  shown in Figures  5  and  6.  At an electron beam  cur-
 rent  of 10  nA,  the shape of  the curve  is  similar  to  that found in previous
 experiments using the parallel plate electrode  system.  The primary ioni-
 zation section  begins at the origin and the  effect of saturation is
 observed at an  electric  field of 0.5 kV/cm.  Secondary  ionization effects
 begin to obscure  the saturation plateau at  approximately 2.5 kV/cm, and
 the curve is terminated  by a slight upturn  at 6.0 kV/cm.

     In contrast,  the curve for 10 yA beam current is rather linear (Figure
6) .  No separate regimes of primary,  saturation, or secondary ionization can
be discerned when the beam current is increased three orders of magnitude.
At this current level the copious primary ionization cannot be completely
separated so no saturation plateau appears.   Secondary ionization effects
contribute to the total measured ion current at higher electric field.

                          SUMMARY AND CONCLUSIONS
     Full utilization of domestic coal resources is one of the top priorities
of the U.S. Department of Energy, but much of the nation's coal has an ash
which is high in resistivity making it difficult to collect at high effi-
ciency for most existing electrostatic precipitators.  Programs for improve-
ment of stationary source control devices will concentrate on the removal of
small, respirable particulate matter as well as high resistivity fly ash.
Enhancing the collection efficiency while reducing the cost of ESP's has
motivated researchers to develop novel devices to charge particles upstream
from new or existing collectors.  DOE is currently supporting a program at
the Florida State University which concerns the development of the new con-
trol concept of electron beam ionization for electrostatic precipitators.
While primarily aimed at particle charging, electron beam precipitation is a
companion effort to DOE sponsored work on SO  and NO  removal by e-beam
treatment.
                                     105

-------
     Electron beam ionization investigated in a series of earlier studies
characterized three ionization regimes in a parallel electrode system.
From these studies emerged the design for an electron beam energized pre-
charger within a large laboratory-scale electrostatic precipitator.  The
Electron Beam Precipitator test system is capable of treating model aerosols
under a wide range of experimental conditions.  Specialized modules sepa-
rated by ductwork are arranged into a closed-circuit or "racetrack" con-
figuration.  Aerosols of several types which resemble high resistivity fly
ash are resuspended at the aerosol injection module by three particle gen-
eration devices.  Two monitoring modules upstream and downstream from the
precipitator accept a host of gas stream, particle detection, and charge
measurement sensing units to quantify EBP conditions.  These are integrated
with a microprocessor-based data acquisition and analysis system which
allows a wide variety of signal information to be processed efficiently.
Particle removal occurs in a typical corona wire-plate collector with SCA of
200 separated from a precharging module by a short transition section.

     The particle charging function is handled by the electron beam pre-
charger in the two-stage system.  An electron accelerator delivers an e-
beam into the precharging module through which the-test aerosol flows.  A
double rack electrode precharger provides an electric field which separates
the ionized plasma for the purpose of charging particles with a net nega-
tive charge.  Unused ions and electrons are removed from the flow because
of their high mobilities.  Ionization experiments using the double rack pre-
charger show that high electric fields (6 kV/cm) and ion current densities
(54 mA/m2)  for particle charging can be achieved.  Plots of precharger ion
current vs. electric field show that as e-beam current is increased, satura-
tion of ion current is not achieved because of the large secondary ionization
effects.

     The measurement of the charging and collection efficiency of resuspended
test aerosols similar to and including high resistivity fly ash will be the
focus of a comprehensive experimental program.  The operation of the EBP and
especially the e-beam precharger will be optimized to a configuration capable
of being applied to a new or existing ESP installation.  Modifications to
the EBP for the investigation of high temperature conditions as well as for
combined SOx/NOx/Particle treatment are feasible.  Field testing of electron
beam precipitation on the pilot scale will follow a thorough investigation
of the concept under laboratory conditions.   Electron beam precharging might
be most appropriate when applied to coal fly ash boilers with high resis-
tivity dust problems as a retrofit installation in a duct adjacent to an
existing collector.

     The work described in this paper was not funded by the U.S. Environ-
mental Protection Agency and therefore the contents do not necessarily re-
flect the views of the Agency and no official endorsement should be
inferred.
                                     106

-------
                             ACKNOWLEDGEMENTS
     The Electron Beam Precipitator program is supported in part by U.S.
Department of Energy Contract #DE-AC21-81MC16229.  The authors would like
to acknowledge the assistance of Kenneth J. Schafer, Stephen J. Stout,
and Robert H. Hart in the design, construction, and operation of the F.S.U,
Electron Beam Precipitator.  We also thank J. Walter Phillips for running
the 3 MeV electron accelerator.
                                REFERENCES
1.  M. H. Anderson, J. R. McDonald, J. P. Gooch, and D. V. Giovanni.
    High Intensity Ionizer Development.  In:  Proc. 3rd Symp. Transfer
    and Utilization of Particulate Control Technology, Vol. II,
    (EPA-600/9-82-005b),  pp. 334-348, 1982.

2.  D. H. Pontius and L.  E. Sparks.  A Novel Device for Charging High
    Resistivity Dust.  APCA Journal 28^(7) , pp. 698-700, July 1978.

3.  G. Rinard, M. Durham, D. Rugg, and L. E. Sparks.  Development of a
    Charging Device for High-Resistivity Dust Using Heated and Cooled
    Electrodes.  In:  Proc. 3rd Symp. Transfer and utilization of Par-
    ticulate Control Technology, Vol. II,  (EPA-600/0-82-005b), pp.  283-
    294, 1982.

4;  S. Masuda, H. Nakatani, and A. Mizuno.  Boxer-Charger Mark III  and
    Its Application in ESP's.  In:  Proc. 3rd Symp. Transfer and Utiliza-
    tion of Particulate Control Technology, Vol. II, (EPA-600/9-82-005B),
    pp. 380-389, 1982.

5.  R. H. Davis, W. C. Finney,  and L. C. Thanh.  Electron Beam lonization
    for Coal Fly Ash Precipitators.  In:  Proc. of the American Nuclear
    Society Topical Conference  on Atomic and Nuclear Methods in Fossil
    Energy Research,  Mayaguez,  Puerto Rico, pp.  481-494,  December,  1980.

6.  W. C. Finney, J. S.  Clements, and R. H. Davis.  Primary and Secondary
    lonization in an Electron Beam Precipitator System.  In;  Proc. 3rd
    Symp. Transfer and Utilization of Particulate Control Technology,
    Vol. II, (EPA-600-9-82-005b), pp. 358-369, 1982.

7.  J. S. Clements, W. C. Finney, O. Tokunaga, and R. H.  Davis.  Stable
    Secondary lonization in a Test Geometry for Electron Beam Precipita-
    tors.  In;  Conf. Record of the IEEE-Industry Applications Society
    Annual Meeting, Philadelphia, PA, pp. 1136-1141, October, 1981.


8.  R. H. Davis, W. C. Finney,  J. S. Clements, and O. Tokunaga.  Secondary
    lonization as Enhanced Radiation Dose.  In:  Conf. Record of the IEEE-
                                    107

-------
 9.
10.
11.
12.
13.
14.
Industry Applications Society Annual Meeting, San Francisco, CA,
1981.

W. C. Finney, J. S. Clements, 0. Z. Tokunaga, and R. H. Davis.
Application of Electron Beam Technology to Particulate Matter
Control.  Paper 82-27.4, 75th Annual Meeting of the Air Pollution
Control Association, New Orleans, LA, June 20-25, 1982.

M. D. Durham, G. A. Rinard, D. E. Rugg, and L. E. Sparks.  Measurement
and Interpretation of Current Density Distribution and Charge/Mass
Data.  In:  Proc. 3rd Symp. Transfer and Utilization of Particulate
Control~Technology, Vol. II,  (EPA-600/9-82-005b), pp. 54-65, 1982.

J. R. McDonald, M. H. Anderson, R. B. Mosely, and L. E. Sparks.
Charge Measurements on Individual Particles Exiting Laboratory Pre-
cipitators.  In:  Proc. 2nd Symp. Transfer and Utilization of Par-
ticulate Control Technology, Vol. II,  (EPA-600/9-80-0396), pp. 93-
113, 1980.

M. K. Mazumder, R. G. Renninger, T. H.  Chang, R. W.  Raible, W. G.
Hood, R. E. Ware, and R. A. Sims.  Simultaneous Measurements of Aero-
dynamic Size and Electric Charge of Aerosol Particles in Real Time on
a Single Particle Basis.  In:  Proc. 3rd Symp. Transfer and Utiliza-
tion of Particulate Control Technology, Vol. II,  (EPA-600/9-82-005b),
pp. 160-168, 1982.

P. J. Heerden.  The Crystal Counter, A New Instrument in Nuclear
Physics.  In;  N. V. Noord Hollandsche Uitgevers Maatschappij,
Amsterdam, 1945.
H. J. White, Industrial Electrostatic Precipitation.
Reading, Massachusetts, 1963.
Addison-Wesley,
                              Electron Beam
                      Vacuum
              Accelerator — — *• —
                       Window
                                                        Separated
                                                        Charge
                                                        Density
                                   Plasma
Figure 1.  Electron beam ionization in air.   An accelerated electron beam
           passes from vacuum into air through a thin foil window and
           ionizes air molecules.  The plasma can be separated using an
           electric field.
                                     108

-------
    Figure 2.


Plan and front  views
of the Electron Beam
Precipitator, showing
the e-beam precharger,
the collector,  and
other specialized
modules.
                               PARTICLE
                               INJECTION
                   WINDOWS
                                      FAN
   f                         FILTER
    3'    PRECHARGER
   j_    ,_/,     COLLECTOR
MONITORS MM
                             i9"--2'
                                       \
                                            PLAN VIEW
                     -7'6"-
                                             WINDOWS
                                                            MONITORS
                                                   VIEW
   MICROCOMPUTER DATA
   ACQUISITION   SYSTEM
   CONTROL    n    TARGET
    ROOM      U     ROOM
             WALL
                 PARTICLE SIZE
                 TEMP/HUMIDITY

                 AIR VELOCITY
                 TEMPERTURE
                    DATA
                 ACQUISITION
                  SYSTEM
                   A PRECHARGER
                     CURRENTS
                     VOLTAGES

                 COLLECTOR
                  CURRENTS
                  VOLTAGES
                      Figure 3.

         Layout of the microprocessor-based
         data acquisition  and analysis  system
         including all electrical inputs  from
         the EBP monitoring systems.
                                  109

-------
                                       Beam Dump
                                                    /Support
                                                      Frame
                 Air     \  (
             Temperature   I
                 and  -
               Velocity
             Particle
             Number
              and
              Size
                        Scattering
                          Foil
                             Electron Beam  e~
                                 Tube
                                                        Moveable
                                                       Beam Current
                                                         Probe
             Figure  4.   Top view of the precharging module  showing
                         the major functional components.  Particles
                         are charged in the e-beam  energized ionization
                         zone which is bounded by the double rack pre-
                         charger.
 24r


 20 •


Sl6-


1.2
         Electron Beam Current = 10 nA
         1/16" Baffle
         S=l5.24cm
                                               2400
0.5 _
  N


0.4 <
                                      0.3 g
                                        e
                                      0.2 £
                                        3
                                      O.I 5
      01     23456
                 E = V/S(kV/cm)

Figure  5.  Plot of anode  ion current
            and ion current density vs.
            applied electric field for
            10 nA electron beam current.
Electron Beam Current =IO/i A
1/2" Baffle
S=l5.24cm
      Figure 6.
                            34
                         = V/S(kV/cm)
   Plot of anode ion current
   and ion current density vs.
   applied electric field for
   10  yA electron beam current
                                          110

-------
  EXPERIMENTS WITH WIDE DUCTS IN ELECTROSTATIC PRECIPITATORS
             by : Ekkehard Weber
                  Helmut Wiggers

                  University of Essen
                  D 43OO Essen 1, West Germany
                            ABSTRACT

     The repeatedly observed enhancement of the effective
migration velocity with increasing duct width holds out the
prospect of less expensive electrostatic precipitators.
Vagueness still exists regarding the optimum of the width and
the influencing factors on it. Based on laboratory experiments
indicating almost a proportionality between migration velocity
and duct width up to 715 mm a pilot precipitator with a maximum
flow rate of 36/OOO m3/h was constructed. Parallel to large
scale conventional electrostatic precipitators for a coal-fired
power station and the romm dedustion of a sinter plant duct
widths up to 1,OOO mm were tested. Besides a far reaching
conformity with the laboratory results an optimum width was
ascertained, which is influenced for instance by the gas
velocity. Moreover, with a theoretical precipitator model it
succeeded to explain the corresponding influences on the
migration velocity.
                              Ill

-------
                            INTRODUCTION
     With increasing requirements regarding the invironmental
protection the interest grows to optimize the electrostatic
precipitators, which are frequently used to dedust industrial
exhaust gases. The operation of an electrostatic precipitator
is often described on the basis of Deutsch's equation  (1)

                 c = c   • exp  ( - w A/V  )

that represents the relationship between the dust concentrations
of the crude and the clean gas, co respectively c, the total
collecting area A, the volume flow rate V and the migration
velocity w. According to the Deutsch theorie the migration
velocity can be considered as the effective velocity of the dust
particles towards the precipitation electrodes. Up to now it is
not possible to calculate the migration velocity from theory.
Therefore it represents a process coefficient defined by the
Deutsch equation.

     The nowadays used electrostatic precipitators are mostly
of the plate typ, where the precipitation plates form numerous
ducts appropriate to their parallel grouping. The gas to be
cleaned flows in proportion through the equal wide ducts.
Sparking electrodes are arranged along the centre line of each
duct. As well for the precipitation electrodes as for the
sparking electrodes rapper mechanisms are installed to remove
in periodic time intervals the precipitated dust from the
electrodes.

     As a measure for the size and for the costs of an electro-
static precipitator the precipitation area of the plate elec-
trodes can be referred to. Therefore, a substantial improvement
of the electrostatic precipitators in regard to their costs
implies to minimize the necessary precipitation area without
producing overproportional additional costs at other parts of
the precipitators. According to the Deutsch equation a reduction
of the precipitation area can be obtained only by a simulta-
neous increase of the migration velocity, for in each applica-
tion of an electrostatic precipitator the volume flow rate and
the dust concentrations of crude and clean gas are defined.

     To design an electrostatic precipitator one has to use
extensively the experiences with already existing installations.
Because fundamental changes of the construction involve an
enormous risk the influences of several design parameters on the
migration velocity yet need to be investigated. This is valid
for instance for the duct width, that in conventional electro-
static precipitators has presently a value between 250 and
30O mm. Several investigations some years ago already indicated
that wider ducts can effect higher migration velocities


                               112

-------
 ( 2 f 3,4 ).  However the results of the corresponding works often
received scepticism since they could not be explained with the
common idea about the motion of the particles in an electro-
static precipitator. The same happened to the repeatedly
observed increase of the migration velocity with increasing gas
velocity.  Considering especially the possible decrease in costs
of the precipitator construction it seemed therefore necessary
to clarify the interrelationship between the precipitation
efficiency, the duct width and the migration velocity.

                       LABORATORY EXPERIMENTS

     To clarify the above mentioned questions first of all
extensive laboratory investigations were performed. Figure 1
shows the outline and the most significant data of the used
laboratory electrostatic precipitator. It contained one duct
with a height of 0.7 m and a length of 3 m. The flat precipita-
tion electrodes without any profil were fixed in a frame per-
mitting to vary continuously the distance of the precipitation
electrodes, that is the duct width, up to 715 mm. The duct was
covered above and below with plates of plexiglass, which were
used simultaneously as electrical isolating mountings for the
sparking electrodes. These electrodes were punched from 1.5 mm
sheet-steel and had rounded emission spikes arranged mutually
in a distance of 25 mm. The distance of the sparking electrodes
from each other was biased by the actual duct width. Usually
the corresponding relation of the distances was 0.96, which
from experience is in the optimum range.

     The laboratory electrostatic precipitator operated with
negative corona. For that a 200 kV high-tension equipment
could be used. The voltage was adjustable by hand and by
automatic.  To smooth the high-tension a 25 nF capacitor was
wired parallel to the precipitator. A high voltage-stable
resistor of 1 kJ2 was in series with the capacitor to allow
only an aperiodic discharge of the capacitor in case of spark
over in the precipitator.

     As the gas to be cleaned the air of the laboratory with a
temperature of 21 °C was used, in which before quartz powder
was injected. The particles of the quartz-powder had a mass
median diameter of 9 pm. 1O % of its mass fraction were larger
than 23 pm and 1O % were smaller than 1.8 pm. The value of the
specific dust resistance was between 1O^ and 101° flcm depending
on the relative humidity of the gas, which was between 43 and
64 %.  Hence normally no back corona was observed.

     The electrostatic precipitator was on the suction side of a
fan that could effect a volume flow rate up to 2,900 m3/h.
A fabric filter between the electrostatic :precipitator and the
fan permitted to operate the electrostatic precipitator also


                               113

-------
      high voltage
precipitation
plates
                                              outlet
                      frame
                  gas flow rate: variable, up to 2,900 m3/h
                  number of ducts:  1
                  duct width : variable, up to  715 mm
                  plane precipitation electrodes
                  height of precipitation plates :   0.70m
                  length of precipitation plates :   3.00 m
                  total collection area   :        4.2  m2
                  voltage: negative corona, up to 200 kV
Figure 1. Fundamental Structure and Data  of
            the Laboratory Electrostatic
            Precipitator
                         114

-------
with reduced efficiency or even without voltage. As precipitator
inlet a 2.5 m long quieting duct without things build in was
mounted, and before the inlet a mass force separator retained
the larger particles of the gas-dust mixture. The laboratory
air was sucked through a broad slit into the separator. Hence
the pressure in the electrostatic precipitator was nearly that
of the atmosphere. Because the laboratory electrostatic
precipitator contained neither a rapper system nor hoppers,
after each experimental setting it was cleaned by hand for the
next measurements.

     The results of the precipitation measurements are represen-
ted in the two following drawings. Figure 2 shows the migration
velocity at 1.0 and 2.O m/s gas velocity, plotted as a function
of the duct width. With regard to the exactness of the single
measuring points the influence can be described as a proportio-
nality between migration velocity and duct width. That is
equivalent to a constant efficiency with varying duct width.
Moreover it has to be noticed that by doubling the gas velocity
from 1.0 to 2.0 m/s the migration velocity becomes also nearly
doubled.
     In figure 3 the migration velocity at gas velocities of
1.5 and 3.0 m/s are outlined, again calculated with the. defining
Deutsch equation from the precipitation results. Once more the
influence of the duct width on the migration velocity can be
described as proportionality. By doubling the gas velocity the
migration velocity increases anew, but now less than before  (5).

     The discovered proportionalities between duct width and
migration velocity as well as the influence of the gas velocity
in the precipitator on the migration velocity could not be ex-
plained with the hitherto existing theories. Besides to the
economic interesting question whether or to what extend these
laboratory results are valid for large scale precipitators under
industrial conditions, acordinly the demand arose to explain
the found dependences.

                              THEORY

     The inability of the Deutsch theory to explain the observed
influences made it necessary to introduce decisive modifiaations.
Measurements of dust concentration profiles in the laboratory
precipitator indicated that the swirling of the dust particles
in the duct is not so complete as assumed before but exists
nevertheless in a certain grade. The thereby produced incidental-
like motion of the particles, which is superimposed to their
drift to the precipitation electrodes, was described with the
model of turbulent diffusion. On simplifying assumptions, for
instance a constant drift velocity of the particles to the


                               115

-------
    100-
   cm/s
Q)
C
.0
•c
I
50*-
                  200
                                     600
mm
800
                                400
                            duct -width
Figure 2. Influence of the Duct-Width on the Migration Velocity
           at Gas Velocities of 1D and 2.0 m/s

-------
    100-
   cm/s
I
.§
50-
     0
       0
                               v=1.5m/s
              200
600
mm
800
                                £.00
                            duct - width
Figure 3. Influence of the Duct-Width on the Migration Velocity
           at Gas Velocities  of 1.5 and 3.0  m/s

-------
precipitation electrodes and a constant diffusion coefficient,
a mass balance for a volume element led to a linear differential
equation of the second order, that is reproduced in figure 4.
The boundary conditions necessary to solve the differential
equation are specified below. Boundary condition I means a
constant dust concentration at the inlet of the precipitator.
Condition II expresses, that there exists no effective mass
flow through the center plane of the duct. Boundary condition
III is a novel one. It originated from observations of preci-
pitated dust at the plate electrodes leading to the assumption
that part of the precipitated dust reentrains permanently into
the gas flow. This was described with a constant dust concentra-
tion at the surface of the precipitation electrodes (6).

     The differential equation was solved numerically and with
the Deutsch equation the migration velocity was calculated. With
acceptable values for the diffusion coefficent, for the drift
velocity and for the dust concentration at the precipitation
electrode it succeeded in reproducing the observed influences
of the duct width and of the gas velocity on the migration
velocity. According to that the measuring results are in a
near relation to the partial reentrainment of already precipita-
ted dust particles into the gas flow.

     The capacity of the theoretical model turned out to be
even much larger. Besides the reproduction of the influences of
duct width and gas velocity on the migration velocity it offered
an explanation for the often observed improvement of the preci-
pitation with  increased dust concentration in the crude gas.
Concerning the specific dust precipitation observed deviations
from the usually logarithmic behaviour could be theoretically
reproduced too. Moreover the theoretical model offered an ex-
plaination for various dust concentration profiles publicated
in the literature.

     Considering the power of the theoretical model also its
predictions were of interest. With regard to the influence of
the duct width on the migration velocity the following was to
see : The increase of the migration velocity with increasing
duct width is initially nearly proportional, then it becomes
weaker and finally a maximum is reached. This maximum occurs
all the sooner as higher the gas velocity is. A restriction
of the fundamental influences to small electrostatic precipi-
tator s could not be seen, therefore the hitherto obtained
experimental results should be valid at least in their tendency
for large scale precipitators.
                               118

-------
     U
             AX
AX   ;
                                                   = c
ucl  n=
   ly=°
                                                    ay
= o
 (D
 IE)
 (M)
                                       a = /?a/^ - w/df/7 of
                                       c = particle concentration
                                       D s eddy diffusivity
                                       u = drift  velocity
                                       v - gas velocity
                                       x - distance from inlet
                                       y = distance from sparking plane
Figure 4. Theoretical  Model of Electrostatic Precipitation

-------
                   PILOT PRECIPITATOR  EXPERIMENTS

     To test the laboratory results with regard to their validi-
ty under industrial conditions and their transferability to
larger precipitators a large pilot electrostatic precipitator
was constructed. Figure 5 presents a view of the precipitator
and the most important data. The pilot precipitator consisted
of two sections each having two ducts. The duct width was
variable up to 1,000 mm. The effective hight of the precipita-
tion electrodes came to 2.5 m, the length in each section was
3.0 m. This results in an effective precipitation area of 60 m2.
The volume flow rate through the electrostatic precipitator was
adjustable up to 36,000 m3/h. As power supply a'280 kV transfor-
mer was installed, its negative voltage being adjusted automa-
tically to the highest value actually possible. Both sections
of the pilot electrostatic precipitator were provided with
rapper systems working independently and automatically, and
the dust in the hoppers was removed automatically too. So the
pilot precipitator could be operated continuously.

     The first application of the pilot electrostatic precipita-
tor was in a coal-fired power station. It operated in parallel
to a large scale electrostatic precipitator that had to dedust
the flue gases of a steam boiler plant of 150 MW fired with
pulverized coal. During the tests the gas in the pilot precipi-
tator had a temperature of about 120 °C. Under these conditions
the specific resistance of the dust was at 4 • 1012 Qcm. Duct
widths from 250 up to 1,OOO mm were tested. The gas velocity
in the pilot precipitator was varied between 0.5 and 2.0 m/s.
Figure 6 shows the influence of the duct width on the migration
velocity as it results from a regression analysis that conside-
red also the influence of the crude gas dust concentration on
the precipitation. To elucidate the nature of the law the
curves are extrapolated beyond the measured duct width interval.
Therewith it can be seen that initially the migration velocity
increases almost proportionally to the .duct width. Then the
increase weakens and following a maximum the migration velocity
decreases with further increasing duct width. The duct width,
at which the maximum of the migration velocity occurs, depends
from the gas velocity. The maximum is reached as sooner as
higher the gas velocity is. Also an increase of the migration
velocity with increasing gas velocity can be noticed, though
at wider ducts a decrease too. In other respects, if the
migration velocity of the large scale precipitator is checked
with the comparable one of the pilot precipitator, correspondence
can be stated.

     As the second application for the pilot precipitator the
room dedustion of an iron ore sintering plant was choosed.
From the crude gas flow of the again already existing large
scale precipitator a part was separated and after dedustion

                               120

-------
                           gas flow rate:
                               variable, up to  36,000 m3/h
                           number of ducts:  2
                           duct-width: variable, up to 1000mm
                           height of duct:  2.5 m
                           number of sections : 2
                           total length of ducts 2x3m = 6 m
                           total collecting area : 60 m2
                           voltage :
                           negative corona, up to 280 kVp
Figure 5: View and Fundamental Data of the
           Pilot Electrostatic  Precipitator

-------
      cm/s
    £
    o
    c
    .o
    I.
to
N)
           V =2.0/77/5
           V =7.5/77/5
           v = 1.0 m/s
           v=0.5m/s
mm
          0                   500                  7,000
                                     duct- width
    Figured. Influence of Duct-Width  on the Migration  Velocity
              at Various  Gas  Velocities ( Fly Ash Collect/on)

-------
was led back to the inlet of the large precipitator. The same
duct widths and gas velocities as before were tested. The gas
temperature in the pilot precipitator was now about 90 °C. The
numerous measured values of the penetration were again analysed
with a multi regression method. The result is represented in
figure 7- The approximate proportionality between duct width
and migration velocity at smaller duct widths can be seen
clearly. The influence of the gas velocity on the position
of the migration velocity maximum is in this case less evident.
But also the increase of the migration velocity with increasing
gas velocity is again very clear. A comparison of the migration
velocity of the large electrostatic precipitator with that of
the pilot precipitator shows again correspondence.

                           CONCLUSIONS

     On the basis of experiments with a laboratory electrostatic
precipitator the duct width of a larger pilot electrostatic
precipitator was varied up to 1,000 mm. The tests were performed
under several industrial conditions. As a result it can be
noticed that at narrow   duct widths like today in use the
migration velocity is nearly proportional to the duct width.
A maximum was found only at wider ducts and dependent from the
gas velocity. To summarize theoretical and experimental
investigations they confirmed the following : By increasing
the usual duct width the necessary precipitation area can be
reduced while the efficiency remains constant. That is aquiva-
lent to a reduction of the costintensive things built in of an
electrostatic precipitator. With this compared the costs for
voltage increase of the high tension equipment can be neglected.
It is imaginable that the position of the maximum is influenced
by the duct length too. But this has yet to be investigated.
Summarizing it can be stated that electrostatic precipitators
with wide ducts of more than 500 mm are applicable and in-
teresting for industrial use. Without reducing their efficiency
they can be constructed and operated with essential less ex-
penditure as that of the nowadays used precipitators.

                           ACKNOWLEDGEMENT

     The necessary financal aids to perform the extensive ex-
periments of the all together two research projects were granted
by the Home Secretary of the Federal Republic of Germany,
Umweltbundesamt.
                               123

-------
10

Ja.
      cmls
    o
    o
c
.o
   I
                                                 v= 1.5 m/s


                                                 v =2.0 m/s
                                                  X  v=1.0m/s
                                                     v =0.5 m/s
          0              500             WOO     mm    1,500

                              duct-width

     Figure 7. Influence of Duct-Width on  the Migration  Velocity

              ot Various Gas Velocities (Room Dedust/on)

-------
                           REFERENCES
1.   Deutsch,  W.
2.   Aureille,  R.;  Blanchot,P.
3.   Gubner,  0.
4.   Heinrich,  D.O.
5.   Weber,  E.;  Wiggers,  H.
6.   Wiggers,  H.
Bewegung und Ladung der Elektri-
zitatstrager im Zylinderkonden-
sator.
Annalen der Physik, IV. Folge,
68, pp. 335-344

Experimentelle Untersuchung ver-
schiedener Parameter auf den
Wirkungsgrad eines Elektrofliters,
Staub - Reinhalt. Luft 31  ( 1971)
Nr. 9, pp. 371 - 375

Einfliisse auf die Wanderungsge-
schwindigkeit, nachgewiesen mit
einem Versuchselektrofliter.
Dissertation, Universitat Essen,
1976

Der groSe Gassenabstand im
Elektrofilterbau.
Staub - Reinhalt. Luft 38  ( 1978)
Nr. 11, pp. 446 - 451

Untersuchungen an einem Labor-
elektrofilter iiber die Einfliisse
der Gassenbreite und der Gasge-
schwindigkeit auf das Abscheide-
verhalten und die Staubkonzen-
trationsverteilung.
Staub - Reinhalt. Luft 4O  ( 1980)
Nr. 11, pp. 469 - 473

Untersuchungen an einem Labor-
elektrofliter, insbesondere iiber
den EinfluB der Gasgeschwindig-
keit und der Gassenbreite auf
das Abscheideverhalten.
Dissertation, Universitat Essen,
1982
                              125

-------
              A RECONCILIATION:  WIDE VERSUS NARROW  SPACED
                  COLLECTING PLATES FOR PRECIPITATORS

                  by:   Dennis G. Puttick
                        Peabody Sturtevant Ltd.
                        London, N.14.
                                ABSTRACT

     MLsaka of Hitachi, amongst others, has presented evidence for the
improvement in electrostatic precipitator performance, using much wider
gas passage spacing than normal (typically 20 in.).  Such designs are
now in general use.

     This appeared to be directly contrary to the experience of Peabody
Sturtevant who, on very large utility ESP plants treating highly
resistive particulates, produce remarkable efficiencies with narrower
spacing than normal (typically 8 in. - 9 in.) using a design first
developed during the mid-1960's.

     In developing his reconciliation, the author uses observations of
the actual internal construction of different designs of precipitators
and the effect that increasing variance from perfect electrode geometry
has upon total power input and field intensity at various plate spacings.

     This return to basic precipitation electrostatic theory demonstrates
not only that Peabody Sturtevant is right to pursue its narrow spacing
design but other companies,  with their own particular standard of
construction, may be equally right in developing wider spacer techniques.
                                    126

-------
     It is becoming established in the electrostatic precipitator field
that increasing the duct spacing, i.e. the distance between adjacent
collecting plates, can result in a higher achieved efficiency.  There
has therefore been a tendency by many companies in the field to put forward
wider plate spacing to achieve economy in total collecting plate area and
more particularly as an answer to problems that have been encountered with
higher resistivity dusts.

     These notes propose the reasons why many plant designs have achieved
better results with wider spacing and make a comparison between the
Sturtevant narrow spacing system with wider spaced systems.

     For many years, the norm for precipitator collector spacing
(including the Sturtevant spacing) was between 10" and 12".  In the early
60's, Sturtevant produced their first 8" narrow spaced precipitator.
Various companies also produced narrow spaced precipitators down to 6"
duct size using the package principle and with only 3 to 4M plate heights.
These have generally worked well.

     However, in the last 20 years or so for reasons both of economy and
ground space limitations, precipitators have got taller.  Plate heights
have increased from typically about 30 ft. (9 metres) to 40 to 50 ft. high
(12 to 15 metres).

     It is also during this period that the vogue for wider duct spacings
has occurred.  Anybody familiar with the internal design of precipitators
particularly with their final practical embodiment will be aware of the
rapidly increasing divergencies from ideal in internal electrical
clearances with increasing plate height.

     Most plate designs are composed of vertical strips fixed to a support
beam at the top and a fixing beam at the bottom with either top or bottom
supports being the point of rapping.

     Most of these plate designs utilise cold rolled sections.  Although
one in 500 is a typical straightness tolerance for most purposes, for
precipitator applications this is improved to a tolerance of about one
in 1,000.  This tolerance means a possible inbuilt bow of up to 15mm in
a 15M high plate at the production stage irrespective of any other
distortions that may be produced during shipping and erection.
Precipitator suppliers always attempt to improve on this but obviously
the inherent problems easily result in the acceptance of plate
straightness very far from the perfect plate alignment situation.

     Until the 1970's, the precipitator world was divided equally between
wire and weight designs and wire in frame designs.  Currently, the vogue
is fairly equally divided between frames and "rigid" masts;  the latter
frequently and often hopefully using some means of promoting a better
corona than round or square wires.  Wire and weight electrode systems
have suffered in reputation because of poor wire designs used by some
manufacturers resulting in excessive operational failures.  However,
                                    127

-------
 Peabody Sturtevant are still using this  design with considerable success
 and a minimal failure rate along with  both frame and mast designs.  The
 simplicity of a mast design should lend  itself to inherently better
 clearances than the more complex frame.   Our  experience is that this is so.

      In an attempt to overcome clearance problems for high plates, frames
 are frequently made in two sections with separate rapping for each
 section.  However, this system requires  further increased complexity of
 support steelwork and when dealing with  many  hundreds of lanes in a large
 installation, the chance of at least one mis-aligned frame occurring
 (thus effecting the whole stage) is considerable.

      For a strip plate frame system, a reasonably "thin"- plate strip will
 be about 35mm.  The outside dimensions of the frame tubing will be similar
 and often a little greater.  This means  that  with a 250mm duct spacing,
 the actual electrical clearances with  perfect alignment will, be 90mm,
 i.e. a clearance reduction of 35mm.  This will obviously occur at many
 points over the plate where the horizontal tubing of the frame passes the
 thickest parts of each plate strip.  Typical  plate sections are as shown
 in Fig. 1.
Various Types Of Collecting Electrodes
                                              Voltage/Current Parameters
            Shielded Plates
                           V-Pockets
                           V-Plates
                           Strip Plates
                    Flat Plats
                                                  Plate
                                                            Duct Spacing
                                                          Si
                                                 Thickness
                                                                Discharge Electrode
                                                                   Thickness
                FIG. 1                               FIGl  2
      With the Sturtevant flat plate/wire and weight system,  a series of
 individual wires pass from a top grid suspension through and below the
 plates to bottom grid tensioning.   The minimum clearance is  at the top
 and bottom of the plates where horizontal  stiffening tubes 38mm O.D.
 occur-  This gives a perfect alignment clearance reduction of 20.5mm.
                                     128

-------
     Between top and bottom stiffening tubes on this flat collecting plate,
horizontal joggles give a minimum protrusion into the gas stream from the
centre line of the plate.  In the worst case, there is only a 10mm
clearance reduction.

     At the top and bottom of a Sturtevant flat plate, a practical on-site
variation between wire and plate will be not more than plus or minus
3-4mm (say worse case situation of 6mm).  This gives a total clearance
reduction of 26.5mm at these points.  With say a 13M high plate, it will
be jig spaced at top, bottom and two intermediate points.  This will
effectively, within 2 or 3mm, ensure the verticality of the plate at all
points.  The clearance reduction at the joggles between top and bottom
stiffening tube will, therefore, be well within 26.5mm.  This means,
therefore, that for a precipitator equipped with Sturtevant flat plates
and wires, the clearance reduction for a practical operating embodiment
of this design should not be more than this 26.5mm.

     With a frame design as stated above, the clearance reduction with
perfect alignment will be about 35mm.  However, of more significance than
this is the much poorer final site clearances that will be achieved
compared to Sturtevant's flat plate/wire systems.  With the best efforts,
practical out of alignment figures with a frame of 10, 15, or even 20mm
compared to the Sturtevant flat plate/wire, 6mm are possible when the
manufacturing and assembling difficulties are considered.

     In fact, these notes came about as the result of much time spent
observing the internal alignment of many competitive makes of precipitator
and the idea was formed that the pressure towards wide spacing was
possibly due to the increasing clearance reduction observed in tall strip
plate/framed type precipitators.

     It was thought worthwhile to look at standard published formulae
relating corona emissions to corona discharge distances and also voltage
flashover distances.  These distances are illustrated diagrammatically
in Fig. 2.

     The corona, current flow for a particular operating kV will be
dependent upon the distance from the point of emission discharge to the
mean centre line of the plate.  In practical terms, half the duct width;
i.e. Si on Fig. 2.  The voltage actually achievable will be dependent
upon the spark gap distance,  i.e. Se on Fig. 2.

     Three standard formulae are relevant all of which for example appear
in somewhat similar form in White - Electrostatic Precipitation;  quoting
earlier sources.

     These formulae are as follows:-

              E   =   Kl.Se                                    	 (1)
              Where   Kl  =  constant
                      E   =  spark over voltage
                      Se  =  spark over distance H.T. to earth.

                                    129

-------
     The second formula is:

              I   =    K2

                      Si
                         E  (E - Eo)

                           Log 4Si
                                                       (2)
              Where I   =  corona emission per unit length of wire
                    K2  =  constant
                    Eo  =  corona onset voltage
                    Si  =  half duct width
                    R   =  radius (or equivalent radius) of emitter.
The final formula is:

         Eo  =   KS.R.Log.
                                                                      (3)
                                 2Si                 >

              Where K3  =  constant.

     KL,  K2 and K3 are constants dependent upon the particular precipitator
     operating characteristics.

     Prom these established formulae, we can proceed as follows:-

              Combine the square of (1) with (2):
Ely =   Kl2.K2.Se2
                             Si'
                               (E - Eo)

                               Log 4Si

                                   XR
                                                                      (4)
              Inserting both (1) and (3) in (4), we have:
              EIW =
KL'
-.K2.Se* .
S?
(Kl.Se-K3 Log 2
Log 4Si
TR
ST)

                                                                 (5)
     This formula is for the power discharge for one discharge electrode
per unit height of this electrode.  To convert this to power per unit
area, we must know the discharge electrode spacing inlet to outlet of
the plant.  The formula (2) is in fact based on an electrode spacing
which is equivalent to duct spacing.
                                    130

-------
     Therefore,  total power per unit height of all discharge
     electrodes                 =    EL,  x  n.


     The effective plate length -    2 x Si x n.

     n  =  number of wires in plate length under consideration.

     Therefore,  power per unit
     area of collecting plate   =    EL,                       .....  (6 )

                                     2Si

     This gives a final equation for total power as follows:-

     El (total power)  =
                                                                      (?)

                               2.Se
                                             7TR


     For comparison between Si and Se with all other things being equal,
it is necessary to assume a typical plant and a radius of a round
collecting wire (or an equivalent radius if a round wire is not used)
and also to determine Kl and K3.  Since we are considering only a ratio
of power with similar conditions, Kl .K2. can be ignored.

     For particular operating conditions, we can obtain Kl from
formula (1) and K3 from formula  (3);  that is,  by inserting E, Eo, Se,
Si and R obtained from a plant operating under reasonably constant
conditions .

     Typical plant operating conditions have been obtained as follows :-

     Operating on pf dust at 200 spacing with a total clearance reduction
of 26mm and with corona onset voltage at 17kV and flashover voltage at
33kV with 1.5 equivalent radius electrode.

     A second condition is under the same operating conditions with 200
duct spacing and 13kV corona onset voltage with 30kV flashover voltage,
31mm total clearance reduction and an equivalent discharge electrode
radius of 1.05mm.

     A further practical example under not too dissimilar operating
conditions is with a framed system with 250mm duct spacing and 19kV
onset voltage with 45kV flashover voltage and a total clearance reduction
of 4 1mm with 1.4 equivalent discharge electrode radius.

     For these particular cases, Kl and K3 can be calculated.  Ratios of
power input can then be obtained for differing total clearance reductions
and increasing plate spacing.  In all three cases, the results were very
similar and the form of the curves produced was the same.  Fig. 3 shows
the resulting curves.

                                    131

-------
        Relative Power Input
               (Ibtal Power)
       Flat plate/wire
              fypical framed system
1-2-


1-1-


1-0-


&9-


0-8-


0-7-


0-6-


05-


0-4-


0-3-
            I    I    I    I
            Plate Spacing mm


              FIG. 3
                                        0-2-
     Relative Power Input
         (Average Field Intensity)
                                                         Wire & Weight
                                                        Framed
       \    \
      25O  300
 I
4OO
 I
450
                          500
           Plate Spacing mm

             FIG. 4
     As can be seen from Fig. 3, there  is  an increase in total power
with increase in plate spacing, under perfect alignment conditions
comparing a flat plate/wire plant with  a typical framed system.  This
also illustrates the difference in total power between the two systems.
The Sturtevant 200mm spacing is equivalent to a 250mm frame system in
terms of total power.

     Also, from this, one might conclude that there is no limit to the
improvement that can be obtained by  increasing plate spacing.

     However, this is not necessarily correct.  For practical purposes,
an empirical factor  "w", the effective  migration velocity is used to
determine achievable efficiency for  a particular plant inlet volume and
collecting area.  In its absolute sense "w" is the velocity of the
average typical particle towards the plate at right angles to the plate.
Obviously, the motive power determining this absolute velocity is the
immediate electrical environment of  the particle.  That is, the average
field intensity, i.e. the power per  centimetre of space gap per square
metre of plate area.

     Any increase in this average  field intensity should produce a higher
actual  "w" and therefore a higher  efficiency.  To obtain this value,  it
is necessary to divide  Equation  (7)  by 2Si.  This gives us a field
intensity equation as follows:-
                                     132

-------
     El (field intensity)   =
                            ?       ?                M
                          Kl  .K2.Se .(Kl.Se-K3.Log 2Si)
(8)
     Using similar plant  factors as before,  this  gives ratios of field
intensities as shown  on  Fig.  4.

     It is seen that  under perfect alignment conditions, there is
considerable difference  in field intensity  between a framed and a wire
and weight system.  Although there is shown an increase in field intensity
with plate spacing, a point is soon reached when  increasing the plate
spacing causes a fall in field intensity.   This occurs earlier with a
wire and weight system than with a framed system.
                                             Relative Power Input
                                                 (Average Field Intensity)
  Relative Power Input
      (Average Field Intensity)
  1-2-


  1-1-


  1-0-


  09-


  0-8-


  07-


  0-6-


  0-5-


  O4-


  0-3-
 0-2-
Perfect Alignment
6mm Discrepancy
                    Total Clearance
                    Reduction mm
   Wire and Weight System
1-2-
1-1-
1-0-
0-9-
0-8-
0-7-
0-6-
0-5-
0-4-
0-3-
\
\
\
\
\
^Alffint^^^ TbtatClearance
//5mmDiscrepancy"^^Reduction mm
/ 10mm i\' ^^^^^^^^33-0
\ . 	 38-0
15mm XT 7^-»°
/ 	 	 \~ ' 48'0
Line of Optimum Plate Spacing
/ ,<*20mm
/ Framed Systems
   200  25O  300  3SO  400  45O
                                          200  250  300  350  400  450  500
            Plate Spacing mm
               FIG; 5
                                              Plate Spacing mm
                                                 FIG.' 6
     Fig. 5 shows the effect on a Sturtevant flat plate/wire  system of
taking into account discrepancies from perfect alignment.   The  average
field intensity falls quite sharply with this misalignment.   The 6mm
discrepancy is  the maximum misalignment to be expected from this plant.
ffowever, it does show that the optimum field intensity is to  be found at
about 250mm.  However, between 200 and 350mm, the field intensity varies
comparatively slightly between plus and minus 3%-  The pointlessness of
increasing the  spacing of such a plant beyond about 250mm is  clearly
illustrated.
                                     133

-------
     In Fig. 6  is shown the perfect alignment situation for  a framed
system along with a series of curves  showing the effect of increasing
misalignment.

     This  shows the devastating effect  on field intensity with discrepancies
which are  not uncommon in framed precipitators.

     The other  important point is that  increasing discrepancies demand
increasing plate spacing for optimum  field intensity and  therefore for
maximum efficiency for the collecting plate area utilised.
      Relative Power Input
          (Jberage Field Intensity)

       Perfect Alignment

                  Wire&Weight
Relative Power Input
                       Total Clearance
                        Reduction mm
           Line of Optimum Plate Spacing
                                           0-9-
                                           08-
                                           0-5-
 0-3-
 0-2.
                                           0-3-
                                            0-2.
      age Field ki
        HghRnisthRyOust
    15mm Discrepancy
    43mm Total Clearance Reduction
           300  350  4OO   450  5OO
            Plate Spacing mm
                FIG. 7
    300  350  400 450
     Plate Spacing mm

        FIG.  8
                                                                  1
                                                                 soo
      In Fig. 7, we effectively combine the field  intensity/plate spacing
 curves for the Sturtevant  flat plate/wire precipitators and framed
 precipitators.  It is  clear how much penalty a  framed system with 15mm
 or 20mm out-of-alignment discrepancy has compared to a flat plate/wire
 system.  Despite the optimisation of the flat plate/wire systems at  about
 250mm, the field intensity at 200mm is still better than even the widest
 spaced frame system.

      Thus, we arrive at a  reconciliation between the anomaly of the
 excellence in performance  of Sturtevant very narrow spaced precipitators
 and the need by many manufacturers particularly with tall plates to  build
 wider duct spaced precipitators to achieve required efficiencies.
                                      134

-------
     One other useful way  to  use Equation (8) is to  compare field intensity
ratios on the same precipitator operating under differing conditions.  On
a pf precipitator where widely differing coals were  burnt,  figures for
corona onset voltage and maximum operating voltage were  taken under low
resistivity dust conditions.   These parameters were  also taken at a later
time when the resistivity  of  the ash was very much higher.   The net effect
in this second case was a  higher corona onset voltage and a lower operating
voltage.  With these inserted in Equation (8), curves were produced as
Fig. 8.

     It must be emphasised that these curves are not a comparison of the
actual field intensity achieved in each case but only the difference in
field intensity with increasing plate spacing for the typical out-of-
alignment indicated.

     This illustrates that a  greater percentage improvement can be
expected in field intensity with increasing plate spacing when a more
highly resistive dust is being dealt with.  It also  illustrates that the
higher resistive dust requires wider plate spacing for optimum performance.

     This conclusion will  probably not come as a surprize to those working
in the field of wide spacing.

                               Relative Power Input
                                  (Average Field Intensity)

                                      Wire&Weight
                          K    ^
                        1-2-
                        1-1-


                        1-0-


                        0-9-


                        08-


                        0-7-


                        06-


                        05-


                        0-4-^


                        0-3-
                        0-2
      SRMast
Typical Operating Clearances
                          20O  250  300  350  400  450  500
                                  Plate Spacing mm

                                    FIG.  9
                                     135

-------
     So far, we have discussed comparisons between wire weight systems
and framed systems.  However, as already stated, there is an increasing
vogue towards mast systems.   These certainly have bigger problems in
initial clearances and final alignments compared to a wire system.
However, typically one can expect a mast system to have far less alignment
problems than a framed system.  In Fig. 9, we compare a wire system with
a Sturtevant S.P. mast electrode system with typical operating clearances
and assuming the same emitting radius in both cases.

     This shows a marked reduction in field intensity using a mast system
although not as great as with a typical framed system.  For any manufacturer
of masted systems  (such as ourselves), it is extremely important that the
emission characteristics of their mast system be very much better than a
wire system to even achieve wire performance let alone surpass it.

     In Fig. 10, we show the difference in operational characteristics
of a Sturtevant S.P. mast electrode compared with a wire system.
             40-r
             30--
             20--
             10-L
             kV
Operating Characteristics
    Wire/SP Mast Electrodes
        Second Electrical Stage
        200 MW Boiler
        200mm Duct Spacing
                                                       mA
                                  FIG  10
     As can be seen at 30kV operating voltage, the S.P. mast achieves 90%
more corona emission than the wire.  However, for the reasons illustrated
in Fig. 9, the actual overall net operational improvement in the field
intensity for the S.P. mast has been found to be not more than about 10
to 20%.  Manufacturers who are moving over from wire electrode systems
to mast systems must bear this carefully in mind.  If their mast system
characteristics compared with their original wire system are not similar
to Fig. 10, then they are likely to find an overall fall-off in
performance compared to their original wire systems.

     With regard to the various clearances chosen, some manufacturers may
argue that the total clearance reduction used for perfect alignment does
not apply to their own framed system particularly perhaps because of the
geometry of the frames and collecting plates.  There is, of course, some
                                    136

-------
merit in this but it is difficult to achieve perfection in geometry for
all the closest clearance points in a precipitator when dealing with the
extremely large number of such points involved.  However, to take some
account of this factor, the perfect alignment situation for frames has
been reduced from a typical figure of 35mm clearance reduction to an
arbitrary 28mm.

     It may also be added that in preparing this paper, a considerable
number of different k factors were obtained from both wire, mast and frame
systems and used in the computation of alternative curves.  Although there
were differences in the results, these were of comparatively low degree
and the general form of the curves produced did not change.

     It was not expected when resorting to established formula in order
to relate duct spacing with flashover clearances that so many interesting
things would emerge which explains to some extent contemporary vogues.

     The reason for example why very narrow ducted Sturtevant
precipitators have been operating so excellently in the last 20 years on
highly resistive dusts and with plate heights up to about 13 metres.
Similarly, pressure from some manufacturers to increase duct spacing
particularly when dealing with highly resistive dusts is explained.  The
particular usefulness of mast electrodes in relation to framed systems
and the particular importance of the emission characteristics of mast
electrodes when comparing them to wire systems has also been indicated.

     However, whilst this paper may explain the excellence of the
Sturtevant wire weight flat plate performance compared to framed systems,
it does not explain the apparent better reputation of European framed
systems used in the U.S. compared to the generality of U.S. designed
wire/weight systems.

     All the Sturtevant plants referred to in addition to being flat
plate specifically designed for perfection in alignment, also use the
multi-rap principle.  That is, rapping at more than one position up the
side of each collecting plate.  With this system, the rapping energy per
square metre of plate area is not only well in excess of all other
designs currently marketed, but the rapping occurs at optimised positions.

     Local U.S. wire and weight designs are almost invariably top rapped
and, as a general rule, the rapping is inferior to European bottom rapped
designs and infinitely inferior to multi-rap.

     All of which illustrates that no single aspect of precipitation
technology can ever be considered in isolation.

     The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the views of the Agency and no official endorsement
should be inferred.
                                    137

-------
                              .REFERENCES
1.     Harry J. White.    Industrial Electrostatic Precipitation.
                          Pergamon Press, 1963.
                                  138

-------
                PULSE CORONA AS ION SOURCE AND ITS BEHAVIORS
                        IN MONOPOLAR CURRENT EMISSION

                by:   Senichi Masuda and Yoshiaki Shishikui
                     Department of Electrical Engineering,
                     University of Tokyo
                     7-3-1, Kongo, Bunkyo-ku, Tokyo, JAPAN 113
                                 ABSTRACT

     The pulse voltage applied to a discharge electrode produces a weakly
ionized plasma with about 5 x 10^ ions/cm^ in ion concentration in a very
short time of about 10 - 100 nano-seconds.  The monopolar ions are extracted
by an externally applied dc field towards the charging area from this plasma,
and the current pulse lasts for a substantially longer time of about 1-3
milli-seconds.  This is due to a long life time of the plasma which disap-
pears by recombination and charge separation.  The expansion of the monopolar
ion cloud during its migration across the charging area also contributes to
the increase in the duration time of the current pulse measured at the count-
er electrode.  The theory explaining the wave form of the ionic current by
the recombination, separation and expansion of ions agree very well with the
experimental results.

                               INTRODUCTION

     The specific advantages of using an extremely short pulse high voltage
with several tens to several hundreds nano-seconds duration time have been
recognized in the pulse-energization of electrostatic precipitators of both
twin-electrode type (1,2) and tri-electrode type (3), and also in the pro-
duction of ions in a precharger called Boxer Charger (4) currently being de-
veloped by the auther.
     The pulse voltage proceeds on a transmission line with a speed of ca.
28 cm per 1 nano-second, so that the geometrical length of this pulse voltage
becomes comparable to the total length of the corona wires.  As a result, it
behaves as a travelling voltage wave to produce a very special type of
streamer coronas (5,6)  along the entire length of the transmission line con-
sisting of the corona and counter electrodes or two parallel corona wires.

     These streamers, although belonging to a weakly ionized plasma, possess
a sufficiently high ion concentration for most of the practical charging pur-
poses,  so that they serve as a good plasma ion source emitting monopolar ions


                                    139

-------
with the aid of an externally impressed extracting  dc  field.
     The advantage of using such an extremely short pulse  high voltage is
two-fold.  One is the ease in construction of its pulse  power ,s>upply with its
low initial cost.  Another is its low running cost  resulted by a very high
power efficiency.  This is due to the fact that  the energy on the corona
transmission line is highly localized in  time and space  on a  travelling wave
packet to be effectively converted into corona energy, and that this conver-
sion is enhanced by the reflection .of the pulse  wave at  the open terminal of
the line, and further that the residual energy of pulse  affer losing its
ability of producing streamers to become  a loss  remains  quite JLow.   It must be
also emphasized that,, in the presence of  the extracting  dc field externally
applied, the travelling wave voltage primarily acts as a trigger and a sub-
stantial part of corona energy can be supplied from the  dc field.
     In the cource of investigations of the coronas induced by such an ex-
tremely short pulse voltage, several important phenomena are  discovered.  In
the case when two parallel wires are used, the streamers are  launched at
first from the negative wire (negative streamer), and, then,  the positive
streamers occur from the positive wire when the  pulse voltage is sufficiently
high, exceeding a certain threshold (5).  There  seemed to  be  no difference in
ion emission ability between the negative and positive streamers, whereas the
latters consume a substantially greater energy than the  formers (5,6).   Al-
though the streamers are formed in a very short  time of  say 10 nanip-seeonds
or less, .the ionic pulse current produced indicates a very long duration time
of about 1-3 milli-seconds.
     In consideration of the importance of better understanding of  the plasma
structures of these -streamer coronas, investigations are made on the time-
dependent ion concentration of these plasma and  mechanism  of  producing such
a long current duration time.  Based on the experimental data a mathematical
model is proposed on the ion separatipn and migration processes, and the nu-
merical results obtained by computer simulation  using  this model are compared
with the experimental results.

     The present paper reports on the results of these experimental and the-
oretical investigations.

                            EXPERIMENTAL  METHODS

     The experimental apparatus used in the first series of experiments  is
shown in Fig. 1  (a).  The upper corona electrode consists  pf  parallel wires
having a square cross-section ;(5 mm x 5 mm), spaced at 23  mm  (face  to face)
and insulated from each other of two adjacent wires to form the mother .and
daughter wires.  The lower grounded plate is an  ion collecting electrode con-
sisting of two concentric disks, insulated from  each other,, where the ianer
disk (diameter:70 mm) serves as a current measuring electrode.  The distance
between the upper and lower electrodes is 60 mm. An extremely short pulse
high voltage with a half duration time of ca. T, =70 nano-seconds is applied
between the mother and daughter wires in  such a  polarity that the latters are
negative in reference to the formers throughout  this paper.  This produces
corona discharge as a plasma ion source.  The corona starts from the negative
daughter wires in a fo/rm pf uniformly distributed streamers (negative stream-
ers) extending towards the positive mother wires. Then,  provided pulse volt-


                                     140

-------
tage is sufficiently high,  the positive mother wires launch much brighter
streamers (positive streamers) towards the negative wires (5,6).  The voltage
difference between the daughter and mother wires disappears very quickly
through the leakage resistance R.   At a certain delay time, TJ, after the
formation of the streamers  a dc voltage is applied in a form of a step-func-
tion between the mother wires and  ion collecting electrode so as to extract
from the plasma of streamers the monopolar ions of a desired polarity.  This
delay time allows at the start of  the ion extraction the variation of the ion
concentration of the plasma undergoing recombination, and thereby the estima-
tion of the initial plasma  ion concentration.
     The monopolar ions, thus extracted, travel across the inter-electrode
gap, inducing a displacement current on the measuring electrode M.  In order
to detect the convection current of ions, which enables the estimation of
local ion concentration of  the moving ion cloud, a grid electrode is located
close to the ion collecting electrode to screen the disturbance in a form of
displacement current.  A bias voltage applied to the grid makes its potential
undistorted in the application of  dc voltage.

     With this experimental setup  the first series of experiments is per-
formed, detecting the current wave form from the measuring electrode which
indicates the variation of  number  concentration of oncoming ions.

     Then, the second series of experiments is made to measure accurately the
total quantity of charge carried by the extracted monopolar ions.  For this
purpose the experimental apparatus is modified as shown in Fig. 1 (b) where
the grid is removed to avoid its charge collection.  The potential across the
capacitor inserted between  the measuring electrode and the ground gives the
total charge.  This quantity measured with variation of the delay time, T,,
provides the information on the decay of ion concentration in the plasma with
time.
     Finally, the distribution of  the plasma density in the transverse direc-
tion is estimated by measuring the distribution on the total charge collected
on the ion collecting electrode.  The ion collecting electrode is modified
for this purpose as shown in Fig.  1 (c) where a series of parallel strip-like
probe electrodes with 5 mm  width is attached on the disk.
                            EXPERIMENTAL RESULTS
WAVE FORM OF IONIC CURRENT
     At first the negative dc voltage is applied to the mother wires to ex-
tract negative ions,  and the effects of the magnitude of the extracting dc
field, E  ,  and delay time,  T,,  on the current wave form are observed.  The
wave form shown in Fig.  2 (a) and (b) represent those observed at different
levels of E,  with zero  delay time,  T,, and a constant pulse voltage Vp= -37
kV.   A shortinitial  pulse,  induced by the dc voltage applied to the wires,
serves as the time origin.  It is followed by a larger pulse of convection
current by the oncoming  ion cloud.  It can be seen that the increase in E,
produces the decrease in the interval between the two peaks as a measure of
ion  transit  time, reduction in width of the ion cloud pulse, the increase in
its  pulse height, and finally the increase in number of total oncoming ions
given by time integration of the pulse wave form.  All these effects are pro-
duced by the increase in both the separation speed of the monopolar ions in

                                     141

-------
the plasma region and the migration speed of the separated  ions  in the gap.

     Next, the effect of the delay time, T,, on the wave  form is observed
with the magnitudes of V  and E,  unaltered (Fig. 3).  The  first spike is
induced by the pulse voltage applied across the wires, and  the second  small
pulse by the step-wise application of the dc voltage.  The  interval between
the second and third pulses as a measure of the ion transit time remains al-
most unchanged with increasing T, when E,  is kept constant.   The time inte-
gration of the third pulse as the total arriving ions, on the other hand,
decreases with T,.
                d
     Then,the effect of the polarity of the dc extracting field  is investi-
gated.  First, the positive ions are extracted in the '.case  when  a suffi-
ciently high pulse voltage V = -47 kV is applied so as to produce both nega-;
tive and positive streamers,-and the dc high voltage is applied  at zero delay
time.  The wave form in this case is shown in Fig. 4 (a).   For comparison,
Fig. 4 (b) shows the wave form when the negative ions are extracted under
otherwise the same condition.  It can be seen that the wave forms of positive
and negative ionic current agree quite well, although the charge carried by
positive ions is slightly less than that by negative ions.   Second, the same
comparison is made with a lower dc voltage possible, to produce only negative
streamers.  The results obtained are shown in Fig. 5.  In this case, the
amount of positive ions extracted becomes much less than  that of negative
ions.  Its difference becomes larger in the case when the pulse  voltage V
is in the lower level to produce only negative streamers.

AMOUNT OF EXTRACTED IONS

     Changing T, with keeping V  and E   constant, the total  charge of the
negative ions arriving at the measuring electrode M in Fig.  1 (b)  is mea-
sured.  The measurements are made for V = -55 kV and V =  -37  kV.   In the
former case both positive and negative streamers occur, as  shown in Fig.
6 (a).  Whereas, in the latter case, only negative streamers  with much less
luminosity appear, as indicated in Fig. 6 (b).

     The measured charge Q. for these cases are plotted against  the delay
time, T , in Figs. 7 and 8^" respectively.

     Then, the same experiment is made for the positive ions  extracted.  The
results obtained at a lower pulse voltage V = -35 kV (only  negative stream-
ers) and E  = 4kV/cm are plotted against T, in Fig. 9.  The results for the
negative ions extracted are also indicated in this figure.   Again in this
figure, it is clearly shown that the number of negative ions  extracted is
larger than that of positive ions, but the rate of its decay with T, is al-
most the same.

     It can be seen from these figures that
(i) the number of negative ions possible to extract from  the plasma drops
rapidly with increasing delay time, T,, probably because  of a continued ion
recombination process, (ii) this number of ions is also dependent upon the
magnitude .of the extracting dc field, E  , especially at  its lower level,
(iii) in spite of a great difference in fuminosity between  the plasma  pro-
duced by both negative and positive streamers (Fig. 7) and  that  by negative
streamers only (Fig. 8), the number of negative ions possible to extract does
not indicates much difference: the former providing only  twice the negative


                                    142

-------
ions by the latter, and (iv) as for the polarity of ions, the negative ions
can be extracted in a greater amount than positive ions, but the dependence
of  the ion number possible to extract is almost the same.

DISTRIBUTION OF IONIC CHARGE ON THE ION COLLECTING ELECTRODE

     The streamers produced by the extremely short pulse voltage between the
wires do not represent a uniform plane-like plasma ion source with a constant
ion density.  Its irregularity will produce non-uniformity in the charge on
the ion collecting electrode in the transverse direction perp.endicular to the
wires.  Fig. 10 indicates the measured distribution of the ionic charge col-
lected.  Fig. 10 (a) and (b) show the results obtained at a high voltage V =-
45 kV (both positive and negative streamers) for the negative and positive
ions extracted, respectively.  The pattern of the distribution does not
change dependent on the polarity of the ions.  In both cases, the charge
becomes minimum at the points just below the wires.  Fig. 10 (c) and (d) are
the results obtained at a low pulse voltage V = -35 kV (only negative stream-
ers) for the negative and positive ions extracted, respectively.  Again in
this case, the distribution is not much dependent on the polarity of ions,
but the total charge of negative ions becomes about twice the charge of posi-
tive ions.  It can also be seen that the distribution is not much dependent
on whether both negative and positive streamers are produced, or negative
streamers only.

                    SIMULATION OF ION SEPARATION PROCESS

     The following mathematical model of ion separation process is proposed
to perform its computer simulation, and its results compared with the experi-
mental data.  The model is two-dimensional, taking two infinitely large par-
allel plane electrodes, A and B, respectively a series of corona wires and
an ion collecting electrode.  The ion separation proceeds in the sequence as
shown in Fig. 11 (a) - (d).

     (i) Stage I (Fig. 11 (a)): At an instant of streamer formation, a uni-
form sheet of plasma with a thickness LQ and ion density ng is produced over
the entire surface of A.   In the later process of ion separation, no free
electrons are considered as this proceeds slowly so that all the free elec-
trons are attached to electronegative gas molecules to form negative ions.
In the subsequent period of delay time, T, , the bipolar ions in the plasma
are allowed to recombine according to:

                         dn                                              f-, \
                         — = - an n                                     (1)
                         dt       p n
where a = recombination coefficient, n , n  = number densities of positive
and negative ions, and n represents either n  or n .
                                            P     n
     (ii)  Stage II (Fig.  11 (b)): Then, the extracting dc field is applied,
and the ions of both polarities start to separate by migration driven by the
time-varying local field, E(x,t).  The velocity of ions is taken as yE where
y represents ion mobility.  The ions of one polarity are collected by the
electrode A, whereas those of another polarity migrate towards the electrode
B.  In this stage there are three different regions of ions; the regions of
positive ions only, bipolar ions, and negative ions only.  The number density


                                     143

-------
of Ions of each polarity is assumed to be uniform  in  each region.   The solu-
tion of Poisson's Equation differs dependent on  the region,  and is  given, by. the
following equations.

   a) Region of positive ions ( 0 <_ x < L3 ) :
               en        1
      E(x) = _
                                          n              p

               en      en      en
                                                                          (2)
               £0      EO      £0

   b) Region of bipolar ions ( 1,3 4 x < L2  ) :
                                                      en
      E(x) =   { V0-     (- n2* nL22+ nL32' )}  -     -.(Li - L2)         (3)
   c) Region of negative ions ( L2 <_ x < LI  ) :
             en        -i                                       en
      E(x) = — x +  [ f { V0- •— (- n Lj2+ n  L22+ n  L32)}  - — LJ  ]    (4)
             eo        L    u  2e0     n *    n *    p d       eo

   d) Ion-free region ( L^ <_ X < L ) :

      E(x) =   { V0-     (- nLl2+ nL22+ nL32 ) }                     ,   (5)
where V0 = dc voltage applied between A and B,  e  =  elementary charge,  EQ =
electrical permitivity of free space, and L,  LQ,  LI ,  L2 ,  L3 are indicated in
Fig. 11.  In the region of bipolar ions, the  recombination continues to occur
according to Eq. (1) to decrease each ion concentration.   The fractions of
monopolar ions separated from the bipolar region  in each  time step of  computer
simulation are added, respectively to the regions of positive and negative
ions in the next time step, and the averaging is  repeated of ion concentration
in each region.  The cloud of separated negative  ions migrating towards B
undergoes an electrostatically induced expansion  with increasing cloud thick-
ness.  This contributes to the increase in the duration  time of the current
pulse measured at the electrode B.  In such a way,  ions  arrive  at the elec-
trode B one after another.

     (iii) Stages III and IV (Fig. 11 (c) and (d)): These are the stages of
ion migration after the bipolar region has disappeared.   In the case when the
electrode gap, L, is small, the stage II is followed by  the stage III'.  The
field intensity in each region is given by Eqs.  (2) - (5).

     Using the model described above, computer simulations of the entire ion
separation process is made for the assumed initial  values of ng and LQ .

     The curves in Fig. 12 show the results of the  simulation using ng= 6 x
109 /cm3, LQ= 1.6 cm.  The measured values in Fig.  1 taken at a high pulse
voltage to produce both positive and negative streamers  are also plotted.
The agreement between the simulation and the  measured values is very good.

     Fig. 13 gives the same results using initial values, n0= 5 x 109  /cm3,
LQ= 1.0 cm.  These again agree quite well with the  experimental plots  trans-r
fered from Fig. 8 measured at a low pulse voltage to produce, only negative
streamers.


                                     144

-------
     Next, the ionic current wave forms are derived by the simulation and
compared with those measured with a grid in Fig. 14 (a) and  (b).  Fig.  14  (c)
shows a similar comparison between simulation and measured for  the case when
the grid is removed so that displacement current is being observed.  Again  in
Fig. 14 a good agreement can be seen between the simulation  and measured.

     The results of curve fitting in Fig. 12, 13 and 14 are  likely to support
the validity of the mathematical model of ion separation process produced
above.   Hence, the initial plasma density is assumed to be 'about 5 - 6  x 109
/cm3 independent of the magnitude of pulse voltage Vp, i.e.  whether both posi-
tive and negative streamers are produced, or only negative streamers are
formed.

     It should be noted in this place that the simulation produces no appre-
ciable difference between the positive and negative ions.  In reference to
Fig. 9, however, the measured results obtained at a low pulse voltage indi-
cate a substantial difference far beyond the estimation by simulation.  This
may be explained by a large segregation of positive and negative ions of
plasma specific to the negative streamers.  It is considered in this case
that negative ions are localized in a region close to the mother wires, and
may be subjected to a strong local extracting field to produce  detachment of
electrons with a very high mobility.  In the case when both  negative and posi-
tive streamers are produced, this strong local field will be screened by the
abandant positive ions existing around the mother wires.

                                 CONCLUSIONS

     The investigations are made on the initial plasma density  of streamers
produced by an extremely short pulse high voltage with ca. 70 nano-seconds
half duration time, and also on the process of ion separation induced by the
dc extracting field.   The results obtained by experiments and simulations lead
to the following conclusions:
     1) The initial number density of plasma estimated from  the comparison
between measured and simulation curves of ionic charge collected as a function
of delay time provides (5-6) x 109 /cm3 independent of whether both negative
and positive streamers or only negative streamers are produced.

     2) The number of ions possible to extract by dc field from the plasma is
about two  times higher for negative ions than for positive  ions in the case
when only negative streamers are produced.

     3) Although the streamers as plasma ion source are produced in a very
short time of about 10 nano-seconds, the current pulse by ions  extracted and
arriving at the ion collecting electrode indicates a much longer duration time
amounting to 2 - 3 milliseconds.  This is due to a fairly long  recombination-
dependent plasma life time and a slow ion separation process.   This duration
time and the ion transit time decrease with increasing extracting field inten-
sity, while the pulse height and total charge of separated ions rise con-
currently.   The increase in the delay time of application of the dc extracting
field causes the lowering of the pulse height and total charge  as a result of
Ion recombination occuring in the plasma.  The plasma density after recombi-
nation  of 2 - 3 milli-seconds becomes one order of magnitude lower than its
initial value.


                                     145

-------
     4) In the mathematical model used for the simulation  of  ion separation
process the time-dependent local value of field intensity  is  derived by solv-
ing Poisson's Equation in different regions of monopolar  and bipolar ions,  and
the local velocity of ions are calculated from this field  intensity using  ion
mobility.  The recombination of ions is considered in  the  bipolar region.  The
electrostatically induced expansion of monopolar ionic cloud  appears in the
collective motion of the ionic cloud.
     5) The computer-simulated curves of the ionic charge  vs.  time delay
assuming suitable value of ng and Lg allow a very good fitting with the meas-
ured values, and this enables the estimation of n0 as  above.   The validity of
the simulation is confirmed also by comparison of the  computer simulated ionic
current wave forms with the measured ones.

     6) The transverse distribution of ionic current on  the ion  collecting
electrode is fairly irregular with a large dip in the  positions  below the
wires.

                                 REFERENCES

1.   Milde, H.I. and Feldman, P.L.  Conf. Rec. of IEEE/IAS 1978  Annual Meet.,
     66.

2.   Masuda, S. and Hosokawa, S.  Performance of Two-Stage Electrostatic
     Precipitators, Conf. Rec. IEEE/IAS Annual Meet. 1982  (to  be published.)

3*   Masuda, S. and Nakatani, H.  Distortion of Pulse  Voltage  Wave Form on
     Corona Wires Due to Corona Discharges, 1982 EPA-Symposium on Transfer and
     Utilization of, Particulate Control Technology (Oct. 1982  to be published)
4.   Masuda, S., Nakatani, H., Yamada, K., Arikawa, M. and Mizuno,  A.   Produc-
     tion of Monopolar Ions by Travelling Wave Corona  Discharge,  Conf.  Rec.
     IEEE/IAS Annual Meet. 1981

5.   Masuda, S., Mizuno, A., Nakatani, H. and Kawahara, H.  Application of
     Boxer-Charger in Pulsed Electrostatic Precipitator, Conf. Rec.  IEEE/IAS
     pp.904-911, 1980

6.   Masuda, S., Nakatani, H., Kawahara, H. and Mizuno, A.  Particle Charging
     with Travelling Wave Corona Discharge, Journal of Electrostatics,  pp.171-
     176, 1981

                  The work described in this paper was
                  not funded by U.S.Environmental Pro-
                  tection Agency and therefore the con-
                  tents do not necessary reflect the
                  views of the Agency and no official
                  endorsement should be inferred.
                                     146

-------
                                                 Extremely Short
                                              Pulse Power Supply
1
t
I spark gap
. 0 . ri '- x-» /-\ -
Corona p
Electrode , 0
^^~ 111 11
^jfel G Ea Q Ea LJ El U ^^
M G
Grid \ X
Electrode , \ /
CICUUU ( ii ' ii - J
^-" ' — | 5 Mfl: -
Ion Collecting zoom; iinF
ElArtrode ; — J

400n I i 5kff 	
II ' '

ij*
3 uF
5 Mn

! T

r£
i 1
^:1700pF ?
t 1 _i
Trigger Signal
rigger-controlled
lain DC Voltage
>upply
- 1 i
               (a)  Current Wave Form Measurement

   c   ,   it
   (b)  Total Ionic Charge
        Measurement
(c)  Plasma  Density Distribution
     Measurement
              m- mother wires         C --condenser for
              D— daughter wires            charge measurement
              G- guard electrode      P--strip-like electrode
              M— measuring electrode
                    Fig.  1  Experimental  Appatatus
                                   147

-------
              (a)
              (b)
Fig.  2  Effectof DC Extracting
  Field, Ed(,,  on Wave Form of
  Negative Ionic Current
   (Td= 0, Vp= -37 kV)
                                        Fig.  3   Effect  of  Delay  Time,.Td)
                                         on  Wave  Form  of  Negative  Ionic
                                         Current
                                           (Vp=  -37  kV,  Edc=  -3  kV/cm)
                                   148

-------
   (a)  positive  ionic  current
(b)  negative ionic current
Fig.  4  Comparison  of  Wave  Form  between Positive and Negative Ionic
    Currents
       (  V =  -47  kV, E ,  = 3 kV/cm,  T,= 0; both negative and positive

         streamers  )
   (a)  positive  ionic  current
(b)  negative  ionic  current
Fig.  5   Comparison of Wave  Form between Positive and Negative Ionic
    Currents
       (  V  = -35  kV, E , = 3 kV/cm, T .= 0; only negative streamers )
                                 149

-------
              kV (both  negative and
     positive streamers)
 (b) Vp=  -37 kV (only  negative
     streamers)

Fig. 6  Photograph of  Streamers
  Observed with Image  Intensifier
  (pulse  duration time  Th.- 70 ns; EMI
   Type 9912)
S,o
5  5
5
,Edc*-6 kV/cm

  -4 kV/cm
                      Vp = -37 kV

                      Th = 70 n»«c
   1
   o

*  £
                                    oc
                                    §
                                 o.s
    0
             1 0
                      2.0
                               3.0
          DELAY TIME Td (m»«c)
Fig. 8   Charge of Negative Ions  Col-
  lected vs.  Delay Time
  (only  negative streamers produced)
                                             20
                                           £ 10
                                    Ul
                                    O

                                    I *
                                    o
                                                         Vp»-66 kV

                                                         Th » 70 ntec
                                             ,Edc«-5 kV/cm
                                               --4 kV/cm
                                                . -3 kV/cm

                                                 ,-2 kV/cm
                                                             -1 kV/cm
                                                                               u
                                                                               5
                                                                     1  5
                                                                            0.5
                                              0.        1.0        2.0

                                                     DELAY TIME Td (msec)
                                                                          30
                                    Fig. 7   Charge of Negative Ions Col
                                      lected  vs.  Delay Time
                                      (both negative and positive
                                       streamers  produced)
                                             20
                                      10
                                    ui  5
                                    O
                                    £

                                    o
                                                             Vp=-35 kV

                                                             Th = 70 nsec

                                                             Edc=4 kV/cm
                                                              o: negative ion

                                                              • : positive ion
   E
   o
2  2-
   03

   tu
   D
                                                                                O
                                                                            0.5
                                        0.        1.0         2.0

                                              DELAY TIME Td  (msec)
                                    Fig.  9  Comparison of Charge Col-
                                      lected as  a Function  of Delay Time
                                      between Negative and  Positive Ions
                                      150

-------
   s

 §4


 83
 i 2
            mother wire
            daughter wire
       D
D
                    Vp=-46 kV

                    Edc--S kV/cm
                              100
                              4 ui

                                g
                              2 i
          DISTANCE x (cm)
(a) negative ionic charge
   (negative and positive streamers)
 o
 £
         Q: mother wire
         D •' daughter wire
       D
D
                   Vp--36 kV

                   Edc = -:5 kV/cm
          E
       10  o

       8  >
                              21
                                o
          246

           DISTANCE X (cm)
(c) negative  ionic charge
   (negative  streamers only)





s
•& 4
CHARGE Ql
-> M U
ffl : mother wire
Q : daughter wire
DSD
Vp- -46 kV
Edc=+5 kV/cm
-
\ A A 1
WVv
2468
DISTANCE x (cm)




-^
10 o
8 >
ro ^ at
CHARGE DENSIT
(b) positive ionic charge
(negative and positive
streamers)





6
I 4
0 ,
i:
g : mother wire
D • daughter wire
D a n
Vp=-35 kV
Edc = +5 kV/cm
•
V\y\/\
2468
DISTANCE x (cm)




~
10 o
8 ~
6 «
N *.
CHARGE DE
                       (d)  positive  ionic charge
                          (negative  streamers only)
    Fig.  10  Transverse Distribution of Collected Ionic Charge
                                     151

-------
L.

U



1



-V0
  t-Lg-l

   — f-
   distribution of

       ion concentration
    (a)  Stage  I

«t
!
I
"v
(—
M.
w
•f-t-
*+
++
*+


"f
'^
h"
1 *
— --
— — -
	
	
- L— — — — }





GO
«
s
+^
U
•2
7
 (b)  Stage II


,








JL,











LI L











i L












    (c)  Stage III
o      L3


 (d) Stage IV
   •••*{<•
   ^•
   ^•

   H(4
    (e) Stage III1



     Fig.  11   Model of Ion Separation  Process
                                152

-------
  20
  10
5
UI
O
cc
Edc--5 kV/om
   -3 kV/cm

   -1 kV/cm
                                              10
                                     CO
                                     5
                                     D
                                     I
                     Ul
                     (9
                     I  •
                                  0.5
    measured
     values
                                                                o—Edc=-5 kV/om

                                                                       -3 kV/cm •
                                                                       -1 kV/cm
    n0-5xlo9/cm3
    L0-1.0 cm



    ^Edc--5 kV/om


       , -3 kV/cm
                                                                  -1 kV/cm
                                                                                  O
                                                                                  io
                                                                                  ui
                                                                                  O
                                                                                  CC
                                                                                  <
                                                                                  o
                                                                               0.5
    0.         1.0        2.0


          DELAY TIME Td  (msec)
                                3.0
                                  1.0        2.0

                                DELAY TIME Td (msec)
                                                                            3.0
Fig. 12   Comparison  between  Computer-   Fig.  13  Comparison  between  Computer-
  Simulated Curves  and Measured Values
  of Charge of Negative  Ions Collected
  vs. Delay Time
  (Vp= -55  kV: negative  and positive
      streamers)
                        Simulated Curves and  Measured  Values
                        of  Charge of  Negative Ions Collected
                        Vs.  Delay Time
                        (VD=  -37 kV:  negative streamers
                          H  only)
                   (""sec)  0.      2.0   .  4.0 (msec)  p.     2.0    4.0 (msec)
     0-      2.0    (msec)   0.     2.0     4.0 (msec)   Q.      2.0     4.0 (msec)
                         (b)
             v
                             V -55
                             E" "2
                             no- 6 x 10' /cm3,"
                             Lo- 1.6 cm
                             ( with grid )
                           Cc)
   -   - v
 Edc= -2 kV/cm.

•n0- 6 x 109 /cm3
 LO* 1.6 cm

 (grid removed )
    Fig.  14  Comparison between Simulated  and Measured Wave  Forms
              of  Ionic Current
                                       153

-------
               A NEW CORRECTION METHOD OF MIGRATION VELOCITY
                      IN DEUTSCH EFFICIENCY EQUATION
                                     FOR
              CONVERSION OF ELECTROSTATIC PRECIPITATOR SIZING
                    FROM A PILOT-SCALE TO A FULL-SCALE

                                     BY:
                                FUMIO ISAHAYA
             Hitachi Plant Engineering & Construction Co.,Ltd.
                            Research laboratory
              1-1-14 Uchikanda, Chiyoda-ku, Tokyo 101, Japan
                                  ABSTRACT

     In order to confirm the validity of the Deutsch efficiency equation,
the deposition rate distribution on the collecting electrode wall of a small
size cylindrical type precipitator having a diameter of 30 mm and length of
400 mm was measured by utilizing a radioactive aerosol tracer technique.

     The Thorium B solid particles as testing radioactive aerosols having
a relatively uniform particle size of approximate 0.1 microns were used.
As a result, the concept of the Deutsch efficiency equation was held tolerably
good in the range of the gas velocity of 0.35 - 1.0 m/s, S.C.A. of 35 - 106
s/m and corona current density of 2.5 -13 uA/cm.

     Accordingly, in order to put the Deutsch migration velocity to practical
use in industry and to predict the required sizing and S.C.A. for a full-scale
precipitator on the basis of the test results in pilot-scale one, such a new
correction formula for the migration velocity in the Deutsch efficiency equ-
ation which can be corrected as a function of the aspect ratio of collecting
electrode, spacing of electrode, treating gas velocity, S.C.A., corona current
density, particle size distribution, a wide spacing effect and re-entrainment
effect, was proposed.  Furthermore,  according to this formula the performance
characteristics curves of collection efficiency versus S.C.A. for a pilot-
scale and full-scale precipitator were given in comparison with the both in
the range of the industrial use such as a coal-fired boiler plant.
                                    154

-------
                                INTRODUCTION

     At the present state, the most reliable source of data for the prediction
 , evaluation and correction of the Deutsch migration velocity for the precipi-
 tator sizing for the new duty can be obtained on the basis of an actual test
 result in a pilot-scale electrostatic precipitator.  However, the Deutsch mig-
 ration velocity in a pilot-scale precipitator represents usually several times
 too high, compared with     one of a full-scale precipitator.  It has been
 living yet as     one of the important problem awaiting solution that the
 reason why such a difference in the Deutsch migration velocity arises from.

     The purpose of this paper is to make clear     the reason for this dis-
 tinct difference, furthermore to propose a new correction method of the mig-
 ration velocity in the Deutsch efficiency equation for the conversion of pre-
 cipitator sizing from a pilot-scale to full-scale.  For instance, a usual
 pilot-scale precipitator has been designed with the collecting electrode hei-
 ght of approximate 1 - 3 m, its length on the direction of gas flow of appro-
 ximate 1 - 5 m and S.C.A of approximate 15 - 100 s/m, in such case, the Deu-
 tsch migration velocity will usually represent several times too high, compar-
 ed with     one of a full-scale precipitator, in spite of the same testing and
 operating conditions between the both, such as the electrode configuration,
 rapping method, electrode spacing, corona current density, treating gas velo-
 city, gas temperature, gas composition, dust particle properties and even
 S.C.A..

     As the first step of this investigation, in order to confirm the validity
 of the Deutsch efficiency equation, the deposition rate distribution on the
 collecting wall of a small model precipitator was measured by utilizing a
 radioactive aerosol tracer technique using the Thorium B solid particles.
 Then the next step one, depending on the basis of the Deutsch migration velo-
 city obtained from the test results in a pilot-scale precipitator, a new cor-
 rection formula for the Deutsch migration velocity and S.C.A. satisfied the
 requirement of collection efficiency of a full-scale precipitator was proposed
   Furthermore, according to this formula the performance characteristics
 curves of collection efficiency versus S.C.A. for a pilot-scale and full-scale
 precipitator were given in comparison with the both in the range of the indus-
 trial use such as a coal-fired boiler plant.

     On a realistic preeipitator design, it has been found widely acceptance
 with an international field of an electrostatic precipitation technology
 that the Deutsch migration velocity can be utilized as a mental image design
 parameter.   It is widely accepted with the practical design of the electro-
 static precipitator that the fundamental characteristics of collection effi-
 ciency under the influence of turbulent flow was given by the following
 classical Deutsch efficiency equation, later refined by White,


                     fj = 1 - exp - (UJ. S.C.A.)                         (1)

where,  ff  is the collection efficiency, £Ji)  the Deutsch migration velocity,
S.C.A.= A/Q the specific collecting area, A the effective area of collecting
electrode,  Q the volumetric flow rate of treating gas.

                                   155

-------
     Hereupon, it is a noticeable matter that  the  so-called Deutsch migration
velocity cannot be measured directly by some instrumentation means, it should
be derived from the equation  (l), giving the S.C.A.  as  a design parameter and
the collection efficiency on  the basis of  the  actual test results in an ope-
rating precipitator.  Also simultaneously,  the physical meaning of such Deu-
tsch migration velocity is macroscopically defined as an average value based
on a time and space- dependence for the charged particles cloud passing
through a precipitator under  the influence of  the  corona electric field in a
turbulent flow in which some  uncertain phenomena for the charged particles
aerodynamic behaviour such as a turbulent  diffusion,  mutual coagulation,  re-
entrainment of the dust layer on collecting electrode and flow disturbance due
to ionic wind are included.   Therefore, a  thoretical migration velocity for a
single charged particle in a  laminar flow  under the  influence of the static
electric field and which can  be derived from a force balance between the
Coulomb's force and the Stokes's force becomes several  times higher than  the
Deutsch migration velocity.   Furthermore,  currently  the migration velocity
measured by a laser Doppler velocimeter for a  horizontal duct type model  pre-
cipitator has been reported and it shows a much higher  experimental value by
a factor of several ten times in comparison with the Deutsch migration velo-
city in industrialprecipitator.  This experimental results suggest that an
instantaneous velocity due to a specified  time and space-dependence in the
precipitation zone is measured with this method.   Thus,  the migration velocity
based on the thoretical calculation or the experimental measurement as men-
tioned before is not meaningful as a means of  predicting or evaluating the
precipitator sizing and S.C.A. to a realistic  designing for an industrial
precipitator.

     On the other hand, depending on the equation  (1),  the mathematical simu-
lation model considered the influence of particle  size  distribution and elect-
ric field pattern has been investigated and the collection efficiency was
computed numerically*  However, in this mathematical model analysis,  since the
influence of uncertain factors in precipitation phenomena such as a particle
coagulation, re-entrainment of dust layer  on collecting electrode, flow dis-
turbance due to ionic wind, aspect ratio and configuration of collecting
electrode are un-accounted, it has never been  reached completion to be uti-
lized as a simulation model,  particularly  the  conversion of precipitator
sizing from a pilot-scale to  full-scale.

     In other words, from a stand point of a realistic  idea on a precipitator
designing, the Deutsch migration velocity  means a  mental image design para-
meter in which the whole of capturing processes in dust particles from an
electro-hydro-dynamics separation in turbulent gas flows until an effective
gravity settling motion toward the hoppers should be included.

     CONFIRMING THE VALIDITY OF  DEUTSCH EFFICIENCY EQUATION BY UTILIZING
    ^ A RADIOACTIVE TRACER TECHNIQUE USING THE THORIUM B AEROSOL PARTICLES5'*

EXPERIMENTAL PROCEDURE

     The radioactive Thorium B aerosol particles were generated by an  atomic
disintegration of the Thoron  (Rn)  gas which were diffused from the Thorium
oxide powders of approximate 2.5 kg- piled on the bottom in the vessel  having

                                    156

-------
an inlet and outlet air valve.  This radioactive gas with the  decay half-life
of approximate 54.6 sec. is  disintegrated into the  Thorium B of a fine solid
particles, the following one another the Thorium B  with the decay half-life of
approximate 10.6 hours is disintegrated into the Thorium D (Pb) and finally
becomes stable.  The radioactive aerosol particles  of Thorium  B generated by
such  procedure were introduced into a small size cylindrical type model pre-
cipitator having a diameter  of 30 mm, its length of 400 mm and a diameter of
corona discharge wire of 0.4 mm, and then they were electrostatically collect-
ed on a stainless steel foil sticked closely on the inside wall of the cylin-
drical collecting electrode.  The corona current density and treating air
velocity as the experimental parameters were varied and for each experimental
conditions the radioactivity of deposition rate distribution of Thorium B par-
ticles versus the down stream  location of collecting electrode were measured
by a  radioactivity detecting instrument.

     As the first step of this experiment, in order to confirm whether Thorium
B would be or not, its decay half-life of the radioactivity, beta-ray and
gamma-ray energy spectrum was measured.  As a result, it was clearly shown
that  it must be Thorium B because of a decay half-life of 10.6 hours, maximum
beta-ray energy of approximate 2.3 MeV and peak gamma-ray spectrum of 0.23 MeV
   Furthermore, Fig.l shows  the photograph of an electron-microscope for the
Thorium B particles having a relatively uniform size of approximate 0.1 pin.
                                            -*.
                   -         .
                  +•.:*••
"•**.*
  •  *<
                                 •••*       ..
     Fig.l  Electron-microscope photograph of Thorium B particles

                                  157

-------
EXPERIMENTAL RESULTS AND ITS INVESTIGATION

     As shown in Fig.2, one of the representative examples  in  these  experimen-
tal results, the Thorium B particles deposition rate per unit  area of collect-
ing electrode toward down stream represents a tendency which increases rapidly
and reaches a peak point, after that, on the contrary, decreased  exponentially
toward the down stream.

     The experimental results with such a tendency were found  in  the data of
Mierdel and Seeliger7 (1935) and Robinson (1967) who had measured the deposi-
tion rate distribution on the collecting wall of a model precipitator using a
usual testing dust.  However, the effect of corona current  density on the de-
position rate was not clearly stated in these references.

     The analysis in this paper were carried out within the zone  showing an
exponential decay of the deposition rate.  Further, the required  time to reach
a peak deposition rate as mentioned above is suited for the so-called Pauthe-
nier saturation charging time for a single sphere particle.

     Now, the collecting electrode is divided into n segments  having an equal
length of A L along the down stream and the inlet particles concentration is
C, volumetric treating gas flow rate Q, furthermore when the probability of
deposition rateAfj per unit length A L is kept constant, thus  the deposition
rate Wn for the nth segment can be given as follows
                       Wn = C Q (1 -AH)  "AT,                            (2)

correspondingly,

                       In Wn = In C QA^ + ( n - 1  ) In  ( 1 -A*])
                             = a + ( n - 1 ) b,
hence,

                       Ar|= 1 - exp - (b  ).                               (3)

     Thus, from the equation (1) and Fig.2, the following relationship can be
deriV6d                b _ in (Wn-1/ Wn )
                                d( n )

                                                                          (4)
where, Vg is the treating gas velocity and R the radius of collecting  elect
rode.  Hence, the Deutsch migration velocity can be stated as  follows

                        i  In ( Wn-1 / Wn  )  Vg. R                           .
                         =         ~~
accordingly, from both the experimental results as shown  in Fig. 2  and the
equation (5), the Deutsch migration velocity can be derived.   Fig. 3  shows
such Deutsch migration velocity for the Thorium B aerosol particles versus  the
corona current density with the S.C.A. as parameter.  It  is clearly shown


                                      158

-------
0

*  3
 .10
EH
H
e
H
^  2
a 10
u
Q
H
Du
O
EH
H

H
EH
U

O
H
  10	
                                                       2,5 jj A/cm
 0                    10                   20


                    DOWNSTREAM LENGTH,  L(cm)

Fig.2 Thorium B particles deposition rate versus  downstream location
     of collecting electrode with corona current density as parameter
                                                                     30
                                 159

-------
that the Deutsch migration velocity is strongly dependent upon the corona cur-
rent density and has a proportional relation,  while it is practically indepen-
dent upon the S.C.A..

     In accordance with the equation (l), the Deutsch migration velocity
should be held an unchangeability independently of the S.C.A. or treating gas
velocity when a particle size is uniform, properly speaking.  Such a relation-
ship is almost consist with this experimental result in the range of the
treating air velocity of approximate 0.35 - 1.0 m/s, S.C.A. of approximate 35
- 106 s/m and corona current density of approximate 2.5 - 13 uA/cm as shown in
Fig.3.

 CORRECTION METHOD OF MIGRATION VELOCITY IN DEUTSCH EFFICIENCY EQUATION
 FOR CONVERSION OF PRECIPITATOR SIZING FROM A PILOT-SCALE TO FULL-SCALE

     In the practical operating condition for an industrial electrostatic
precipitator,
   15
  wlO
 \
  o
 EH
 H
 U
 O
 p-1
 o
 H
 EH
 o
 H
 35.4
 67.0
102.0
106.0     767
 55.3    K52
 36.4   2210
                                 1
        1
                       2                      5
                    CORONA CURRENT  DENSITY,  I(uA/cm)
                               10
                                                       15
     Fig.3  Deutsch migration velocity of Thorium B particles versus
            corona current density with S.C.A. as parameter
                                   160

-------
(1) the particles cloud has an wide range of size distribution, appoximated to
    a logarithm normal,
(2) the Deutsch migration velocity is decreased with increasing the S.C.A. or
    length of the collecting electrode,
(3) a re-entrainment occurs when rapped  the collecting electrode or because of
    the influence of aerodynamics flow disturbance and this also is under
    a strong influence of the gas velocity and aspect ratio of collecting
    electrode,
(4) an aerodynamic behaviour of the charged particles cloud is influenced by
    a particle coagulation, degree of flow turbulence and ionic wind,
(5) an average corona electric field intensity is affected by the geometric
    configuration and spacing of the collecting and discharging electrode,
    furthermore a corona current suppression caused by the space charge effect
    of charged particles cloud,
(6) due to a high resistivity dust layer deposited on the collecting electrode
    , a back corona, irregular corona discharge and reducing of space poten-
    tial occurs,
(7) a treating gas velocity in practical use of precipitators should be limit-
    ed to the approximate  0.6 m/s or more because of the avoidance for a free
    oonvectional  flow, furthermore the approximate 1.6 m/s or less which is
    so-called a critical velocity because of the avoidance for an aerodynamic
    re-entrainment on the collecting electrode.

     Considering  these matters as stated above, the Deutsch migration velocity
should be corrected with the following concepts,

(1) in the items  of (1) and (2),  the coefficient "$" as a function of the gas
    flow Reynolds number acting on the aerodynamic behavior of particle repre-
    sented in  a  geometric mass median diameter and the coefficient "^" as a
    function of the geometric standard deviation for the particle size distri-
    bution at the inlet of precipitator  was introduced as shown in Fig.4 and
    5'                                                                 10
(2) in the item of (3), the re-entrainment coefficient given by Isahaya (1979)
    which was defined as a function of the aspect ratio of collecting elect-
    rode,  treating gas velocity and effective gravity settling velocity of
    dislodging dusts from collecting electrode was introduced and this settl-
    ing velocity  is represented as a function of the S.C.A. as shown in Fig.6,
(3) in the item of (4), it is not necessary to correct with each other because
    these aerodynamic situations  between a pilot-scale and full-scale are
    assumed to be in the same conditions,
(4) in the item of (5), applying  the concept for a spread angle of the corona
    electric field model given by Isahaya (1963, 64), an average corona elect-
    ric field intensity was corrected with the collecting and discharging
    electrode spacing, furthermore the space charge effect of particles cloud,
(5) in the term of (6), the coefficient  "JP " as a function of the resistivity
    of dust layer on the collecting electrode,  dust layer thickness and app-
    lied voltage  was introduced as shown in Fig.7,  then the corona current
    density was corrected with this coefficient "J*" .

     In accordance with the concepts as  mentioned above,  a new correction
formula for the Deutsch migration velocity was proposed as follows,
                                     161

-------
              \*r/s.c.A.p\f  ypf+ Pf2  KPO + kfSfCfDf)  /if_\ r  "I
f /Df+ Pf?  Kp(3 +
WDP+ P^  Kf (3 +
                                           1 - exp  -
                                     i
                                                    Wn
                                           1 - exp  -
     1.0
   ^

   EH%
     0.5
   o
   H
   W
   8
      0
        I    i   I   I  I  i I I I
                                                   I    i
         10                          10
                 GAS  FLOW  REYNOLDS NUMBER,  Re
Fig.4 Coefficient (X as a function of gas flow Reynolds number
      in parallel plate collecting electrode
                               162

-------
 fltf fS.C.A.p^
'Vdpj   ^S.C.A.f
                        DfZ+ Pf2 KP(3  + kf SfOfDf )
                        Dp
                     (3
kpSpCpDp)J

Particle  Size Distri-  Spread Angle  Space  Charge
bution and  Aerodynamic of  Corona     Effect
Behaviour Effect       Electric
                      Field Effect
                      \	__
                                               I
                                        Corona Discharge Mode
                                        Effect
     _  exp  _
                    _
             V9f Hf I
   1 - exp  -
VGP
vgp
 Re-entrainment Effect
                                    Wide Spacing Effect
                                                                         (6)
             0.8
              0
                    GEOMETRIC STANDARD DEVIATION, 0T
    Fig.5 Coefficient  Q as a function of geometric standard deviation
          of particle  size distribution
                                     163

-------
where sub-p is the basic design parameter of a pilot-scale precipitator;  sub-
f the basic design parameter of a full-scale precipitator, 
-------
furthermore,  when the spacing of collecting and  discharging electrode of a
pilot-scale precipitator is  designed  to  be  the same  as  one  of a full-scale,
the formula  (6)  can be simplified as  follows,
JL.C.Ay
S.C.A.f
$
                                                                        (8)
                                    1  -  exp
_(Yep_
   V
     From  the  formula  (6)  or  (8)  and on  the basis  of  test results  in  a pilot-
scale precipitator,  the representative  examples of the performance charac-
teristics  curves  of  collection  efficiency  versus the  S.C.A.  for  the pilot-
scale precipitator in  comparison  with  the  full-scale  one are shown in Fig.8
and Fig.9.


                                  CONCLUSION


     The investigation described  in this paper have led to  the following
conclusions,
       l.Q
   §0.
   H
   u
   H
   PM
   Pn
   W
        0
         10-2
                                                      10"
                                      Ifttd/v
    Fig.7  Coefficient J*  as a function of ratio of dust layer potential
           on collecting electrode to applied voltage
                                     165

-------
(1)  The Deutsch migration velocity for a pilot-scale precipitator will be
different from     one of a full-scale precipitator, when the  S.C.A. is even
the same between  the both while the coefficient of re-entrainment is the dif-
ference between the both.  In general, the Deutsch migration velocity for
a full-scale  precipitator will become to be lower, compared with one of
a pilot-scale, since the aspect ratio of the full-scale precipitator is
smaller than  one  of the pilot-scale, particularly when the bulk density or
electric resistivity of dust layer on collecting electrode is  a relatively low

(2) The Deutsch migration velocity is decreased with increasing the S.C.A.
because of the influence of the geometric standard deviation of particle size
distribution  at the inlet of precipitator, however the Deutsch migration velo-
city will be  kept constant in independent of the S.C.A. when the particle size
     99
   u
   £;
   W
   H
   U
   H
   PL,
   fn
   W
   8 90
   u
   W
   o
   u
     50
                            PILOT-SCALE ESP,
                            2H/4N/1F/0.3D/2.7 5L/H
                  I
                           FULL-SCALE ESP,

                           12H/30N/2F/0.3D/0.86L/H
I
1
                                                                120
        0                   40                80
             SPECIFIC COLLECTING AREA,  S.C.A.(s/m)
Fig.8 Representative example for collection efficiency performance
      characteristics curve as a function of S.C.A.  in correspondence
      of a full-scale to pilot-scale
                                    166

-------
is almost uniform.

(3) The Deutsch  migration velocity is increased with increasing the  spacing of
the collecting electrode.  A so-called wide spacing effect is  caused by the
enhancement of an average corona electric field intensity as a predominant
factor.  Therefore,  the validity of an wide spacing effect should  be assessed
in consideration of  a space charge effect, corona current density, resistivity
of dust layer  on collecting electrode, coefficient ft ,  ^  and  both a spacing
of collecting  and one of discharging electrode.

(4) Applying the formula (6)or(8) and on the basis of the Deutsch  migration
velocity obtained from the test results in a pilot-scale precipitator,
the performance  characteristics of collection efficiency versus the  S.C.A.
for a full-scale precipitator can be predicted and evaluated.
      99
    U
    H
    u
    H
    W
      90
   O
   H
   EH
   U
   W
   O
   U
      50
              PILOT-SCALE ESP,

              2H/4N/3F/0.2 5D/3L/H
I
        FULL-SCALE ESP,

        11H/18N/2F/0.3D/1.07L/H
I
I
        0                  40                  80
               SPECIFIC COLLECTING AREA,  S.C.A.(s/m)
                                              120
Fig.9 Representative example for collection efficiency performance
     characteristics curve as a function of S.C.A.  in correspondence
     of a full-scale to pilot-scale
                                    167

-------
                                 REFERENCES

1. Deutsch,  W.   Bewegung und Ladung der ElektrizitMtstrager im Zylinderkonden-
   sator.   Annln.  Phys.  68,  1922. p.335-344.

2. White,  H.J.   Industrial Electrostatic Precipitators, Chap. 6, Addison
   Wesley, New York,  1963.

3. Masuda, S.   Motion of Charged Particles with Difference Sizes Inside
   an Electrostatic Precipitator.  In;  Proceedings of the Fourth Conference
   of the Institute of Eletrostatics of Japan, Tokyo, 1980. p.97-99.

4. McDonald, J.R.    A Mathematical Model of Electrostatic Precipitation,
   Revision 1,  Vol.1,  Modeling and Programing. EPA-600/7-78-lllb, 1978.

5. Isahaya,  F.    Analysis of Corona Electric Field for Electrostatic
   Precipitator and its Industrial Application.  Doctor Thesis of Tokyo Univ.,
   1961.  p.116-190.

6. Isahaya,  F.    Analysis of Efficiency Performance Characteristics of
   Electrostatic Precipitator by Utilizing Radioactive Tracer Technique
   Using Thorium B Aerosol Particles. In: Proceedings of Conference of Inst.
   Elect. Eng.  of Japan, Tokyo, 1961.  p.415-416.

7. Mierdel,  G.  und Seeliger, R.  Untersuchungen fiber die Physikalischen
   Vorgange bei der Elektrofilterung.  Arch Elektrotech., 29, 1935. p.149-172.

8. Robinson, M.  A Modified Deutsch Efficiency Equation for Electrostatic
   Precipitation.   Atom. Envi., 1, 1967. p.193-204.

9. Cooperman, P.  Efficiency Theory and Practice in Electrostatic precipi-
   tation.  In: Proceedings of the Fourth International Clean Air Congress.
   Tokyo, Japan, 1977.  p.835-838.

10.  Isahaya, F.  Electrostatic Precipitator by Using Ionic Wind for Very Low
     Resistivity Dusts from High Temperature Flue Gas of Petroleum-cokes
     Calcining Kiln.  The First Symposium on the Trasfer and Utilization
     of Particulate Control Technology. EPA-600/7-79-044a,  1979. p.453-466.

11.  Isahaya, F.  Analysis of the Corona Field Intensity Distribution in
     Electrostatic Precipitator by a Steel Ball Dropping Method.
     Elecrtrotechnical Jl" of Japan, Vol.8, No.3/4, 1963.  p.151-157.

12.  Isahaya, F.  Study on New Air Cleaning and its Testing method for
     Cleaning Effects. Special Publication of the Hitachi Hyoron.
     1964.  p.44-48
     The work described in this paper was not funded by the U. S. Environmentaj-
Protection Agency and therefore the contents do not necessarily reflect
the view of the Agency and no official endorsement should be inferred.


                                      168

-------
           DISTORTION OF PULSE VOLTAGE WAVE FORM ON CORONA WIRES
                          DUE TO CORONA DISCHARGE

           by: Senichi Masuda and Hajime Nakatani
               Department of Electical Engineering, University of Tokyo
               7-3-1, Kongo, Bunkyo-ku, Tokyo, JAPAN 113
                                  ABSTRACT

     A very short pulse voltage travelling on a corona transmission line pro-
ducing streamer coronas is subjected to a time-dependent distortion of its
wave form owing to energy consumption by corona.  There are two different
streamers in the case of a parallel-wire transmission line.  The first stream-
er is launched from the negative wire at the leading part of the pulse crest
voltage, while the second streamer emitted from the positive wire at its rear
part.  The first (negative) streamer produces a small distortion in the lead-
ing part, whereas the second (positive) streamer causes a great lowering of
voltage wave form in the rear part.  The time-dependent equivalent surge im-
pedance of the line also shows a concurrent lowering owing to these streamer
activities.

                               1. INTRODUCTION

     The specific advantages of using an extremely short pulse high voltage with
several tens to several hundreds nanoseconds duration time have been recognized
in twin-electrode type ESP's (1,2) and tri-electrode type ESP's (3), and also
a precharger called Boxer-Charger developed by the authors (4).   Such a short
pulse voltage proceeds in a form of a travelling wave along a corona trans-
mission line consisting of  either a corona wire and a collecting electrode,
a corona wire and a third non-corona electrode, or two parallel corona wires,
and produces streamer coronas as a plasma ion source uniformly distributed on
the line.  One of its largest advantages is a simplicity in the construction of
its pulse power supply.   Another large advantage is its low power consumption
compared with that producing a long duration pulse voltage.  This is due to an
effective conversion into corona discharge of the pulse energy,concentrated in
a travelling wave packet and not distributed along the total length of the
wire in a form of capacitive energy.   The residual energy after losing its
power to produce coronas remains also small.   In a parallel corona wire trans-
mission line,  two different streamers appear.  The first (negative streamer)
is launched from the negative wire towards the positive one.  In the case when


                                     169

-------
the pulse voltage is sufficiently high,  the  second streamer (positive stream-
er) is subsequently emitted from the positive wire in  the opposite direction
to bridge across the gap.  A more detailed description on these streamers is
given in the separate reports  (5,6).  The travelling pulse voltage loses its
energy as a result of corona production  and, partly, of the skin-effect en-
hanced resistance loss in the  line.  Hence,  ite wave form undergoes a gradual
modification and attenuation,  finally to become inactive so as not to produce
coronas.  The active length of the  transmission line producing negative stream-
ers is greatly increased with  the aid of a dc field externally applied to the
corona wires (5).  In other words,  more  energy for negative streamer formation
can be squeezed out of the pulse by the  aid  of dc  field in this case.

     The largest matter of concern  in the use of an extremely short pulse high
voltage is to understand the role of each streamer in  the modification and
attenuation of the voltage wave form, and in the concurrent change in the
surge impedance of the transmission line.  In this paper are reported the
results of the investigation directed to this target.

                         2. EXPERIMENTAL APPARATUS

     Fig. 1 illustrates an experimental  set-up used in this experiment consist-
ing of 35 m long parallel corona wires shaped in a form of a double-helix
transmission line with 10 m in length.   This is located horizontally between
a long grounded (lower) plate  and an isolated (upper)  high potential electrode.
The dimensions of the double-helix  are given in the figure.   The position on
the double-helix line is expressed  in terms  of its distance,  x, from the inlet
as shown in Fig.l-(a).  Hence, the  actual wire length  from the inlet is L =
3.5x.   One of the two wires (mother wire) of the  double-helix is connected to
the grounded plate at each 0.5 m distance, and the pulse voltage is applied
through a feeder cable between the  two wires in a  polarity that another wire
(daughter wire) is negative in reference to  the mother wire.   The main voltage
wave travels along the helix wires, while two parasitic waves along the trans-
mission lines (daughter wire /  grounded plate) and  (daughter wire / upper plate).
The magnitude of the pulse current  of the former parasitic wave amounts to
37.5 % of that of the main wave.  The main wave proceeding along the double-
helix line produces streamer coronas along its wires,  with concurrent distor-
tion and attenuation in its wave form.   Two  transmission lines, (double-helix)
and (daughter wire / grounded plate), are terminated respectively by the match-
ing resistances RI = 260 ohm and R£ = 690 ohm.  The ratio of R^ and R£ is
0.377, exactly equal to the inverse of the current ratio.   The surge impedance
of the feeder cable is either  50 ohm (section 3.1)  or  75 ohm (sections 3.2, 4
and 5 ), and its length 50 m except in the section 4.2.   A positive dc voltage,
V^c, is applied to the upper plate  to form a dc field, E,  between the double-
helix and upper plate for extracting negative ions from the streamers to the
upper plate.  The magnitude of this field is expressed in terms of an average
value, E = Vdc/80mm,  throughout this paper.

     The voltage wave form is  measured by a  high voltage probe (Tektronix: P
6015) in combination with a high-speed dual-beam oscilloscope (Tektronix:7844),
and the current wave form by a Rogowski-coil having a  sensitivity of 2.2 mV/A
in a passing band between 80 kHz and 30  MHz.  The  development of streamers
between the double-helix wires is observed by a streak photograph taken
through a 7 mm slit with an image converter  camera (John Hadland:-HE 700)

                                      170

-------
combined with an image intensifier (EMI  : 9912)  (Fig.l-(b)).   The  light  signal
from the corona wires is detected with a photomultiplier  tube  (Hamamatsu :  R
212) and compared with the wave form of the pulse voltage applied  to  the wires,
so as to determine the starting point of the pulse voltage on  the  streak
photograph.

     Fig.2 shows the pulse generator used, consisting of a coaxial cable A.
The cable A is charged from a dc high voltage source, B, through a resistance,
R, and, then, discharged through a sphere gap to an initially  uncharged  feeder
cable, C, connected to the double-helix wires.  The duration time  of  the pulse
voltage produced is twice the transient time of a travelling wave  voltage
across the cable A.  The pulse wave form at the inlet of the double-helix is
approximately rectangular, as shown in Fig.3-(a).  The surge impedance of the
cable A is 50 ohm and its length is 20 m.  A ceramic capacitor of  1700 pF is
also used in place of the cable A.  The voltage wave form in this  case is as
shown in Figs. 6-(a) and 7-(a).  The variation of the pulse rise time is made
by inserting an inductance after the spark gap.  The rise time and duration
time of the pulse voltage wave are defined, respectively, by Tr =  time of rise
from 10 % to 90 % of the crest voltage, Vp, and Tft = half-peak duration  time.

                   3. DISTORTION OF PULSE VOLTAGE WAVE FORM

     Prior to this experiment the exact values of R^ and R2 are determined by
supplying to the double-helix a very short pulse voltage with  a duration time
less than half of the time of reflection and observing the reflected wave at
its inlet.  The values of RI and R2, thus obtained, are 260 and 690 ohms,
respectively.  The travelling speed of this voltage wave is 0.28 m/ns, 93 %
of the light velocity in this particular case.

3.1 DISTORTION OF RECTANGULAR PULSE VOLTAGE WAVE FORM

     A rectangular pulse voltage with Tr = 70 ns and T^ = 200  ns produced
using the cable-type pulse generator is applied at the inlet of the double-
helix.  Fig.3-(a) shows the pulse voltage wave forms in the absence of corona
activity measured at the inlet, center and terminal of the double-helix  wires
with a low inlet pulse voltage (Vp = -29 kV) and zero dc field (E= 0).   The
wave form remains almost unaltered, with a slight attenuation  of ca.  3.5 % at
the terminal.  This is ascribable to a skin-effect enhanced resistance loss
in the wires, specific to a fast rising pulse.

     When E  is raised from 0, keeping Vp = -29 kV, the negative streamers
begin to develop at E = -2 kV/cm from the upper edge of the daughter wire in
a region close to the inlet towards the upper electrode (Fig.4-(a)).  As E is
further raised, the negative streamers are enhanced in length, appearing
finally on the whole length of the double-helix.  The streak photograph  of
this negative streamer taken at x = 0.27 m and E = -5 kV/cm is shown  in  Fig.
5-(a), indicating the existence of a time lag T_ = 50 ns.  The distortion of
the voltage  wave-forms in this case is indicated in Fig.3-(b).  It can be
recognized that the negative streamer erodes the region ahead  of the  pulse
crest voltage from t = T_.

     When Vp is raised to -36 kV with E = 0 kV/cm, the positive streamers
begin to develop to bridge across the two wires as shown in Fig.4-(c).   In
this case the voltage wave form becomes greatly distorted even at  the inlet

                                     171

-------
with the rear portion of its crest  greatly eroded as shown in Fig.3-(c).  It
rapidly attenuates to Vp = -33.5 kV at  x = 1.5 m and the positive streamers
disappear beyond x = 1.5 m.  Between x  = 1.5  m and 10 m, only negative stream-
ers, weak in luminosity and short in length,  develop from the daughter wire
with substantially smaller attenuation  of the voltage (Fig.3-(c)).

     Fig.3-(d) indicates the distortion occuring at a high inlet pulse voltage
(Vp = -35 kV) and a high dc field (E =  -5 kV/cm).   The phenomena occuring in
the region x g 1.5 m are almost the same as in the case of Fig.3-(c) where
E = 0.  Beyond x = 1.5 m, the positive  streamers again disappear, leaving only
the negative streamers spreading along  the entire length of tfre daughter wire.
Hence, the positive streamers are not enhanced by the dc field.  The differ-
ence exists, however, in a stronger luminosity and increased penetration
length of the negative streamers (Fig.4-(b)), resulting in an enhanced energy
loss.  The negative streamer in this case causes the erosion not only in the
region ahead of the crest voltage,  but  also in the crest itself.  As a result,
the pulse crest voltage attenuates  more rapidly than in Fig.3-(c).  This also
leads to a change in the appearence of  corona from Fig.4-(b) to Fig.4-(a) in
a region near the terminal.
     Figs.S-(b) and (c) are the streak  photographs of the negative and positive
streamers occured at a high pulse voltage in  Figs.3-(c) and (d), respectively.
The time lags of the negative and positive streamers, T- and T+, are 70 ns and
110 ns for E = 0 kV/cm, and 40 ns and 100 ns  for E = -5 kV/cm.
     It is recognized in Fig.3-(c)  that the negative streamer starting from
t = T_ causes a small but continued erosion in the region ahead of the crest
voltage.  Whereas the positive streamers starting from t = T+ first erode
rapidly the subsequent region behind the crest to produce a very large dip.
This dip formation is likely to be  due  to a large energy consumption required
for formation of the second streamer plasma channel in its propagation process
(leader formation).  This is followed by a steady after-glow accompanied by a
subsequent recovery in voltage up to a  level  slightly lower than -30 kV
(plasma sustaining voltage).  In other  words, a large dip in a wave form sub-
sequent to a high crest voltage (initiation voltage) and followed by a voltage
recovery (plasma sustaining voltage) generally represents a distinct sign in-
dicating the occurrence of positive streamers.  A large distortion of the wave
form occurring at x = 0 from the ideal  lossless wave form depicted in this
figure is considered to be caused by the time-dependent decrease in the volt-
age transmission rate of the double-helix line due to the corona-induced low-
ering of its effective surge impedance  as described later.
     A drastic change occurs in the wave form in the presence of an adequately
high externally applied dc field (Fig.3-(d)).  In accordance with the in-r
creased activity of negative streamer enhanced by E (see Fig.5-(c)), the ero-
sion of the region ahead of the pulse crest starts at a lower voltage with a
smaller time lag T_.  The erosion continues successively with an increased
rate  even at the stage when the pulse  voltage has dropped to a much lower
level.  In other words, the externally  applied dc field enhancing the local
field around the negative wire assists  the extraction of the more energy from
the leading part of the pulse voltage to convert into negative streamers.  In
contrast, the dc field in this particular case has an effect of relaxing the
pulse-induced local field around the positive wire.  As a result the positive
streamers are somewhat hampered (see Fig.5-(c)), and the erosion starting at *

                                      172

-------
t = T+ by these streamers at the subsequent region behind  the  crest  is  also
diminished.   It should be added that the undistorted wave  forms  depicted  in
Figs.3-(c),  3-(d), 6-(b) and 7-(b) are estimated from  those at the output of
the pulse generator by inserting a 50 m long feeder cable  (delay line) , and
thereby eliminating the interference by coronas in the double-helix  complete-
3.2 EFFECT OF RISE TIME AND CREST VOLTAGE

     The capacitor type pulse generator with a capacitor C =  1700  pF  is  used.
A fast rising, narrow pulse wave form (sharp pulse) with Tr = 38 ns and  Tft =
150 ns is produced by removing the inductance Lr  (Fig. 2).  A  slow  rising,
wider pulse (moderate pulse) with Tr = 100 ns and T^ = 320 ns is produced  by
inserting Lr = 10 ;iH.  These values of Tr and T^ are those measured at the
output of the pulse generator.

     Figs. 6 and 7 show the distortion of the pulse wave form  at E  = -5 kV/cm
for these sharp and moderate pulses, respectively.  The pulse crest voltage,
Vp, undergoing attenuation in these figures is plotted against x in Fig. 8,
where two solid curves show the results for the moderate pulse and two dotted
curves for the sharp pulse.

     When the inlet pulse voltage is low (Vp = -32 kV; Fig.6-(a) and  Fig. 7-
(a)), only negative streamers are produced along the whole length  of  the
double-helix, and they erode the front side of the pulse crest, independent
of the pulse wave form.  However, the sharp pulse voltage is  more  easily sub-
jected to the erosion of the pulse crest because of its narrow pulse  width
(Fig.6-(a)).  It can be recognized from Fig.7-(a) that the moderate pulse
wave has enough energy distributed in a time-domain to produce coronas even
after x = 10 m.  This results in a great difference in the attenuation rate
Vp between the sharp and moderate pulses (see the two curves  starting from
Vp = -32 kV at x = 0 in Fig. 8).
     Fig.6-(b) and Fig.7-(b) show  the distortion occurring at a higher pulse
voltage to produce active positive streamers bridging across  the double-helix
wires in the region close to x = 0.  After the pulse voltage  travels  a certain
length (1 m for the sharp pulse and 1.5 m for the moderate pulse), the posi-
tive streamers disappear because of the decrease in pulse voltage.  In both
the sharp and moderate pulses, the effective pulse duration time becomes
decreased at x = 0, owing to the drop ,in transmission rate caused  by  the posi-
tive streamer induced lowering of the effective surge impedance.   This dimi-
nished pulse duration time causes in turn a continued rapid attenuation of
Vp after the positive  streamers disappear.  Thus, in the case of  moderate
pulse, its crest voltage attenuates from -38 kV at x = 0 to -30 kV at x =  10
to meet another solid curve starting from a lower level of Vp at x =  0 (Fig. 8).
In the case of the sharp pulse, even the reversal of its crest level  occurs
as indicated by two dotted curves in Fig. 8 crossing each other at  x = 5 m.
These erosions lead to the flattening of the pulse voltage and lowering of
its rise time.

     Hence, the selection of the pulse wave form and its crest voltage must be
made in consideration of its overall corona producing performance  in  its
entire propagation process.  In the case of a long transmission line, the
selection of a too sharp pulse voltage loses its meaning, since it rapidly
loses its peak region to turn into a broader, inactive pulse.

                                     173

-------
          4. EFFECT OF REFLECTION OF TRAVELLING WAVE AT  TERMINAL

     When the terminals of the double-helix wires are open  instead of being
terminated with matching resistances, a reflection  of  pulse  occurs at the ter-
minal.  As a result the terminal voltage is  theoretically  doubled when no
corona loss exists, and this reflected wave  is superimposed  to the oncoming
wave to enhance the overall apparent pulse voltage  on  the  line.   The effect
of this reflection is investigated with a special attention  to its technical
importance, using the moderate pulse voltage  (Tr =  100 ns  and Tfo = 320 ns)
with different levels of its crest at the inlet, Vp, and with effectively open
terminals (RI = 100 kilo-ohm and R2 = oo ) .
4.1  CHANGE OF OVERALL PULSE CREST VOLTAGE

     The overall apparent pulse crest voltage, including both  the  forward and
reflected travelling waves superimposed, is plotted  against  the  distance, x,
in Fig. 9.  In the case when no corona occurs  (bottom line),  the  overall  crest
voltage continues to rise to become almost doubled at x =  10 m as  expected.
Fig.lO-(a) indicates one of the examples of wave  form modification in  the
similar corona-free case but with E = -5 kV/cm.   The reflected wave can  be
clearly distinguished from the forward wave at x  = 0 and x = 5 m.

     When the inlet pulse crest voltage is raised, the positive  streamers
begin to occur at the terminal region, and the wave  forms  at x = 5 and 10 m
indicate a typical dip specific to the positive streamers  (Fig.lO-(b)) .  The
crest levels prior to these dips at x = 0, 5  and  10  m are  limited  to -35 kV
which is the initiation threshold of the positive streamers  in this case.
This voltage limitting effect is clearly indicated also in Fig. 9.   The  upper
three curves starting from different levels of Vp at x = 0  come to meet at
around Vp = -35 kV at x = 5 m, and remain flat up to x = 10  m.   As soon  as the
voltage, enhanced by reflection, exceeds this threshold, the positive  stream-
ers are produced to cause a large dip, and a  subsequent recovery in voltage,
as described before.

4.2  RELATIONSHIP BETWEEN RESIDUAL DC VOLTAGE AND POSITIVE STREAMERS

     In an open-ended transmission line, the  voltage wave  propagates back and
forth many times between the pulse generator  and  the open  terminal of  the line
through the feeder cable, undergoing multiple reflections.   The  wave dissi-
pates its energy owing to the corona loss, skin-effect enhanced  resistance
loss, and the loss at the spark gap by passing through its arc plasma.   The
wave is gradually broadened by the energy dissipation to lose  its  nature as
a travelling wave, finally to be converted into a dc residual  voltage  remain-
ing on a transmission line.

     This dc voltage is measured at t = 6 ys  to estimate approximately the
conversion ratio of pulse energy into corona  energy.  In Fig. 11  the residual
dc voltage is plotted against the pulse crest voltage, Vp, at  the  inlet  of the
double-helix line where the solid curve is for E  = 0 kV/cm and the dotted
curve for E = -5 kV/cm.  Both solid and dotted curves coincide very well to
each other, indicating a linear rise in the residual voltage with  Vp up  to
vp = -20 kV.  With E = -5 kV/cm the negative  streamer begin  to occur at  Vp =
-14 kV, whereas no streamers occur with E = 0 up  to  Vp = -20 kV.  Thus,  it can
be seen that the energy consumption of negative streamers  is so  small  that no


                                      174

-------
appreciable change occurs in the residual dc voltage.  At Vp = -20 kV  the
positive streamers begin to occur, with both E = 0 and 5 kV/cm,  in the termi-
nal region of the double-helix line.  With increasing Vp from -20 to -41 kV,
the positive streamer region spreads gradually from the terminal to the whole
length of the line.   Concurrently, the curves begin to deviate from the
straight line,  taking maximum at Vp = -25 kV, and drop rapidly to become zero
at Vp = -41 kV.   This again indicates a large energy consumption of the posi-
tive streamers,  probably because of their very conductive plasma channels
bridging across  the wires to bleed the charge and energy of the  pulse  voltage
wave.

  5.  SURGE IMPEDANCE OF DOUBLE-HELIX WIRES AND ITS MODIFICATION BY CORONAS

     The surge  impedance, Zo, of a transmission line is modified by corona
loss during the  propagation of the pulse voltage.  The effective value of Zo
can be obtained  by measuring the lacal time-dependent values of  the pulse
voltage V(x,t)  and current I(x,t) as

          Zo(x,t) = V(x,t)/I(x,t)                          	(1)

     The pulse voltage and current wave forms at the inlet, V(0,t) and 1(0,t),
are measured with a dual-beam oscilloscope for different levels  of inlet
pulse  crest voltage to obtain Zo(0,t).  A square-waved pulse voltage  (Tr =
50 ns and T^ =  240 ns) is used, and the end of the double-helix  line is termi-
nated with the matching resistances.  To eliminate the current of the  para-
sitic pulse described in section 2, the Rogowski coil (current probe)  is wound
around the mother wire.  The surge impedance of the feeder cable used  is 75
ohm and its length 50 m.

     The wave forms of V(0,t) and 1(0,t) thus obtained are shown in Fig.12.
In the case when no corona occurs, V(0,t) and I(0,t) indicate the same wave
form as shown in Fig.12-(a).  With  Vp kept constant and increasing E from 0 to
-6 kV/cm, the negative streamers occur along the whole length of the double-
helix wires.  The wave forms indicate a slight concurrent change (Fig.l2-(b)).
The voltage crest decreases from -34 kV to -33 kV and the current crest in-
creases from -133 A to -140 A.   In the case when the positive streamers occur,
the voltage indicates the typical dip, and current shows a concurrent  abrupt
rise (Fig.l2-(c)).

     The time-dependent surge impedance Zo(0,t) is calculated by Eq.(1) from
measured wave forms  of V(0,t) and I(0,t) at different levels of  Vp, and
plotted against  time in Fig.13.  In the case when no corona activity takes
place (Vp = -26  kV)  the value of Zo(0,t) remains almost constant at about 270
ohm during the entire pulse duration time.  In the case when only, the  nega-
tive streamers occur (Vp = -33 kV and -35 kV), it shows a very small time-
dependent drop from 250 ohm to 240 - 200 ohm.  In the case when  the positive
streamers occur  (Vp  = -39 kV and -41 kV), Zo(0,t) drops sharply  after  a cer-
tain delay time  from 250 ohm to 80 - 100 ohm.  This delay time coincides with
the time lag for positive streamer initiation.
     The drop of the effective surge impedance does not mean the abrupt jump
of the pulse  current from the pulse generator, but it produces a local jump
of current in the cost of local attenuation of the pulse voltage.
                                     175

-------
                      3. CALCULATION  OF CORONA LOSS

     We consider a simple case when no  reflection occurs.  The power of a for-
ward travelling wave at an arbitrary  position x and time t is given by:

          P(x,t) = V(x,t) x  i(x,t)                             	(2)

Therefore, the pulse energy,  E(x), carried by the wave in the forward direc-
tion is

          E(x) = /P(x,t)dt                                    	(3)

               = /V(x,t) x  I(x,t) dt                         	(4)

               = /V(x,t)2/Zo(x,t) dt                         	(5)

The energy converted to the  corona loss in a region between x-£ and X£ is

          Corona Loss = E(XI) -  E(x2>                         	(6)

     This corona loss is estimated for  the negative and positive streamers
separately from the measured voltage  wave forms and time-dependent change of
surge impedance by using the Eqs.  (5) and (6).   The following simplified con-
ditions are taken for this estimation:
Negative streamers:
     1) xi = 0 m and x^ = 10 m.
     2) Zo(x,t) = 240 ohm (constant).
     3) In the absence of positive streamers, the integration of Eq.(5) is
        made only in a time  interval  for the corona-induced wave distortion
        to appear.
     4) In the presence of positive streamers in the inlet region of the line,
        the integration is made  from  t  = 0 to the time of the crest voltage.
Positive streamers:
     1) x-^ = 0 m and X£ = position at which positive streamers disappear.
     2) Zo(xi,t) = 100 ohm and  Zo(x2,t) = 240 ohm.
     3) The integration of Eq.(5)  is  made from the time of crest voltage to
        the time at which positive streamer induced wave distortion disappears
        at x = X£, i.e. V(xi,t)  =  V(x2,t).
     The estimated value of  corona loss by negative streamers per unit length
of the double-helix, VE_/Vx,  ranges from 0.005 to 0.010 J/m and increases with
increasing inlet pulse voltage.   The  corona loss per unit length of wires is

          VE_/VL = 0.010/3.5 [J/m] =  0.0029 [J/m]             	(7)

     On the other hand, the  corona loss by positive streamers per unit length
of the double-helix, VE+/Vx,  is  0.40  -  0.58 J/m, about fifty times as high as
VE_/Vx.  The loss per unit  length of  the wires is VE+/VL = 0.11 - 0.17 J/m.
It should be noted that 27 - 44  % of  the inlet pulse energy is consumed by the
positive streamers after propagation  of only 1.0 or 1.5 m from the inlet.

     It is confirmed by other experiments that, in spite of this high energy
consumption of positive streamers,  their emission ability of ionic current is


                                      176

-------
almost equal to the negative streamers (5,6).  Hence, the occurrence of the
positive streamers should be avoided in order to produce as much monopolar
ions as possible along a longest possible corona transmission line with a
limited pulse energy.

                            8.   CONCLUSIONS

     The corona-induced wave-form distortion of an extremely short pulse volt-
age travelling on a long transmission line is investigated, and the following
conclusions derived:

1)  The negative streamer is launched at the leading part of the pulse crest
    voltage, and erodes its front side.  Whereas the positive streamer, sub-
    sequently emitted at a higher pulse voltage, bridges across the wire gap,
    and causes a great dip in the voltage wave form in the region behind its
    crest.
2)  The positive streamers occur primarily in the inlet region of the line, as
    it rapidly disappear after propagation of say 1.0 or 1.5 m, because of
    their large energy consumption to produce a sharp attenuation of the pulse
    voltage.  The negative streamers consume much less energy, and can develop
    along a long distance of the line when assisted by an externally applied
    dc negative field for extraction of negative ions.  This dc field plays
    a role of squeezing more negative corona energy out of the pulse voltage
    wave, but produces no  enhancing effect for the positive streamers.
3)  Even at the inlet of the line,  the region behind the crest is greatly
    eroded when the positive streamers occur.  This is because of a concurrent
    abrupt drop of the line surge impedance.
4)  The shorter the pulse duration time,  the faster the pulse crest voltage
    attenuates.
5)  In the case when the pulse voltage is high enough at the inlet so as to
    produce intensive positive streamers  at the inlet region, the attenuation
    of pulse crest is more rapid than in  the case when only negative streamers
    occur.  This is because of a great erosion in the region behind the crest
    by positive streamers to produce also a great decrease in an effective
    pulse duration time.
6)  The corona losses per unit length of  the double-helix line due to nega-
    tive and positive streamers are 0.005 - 0.010 J/m and 0.45 - 0.58 J/m,
    respectively.   In the case when positive streamers occur, 27 - 44 % of the
    inlet pulse energy is consumed after  propagation of only 1.0 or 1.5 m from
    the inlet.   Therefore, the occurrence of positive streamers must be avoid-
    ed in order to provide as much monopolar ions as possible along a longest
    possible length of a transmission line with a limited pulse energy.
7)  When the line  is open-ended, the reflection of the wave can raise the
    overall pulse  voltage.   But it dose not exceed the positive streamer ini-
    tiation threshold as a result of a strong crest-trimming effect of the
    positive streamers consuming large energy.  In the case when the pulse
    voltage is  very high to produce positive streamers along almost whole
    length of the  double-helix line, the  pulse energy can be completely con-
    verted to corona loss after the multiple reflection of the wave on the
    line.
8)  The base line  surge impedance, Zo, at the inlet of the double-helix is
    270 ohm.   The  negative streamers produce a slight time-dependent decrease
                                    177

-------
    of its value from 250 to 240 - 200 ohm.  The positive  streamers,  on  the
    other hand, produces after a certain delay time a sharp abrupt  drop  from
    250 to the level of 80 - 100 ohm, as a result of concurrent  drop  in  volt-
    age and rise in current.
                                 REFERENCES

1  Milde,  H.I.  and VanHoesen, H.E., Application of Fast Rising Pulses to
   Electrostatic Precipitators,  Conf. Rec. IEEE/IAS Annual Meet. 1979,
   pp.158-162  (1979).

2  Masuda, S. and Hosokawa, S.,  Corona Characteristics of Travelling Waves
   in the  Pulse Energized ESP,  Conf. Rec. of the Institute of Electrostatics
   Japan Annual Meet.  1982 (Nov. 1982 to be presented).

3  Masuda, S. and Hosokawa, S.,  Travelling wave Pulse Energization in A Tri-
   Electrode Type Electrostatic Precipitator, to be published separately.

4  Masuda, S.,  Mizuno, A., Nakatani, H. and Kawahara, H.,  Application of
   Boxer-Charger in Pulsed Electrostatic Precipitato-r, Conf. Rec. IEEE/IAS
   Annual  Meet. 1980,  pp.904-911 (1980).

5  Masuda, S.,  Nakatani, H., Yamada, K., Arikawa, M. and Mizuno, A.,  Produc-
   tion of Monopolar Ions by Travelling Wave Corona Discharge, Conf. Rec.
   IEEE/IAS Annual Meet. 1981, pp.1066-1073 (1981).

6  Masuda, S. and Shishikui, Y.,  Pulse Corona as Ion Source and Its Behav-
   iours in Monopolar Current Emission,  The Fourth EPA-Symposium on the
   Transfer and Utilization of Particulate Control Technology (Oct. 1982
   to be presented).
               The work described in this paper was not funded
               by the U.S. Environmental Protection Agency and
               therefore the contents do not necessarily reflect
               the views of the Agency and no official endorsement
               should be inferred.
                                     178

-------
                                                    X •*-
                                10 m
                                           I
                           80 mm
                                  Upper  P.lte Electrode
                                               Double-Helix
                                               Wires
                          r
                     |     TeOmm
        Terminal
         (x = 10 m)
                 Lower Plate
                     Electrode
         5 nun
               20 mm
(a)  side view  - '

             (b)  sectional  view
Fig.l
Experimental  Set-up of  Double-Helix
Corona Transmission Line
                                                               .
                                                              /
                                                              1
                       Daughter Wire
                         Mother wire
                       /
                                                                         Feeder
                                                          Inlet   Rogowski
                                                         (x = 0 m)       Coil
                                                         80 cm
(7 mm)
                                                     Upper Plate
                                                      / Electrode
                                                   Double-Helix
                                                      Wires
                                                           100  cm
                                             *  \
                                           Lower Plate
                                           Electrode
                 I.C.  :  Image Converter
                         Camera
                 I.I.  :  Image Intensifier
                 P.M.  :  Photomultiplier
                         Tube
                         I—* to trigger of Image Converter
                                         Camera
                       |~| photo-cell
                           Lr      C
 SO Hz
                                                         Fig. 2
                                                         Pulse Generator
                       (a)
                       only negative
                       streamers
                       (Vp = -28 kV)
                       (b)
                       only negative
                       streamers
                       
-------
        -30
       (a)  Low Pulse Voltage and Zero DC Field
           (no corona discharge; Vp = -29 kv
            at x =  0)
                                                     -30
                  (b) Low Pulse Voltage and High DC Field
                     (only negative streamer; Vp = -28  kv
                      at x = 0; E = -5 kV/cm)
        -40
        (c)  High Pulse Voltage and  Zero DC Field
           (positive streamers only in the
            region: xs 1.5 m; Vp = -36 kv
            at x = 0)
                                                      -40
                  (d) High Pulse Voltage  and High DC Field
                     (negative and positive streamers in
                      region: x i 1.5 m  ; Vp=-35 kV at
                      x = 0 ; E = - 5 kV/cm)
            Fig.3   Wave  Form Distortion  of Rectangular Pulse Voltage
                     (Tr = 70  ns,  Th = 200 ns)
                                              MS)
(a) Vp - -28 kv (at x-0
   E = -5 kv/cn
   E - 0 kv/cm
   E = -5 kv/c
56     100                    I Fig. 5
                         •
                              Streak Photographs of
                              Streamers  Taken at x =
                              0.27 m
                              (square-waved pulse
                                voltage  : Tr = 70 ns,
                                Th =  200  ns)
                                             180

-------
         100   200   300    400   500
 -40
                                                     100    200   300   400   500
                                                                 undistorted
                                                                 wave form
     (a) negative streamer only
                                            -40
(b)  negative and positive streamers
               Fig.6  Distortion  of Voltage  Wave Form
               (sharp pulse voltage : Tr =  38 ns,
                Th  =  150 ns  ; E  = -5 kV/cm)
                 t (ns)
         100    200    300    400    500
-40
        100    200    300   400    500
                                           -40
       (a) negative streamer only
(b) negative and positive streamers
               Fig.7  Distortion  of Voltage  Wave Form

               (moderate pulse voltage  : Tr = 100 ns,
                Th  =  320 ns  ; E  = -5 kV/cm)
                                    181

-------
       -40
       -30
     o -20
     2
     u
     8 -10
Moderate Pulse
   Tr = 100 ns,
      Th =  320 ns)
             Sharp Pulse
             (Tr=38 ns, Th"150 ns)
               no corona discharge
          0    2    4   6    8    10
                Distance  x (m)

Fig. 8
Effects  of Inlet Crest  Level of Pulse
Voltage  and its Sharpness  on Attenua-
tion of  Pulse Crest Voltage
(E = -5  kV/cm)
                                               ~ -40
                                               8, -30
                             o
                             £
                             3
                             a.
                                                 -10
                                                           with corona discharge
                                                          , co'°
                                      2468
                                         Distance x (m)
                                                                         10
                          Fig. 9
                          Change  of Overall Pulse Crest Voltage,
                          Vp, with  Open Terminal
                          (moderate pulse  : Tr  = 100 ns, T^ =
                          320 ns  ;  E = -5 kV/cm)
                   t   (ns)
       0    100   200   300   400   500
           (a)  without coronas
               (E = -5 kV/cm)
                                            t   (ns)
                                0    100   200   300    400   500
                                              -30
                                              -40
                                     (b)  with coronas
                                         (E = -5  kV/cm)
         Fig.10  Distortion of Voltage  Wave Form with Open Terminal
                  (moderate  pulse : Tr = 100 ns, Th  =  320 ns)
                                        182

-------
      -12
      -10
       -6
       -4
       -2
                —O- 2 = 0 kV/cm
             -*	*-E = -5 kV/cm
             I      I     I
         0   -10    -20  -30   -40
        Inlet Pulse Crest Voltage Vp (kv)

Fig.11
Residual DC Voltage after Multi-
Reflection  of Travelling Wave
Pulse Voltage on an Open-Ended
Line
(pulse generator :  capacitor  type
(C = 2500 pF)  ; feeder:75 n cable
(6 m long)  ; residual  dc voltage :
measured at t = 6 us)
 (a)
no corona
discharge
 (E = 0 kV/cm)
(b)
only negative
streamers
(E = 6 kV/cm)
(c)
both negative
and  positive
streamers
(E = 6 kV/cm)
V: lOkV/div
  : 23A/div

t: 50 ns/div
V: 10 kV/div
I: 23 A/div

t: 50 ns/div
V: lOkV/div

I: 92 A/div

t: 50ns/div
 Fig.12
 Distortion of  Voltage  and Current Wave
 Forms  Due To Streamer  Coronas
 (Tr  =  50 ns, Th = 240  ns)
^+n
<
~ 300
N

0>
O
c
to
"S 200
a
E

3
co 100
£
c




O—o




No Corona

, (E = 0 kV/cm)
/

/ 	 ,
o "~°— •o-"°^"""**o.^_/' "
£>"3!v. 	 n_
a r

Both Nee
and Posi
w

>-


Tfc£x— A °^-a—<> 1
\^
ative ^S\
tive i
Streamers








'^""A~AV
AJ
• V
\ -
\
X
Tt X '••



<"


Vp =
""^o-o -26 kV



ily Nega
Stre

' -33 kV
' -35 kV
;ive
amers

S9— V«*«*
»--
—&—X—~t~
— *—*— *^.

T -39 kV
-41 kV












0 50 100 150 200 250
t (ns)
Fig. 13 Time- Dependent Change in Inlet
for Different Streamer Modes

Surge Impedance, Zo(0,t),

(square-waved pulse voltage : Tr = 50 ns, Ti, = 240 ns ;
                 E = -6 kV/cm)
                                        183

-------
                         ELECTROSTATIC PRECIPITATOR
                           ANALYSIS AND SYNTHESIS

                by:  Ta-Kuan Chiang
                     Thomas W. Lugar
                     General Electric Environmental Services, Inc.
                     Lebanon, Pennsylvania 17042
                                  ABSTRACT
     An analytic model using the modified Deutsch approach has been developed
to describe the overall performance of an electrostatic precipitator comprising
two different configurations or energization methods in tandem.  Measurements
of the overall collection efficiency and the collection efficiency of either
configuration provide sufficient information to isolate the performance con-
tributed by the other.  Field data, obtained with pulse energization at the
outlet half of the plate area and conventional dc energization at the inlet
half of a full-scale utility precipitator, are presented to illustrate the
method.
                                    184

-------
                                INTRODUCTION
     In electrostatic precipitation technology, pilot-scale testing under
actual flue gas conditions has been used for obtaining full-scale sizing data
for new applications or fuel sources.   It has also been used for field demon-
stration of new improved electrostatic precipitator designs.  Typically, a
scaled-down version of the conventional or improved design is set up in para-
llel with the existing flue gas control device.  This parallel arrangement is
often referred to as a slipstream pilot.  A slipstream pilot is generally
small and compact in size and thus transportable from site to site.  To obtain
meaningful test results, conditions of gas flow, particle size distribution
and electrical sparkover levels experienced in a full-scale installation must
be closely duplicated in the slipstream pilot.  Even if these conditions are
satisfied, prior operating experience  of full-scale installations is required
to provide empirical guidelines to correlate slipstream pilot observations to
full scale with acceptable accuracy.  Thus, unless reliable scaling means
exist for direct extrapolation of slipstream pilot performance to full-scale
design, the method of slipstream pilot testing cannot be extended without
reservation for testing a new or improved precipitator concept since prior
experiences on full-scale installations would not exist.

     This paper discusses the methodology of evaluating an alternate on-stream
full-scale pilot test wherein a portion of the existing full-scale precipita-
tor is modified with a new or improved precipitator design or energization
concept.  Where space is available, an electrical section could be added in
series to test a new design concept.  Following the modified Deutsch approach
(1,2) for both configurations, an analytic model capable of separating precip-
itator performances of each individual configuration can be established.  With
the assistance of this analytic model, full-scale sizing information for a
new or improved precipitator can thus  be obtained directly without the pre-
requisite of empirical guidelines derivable only from existing full-scale
installations.  Obviously, a properly  selected testing site on a small full-
scale unit would be ideal to minimize  on-stream full-scale pilot testing costs
in the case where a new precipitator design is to be evaluated.  The size of
the full-scale unit would not be a critical consideration in testing of new
methods of energization such as pulse  powering which requires no internal
modifications to the precipitator.

                            STATEMENT  OF PROBLEM

     Similar to the circuit analysis and synthesis in electrical engineering
practices, on-stream electrostatic precipitator pilot testing comprising two
different configurations or energization means in tandem is also a problem of
analysis and synthesis.   Following the modified Deutsch representation, the
input and output of the problem are the corresponding volumetric gas flow rate
v, the particle loadings or the collection efficiency n, and k values for
mainly particle dispersion characterization.  The precipitators are completely
and uniquely described by modified migration velocity tofc, collection area A,
or the plate area to flow rate ratio A/v.
                                     185

-------
     The problem of electrostatic precipitator  synthesis  is  to  find  two
precipitator configurations in tandem characterized with  a different (o^ and
A/v to provide a resultant efficiency of  r\,  or  a combined modified migration
velocity of (% at the conditions  of  flow  rate v,  and  particle dispersion k.
To state briefly, the problem of  synthesis becomes:

     Given:  Particle dispersion  k;  upstream precipitator characterized
             by (%, AI/V;  downstream precipitator characterized by o%,
             A2/v.

     Find:   The required  performance,  (% or ri,  has indeed been achieved.
Schematically, the problem can be represented conveniently in a block-diagram
form as shown in Figure 1  where size of each block indicates the relative
size of each precipitator  configuration with parameters of interest  contained
inside each block as illustrated.
                   A)Synthesis problem,  cj'k/o;k = ?
m/"\_

k
v
^

"A
k
v
Wk'2,A2
Wk,^
V^+Aa
                   B) Analysis problem, A' = ? to meet >?'
    in
» k n
1 v *
"'k,A'

V, A'
     Figure 1.  Block-diagram representation of on-stream pilot  testing.

     The problem of electrostatic precipitator analysis  of particular  interest
here is to scale up the pilot data of an on-stream full-scale  pilot  testing
directly to a full-scale design.   Briefly stated the analysis  problem  is:

     Given:  Downstream precipitator characterized by w£, A2/v.

     Find:   A'/v size of downstream precipitator required to  achieve
             a design efficiency of n",  using the downstream precipitator
             only.
Block diagram representation of the corresponding analysis problem is  also
illustrated in Figure 1.

                                     186

-------
                   EXPERIMENTAL AND ANALYTICAL PROCEDURES

     To  implement an  on-stream full-scale pilot  testing, a pilot  designed with
all aspects  of  a full-scale unit except a much reduced  length  in  its  flow
direction  is to be  installed directly downstream following an  existing  full-
scale conventional  configuration or installed to the  outlet  section of  the
existing unit by retrofit.

     Utilizing  correct analytical procedures, the pilot performance thus
obtained can be used  directly to size a full-scale unit.  Experimentally,
performances of the conventional unit alone and  together with  the pilot com-
bined will be used  to establish the baselines respectively.  Analytically,
normalization procedures will be followed to reduce performance of the  com-
bined configuration at an equal footing.  These  experimental and  analytical
procedures are  best illustrated via block-diagram illustrations as shown in
Table 1.

   TABLE 1.   PROCEDURES OF PRECIPITATOR PERFORMANCE SYNTHESIS  AND ANALYSIS
Step Block Diagram
1 0^>

»K,A,
Remark
^k, HIO Baseline wk)
V1 A., Conventional
A = A1
                              , = fA, A2 =
                                                     Basel! new
                                                               k,
                                                     New/improved
                                                     Normalization, A'.,
      IV
               v?
*XA,
k'
v?
^.(A'-I-A!)
                              VZ.A',
                                    Normalization,i
      VI
                                                     Normalization, A'
                                    Normalization, w'k/wk
                                    187

-------
     Beginning at Step I, Table 1, the pilot is deenergized.  A baseline  of
the existing full-scale conventional precipitator u^, AI/VJ, TIJ,  is  first
established experimentally at an operating condition of k and vj-  Similarly,
Step II establishes experimentally a baseline of the combined performance,
Wfc, A/V2, ^2, at an operating condition of k and v2.  Steps I and  II thus
complete the experimental procedures.

     Step III starts the normalization procedures.  Based on (% established
in Step I, it proceeds to find out the increased plate area to flow  rate
ratio, Af/V2, that is required to generate equal performance r\2 as obtained
at the combined baseline condition.  At Step IV, the actual normalization
begins by introducing an artificial new particle dispersion value k' immedi-
ately following the conventional precipitator.  The k' introduced in Step IV
is made to equal the k' contained in Step II.  By accomplishing these steps,
performance of the conventional precipitator has thus been separated into two
parts consisting of an upstream section and a downstream section, the down-
stream section is normalized in such a way that a comparison of performance
with the new improved electrostatic precipitator can be made directly.

     Similarly, assuming that the new or improved electrostatic precipitator
is to take the place of the existing full-scale conventional precipitator as
well, Step V starts the normalization at the inlet by reducing artificially
the plate area to flow rate ratio such that the overall performance  is equal
to that in Step II.  With conditions at the outlet section remaining unchanged,
Step VI concludes the analytical procedure with a direct comparison  of
performances, oo^/w^.

                         MATHEMATICAL REPRESENTATION
     Using modified Deutsch equation, precipitator performance is described
in terms of a modified migration velocity (%, and a k value varying from 0.4
to 0.8 to account for particle size dispersion, electrical energization, non-
uniformity in gas flow, and others representative of a typical operation of a
conventional precipitator.  To simplify the mathematical representation of an
on-stream pilot performance, k value here is assumed to be predominately
particle size dependent whereas other factors are of minor concern.  Typi-
cally, if the existing full-scale conventional precipitator is represented
with value k, then the pilot improved precipitator which is downstream of the
existing conventional -unit has to be represented with a different value k" to
describe the effect of reduced particle sizes.

     In Figure 1, let w^ be the modified migration velocity of the conventional
unit having plate area AI, and also for that particular application, particle
size dispersion is represented by value k.  If the overall collection effici-
ency with the on-stream pilot unit is r)2 at flow rate of V2,  then, from the
conservation of mass, and in accordance with modified Deutsch approach,


     D2 '= l-EXP{-(wkA1/v2)k} EXP{-(o)k'2A2/v2)k'}                           (1)
                                    138

-------
also,
        = l-EXP{-(aJkA/v2)k}                                                (2)
where ^"2 corresponding to value k' is the modified migration velocity of
the pilot unit having plate area A2, Wfc corresponding to value k is  the com-
bined modified migration velocity having a combined plate area of A  = A^ + A2-
Note that the full-scale improved precipitator, i.e., without the upstream
conventional unit,  would have a modified migration velocity w£ corresponding
to value k.  It is  invariably true that physically the modified migration
velocity 
-------
                                     A
     Comparison of Equations (5) and  (8) results  in,


     Wk'2/Wk2 = TIf (    - 1)                                              (9)
where quantities at the right-hand side of Equation  (9)  are  all  known.

     Similarly, Step V establishes the required size of  an improved  precipi-
tator for achieving equal efficiency as the combined unit.   Let  U)£ be  the
modified migration velocity of a full-scale improved precipitator, and  also
defining an enhancement factor £=0)^/0)^, the required size AVv2,  can be found
from,

     n2 = i-EXp{-(\A/v2)k>                                              do)


        = l-EXP{-(ttkA'/v2)k}

or,

     A'/v2 = (wk/ewk) (A/v2)                                              (11)

    With Equation (11) , performance of the new improved  precipitator can also
be expressed in the form,
Comparison of Equations (5) and (12) establishes,

     e = Wk/tOk


                 -»

     From Equations (9) and (13), it is also shown  that  the  enhancement  factor
e is invariant with respect to its positions in a setup  as long as  the inlet
conditions, e.g., flow rate v, and dispersion value k, remain the same.   This
conclusions is in agreement with the fundamental precipitation process for
which the precipitator performance is known to be characterized with flow rate
and dispersion value k.

                            ILLUSTRATIVE EXAMPLE
     On-stream full-scale pilot testing examined  in  this  paper is new,  and
thus it has not been actually carried out  in  the field.  However,  the recent

                                     190

-------
demonstration of pulse energization on a full-scale utility electrostatic
precipitator (3) provides an illustrating example of on-stream full-scale
pilot testing.   During that field demonstration, pulse energization was only
applied at the outlet half of the plate area of an existing conventional pre-
cipitator whereas the inlet, or the upstream, half of plate area operated as
a conventional precipitator.  Thus, f in this case was 0.5.  Use of Equation
(13) and assuming k = 0.5, the pertinent data contained in Table  I  of
Reference (3) are reduced accordingly and are tabulated in Table II

k = 0.5, i
Date
11/12/80
11/13/80
11/20/80
11/20/80
11/24/80
11/24/80
11/25/80
11/25/80
TABLE II.
5 = \/V ]
^k (m/s)
.1576
.1576
.3480
.3480
.2282
.2282
.2292
.2292
FIELD RESULTS (3)
F " W0)'
S (m/s)

.2484

.3905

.3027

.2999
V°> =
fv

1.5761

1.1221

1.3264

1.3086
.7092
e

2.15

1.24

1.65

1.62
m/s
1/F
4.50
4.50
2.04
2.04
3.11
3.11
3.09
3.09
               Modified migration velocities:

                   w (0),  dc,  conventional,  washdown baseline

                   w, ,     dc,  conventional,  operating
                    .*
                   w, ,     pulse energization,  operating

     To avoid  the  ambiguity in the enhancement factor e, a dc performance
factor F,  defined  as the ratio of modified migration velocity at the opera-
ting condition over the maximum possible modified migration velocity at the
unseasoned condition or the   washdown condition, is also introduced in Table
II.   Figure 2  is a plot of theevs.  1/F.  Surprisingly, experimental data thus
reduced and plotted fit almost exactly according to a straight line inter-
cepting the line e = 1.0 at (1/F) =  1.361 or F = 0.735.  From Figure 2, it
is seen that the on-stream full-scale pilot  testing was indeed an effective
means for  obtaining directly the full-scale  sizing data for new improved
precipitators.
                                     191

-------
 5.0



 4.0

1/F


 3.0




 2.0



 1.0
                                       iii
_  F =
   £ E
Mean Particle Migration Velocity
K=0.5
   Wk (0), dc, after washdown
         0.7092 m/s, March 1980
         or an unseasoned
         mean.
   0>k    dc, operating.
         pulse energized.
                                                       No Sodium
                                                       Sulfate
                                                       Conditioning
                                                       (Nov. 1980)
                                  Sodium Sulfate Conditioning
                                  to 0.7% Na-O, total
                                  (Nov. 1980r
              Right after Sodium Sulfate
              Conditioning to 1 % Na«0, total
              (Nov. 1980)         *
     1.361(F = .735)
                            1.5
                                                2.0.
                                                                    2.5
                        Figure 2.  Field  results (3).

                         DISCUSSIONS AND  CONCLUSIONS
     Although a great deal of research  effort has been devoted in the past on
the subject of electrostatic precipitation modeling, reliable scaling means
still do not exist for direct extrapolation of laboratory results to pilot-
plant performance, and pilot-plant performance to full-scale design.  It is
generally observed that a laboratory precipitator outperforms a pilot-plant
precipitator, and a pilot-plant precipitator outperforms a full-scale instal-
lation.  Direct scaling factors, however,  have not been found.  Thus, to avoid
pitfalls inherent in arbitrary and improper interpretation of slipstream
pilot-scale observations, empirical guidelines derived from operating experi-*
ences of existing full-scale units of similar application or fuel source are
provided, and are strictly observed for obtaining sizing information with
acceptable accuracy.  With diversified,  long-term experiences existing for
full-scale conventional wire-duct precipitators, this empirical approach for
sizing data has been successful for introducing the conventional precipitator
for new applications or fuel sources, but, definitely, it has failed to
realize new improved precipitator for full-scale applications.

     To promote advancement in modern electrostatic precipitation technology,
new ideas of improved precipitator need to be encouraged and introduced timely
and economically.  To accomplish this goal, this paper investigated the pro-
blems of an on-stream full-scale pilot  testing scheme in which a one-to-one
scale pilot with a short axial length in the flow direction is introduced or
retrofitted downstream of an existing full-scale conventional precipitator.
Because the size of this pilot is nearly a full-scale replica, pilot data of
the new improved precipitator if separable from the combined performance would*
                                      192

-------
thus unquestionably represent a full-scale design;  no prior experience on an
operating full-scale unit would thus  be needed.   Following the modified
Deutsch approach,  the method of analysis and synthesis of the combined on-
stream full-scale  pilot  testing has been indeed  established providing means
to separate the combined performances.   The combined field data taken from
recent development on pulse energization comprising a conventional dc ener-
gization upstream  and a  pulse energization downstream were used to demonstrate
the utility of  this analytical method.

     To assure  practical and economical on-stream full-scale pilot testing,
the importance  of  the testing site selection must be emphasized.   An ideal
site for on-stream full-scale pilot testing would thus satisfy all or at
least one of the conditions listed below:

     •    An operating full-scale conventional precipitator with
          an adequate flow capacity.

     •    An operating full-scale conventional precipitator requiring
          upgrading to meet emission  standards.

The former condition prevents an unnecessarily large,  and consequently costly,
pilot fabrication  and erection;  whereas the latter  justifies a full-scale
installation of an improved precipitator.

               The work  described in  this  paper  was not  funded by
               the U.S.  Environmental Protection Agency  and there-
               for the contents  do not  necessarily  reflect the
               views of  the Agency and  no  official  endorsement
               should be inferred.

                                 REFERENCES
1.    Allander, C. and Matts,  S.  Einwirkung  der Korngrossenverteilung  auf
     den Abscheidungsgrad von Elelctrofiltern.  Staub,  52:   738,  1957.

2.    Matts, S. and Ohnfeldt,  P.O.  Efficient Gas  Cleaning with  SF Electro-
     static Precipitators.  Flakt Review,  1963-64.

3.    Piulle, W., Marchant, G.H. and Gooch, J.P.   Evaluation  of  Performance
     Enhancement Obtained with Pulse Energization Systems on a  Hot-side
     Electrostatic Precipitator.  Paper presented at  the Third  Symposium
     on the Transfer and Utilization of Particulate Control  Technology,
     U.S. Environmental Protection Agency, Orlando, Florida,  March 9-12,
     1981.
                                    193

-------
                 COMPUTER MODEL USE FOR PRECIPITATOR SIZING

               by G. W.  Driggers
                  A. A.  Arstikaitis
                  L. A.  Hawkins
                  Environmental Systems Division
                  Combustion Engineering, Inc.
                  Birmingham, Alabama 35243
                                  ABSTRACT

     The design of utility industry precipitators is typically based on a
historical data base of performance and operating characteristics gathered
from full scale units.   Development over the last ten years of first principle
analytical models for performance prediction offer new tools for the
precipitator designer.   The EPA/SORI model has been selected by the Combustion
Engineering Environmental Systems Division for refinement to the C-E
precipitator design.  The current historical data sizing procedure and
performance prediction approach is described followed by a discussion of the
EPA/SORI model and refinements required to make it fit the C-E design.

                                INTRODUCTION

     The selection of specific collection area (SCA) for utility precipitators
is based throughout the industry on extrapolation from existing data.^'^^
Information collected on operating full scale units, including net penetra-
tion, electrical characteristics, type of coal fired, type of boiler, flyash
chemistry, and resistivity, are typically used to draw correlations to new
utility requirements.  It is normal in the analytical design of a precipitator
to use the traditional or modified Deutsch-Anderson equation relating
precipitator size and collection efficiency to a known or extrapolated
precipitation parameter ("migration velocity") to predict SCA needed to meet
a specified level of performance.^'  The Environmental Systems Division of
Combustion Engineering, Inc. has been able to use this approach with success
when designing new units.  However, changes in the industry—requirements for
very high efficiency, new sources of coal, power consumption guarantees, and
stricter opacity limits—have made it necessary to look at more sophisticated
analytical techniques for predicting operating characteristics.  Expanding
the data base to cover fully all these variables through pilot, prototype,
and full scale testing, would be prohibitively expensive.
                                     194

-------
     In the last ten years,  under Environmental Protection Agency sponsorship,
Southern Research Institute  (SORI) has developed a mathematical model of the
precipitation process and computerized it.(3)  Use of the current model to
synthesize a particular precipitator is limited, however, by certain implicit
assumptions that were necessary to make the model a general process simulation.
To make the model fit a specific precipitator manufacturer's design requires
that modifications be made and certain input variables specified.  C-E/ESD
has identified these required changes and developed predicted values for the
input variables based on a variety of tests on C-E precipitators.  The
refined model is being used  to assess predicted efficiency, fractional
penetration (which is, in turn, input to an opacity prediction program), and
power consumption on new and proposed units.

     The following sections  discuss the existing data base and sizing
procedure, the basic EPA/SORI computer model, required changes to the model,
how these were implemented,  and an example of results that have been obtained
analyzing operating units.

                       DATA  BASE AND SIZING PROCEDURES

     The C-E data base consists of design and performance information on over
twenty-five installations with the number growing yearly as new units are
commissioned and tested.  These test data have been used to develop sizing
guidelines based on use of the calculated precipitation parameter WK (commonly
called the "modified migration velocity").   These Wj^ values are cross
referenced to coal source and chemistry, particularly sulfur, moisture and
sodium content.   Other parameters, such as inlet dust loading, operating
temperature, ash resistivity, gas analysis and particle size distribution,
are also correlated to W^.  Boiler operating parameters, geometry of the
precipitator and a historical chronology are, of course, also included in the
computerized data base management program.

     To evaluate sizing on a new application the first step is to select a
generic coal type and/or chemistry in the data base that best approximates
the new specification.  A value for WK is drawn specifically from the data.
Subsequently,  if an inlet total dust loading is not specified, one is cal-
culated based on given ash content and Btu values for the coal, the plant
size and an assumed carryover value.  Pitot correction factors from the
published ASME literature and C-E data base are applied to the stoichiometric
calculations to  determine the total flow anticipated for the new precipitator.
From these values a specific inlet dust (or "grain") loading can be determined
and a total efficiency requirement calculated for a specified outlet concen-
tration.   This is compared to the efficiency requirement (if one is specified)
and the more stringent value is used in subsequent calculations.

     The parameter WK is then modified by correction factors based on the
comparison of  the specific values for sulfur, sodium, temperature, moisture,
and inlet  dust loading in the data base vs.  the specification.  The final
value for  Wj^ is  used  in the  equation
                                     195

-------
               SCA
(where r\ is the required efficiency and x is the exponent of the Matts-
Ohnfeldt modification to the Deutsch-Anderson equation) to determine the
required collection efficiency.  With the specific collection area known, the
remaining geometrical parameters can be calculated.

     The approach being used by C-E to improve on this process is twofold.
First, the correction factors used to modify WR are being defined based on
theoretical modeling as well as empirical data to allow better interpolation
and extrapolation.  Second, the theoretical model is being coupled with
empirical data to derive a more rigorous predictive technique.  The basis for
these efforts is the EPA/SORI model.

                           EPA/SORI COMPUTER MODEL

     Over the past decade, Southern Research Institute (SORI) has been
developing a computer-oriented mathematical model of the electrostatic
precipitation process. (3)  The model is necessarily general so that it can be
applied to most wire-plate geometries.  The model can be a powerful tool for
the ESP manufacturer, particularly when modified to apply to a specific
precipitator design.

     Required inputs to the SORI ESP model include important dimensions of
the precipitator internals, such as wire-to-plate spacing, wire-to-wire
spacing, and collecting surface per electrical section; operating voltage and
current in each electrical section; and gas stream parameters, such as inlet
dust loading, particle size distribution, gas volume flow rate, gas viscosity,
and gas temperature.  The model gives an ideal prediction of collection
efficiency and outlet particle size distribution.  The model also predicts
ESP performance corrected for nonideal conditions, such as non-uniform gas
flow distribution, gas sneakage, and reentrainment with and without rapping.

     The model is built around the traditional Deutsch-Anderson equation,
               n = Le

where n = collection efficiency,
      W = "traditional" migration velocity near collecting surface,
    SCA = specific collection area

The derivation of this equation assumes (among other things) that the particle
migration velocity (W) is constant near the collecting surface since the
migration velocity is calculated from
               w =
                    6 n ay
                                     196

-------
where q = particle charge,
     Ep = electric field at the collecting plate,
      C = Cunningham correction factor,
      a = particle radius,
and   V = gas viscosity.

The assumption implies that the particles are all  of the same size and that
electrical conditions are constant.   To  satisfy the first condition, the
particle size distribution is divided into discrete size bands, each having
a representative particle size and all calculations are performed for each
particle size band.  To satisfy the second condition, the precipitator is
divided into length increments that are  small enough that electrical conditions
remain essentially constant over the increment.

     In order to calculate the particle  migration  velocities, the electric
field at the collecting electrode and the charge on each particle size must
be determined in each increment.   Particle charge  is calculated from the
unipolar, ionic-charging theory formulated by Smith and McDonald.(^'  The
electric field distribution in the interelectrode  space is determined from
classical equations,  assuming a round wire discharge electrode and flat plate
collecting electrode geometry.  The results are used to calculate the average
electric field at the collecting electrode in a particular length increment.
However, few precipitators  have round wire, flat plate geometries and for
enhanced accuracy the model must be modified to account for actual voltage
and current conditions.  This can be accomplished  by determining the corona
current distribution at the collecting plate as a  function of the discharge
and collecting electrode geometries.  This distribution is used in the computer
model as a boundary condition for calculation of the interelectrode electric
field and the associated gradient near the plate.   A theoretical correction
is also required for space charge density near the DE if a non-round wire
geometry is used.

     Also, it is necessary to collect data sufficient to predict particle
size distribution,  voltage-current characteristics and non-ideal parameters
(sneakage per field,  reentrainment and gas flow distribution).  Through
various measurements  on utility precipitators by C-E and others, sufficient
data have been accumulated  to make the computer model highly accurate with
regard to performance correlation.

                      COMPUTER MODEL CHANGES AND INPUT

     The EPA/SORI precipitator performance prediction program has been
modified to model the specific geometry  and electrical characteristics of the
G-E precipitator.  Non-ideal factors have been determined based on the data
base discussed earlier after modeling of the electrical characteristics.
Since the majority of the utility data base was obtained on designs with
identical mechanical  characteristics, (plate spacing, sneakage baffles,
aspect ratio,  rapping system,  etc.), the model can be calibrated very
accurately.   The key  element in assuring that the  values determined for non-
ideal parameters were not just "Kluge factors", was establishing real
electrical operating  characteristics and particle  size distributions.  The
former came from specific measurements,  the most involved of which was corona

                                     197

-------
current distribution on the collecting plate.   The latter was based on
measurements or the existing data base.

CORONA CURRENT DISTRIBUTION DETERMINATION

     Combustion Engineering uses the collecting electrode (CE) configuration
shown in Figure 1.  The contours provide rigidity  and  help to reduce reentrain-
ment.  Two discharge electrodes  (DEs) are  currently in use,  a barbed electrode
and a "star" electrode.  The barbed electrode  is used  extensively in inlet
fields when an ash with high resistivity is being  collected.   The star
electrode is used to achieve higher interelectrode voltage for a given
corona current.

     The points on the barbed electrode are corona current sources and the
collecting surface is irregular, thus current  density  distribution different
from that produced by the round wire on a  flat plate is expected.   This
distribution was measured for comparison to the round  wire distribution.
These data were then used to develop a correction  to the predicted electric
field at the plate as discussed earlier.

     The current density distribution at the collecting electrode surface for
the C-E wire-plate geometry was measured using the apparatus  shown schematical-
ly in Figure 2.  An array of metal squares arranged in a  horizontal
                  DEWIRE
                 LOCATION
                         TYPICAL GAS PATH SHOWING COLLECTING
                         ELECTRODE CROSS SECTION AND DE WIRE
                         LOCATION.
                                               CROSS SECTION
                       BARB ELECTRODE    STAR ELECTRODE
        Figure 1.  Discharge and collecting electrode  configurations

                                     198

-------
                                                       PICOAMMETER
                SENSOR PANEL
          CE-
                   DE-
                         -50KVDC
                        SENSOR
                                   SCHEMATIC
                                    CE
                                                       PICOAMMETER
                                TYPICAL FOR EACH SENSOR
              Figure 2.   Current density measurement apparatus.

row across the CE panel  and a horizontal strip served as current sensing
elements.   A sheet of plastic formed to the shape of the CE panel served as
a movable  mount for the  sensors and as a dielectric to isolate the sensors
from the grounded CE panel.  Each sensor was connected by shielded cable to
a switch panel located outside the precipitator.  The panel was connected by
shielded cable to a picoammeter which was grounded directly to the CE panel.

     The panel was attached to a CE plate in a utility precipitator so that
the sensors were initially in the same horizontal plane as a barb on the DE
wire.   A voltage-current curve was obtained and the current at each sensor
was recorded at each point on the curve.  The sensor panel was subsequently
moved  down five times in one-quarter inch increments and a V-I curve was
taken  at each position.   The last increment placed the sensors in a horizontal
                                    199

-------
plane halfway between two DE points.
are shown in Figure 3.
Specific locations for measurements
             PLATE CENTERLINE
                              DISCHARGE ELECTRODE LOCATION
                                                     DISCHARGE ELECTRODE
                                                     CONFIGURATION
             Figure  3.   Location of  Current density measurements.

      Based  on these measurements, it  was observed that the current density
 is  about  twice the  average in the region around the barb on the wire.  Moving
 away from the wire  in  the horizontal  plane, the current falls off rather
 slowly at first and then drops quickly near the edge of the plate in a
 fashion similar to  the round wire/flat plate geometry.  In the vertical
 direction,  the current is about constant in the first quarter distance
 between the points  and then falls off to a minimum halfway between the barbs.
 It  is the latter contour that differs most significantly from the assumptions
 of  the basic model. The model uses three different techniques to calculate
 the electric field  at  the collecting  plate:
      (1)  an estimation based on the average electric field across the
          interelectrode space;
      (2)  a  rigorous calculation of  the electric field distribution in the
          interelectrode space, used when operating voltage and current
          are known; and
      (3)  a  calculation procedure similar to technique (2) used when operating
          voltage and current are unknown.

 The third technique is the only one sensitive to the geometry of the DE wire.
 It  was used to find the equivalent  round wire radius necessary to produce
 the measured current density in each horizontal section at the applied
 voltage.  An overall equivalent radius was then found to give the same field
 at  the plate as the average from the  horizontal sections.  After establishing
 an  equivalent round wire medium, correction factors were developed to modify
 the predictions of  electric field at  the plate made by techniques (1) and
 (2).

      The  computer model was subsequently modified to take these corrections

                                      200

-------
into account.   The modified computer model was then used to quantify  non-
ideal factors  on a consistent basis using the C-E data base.

VOLTAGE-CURRENT MODEL

     Voltage-current (V-I) characteristics of a specific wire and plate
geometry will  vary as the dust loading varies and as the resistivity  of  the
dust layer changes.  To analyze an existing precipitator, the measured
values from transformer-rectifier (T-R) set instruments can be used.  For
predicting performance of a new unit, some estimating procedure must  be
used.  For a new unit, C-E has developed and is refining an empirical model
based on data  collected from operating units with various inlet dust  loading
and ash resistivity.  Ash resistivity is being characterized by laboratory
measurements and prediction using the computer program of Bickelhaupt.(5)
Correlations between ash chemistry using ASTM coal ashing techniques  and
hopper ash samples are being developed.  Thus, the very extensive coal ash
data base at C-E can be used to predict flyash resistivity for new applica-
tions.

     V-I curves from several paralled electrical cells at a facility  are
plotted together (as illustrated in Figure 4) and a best fit is established.
                                  POINTS ARE ACTUAL DATA
                                  LINES ARE BEST FIT
                 AVERAGE
                 PLATE
                 CURRENT
                 DENSITY
                            . FOURTH FIELD

                              SECOND FIELD

                                 • FIRST FIELD
                                SECONDARY VOLTAGE
     Figure 4.
Voltage-current model for a C-E precipitator  with barbed
electrodes operating on a IQll ohm-cm  lignite ash and 2.5
grain/ACF inlet loading.
These data are then correlated to dust loading and resistivity as a  function
of electrical  cell location (inlet, second, third, etc.).  Thus, using a
specified  flyash chemistry and inlet dust loading, a prediction of the
operating  electrical characteristics of a particular section can be  obtained
by working from the first field to subsequent fields.  Special attention
                                    201

-------
is given to recording flashover (sparking) points and conditions, particularly
repetition rate.  This has only been significant on inlet fields, since second
fields typically experience only occasional sparking and all subsequent cells
operate at current limit for the T-R set in a well constructed unit.
Tolerances are tightly controlled on new units to assure this condition.

     For a new application, an operating point resistivity is obtained from
the data base or predicted using the program of Bickelhaupt.  This is used
in combination with the predicted inlet dust loading and anticipated electrical
section size to project the probable voltage-current characteristics and
operating point.

PARTICLE SIZE DISTRIBUTION MODEL

     The statistical model developed by SORI under EPRI sponsorship is used
to estimate the inlet particle size distribution for the new application
where a coal type is specified.  Nominal and worst case distributions are
analyzed on Western sub-bituminous coals to assess possible impacts to both
total collection efficiency and opacity prediction.  This portion of the
analysis is gaining significance as fractional efficiency guarantees are
being specified by some customers.

                         COMPUTER MODEL APPLICATION

     The modified computer model is currently being used in conjunction
with the data base design procedure both as a cross check and a predictor of
performance parameters not otherwise obtainable.  In particular, fractional
penetrations are being computed and used as input to a numerical opacity
prediction program.  The program is also being used in conjunction with the
flyash resistivity predictor and V-I model for determination of power
consumption expectations, precipitation parameter correction factor computation,
and energy management system control algorithm development.

     Total collection efficiency predictions and measurements made during
past performance tests agree to within 0.05 percent for three cases examined
in detail.  Performance projections for new units being brought on line in
late 1982 and through 1983 will be made to enhance confidence in the models
and refine the predictive techniques.  Continual development of various
aspects of the C-E specific model are anticipated leading ultimately to a
design approach based on process understanding; integrating the electrodynamic,
gas dynamic, chemical and mechanical variables to obtain the most cost-
effective precipitator.

     The work described in this paper was not funded by the U.S. Environ-
mental Protection Agency and therefore the contents do not necessarily
reflect the views of the Agency and no official endorsement should be
inferred.

                                 REFERENCES

1.  White, Harry J. Review of the state of the technology.  In;  Proceedings
    of the International Conference on Electrostatic Precipitation, October

                                     202

-------
    1981; published by APCA,  Pittsburgh,  PA.

2.   Nichols,  Grady B. Electrostatic precipitator  design.   In;   Proceedings
    of  the  International Conference on  Electrostatic  Precipitation,  October
    1981; published by APCA,  Pittsburgh,  PA.

3.   McDonald, Jack R.  A mathematical model of  electrostatic precipitation.
    EPA-600/7-78-lllb, U.S. EPA, Washington,  B.C.  1978.

4.   Smith,  W. B. and McDonald,  J. R.  Development  of  a  theory  for the
    charging  of particles by  unipolar ions.   J. of Aerosol Science,  7:151-166,
    1976.

5.   Bickelhaupt, Roy E., A  technique for  predicting flyash resistivity,
    EPA-600/7-79-204, U.S.  EPA, Washington, DC, 1979.
                                    203

-------
                IMPROVEMENTS IN THE EPA/SRI ESP PERFORMANCE MODEL

                                       by

                                 M. G. Faulkner
                           Southern Research Institute
                            Birmingham, Alabama 35255

                          R. B. Mosley; J. R. McDonald
                           Crestmont Associates,  Inc.
                          Central City, Kentucky  42330

                                       and

                                  L. E. Sparks
                  Industrial Environmental Research Laboratory
                      U.S. Environmental Protection Agency
                  Research Triangle Park, North Carolina 27711
                                    ABSTRACT

     Revision 3 of the ESP performance model developed  for  the  EPA  at  Southern
Research Institute has been completed.  This version  features a reduction in
required computer time of about a factor of 10 over revision 1  for  the rigorous
calculation of collection efficiency.  In addition, several new procedures have
been added to the model.  One of these allows the calculation of plume opacity.
Another calculates the effects of rapping reentrainment  through a dynamic
process which examines the results of each rap as it  occurs.  These changes and
others will be described.
This paper has been reviewed in accordance with  the U.S.   Environmental
Protection Agency's peer and administrative  review policies  and approved for
presentation and publication.
                                     204

-------
     Revision 3 of the electrostatic  precipitator (ESP) performance model
developed by Southern Research Institute  (SRI)  for the Environmental Protection
Agency (EPA) has been completed.  Since  the  model was released in 1975 (1),  it
has been widely used to study and troubleshoot  existing precipitators and to
validate proposed new precipitators.  The  revisions  (2, 3,  4) to the original
model have increased its usefulness and  convenience  of operation.  Revision 3
further increases the model's utility by  offering a  significant reduction in
required computation time plus several new features  which are described in the
following paragraphs.

     The ESP model predicts precipitator  performance by first performing the
collection efficiency calculation under  ideal conditions and then correcting
the results for non-ideal conditions and  unmodeled effects.   For calculation
purposes, the precipitator is divided into sections, permitting different
electrical conditions in each section, and then further divided into a maximum
of 45 incremental lengths on the order of  the ESP's  wire-to-wire spacing.  The
efficiency calculation is performed sequentially  on  each incremental length to
determine the total collection efficiency.   To  compensate for the different
charging and collection rates of large and small  particles,  the input dust load
is divided into a maximum of 20 size bands,  each  of  which is handled separately
in the efficiency calculation.

     The ESP model uses the Deutsch equation to calculate the collection
efficiency for each particle size band in  each  incremental  length.   The Deutsch
equation is given by

                       n = 1 - exp (-Aw/Q)                            (1)

where

     n = the fractional collection efficiency,
     A = the collection plate area,
     Q = the gas volume flow rate,  and
     w = the migration velocity of the charged  particles.

The migration velocity is calculated from
where

     q = the charge on the particle,
    E  = the electric field at the plate,
     C = the Cunningham slip correction  factor,
     a = the radius of the particles, and
     u = the viscosity of the gas.

Thus the calculation of collection efficiency requires  the  calculations of the
charge on each  size particle in each incremental  length and of the electric
field at the plate.  This can be done simply by estimation  formulas or more
rigorously by using a more complete charging theory  developed by Smith and
McDonald (5) and  a  detailed analysis of  the field  at  all points in the

                                     205

-------
precipitator .  The more rigorous  process  yields better accuracy over  a  wide
range of applications and  is  the  more frequently used routine.  However,  it
involves many lengthy calculations  and requires much more computer  time than
the estimation process.  The  changes  resulting in reduced computer  time for
Revision 3 affect the electric  field  and  particle charge calculations of  the
rigorous calculation procedure.

     Table 1 contains a summary of  ESP model features including the changes
due to the various revisions.  These  features are discussed below in  the  order
in which they are presented in Table  1.

     1.  The estimation procedures  for calculating particle charges and the
electric field at the plate remain  unchanged.

     2.  The rigorous calculation of  the  electric field at the plate has  been
changed in Revision 3.  Although  the  general procedure used to solve  the  field
equations remains the same, the convergence of the iterative technique  used to
achieve a solution is accelerated using the dominant eigenvalue method.   This
method has previously been applied  to various chemical processes and was
applied to the electric field calculation by Felder- and Arce-Medina.  (6)

     The procedure used in the rigorous calculation of the field at the plate
consists of simultaneously solving  two field equations relating the electric
potential and the space charge density at every point inside the precipitator.
These are:
                                -  2- ,and                              (3)
                   Ax     Ay       e0
where

     p = the space charge  density,
     V = the potential,
     x = the coordinate  from wire  to plate,
     y = the coordinate  along the  gas flow, and
    e  = the permittivity  of free  space.
     0

In the model, the calculation is performed at every point on the  grid  shown in
Figure 1.  The  initial estimate  of  the potential at each grid point, Vj;, is
made using Cooperman's solution  to  the field equations. (7)  Following this,
the space charge density at  each grid point PJ;, is calculated  from V^;.
Vij is then recalculated from Pij.   The process of alternately  calculating
p^j and V-jj continues until  the  change in Y£J between  iterations  is less
than a preset value  at every grid  point.   At this point the current density
at the plate is calculated and compared to the measured value.  If the values
differ by more  than  1 percent,  the  space  charge at the corona is  adjusted and
the calculation of p^j and V^j  begins again.  Otherwise the potential  solu-
tion is considered to have converged, and the value of the electric field
at the plate is obtained.
                                     206

-------
            TABLE .1.  ESP MODEL  REVISION SUMMARY
Process                                                  Revision




Estimation procedure                                      1>2,3




Electric field calculation




   rigorous calculation                                   1,2




   same with dominant eigenvalues  (faster)                    3




   analytic approximation                                   2




Particle charging calculation




   Simpson's rule integration                             1




   Gaussian integration  (faster)                            2,3




Correction for non-ideal effects




   sneakage, velocity distribution                        1,2,3




   empirical rapping calculation                          1»2,3




   dynamic rapping calculation                                3




Opacity calculation                                           3




Option of metric input data                                   3




Option to reduce printed output                               3




Internal data set                                             3




VI curve generation                                       1»2,3




   option to stop after printing VI  curve                     3




Routine to check input data                                   3
                                207

-------
                                           = Vo ON WIRES
                                                  V = 0 ON PLATES
A



4,







x
X
D
S
21
31


_ _







/

>K
22
32
42
*








f

} •
23
33
43

-^









* !
24



-yi









5 / \ \
,15 / » fc >
i___ y AXIS
I '
4.--
i
+ ^_
1



\
\
\

\
\
fi.. 	 »
B
i
s
'



3,






C
                                                        AREA OF
                                                        INTEGRATION
                  Sy = ONE  HALF WIRE TO WIRE  SPACING

                  Sx= WIRE TO  PLATE SPACING
             Qxay  = INCREMENT SIZES FOR  INTEGRATION

                  V0 = APPLIED  VOLTAGE
                  Ex = COMPONENT OF ELECTRIC  FIELD PERPENDICULAR
                      TO PLATE

                  Ey = LONGITUDINAL COMPONENT OF ELECTRIC  FIELD
                                                                  4102-92
Figure I.  Nomenclature used in the numerical analysis of the  electrical
          conditions in wire-opiate precipitators.
                                    208

-------
     In Revision 3, the convergence rate has  been accelerated by the dominant
eigenvalue method.  Using the new procedure,  two criteria are applied to the
calculated potential after each iteration.  These are (1) at least five
iterations must have taken place since  the  last  acceleration step, and  (2)  the
change in the potential between the last two  iterations must be less than a
preset value at every grid point.  These criteria ensure that each acceleration
step has sufficient time to produce a stable  solution before a new acceleration
step is applied.  If both of these criteria are  met,  the potential is adjusted
at every point by a factor derived from the potential changes due to the last
two iterations.  The potential adjustment has  the form:
where

     V.-     = the potential at point  i,j calculated  in the
       J       last iteration,

     V>n   '  = the potential calculated' in  the  previous
       J       iteration,

     Vi7     = £^e adjusted potential  for use in  the  next
       J       iteration,

          a  = a damping factor, 0 < a  < 1, used  to prevent
              instability, and

          X  = the dominant eigenvalue.

X is given by

                           (n)        (n-1)
                   X = II AV    11/11 AV      II ,                        (6)

where
                     AV(n) II = (-.  [v    - V0]2).             (7)
Following the adjustment, the mutual calculation  of  p^;  and  Vj; is
resumed until either the potential converges  or the  conditions for another
acceleration step are met.  Since the convergence criteria are not altered by
this process, the accuracy of the calculation remains  high.  The effect of a
potential adjustment is greater than the effects  of  many iterations, as shown in
Figure 2.  (8)  Note that after the fourth acceleration step in Figure 2, the
error increases for one iteration, after which it drops  sharply and stabilizes
before the next acceleration step.  The reduction in the number of iterations
required for convergence is clearly displayed in  Figure  2.  In the ESP model,
the inclusion of the dominant eigenvalue method reduced  the number of V-p
iterations required for the complete efficiency calculation from 16,000 to 1,400
in a study of a power plant.  This is a 91 percent reduction in the number of
iterations.
                                     209

-------
           CC
           O

           CC
           O
           CC
           CC
               ID'3
               10-4
10-5
               10'6
                           20        40
                         NUMBER OF ITERATIONS
                                 60
                    D WITHOUT DOMINANT EIGENVALUE
                    O WITH DOMINANT EIGENVALUE
                    f APPLICATION OF ACCELERATION STEP
                                                4102-187
Figure 2.  Effect of dominant  eigenvalue convergence acceleration.
                                210

-------
     Table 1  shows the inclusion of an analytic  approximation to the electric
field in Revision 2.  This approximation offered almost  the same calculation
accuracy as the rigorous calculation while using much  less  computer time.  With
the incorporation of the dominant eigenvalue  procedure in Revision 3, the
rigorous field calculation is now as fast as  the approximation.   Consequently,
the approximation procedure was not included  in  Revision 3.


     3.   The  speed of the particle charging routine  is also reduced.  In order
to calculate  the charge on each size particle  in each  incremental length using
the more exact ionic charging theory, many integrations  are required.  The
change to the Gaussian integration scheme in  Revision  2  greatly  reduced the
integration time.

     The net  effect of the changes to the electric  field calculation and the
particle charging routine is a reduction in computer  time of about 90 percent
with no  loss  in accuracy.

     4.   As previously mentioned, after the ideal collection efficiency is
calculated, the model corrects the efficiency  for various non-ideal effects.
Corrections for sneakage, non-rapping reentrainment,  and non-uniform gas
velocity are  applied by adjusting the migration  velocity for each size particle
by a factor based on theoretical and experimental studies of a pilot scale
precipitator. (1)  This method of correction  has not been changed in Revision
3.

     Also unchanged is the empirical calculation of  losses  due to rapping
reentrainment.  This efficiency correction is  based  on the  amount of mass
collected in  the last section of the ESP.  A  fraction  of the collected mass is
fit to a log-normal size distribution and added  by  size  bands to the ESP
penetration dust predicted by the ideal efficiency  calculation.   The collection
efficiency is then recomputed based on the combined  effluent. The log-normal
size distribution used to describe the reentrained  particles has default
parameters of a 6 um mass median diameter with a standard deviation of 2.5.
Other size distribution data may be used in addition  to  these values if desired.
However, the  total mass of the reentrained particles  is  fixed by a relationship
based on data derived from a study of rapping  reentrainment in six power plants.
(9)  The subroutine which performs these operations  requires a negligible amount
of computer time relative to the ideal calculation  and occurs automatically
unless the new dynamic rapping routine is selected.

     The new  dynamic rapping routine utilizes  a  different process to calculate
the effects of rapping reentrainment.  This routine  keeps track  of dust layer
growth at every point in the ESP as a function of time.   At a user specified
time of  rap,  the collected dust is removed from  the  rapped  precipitator
increments.   A specified fraction of the removed dust  is fit to  a log-normal
size distribution and stored for the calculation of  reentrainment.  The entire
collection efficiency calculation is then repeated  with  the difference that the
reentrained dust particles are reintroduced into the  gas flow during the
calculation.   The  mechanism for this is that  the number  of  particles in each
size band per nr  of gas which are to be reentrained  due  to  the rap of increment
i  is added to number of particles/size band/m3 already present in the gas flow

                                     211

-------
immediately before the calculations  for increment i+1 begin.    Thus  the
efficiency calculation in  increment  i+1 is performed on an  increased amount  of
dust due to the reentrainment  from the preceding increment.   Inherent  in this
mechanism is the assumption  that  the reentrained particles  instantaneously
acquire the charge found on  particles of the same size at that  point in the  gas
flow.

     The calculation of rapping  reentrainment effects using  the  dynamic rapping
routine is a very  lengthy  procedure  due to the fact that the  ideal collection
efficiency calculation must  be performed at least twice for  each rap.   The first
time is to find the dust layer thickness at the time of the  rap,  and the second
is to calculate the effects  of the reentrainment.  Depending  on  the  complexity
with which the reentrained dust  is specified, up to seven additional efficiency
calculations may be required.  Therefore it is recommended  that  the  estimation
procedure be used  to reduce  computer time.  The dynamic rapping  calculation
process is very flexible in  that  the user may vary the times  of  the  raps, the
rapping sequence,  the  fraction of the collected dust that gets  reentrained,  the
size distribution  and  the  density of the reentrained dust,  and  the duration  of
the rapping puff.  This flexibility  should be of use when comparing  different
rapping schemes.   As of this  time, no experimentally supported  data  set,  such
as that used in the empirical  rapping correction, exists for  the dynamic rapping
calculation.  A study  should  be  performed to determine whether  such  a  data set
can or should be developed.   Currently, the data for dynamic  rapping must be
matched to each application.   Figure 3 shows the results of  using dynamic
rapping on power plant data.

     5.  A plume opacity calculation procedure has been added to Revision 3  of
the model.  This routine calculates  differential and total  extinction
efficiencies and the total opacity of the ESP outlet plume  calculated  by the
model.  These are  related  by:

          Opacity  = 1  - exp  (-EL) = 1 - exp (-NAQextD,              (8)

where

     E = the extinction coefficient,
     L = the pathlength of the  light beam,
     N = the particle  concentration,
     A = the projected area  of the particle, and
     Q = the extinction efficiency.

Figure 4 shows the extinction efficiency as a function of the particle size
parameter, x = irD/X, for four complex indices of refraction.  D and  X  are the
particle diameter  and  the  wavelength of the light beam, respectively.   The
calculation, which is  based  on Mie theory, (10) is performed at  10 wavelengths
weighted according to  the  photopic response of the human eye.   EPA-approved
opacity monitors must  simulate this color response, which is maximum at X =  0.55
Urn and has a width at  half maximum of 0.1 ym.  The required input data are
stack diameter and complex index of refraction.  A single wavelength-dependent
index of refraction or up  to 10  values of wavelength-independent indices of
                                      212

-------
O


<
OC

UJ

UJ
Q.
UJ
o
oc
UJ
0.
           A

           D
                0*
                    '.6
 10°
   10-1
10-2
     1Q-1
                     • EXPERIMENTAL

                     D DYNAMIC RAPPING

                     A EXAMPLE 3, NORAP + RAP, ag = 0, S = 0
                       1
                                       1
                                                         0.0
                                                         90.0
                                                         99.0
                                                               >
                                                               u

                                                               UJ
                                                               O
                                                               il
                                                               u
                                                               cc
                                                         99.9
                                                         99.99
                      10°
                                    10
                     PARTICLE DIAMETER, /
   102


4102-126
    Figure  3.   Comparison of  standard and dynamic  rapping
                reentrainment  calculations.
                             213

-------
25
U U
PC
cc u-
                       PARTICLE DIAMETER Own) FOR WAVELENGTH 0.55 urn
                  0.5         1.0         1.5         2.0         2.5
                                      3.0
                              T
                    (b) n
     I           I          I

(a) n =1.33 (H2O)
  !d) n= 1.5-0.1!
       (c) n = 1.96-0.661 (carbon)
                                              SCATTERING COMPONENT OF (c)
                                             • *^» O^K M^ ^fm ^^^ «i^»	
                                                          ~ ^^^ ^" ^^ ^""» ^"» ^"^«
                                             ABSORPTION COMPONENT OF (c)
                                                                 I
                                             10                 15
                               PARTICLE SIZE PARAMETER x = 7rD/X
                                             20
                                          41TI-4E
  Figure 4.  Extinction efficiency as a  function of  particle  size parameter.
                                         214

-------
refraction may be entered.   Alternatively,  the  refractive index may be omitted,
in which case the calculation is performed  using the default values of 1.5  and
1.5-0.1 i.

     6.  To facilitate the use of the model  in other countries, the data entry
routine has been modified to receive data  in metric  units if desired.  However,
unless all-metric data is specified, the data must  be in the mixed metric/
English units used in previous versions of  the model.

     7.  The option of reducing the amount  of printed output data is provided.
This will be of use in cases where only data summaries are of interest.

     8.  A built-in data set is provided in  Revision 3.   This data set contains
nearly complete data based on a "typical"  power  plant.  The only input data
required are the resistivity of the collected particle and the collection plate
spacing.  From this, the model calculates  a  set  of  V-I curves and operating
points for an ESP based on a study of 17 cold-side  utility fly ash precipi-
tators. (11)  The length of the ESP is varied so that efficiency calculations
may be performed for SCA values of 40 to 158 ra2/m3/sec (200-800 ft2/1000 acfm).
This routine will be of use for verifying  the operation of the model and for
establishing a starting point for a more detailed examination of ESP design.

     9.  The model may be used as a V-I curve generator.  The ability to
calculate V-I curves has been in the model  since Revision 1 as part of a rou-
tine which calculates the operating conditions and  electric fields in the
precipitator if the actual operating conditions  are  not  available.  This is a
very lengthy calculation which may be more  frequently used in Revision 3 due
to the significantly reduced computer time  resulting from the dominant
eigenvalue procedure.  In addition, there  is now a  switch which causes the
program to terminate after the V-I curve is  printed, in the event that only V-I
data are required.  Alternatively, if the ESP for which  V-I curves are desired
is similar to a utility fly ash precipitator, the V-I curves may be generated in
the routine which calculates the internal  data set.   Again, there is a switch
which causes the program to terminate after  the  V-I  curves are printed.

    10.  A routine for checking the input  data has  been  included in Revision 3.
This routine checks each it£m of input data  for  variables which are out of  the
allowed range,  such as having too many size  bands,  and for operating conditions
which have allowed but unusually large or  small  values,  such as found in a
laboratory scale ESP.  If anomalies in the data  are  discovered, diagnostic
messages will be printed stating that an error has  been  made, or, in the second
case,  that  certain values are unusual and  should be  checked for correctness.

     Revision 3 requires only 12 percent more memory than Revision 1.  On the
SRI Digital Equipment Corporation (DEC) System 20 computer, Revision 3 uses 206
kilobytes of 7  bit memory.  To facilitate model  usage on computers having less
memory available,  the code is marked so that some of the less essential
functions may be easily removed.  The minimum program length, with all of the
marked statements  removed, is 135 kilobytes,  which  is approximately 65 percent
of the entire program.
                                     215

-------
     The data format  for Revision 3 is  largely the same as  for  previous  versions
of the model.  Old data sets may  be used without changes to yield  the  same
results in less time.  Only  the additions,  such as opacity  and  dynamic rapping,
require additional input data.

     The EPA/SRI ESP model has proven to be a useful tool in the design  and
evaluation of electrostatic  precipitators.   Revision 3 of the model  offers
greater utility through greater flexibility and added features  such  as the
opacity calculation and dynamic rapping.  Revision 3 is also much  faster  than
previous model versions, which results  in significantly reduced computer  costs.
This should encourage wider  use of the  model.  As with Revision 1, Revision 3
will be complete in two volumes.   Volume 1  will contain a description of  the
physical processes being modeled,  the algorithms used, a description of  the
necessary input data, and a  listing of  the  FORTRAN code.  Volume 2 will be a
user's manual which will contain  complete explanations of the input  and output
data, examples of programs which  illustrate the various options available, and a
tutorial of the effects of varying some of  the computer operating parameters,
with supporting examples.
                                     216

-------
                                   REFERENCES
 1.   Gooch,  J.  P.,  J.  R.  McDonald,  and S.  Oglesby, Jr.  A Mathematical Model of
     Electrostatic  Precipitation.   EPA- 650/ 2-75-037 (NTIS PB246-188), U.S.
     EPA,  IERL,  Research   Triangle  Park,  NC, 1975.

 2.   McDonald,  J. R.   A Mathematical Model of Electrostatic Precipitation
     (Revision  1):  Volume I.  Modeling and Programming.  EPA-600/7-78-llla  (NTIS
     PB284-614),  U.S.  EPA,  IERL,  Research Triangle Park, NC, 1978.

 3.   McDonald,  J. R.   A Mathematical Model of Electrostatic Precipitation
     (Revision  2):  Volume II.  User  Manual.  EPA- 600/ 7-78-lllb (NTIS PB284-
     615), U.S.  EPA,  IERL,  Research  Triangle Park, NC,  1978.

 4.   Mosley,  R.  B., M.  H.  Anderson, and J. R. McDonald.   A  Mathematical Model
     of  Electrostatic  Precipitation (Revision 2).   EPA-600/7-80-034 (NTIS PB80-
     190994), U.S.  EPA, IERL,   Research Triangle Park, NC, 1980.

 5.   Smith,  W.  B.,  and J.  R.  McDonald.  Development of a Theory  for the
     Charging of Particles  by Unipolar Ions.  J. Aerosol Sci.,  7:151-166, 1976.

 6.   Felder,  R.  M., and E.  Arce-Medina.  Improvements in an Algorithm for
     Calculating the  Potential Field Distribution and Current  Density in a
     Parallel Plate Electrostatic Precipitator.   J. Electrostatics.  (To be
     published.)

 7.   Cooperman,  P.  The Dependence  of the Electrical Characteristics of Duct
     Precipitators  on  Their Geometry.   Progress Report  No. 46, Research Corp.,
     Bound Brook, NJ,   1952.

 8.   Orbach,  0.,  and  C. M.  Crowe.   Convergence Promotion in the Simulation of
     Chemical Processes with  Recycle:  the Dominant Eigenvalue  Method.  Can. J.
     Chem. Eng.  49:509,   1971.

 9.   Gooch,  J.  P.,  and G.  H.  Marchant, Jr.  Electrostatic Precipitator Rapping
     Reentrainment  and Computer Model Studies.  EPRI Contract  RP413-1, The
     Electric Power Research  Institute, Palo Alto, CA, 1978.

10.   Mie, G.  Beitrage zur  Optik  truber Medien,  Speziell Kolloidaler
     Metallosunge.  Ann.  Physik.  25:377-455, 1908.

11.   DuBard,  J.  L., and R.  F.  Altman.  Prediction of Electrical Operating Points
     for use  in  a Precipitator Sizing Procedure.  Presented at  the EPRI
     Conference  on Electrostatic Precipitator Technology for Coal-Fired Power
     Plants,  Nashville, TN, July  14-16, 1982.
                                     217

-------
       NUMERICAL SIMULATION OF THE EFFECTS OF VELOCITY FLUCTUATIONS ON
                   ELECTROSTATIC PRECIPITATOR PERFORMANCE

              by: Eric A. Samuel
                  General Electric Environmental Services, Inc.
                  Lebanon, Pennsylvania 17042
                                  ABSTRACT

     A numerical scheme for including the effects of velocity fluctuations
in plate-wire electrostatic precipitators arising from turbulent diffusion,
within the framework of the already developed trajectory method for precipit-
ator performance evaluation, is described.  The predictions of the scheme
are shown to be in agreement with classical solutions based on Fickian diff-
usion for some simple configurations.

                                INTRODUCTION

     The turbulent flow in plate-wire precipitators may be regarded, in a
restrictive sense, as comprising eddies of a range of diameters, with the
number of eddies having diameters comparable to the plate spacing being small.
The time fluctuations observed in the velocity distributions within gas
passages may be explained within the above model as due to the random
movement and relocation of the eddies.  The motion of particles in a
turbulent velocity flow field may be likened to the Brownian motion of gas
molecules which is governed by random collisions.  The above analogy suggests
that particle motion in a turbulent field may be regarded as governed by
random collisions, with the mean-free path between collisions being of the
order of the average eddy diameter.

     The microscopic motion of gas molecules have been shown to be described
by a macroscopic diffusive transport operation known by Fick's Law(l), which
relates the gas molecule flux to the gas molecule concentration gradient
through a proportionality constant known as the diffusion coefficient.
Analytical descriptions of the effect of turbulence on the precipitator
collection efficiency have centered on an analogous diffusive transport
equation(2).  The eddy diffusivity model for treating particle motion in
turbulent flows yields analytically tractable results for only idealized
precipitator configurations.
                                     218

-------
     The purpose  of  the present stildy is to demonstrate an alternative
computational  method for including the effects of turbulence in theoretical
predictions  of the performance of precipitators,  having realistic configur-
ations,  within the framework of an existing numerical simulation method based
on particle  trajectory  calculations(3,4,5).  This simulation method has been
used previously to study theoretically the  effect of precipitator geometrical
parameters  (plate spacing,  wire spacing, and wire diameter) on the scaling
of precipiator performance.   The central idea in this alternative formulation
is to superimpose a  random  component,  attributable to the fluctuating comp-
onent of the velocity flow  field, on the particle trajectories calculated for
the mean velocity field.  The random component of the particle motion is
calculated in  incremental steps using a simple solution to the convective
diffusion equation.

     The present  study  is devoted to demonstrating the favorable comparison
between  the  simulation  method and analytical results for idealized precipit-
ator configurations.

     The present  description of the above study is structured as follows.
The bext section  will contain a description of the alternative scheme for
factoring the  effects of turbulence into the already developed numerical
simulation code based ori the trajectory method.  The following section will
compare  the  predictions based on the tarjectory method with analytical
results  for  turbulent flows in simple configurations.  Concluding remarks are
given in the final section.

                              NUMERICAL SCHEME

     The numerical simulation code based on the trajectory method has been
described elsewhere(3,4,5).   Briefly,  the method consists first of obtaining
the electrical characteristics of the precipitator by simultaneously solving
Poisson's Equation and  the  Continuity Equation together with the corona
condition suggested  by  Leutert and Bb"hlen(6), and by McDonald, Smith, Spencer,
and Sparks(7).  The  performance of the precipitator is then evaluated by
following the  trajectories  of particles of  known size distribution entering
uniformly across  a gas  passage.   The computer model as described above has
been used to study the  effect of varying the geometrical parameters which
characterize a plate-wire precipitator; namely, the plate and wire spacings,
and the  wire diameter,  on precipitator performance.  The model predicted
improved performance  with decreasing wire spacing and increasing plate spacing
and wire diameter.   Configurations which increased corona power appeared to
also yield improved  collection efficiencies.

     If  D is the  eddy diffusivity, and L the mean-free path (or mean eddy
diameter), it  can be  shown  that(l):

                     D  = L  v/3                                            (1)

where v  is the  mean  flow velocity.  Also, by solving the Diffusion Equation,
it can be shown thatthe mean square displacement between the positions r and
r0 of a  particle  seperated  by a time increment, At, is given by(l):


                                     219

-------
                    <(r - r0)2> = 2 D At                                  (2)

The effect of turbulence is included within the trajectory method by  super-
imposing on the trajectory calculated for the mean velocity  flow field a
random component attributable to the velocity fluctuations.  The random
component is calculated as follows:

     1.  Assume that the motion of a perticle within an eddy is governed
         by Equation 2, which gives the random displacements in the x and y
         directions for motion within the eddy, as:

                 Ax =  (2 D At)15 sin 9                                     (3)

                 Ay -  (2 D At)^ cos 6                                     (4)

         where 6 is the angle between the direction of particle motion within
         the eddy and the direction of mean particle flow.   The angle, 8,
         lies in the range, 0 £ 6 £ 360, due to the requirement that  the time
         average of the fluctuating component of the velocity flow field is
         zero.

     2.  The angle, 6, is initially selected arbitrarily.  The motion of a
         particle between eddies, however, is assumed to result in a  random
         change in the value of 6.  Once the local eddy size, L, is known,
         the effective local eddy diffusivity may be calculated on the basis
         of Equation 1 using the local mean velocity, v.  The time increment,
         At, for calculating the random motion is selected such that:

                 L = v At                                                 (5)

         The time increment, Atm, for calculating the mean particle motion is
         selected such that:

                 At = N At,,                                               (6)

         N being a sufficiently large integer.  A random direction change
         and displacement due to turbulence is calculated once in N displace-
         ments calculated for the mean particle flow.  Isotropic turbulence
         is invoked by assuming the mean eddy diffusivity to be the same
         everywhere within the precipitator gas passage.

     In the absence of turbulence, it is necessary to calculate the trajectory
of a particle of given diameter only once for each entrance  position  in the
precipitator.  In the presence of turbulence, however, it is necessary to
calculate several trajectories for each entrance position, since no two random
trajectories will be exactly the same.

         COMPARISON OF COMPUTER PREDICTIONS WITH ANALYTICAL  RESULTS

     Table 1 describes four simple configurations in which the above  method
for including the effects of turbulence on particle trajectories was  applied.


                                     220

-------
The configurations  were chosen because analytical solutions for the particle
concentration profiles  in the presence of diffusion are available in the
literature.   The four configurations are:

     1.   Diffusion  of particles originating from a point source in a flowing
         gas,  ignoring  wall  effects.  This case to the effect of turbulence
         in  the  central region of  a precipitator gas passage in the absence
         of  electrical  energization.

     2.   Diffusion  of particles originating from a point source in a flowing
         gas close  to a plate boundary,  the plate being assumed to be
         perfectly  reflecting.   This case corresponds to diffusion of
         particles  close to  a collecting plate, in the absence of electrical
         energization,  and assuming complete specular reentrainment of the
         particles  arriving  at the plate (complete absence of collection).

     3.   Same as Case 2 above,  but assuming the plate boundary to be
         completely absorbing.   This case corresponds to diffusion of
         particles  close to  a collecting plate, in the absence of electrical
         energization,  and assuming complete retention of particles which
         impact  the plate (complete absence of reentrainment).

     4.   Diffusion  of charged particles,  all of the same diameter and carry-
         ing the same charge,  in a gas passage across which a uniform static
         electric field is maintained.  This case corresponds to the effect
         of  diffusion in the collecting stage of a two-stage precipitator,
         the first  stage of  which  is the charging stage.

     Figures 1 to 4 summarize the  comparison between predictions of the
computer model and  analytical theory based on Pick's Law.  The agreement
between  the  two  methods are  seen to be reasonably good.   Two sources of
discrepancy  between the analytical and numerical methods are: (i) finite
number of test particles in  the numerical scheme, as opposed to limit of
infinitely large particles assumed in the analytical method, and (ii) the
lack of  complete randomness  in the random number generator used in the
computer method  (see Figure  5).  The good agreement between the analytical
and numerical methods,  especially  in the case of the two-stage precipitator,
demonstrates the validity of the numerical scheme for calculating the
effect of turbulence on precipitator performance.

                                CONCLUSION

     Through comparison with classical solutions to the Diffusion Equation,
the proposed numerical  scheme for  including the effect of velocity fluctuat-
ions on  calculations of precipitator performance appears to have been
validated.   Use  of  the  scheme to study the scaling of precipitator performance
on its goemetrical  parameters,  including the effect of turbulence, is planned.

                                REFERENCES

1.  Chandrasekhar,  S.   Rev.  Mod. Phys.  15, 1, 1943.


                                     221

-------
2.   Leonard,  G.  L.,  Mitchner,  M.,  and Self, S. A.  Experimental Measurements
    of Turbulent Diffusion on Precipitator Efficiency.  In:  Proceedings of
    the Third Symposium on the Transfer and Utilization of Particulate Control
    Technology,  Volume II.  Electrostatic Precipitators.  EPA-6QO/9-82-005b.
    U. S.  Environmental Protection Agency, Research Triangle Park,
    N. C.  27711, July 1982.   p 120 - 129.

3.   Kim, Y.  W.  and Samuel, E.  A.   Electrostatic Precipitators II.  The
    Efficiency and Wire-to-Plate Spacing Ratio.  Physics of Fluids Technical
    Report No.  27,  Department of Physics, Lehigh University, Bethlehem,
    Pa. 18018,  1978.

4.   Samuel,  E.  A.  Collection Efficiency of Electrostatic Precipitators by
    Numerical Simulation.  In:  Proceedings of the Second Symposium on the
    Transfer and Utilization of Particulate Control Technology, Volume II.
    Electrostatic Precipitators.   EPA-600/9-80-039b.  U. S. Environmental
    Protection Agency, Research Triangle Park, N. C. 27711, September 1980.
    p 1 - 30.

    Samuel,  E.  A.  Collection Efficiency of Electrostatic Precipitators by
    Numerical Simulation.  Environment International.  6, 137 - 152, 1981.

5.   Samuel,  E.  A.  Computer Simulation of the Wide Plate Spacing Effect. In;
    Proceedings of the Third Symposium on the Transfer and Utilization of
    Particulate Control Technology, Volume II.  Electrostatic Precipitators.
    EPA-600/9-82-005b.  U. S. Environmental Protection Agency, Research
    Triangle Park, N. C. 27711, July 1982.  p 149 - 159.

6.   Leutert, G. and Bbhlen,  B.  The spatial Trend of Electric Field Strength
    and Space Charge Density in Plate-Type Electrostatic Precipitators.
    Staub-Reinhalt. Luft.  32(7),27 - 33, July 1972.

7.   McDonald, J. R., Smith,  W. B., Spencer, H. W. and Sparks, L. E.  A
    Mathematical Model for Calculating Electrical Conditions in Wire-Duct
    Electrostatic Precipitation Devices.  J. Appld. Phys. 48(6), 2231 - 2243,
    June 1977.
                                      222

-------
           Table  1.   Simple  configurations  used  for  comparing  the numerical method for calculating
                      the effects  of velocity  fluctuations with the analytical method based on
                      Fickian diffusion.
          Configuration
                            Schematic
Analytical Expression
Reference
     1.  Point source in flow-
         ing gas without bound-
         ary effects  (diffusion
         in center of unenerg-
         zed gas passage).
                                             W(x,t)
                                                        exp {-x2/(4Dt)>
                                                                               Chandrasekhar(1)
to
NJ
CO
Point source in flow-
ing gas near reflect-
ing wall (diffusion
near collecting plate
of unenergized gas
passage with complete
specular reentrain-
ment at plate).
                                                      W(x,t;Xl)
                                                                tt
                                Chandrasekhar(1)
                                                                 exp  [-x/(4Dt)J
                                                      - _
                                                      2(7TDt)'5

                                                      + exp  [-(2Xl-x)2/(4Dt)]
         Point  source  in flow-
         ing  gas  near  absorb-
         ing  wall (diffusion
         near collecting plate
         of unenergized gas
         passage  with  complete
         collection at plate).
                                                               .tt
                                             W(x,t;x1)  ' =                     Chandrasekhar(1)

                                                      [ exp  [-x2/(4Dt)]
                                             - exp  [-(2xrx)2/(4Dt)]

-------
           Table  1  (Continued)
      Configuration
Schematic
Analytical Expression
Reference
    Uniform source in
    parallel plate coll-
    ecting stage with
    uniform static
    electric field in
    two-stage precipit
    ator.
                          f exp{-[(y'-a)
                         0
                                 Leonard,
                                 Mitchner, and
                                 Self(2)
                                                 dy'
   W(x,t) is the normalized probability of finding a particle at position x and time t, knowing that
   the particle source is located at the origin (x = 0) and emits particles at time t = 0. The time
   t is given by: t = x/v, v being,the mean gas velocity.
tt
   W(x, t;x^) is the normalized probability of finding a particle at position x and time t, knowing
   that either a reflecting or absorbing wall is located at x = xj, and the point source is located
   at the origin, and emits particles at time t = 0.

   W(y,t;Uy) is the normalized probability of finding a-particle at coordinate position (x,y),
   knowing that a uniform flux of particles having the same drift velocity, uy, entered the collecting
   stage at x = 0 and time t = 0.  The time t is given by:  t = x/v = y/o)y.

-------
     0.25
 j:  0.15
 CD
 
-------
     8.25
     0.20 —
 in
 ^  0.15 —
 
-------
     Q.25
     0.20
  S  0.10
  
-------
     1.5
  Of.
      1.0
      0.5 —
      0.0
               I  I  I   I  I  I   I  I  I  I  I  I  I  I  I  I   I  I  I     Tl  I  I
           0.0        1.0        2.0        3.0        4.0        5.0

                   DISTANCE FROM NON-COLLECTING WflLL,  CM
Figure 4.  Comparison of the predictions of the numerical  scheme for the
          inclusion of turbulence effects with the analytical result from
          diffusion theory for  the case of a uniform flux of identically
          charged particles entering a parallel plate collecting stage
          in which a uniform electric field is maintained (Case 4).
                                    228

-------
     2.Q
     1.5 —
  CO
  
-------
                         CORONA - INDUCED TURBULENCE

                                      by
                 M. Mitchner, G. L. Leonard1, and S. A. Self
                   High Temperature Gasdynamics Laboratory
                      Mechanical Engineering Department
               Stanford University, Stanford, California 94305
                                   ABSTRACT
     The results of previous experiments with a bench-scale precipitator (5 cm
plate-to-plate spacing)  have shown that moderate levels of turbulence can be
maintained in the presence of a corona discharge, and that in accord with
theory, significant gains in precipatator efficiency are possible.  In this
paper hot-wire anemometer measurements in a laboratory-scale precipitator
(25 cm plate-to-plate spacing) are presented that support the previous results
and suggest that similar improved performance is possible in larger-scale
commercial precipitators.  The measurements also provide a further explanation
for the reduced migration velocity previously reported by other workers at low
gas velocities.

                                 INTRODUCTION


     In a theoretical analysis of the effect of differing levels of turbulence
on the performance of electrostatic precipitators (1) it was shown that even
for moderate levels of turbulence, efficiences significantly in excess of that
predicted by the Deutsch-Anderson (2) formula should be possible.  Experiments
conducted in the collection stage (no corona discharge) of a bench-scale
(5cm plate-to-plate spacing) two-stage precipitator yielded results in excel-
lent agreement with the theory (3).  With no corona discharge present, and
therefore no possibility of corona-induced turbulence, the level of turbulent
mixing could be controlled with standard flow-conditioning elements such as a
well-designed entrance nozzle and turbulence-producing grids.

     In a single-stage precipitator, particle collection occurs in the pre-
sence of a corona discharge, and the question arises as to the level of tur-
bulence produced by the corona wind.  In a study of the basic fluid mechanical
effects of the corona wind (4) it was shown that a uniform positive corona
  'Currently at General Electric Corporate Research and Development,
  Schenectady, N.Y. 12345

                                     230

-------
discharge does  not  generate turbulence (5)  but that turbulence could result
from the instability  of  a  pair  of  counter-rotating vortices  produced at each
end of the discharge  wire,  that propagate into the main flow.
     For a negative corona,  because  of the  characteristic non-steady tuft-like
discharge structure,  it  was shown  (4)  that  turbulence was generated throughout
the precipitator  volume.   The intensity of  the negative corona-induced turbu-
lence was comparable  or  less than  the  background turbulence  present without
electrical energization  for gas velocities  exceeding approximately 1.5 m/s but
increased rapidly for decreasing gas velocities.
     The experiments  to  be  described were undertaken to examine how the afore-
mentioned corona  wind effects scaled in a larger facility (25  cm plate-to-
plate spacing), more  nearly approximating the dimensions of  commercial
precipitators.  The results, like  those obtained for the bench-scale precipi-
tator, show that  moderate  turbulence-level  precipitator flows  are achievable
in the presence of  a  corona-discharge.  In  addition, results obtained for low
gas velocities  appear to offer  a further explanation for the decreased migra-
tion velocities observed by other  workers (6,7) for these conditions.

                            EXPERIMENTAL APPARATUS
     Experiments were performed in a laboratory-scale electrostatic precipita-
tor test facility described  more fully by Pejack (8).  The precipitator duct
used in these experiments was 0.75 m high, 1.75 m long,  and had seven 2.77 mm
diameter uniformly spaced discharge wires.  The mean gas velocity at the pre-
cipitator inlet plane was uniform to within ±2% and the rms of the free stream
turbulent velocity was less  than 2% of the mean velocity.
     Hot-wire anemometer measurements were made 0.40 m downstream of the last
corona wire.   Time and space resolved values of the gas velocity component in
the flow direction,  u(x,t) = u(x") + u'(^»t), were made with a Sum diameter
tungsten hot-wire sensor operated with a TSI model 1050 anemometer and linear-
izer.  (Here  u(x") denotes the local mean velocity and u'(£,t) is the local
fluctuation velocty.   We shall denote by UQ = U0(x) the value of u(£) averaged
over a transverse plane located at the streamwise location x.  The rms value
of u'(£,t) will be written as u'(^) or as uf (without an argument), and the
turbulence intensity will refer to either u'/uXx') or to U'/UQ.)  Values of
u = u(&) were measured with  a Hewlett Packard integrating digital voltmeter.
The fluctuation velocity (not linearized) was processed with a TSI model 1076
true rms voltmeter to obtain values of u'.

                             EXPERIMENTAL RESULTS
     Transverse  profiles  of  u(y)  and  u'(y)>  measured at mid-height downstream
of the unenergized  precipitator,  are  presented in Figure 1.   These data show
the turbulent  boundary  layer extending  approximately 7  cm from each wall,  and
a turbulent wake with a centerline turbulence intensity of about 5%.  Voltage-
current characteristics for  both  negative  and positive  corona are given in
Figure 2.
                                     231

-------
   2.0
O  i c
g  '•=>
Si
   0.5-
LU
UJ 1°

1 1 1 1 1 1 1 1 1 1 1 1
naaaaa A A a a a ^ A
i i i i i i i i i i i i
JCE INTENSITY PROFIL
u'(y)/u(y), °/0
4^ 0) 00
LU
_I
m 2
CE
1-
1 ) i 1 l l i i i l 1 i l
"T t"
^
^\ ^ /
l\ f\ /
- \ / '\ / -
\ ' \ ^
W3 V^
i i i i i i i i i i i i
0 2  4 6  8 10  12 14 16 18 20 22 24
  DISTANCE FROM COLLECTING WALL y, cm
                                                2  4 6  8 10 12 14 16 18 20 22 24
                                               DISTANCE FROM COLLECTING WALL y, cm
      (a) Mean Gas Velocity Profile
                                        (b) Turbulence Intensity Profile
      Figure 1.   Transverse mean velocity and  turbulence intensity profiles at
      the exit of the unenergized precipitator.
<4.0
£
0:
o:
13
O
<2.0
o
"1.0
     o NEGATIVE DISCHARGE
     * POSITIVE DISCHARGE
         JS*
       40
              50
         CORONA VOLTAGE, kV
                             60
      Figure 2.   Voltage-
      current corona characteristics.
      Turbulence intensity profiles
 measured at the centerline of the
 exit of the precipitator for several
 corona voltages and for gas velocit-
 ies of 3,2, and 0.9 m/s are shown in
 Figure 3 for negative corona, and in
 Figure 4 for positive corona.  The
 voltages correspond to current
 densities at the collector in the
 range from 0.08 to 1.3 mA/m2, and
 approximately span the operating
 range of most power-plant
precipitators  (0.05  to 0.7 mA/m ).
For UQ = 3 m/s,  these results show
relatively little  change for
negative corona, and essentially no
change for positive  corona, from the
corresponding  turbulence intensity
profiles for the unenergized
precipitator.  When  the gas velocity
is reduced to  1.5  m/s, the
turbulence intensity for the
negative corona  becomes comparable
to that produced by  the turbulent
wake from the  unenergized discharge
wires.  This increased turbulence
appears to result  from the
instability of localized vorticies
generated by large localized body
forces induced by  the discrete
discharge tufts  (9), and by the
nonuniform and nonsteady behavior of
these tufts.   For  u   = 1.5 m/s, very
little change  is produced in the
turbulence intensity profile for the
positive corona  case.

     For a gas velocity of 0.9 m/s
both the negative  and positive
corona produce very  marked increases
in turbulence  intensity.  In the
negative corona  case this additional
                                       232

-------
z  c
UJ^oD
Z -
m
ct
                   so  8
D    O °
X    A
     o

  *
I  I   I  I   I
                            a
                         2  2
                         CD  O
    0 2  4 6  8 10  12 14 16 182022 24
       TRANSVERSE DISTANCE y, cm
      (a)   u0 = 3 m/s

H
-
LU *^
D"3
m
o:
D


8

6


4


2


Q
—i — i — i — i — i — i — i — i — i — i — i — r-
- 0

o
" <£& o
o o o o
0 0 0 0 & A
A
D
on o
o
0
_ o
o _
1 1 1 1 1 1 1 1 1 1 1 1
^ n /. c o ir\ioixicioor\oo o/.
       TRANSVERSE DISTANCE y, cm
      (b)   u0 = 2 m/s
I/
> 10
1
1 —
(/)
z
l_ ~5
Z .
UJI3 6
LU ***
_|-v^
m
Q:
•- 2
n
— i — > — i — i — i — i — i — i — i — i — i 	
- 01 o 2 .
D A & o g ao o 8
4 2o.
A D
a
D O
a
D a a a a
o o o o
a n
; ° ° o °
0 0 0
                                               TRANSVERSE DISTANCE  y, cm

                                               (c)    UQ = 0.9 m/s
                                      Figure 3.    Dependence of
                                      transverse turbulence
                                      intensity  profiles at the exit
                                      of the precipitator on gas
                                      velocity for negative
                                      corona.   The symbols
                                      correspond to the following
                                      voltages:*"?   OkV,  D  -40 kV,
                                      A-50kV,   O-60kV.
increase can be attributed to an
increase in the relative strength of
the discharge tufts and to the
accompanying increase in localized
flow instabilities produced by
corona wind effects.  In the posi-
tive corona case,  the very large
increased turbulence level is
believed to result from the spread
to the precipitator centerline of
unstable axial vortex pairs
generated near the ends of the
discharge wires.
     The turbulence-producing mecha-
nism at low gas velocities for
positive corona would appear to be
dependent on the nature of the
discharge wire termination, as well
as on the height or perhaps aspect
ratio of the precipitator, whereas
                                for negative corona the dependence
                                on these features  should be much
                                weaker or absent.   Further measure-
                                ments  at low velocities with changes
                                in these aspects of the precipitator
                                design are necessary to examine this
                                hypothesis.
                                     In Figure 5,  the turbulence
                                intensity u'/u  measured at one exit
                                location (7 cm from one wall) is
                                shown  plotted as a function of
                                linear current density for two
                                different values of mean gas
                                velocity, for each type of corona
                                discharge.  For negative corona, the
                                turbulence intensity increases
                                rapidly with increasing current for
                                uQ = 1.5 m/s, but  increases slowly
                                for uo = 3 m/s. For positive corona
                                the turbulence intensity increases
                                rapidly with increasing current for
                                     233

-------
NTENSITY
1, %
O) 00
it4
TURBULE
u'(y
3 N>
— i — i — i — i — i — i i i i ' '
~ *8 [£>
an & x
* ft. ff ff
. ii-il 	 1 	 1 	 1 	 1 	 1 	 1 	 L-
       TRANSVERSE DISTANCE y, cm

      (a)   u. = 3 m/s
                                           12


                                        >  10
                                        tiir-
                                        GO
                                        Ct
                                           Q
                                  o  o       o
                                       o
                                                     o      o
                                                  o    o
                                            D  2  4  6  8  10 12  14 16 182022 24
                                                TRANSVERSE DISTANCE y, cm
   8-
   '6-
UJ,
co   2
cc
                                 a
                                 a  -
                    *
       2 4  6 8  10 12  14 16 182022 24
        TRANSVERSE DISTANCE y, cm
       (b)
UQ = 2 m/s
 u  = 0.9  m/s,  but  increases  slowly
 for u  =  1.5 m/s.

    INTERPRETATION OF THE RESULTS
      The experiments  show  that
 corona wind effects become important
 for larger currents and  for smaller
 gas velocities.   This behavior  can
 be understood in terms of  a dimen-
 sionless number  which characterizes
 the strength of  the interaction of
 the corona wind  relative to the
 inertial force associated  with  the
 motion of the gas.
                                               (c)   u- = 0.9 m/s
                                   Figure 4.    Dependence of
                                   transverse turbulence
                                   intensity  profiles at the  exit
                                   of the precipitator on gas
                                   velocity for positive
                                   corona. The symbols corre-
                                   spond to the following
                                   voltages:O  OkV,  G  40kV,
                                   A 50kV, O 60kV.
     The force per unit volume that
acts on the fluid to produce the
corona wind is p^E, where p.. and E
are characteristic values of the ion
charge density and electric field,
respectively.  The electric field is
related to the ion current density J
by the relation J = p^bE, where b is
the ion mobility.  In terms of the
total ion current i from a discharge
wire of length 5L and wire-to-plate
spacing d, J = (i/A)/2iid.  We may
therefore write p±E«(i/J!.)/bd.
     The inertial force per unit
volume acting on the fluid over a
characteristic distance d is of
order puo /d, where p is the mass
density and UQ is the velocity of
the gas.  A measure of the relative
                                      234

-------
                                           10.0 h
-S 8.0
8

  4.0
  2.0
D
1-
                    NEGATIVE CORONA
                    o u0= 150 cm/s
          0.2    0,4    0-6     08

        LINEAR CURRENT DENSITY (mA/m)
                  POSITIVE CORONA

                   o  U0= 90 cm/s
                     "o= 150 cm*
        0,2    0.4    0.6     OB
      LINEAR CURRENT DENSITY (mA/m)
      (a)   negative corona
(b)   positive corona
      Figure 5.   Effect of  gas  speed on turbulence intensity  (7  cm from wall)
      as a function of linear  current density.
importance of corona wind effects  is provided by the dimensionless  parameter


           electric body force/unit volume _ (i/JQ/bd  ^_
           inertial force/unit volume2/,
                                              P uo/d

The experimental results reported  here behave qualitatively in accord  with
this number, and in fact several other features of corona wind effects (9) may
be correlated with the aid of this parameter.

                           LOW GAS VELOCITY EFFECTS
     The Deutsch theory (2) of  electrostatic precipitation does not  predict
any variation of migration velocity with gas speed, although it is generally
appreciated that there will be  a  progressive deterioriation at high  speeds
owing to reentrainment.  What is  not well understood is the reduction in
effective migration velocity observed by many workers for lower gas  speeds.
Examples of previously reported results are shown in Figure 6.
                                     235

-------
                           VAIUCS F0» lAHOC MO
                                or TISTS
                                 6*1 VELOCITY IN MCCIMTATO*
                 (a)   Results of Busby and Darby (6)
           u
           V
           5 30
           ~* 20
           1>0
                           Sip dio.
                           I Ip did
                            •5p dio.
           ^ O            I            2           3           4
           *Z                   Coi velocity.m/jec.


           (b)  Results  of Dalmon and Lowe (L/d a 20)  (7)


Figure 6.   Example of measurements of the variation of the effective
migration  velocity with gas speed reported by previous workers.
                                     236

-------
     There  are  at  least  three different mechanisms  that  have  been suggested  to
explain  these observations.  The  Deutsch  theory  is  strictly applicable  to
particles of one size whereas most  observations  have  been made for polydisper-
sions.   In  qualitative terms one  might expect  that  at higher  gas  speeds the
precipitator electric field acts  on the average  on  particles  of larger  average
size (and charge)  and should therefore exhibit a larger  effective migration
velocity.   A second explanation proposed  by  Cooperman (10) is that increasing
axial turbulent diffusion relative  to convection at decreasing gas speeds
should reduce the  effective migration velocity.   A  third explanation offered
by Busby and Darby (6) is that  the  effect results from transverse turbulent
diffusion (although the  conjecture  that the  turbulence intensity  increases
with gas speed  is  not in accord with measurements reported in this paper).

     In  what follows, each of these mechanisms will be examined separately.
The central feature of the experimental results  on  which we shall focus, is
that the effective migration velocity at  lower gas  speeds is  approximately
increased by a  factor of two for  a  doubling  of the  gas speed.

     If  f(x)dx  denotes the fraction of particles in a polydispersion with
diameters between  x and  x + dx, then using the Deutsch formula, the penetra-
tion from a precipitator having a ratio of collection area to volume flow rate
(A/V) is
                        •5
                         exp[-(A/V)w(x)]  f(x)dx.


The apparent migration velocity w  of the polydispersion is defined by  the
equation Q  = exp[-(A/V)w ].  Following the procedure  of  Allander  and Matts
(11), if we assume that  field charging is dominant  so that w(x)«  x and  that
f(x) is  a log-normal distribution,  the integral  may be evaluated
numerically. ,-The  result is a function of two  dimensionless parameters, the
Deutsch  No. ^ =  (L/d)(w /u )  and  the geometric standard deviation a . Here
L is the precipitator length and  w   = w(xg)  is the  migration  velocity of a
particle having a  diameter equal  to the geometric mean  x (equal to the
median)  of  the  distribution.  For the apparent migration velocity one then
obtains  the  formula w  /w  =  [An  (1/Q)]/
                  ^~~^ 1
     The quantity  J^'L" = (d/L)(uo/wg) may be viewed  as dimensionless  gas
speed.   Shown  in Figure 7a are values of wa/wg plotted as  a  function  of the
dimensionless  gas velocity (or equivalently,  J^l) ,  for some representative
flyash values  for An o_.  For a  fixed value of An  ag if wa^  denotes the
apparent migration velocity  corresponding  to a gas velocity  u^,  then  the
apparent migration velocity  wfl£  corresponding to a gas velocity  u« =  2u, may
be obtained  directly from  this  figure, as a  function of  j^L  =  (L/d)(wg/u1)
and An ag.   The results of such a  calculation  for wa2/wai  are shown  in  Figure
7b.  Laboratory precipitators usually have values of $a  °f  tne  order  of  one
or larger, whereas large efficient commercial  precipitators have  values of
j^L at the  upper end of the scale.  The main  conclusion to be  drawn from  this
figure is  that polydispersion effects alone  are  too  small  to  be able to
account for  the observed dependence of the effective migration  velocity on gas
speed.

     To assess the possible effects of axial turbulent diffusion, we may
employ the theoretical result (1)  that the penetration of  raonodisperse
particles, for the dominant mode (i.e. the mode  that decreases  least .rapidly


                                    237

-------
          DEUTSCH No., 3g
    5  2.5 . 1.67 1.25  1.0 , 0.83 0.71, 0.63 055
      Q =/°exp[-(A/V)w(x)]f(x)dx s exp[-(A/V)wJ
                                          2.0
                                     $0
 0  0.2  0.4 0.6  0.8 1.0  1.2 1.4  1.6  1.8
DIMENSIONLESS GAS SPEED, 2>g' = (d/L)(u0/wg)
                                          1.0,
                                                      i   6   8   10  12  14  16
                                                      DEUTSCH No., 3>a
   (a)   Dependence of apparent
   migration velocity w& on gas
   velocity UQ, for a log-normal
   polydispersion of particles
   f(x) having a geometric mean
   diameter x  and a geometric
   standard deviation a&.  Here
   w  = w (x ) is the actual
   migration velocity for a
   particle of diameter x , and
   J^g = (L/d)(w /UQ) isSthe
   corresponding Deutsch No. for
   a precipitator length L and
   wire-to-plate spacing d.  The
   ratio of plate area to volume
   flow rate is A/V and Q denotes
   the penetration.
                                                (b)   Effect of  a
                                                polydispersion of  particles
                                                when the gas velocity is
                                                doubled (U2 = 2u^)  on the
                                                ratio wa2/wal =  wa(u2)/wa(Ul)
                                                of the apparent  migration
                                                velocities.  Shown is the
                                                dependence on the  Deutsch No.
                                                ^  = (L/d)[w (x  )/Ul]  for a
                                                log-normal distriBution< with
                                                geometric mean diameter  xg,
                                               _ for several values  of the log
                                                of the geometric standard
                                                deviation, Jin a  •
                                                               O
             Figure 7-    The  gas speed effect of a polydispersion on
                          apparent migration velocity.

with precipitator length),  is given by the modified Deutsch formula Q = exp[-
(A/V)wa] ,  where
                                          F, (p)     21/?

                          „   u  2
                 •>„> -    12-)
                                   238

-------
Here P = wd/D,  the electric Peclet
No., provides a measure of the ratio
of particle transport by the
electric field to that by turbulent
diffusion.  The quantity D is the
turbulent diffusivity and  06? =
(L/d) (W/UQ) is the Deutsch No.  The
function F^P)  must be determined
numerically, but is given to a good
approximation by the expression
FX(P) = 1 + 0.2P.

     If w ,  and w 2 denote the
apparent migration velocities corre-
sponding to gas velocities u, and ^
                                            2.0
= 2u-
       the ratio Wa2/wai
calculated directly from the formula
w (2u,)/w (u,).  This ratio is shown
in Figure 8 as a function of
(w/ui).  The effect of increasing
turbulent diffusivity D-^ is
expressed through the effect of
decreasing P^ = wd/D^.  Curves are
shown  for   P^ = 1, 10" , and
10~ .  The corresponding values, D^
= 10~2 m2/s, 1 m2/s, and 102 m2/s
are obtained using the typical
values w = 0.1 m/s and d = 0.1 m.
Typical values of the abscissa fall
in the range from about 0.1 to 0.2.

     These values of D-^ may be
compared with values found in fully-
developed turbulent pipe flow, where
the correlation u d/D = 1000 has
been developed (12).  Using as
typical values u  =1 m/s and d =
0.1 m, we obtain D =10  m2/s.  Thus
in relative terms, the values of D,
shown in Figure 8 that are needed to
obtain significant values for
Wa2/wal are extremely large.
Futhermore, the maximum value of
Wa2/wal that can ^e obtained
corresponding to D^-*- <*>, is 2.  It
therefore seems highly unlikely that
axial diffusion effects alone could
be responsible for the experimental
data shown in Figure 6.

     A third possible mechanism is
based on the effects of transverse
                                              0  0.05 0.1  0.15 0.2  0.25 0.3  0.35 0.4
                                               RATIO OF MIGRATION TO GAS VELOCITY, w/u,
Figure 8.  Effect of the axial
turbulent diffusivity D when
the gas velocity is doubled
(U2 = 2u^) on the ratio
Wa2/wal = V^/W of the
apparent migration velocities,
for the dominant mode.
       Shown is the dependence
on the ratio (w/u-^) of the
migration to the gas velocity,
for several values of the
electric Peclet No. P± =
wd/D^.  Here d is the wire-to-
plate spacing.  Values shown
for D^ are for w = 0.1 m/s and
d = 0.1 m.
                                         turbulent diffusion.  But in
                                         contrast to Busby and Darby (6), we
                                         propose an explanation based on the
                                         very large increase in the rms value
                                         of the turbulent velocity u1 that
                                         the previously described experiments
                                         show occurs at lower gas speeds.
                                         Since the turbulent diffusivity is
                                         approximately proportional to u'L*
                                         (where L* is a measure of the size
                                         of the larger energy-containing
                                         eddies in the flow), if follows that
                                         D is significantly increased and
                                         therefore the electric Peclet No. is
                                         significantly decreased, at lower
                                         gas speeds.  From theory (1) the
                                         penetration, and therefore the
                                         apparent migration velocity, is
                                         significantly decreased.
                                    239

-------
     According to the theory ofthe effect of turbulence, the penetration Q is
a function of the Deutsch No. ^ and the Peclet No. P — i.e. Q = QCJ^.P).
The apparent migration velocities wa^ and wa2 corresponding respectively to
                  and to the Peclet Nos.
                                                              and P, are
the gas velocities u^ and u2
given by the equations Qx = Q ( <^i, PI> = exp[- (A/V1)w&1] and Q2
= Q ( J^/2, P2) = exp[- (A/2V15wa2j and therefore wa2/wal = 2 An (l/Q2)/An
     Performing this calculation one obtains the results shown in Figure 9,
where the effects of gas speeds are shown as a function of ®$ \ ~ (L/d)(w/u..),
for several values of P2/pi = D]/D2>  and for PX = 1.25 or ?^ = 2.5.  The
values selected for DI/DO  approximately span the range inferred from the
experiments.  The main uncertainty hinges on the values for P,, but the values
selected are believed to be not unreasonable.  It is apparent from these
results that the proposed mechanism is certainly capable of explaining the
magnitude of the reported effects of gas speed on migration velocity.  Further
work on this question is in progress.
            4   6   8  10  12   14  16
            DEUTSCH No.,3g
         (a)
= 1.25
                                                   4   6   8   10
                                                    DEUTSCH No.,
                                                12  14  16
                                                 (b)
= 2.50
Figure 9.   Effect of the transverse turbulent diffusivity D when the gas
velocity is doubled (u2 = 2 UL) on the ratio w&2/w . = wfl (u2)/wa(u1) of the
apparent migration velocities.  Shown is the dependence on the Deutsch No.
j^L = (L/d) (w/uj^), for several values of the turbulent diffusivity ratio
D1/D2 = D(ui)/D(u2).  Here w = w(x) is the migration velocity, d is the wire-
to-plate spacing, and P1=wd/D1 is the electric Peclet No.
                                     240

-------
                                 CONCLUSIONS
     For  gas  speeds  in  excess  of  about  3  m/s  and  currents of practical
interest,  corona-induced  turbulence appears to have little influence on the
background fluid mechanical  turbulence  generated  in the boundary layer and by
the discharge wire wakes.  For gas  speeds of  about 1.5  m/s,  negative corona-
induced turbulence appears to  be  of the same  magnitude  as the background
turbulence; positive corona-induced turbulence appears  to be small except
possibly  near the top and bottom  of the precipitator.   At lower gas speeds
negative  corona-induced turbulence  becomes increasingly intense with decreas-
ing gas speeds, and  methods  for achieving moderate turbulence levels in
single-stage  precipitators are unlikely to be effective.  To achieve improved
precipitator  performance with  negative  corona through turbulence control, it
appears that  gas speeds need to be  approximately  equal  to or larger than 1.5
m/s.  This value is  reduced  somewhat for  smaller  corona currents.   Attainment
of moderate turbulence  levels  at  lower  gas speeds with  positive corona may be
possible.   An explanation of previously reported  reductions  of effective
migration velocity at lower  gas speeds  is suggested by  the large increase in
turbulence intensity observed  at  lower  gas speeds.
                               ACKNOWLEDGEMENTS
     This  work was  supported  by  the  National  Science  Foundation under  Grant
Number CPE-7926290  and  by  the Electric Power  Research Institute under  Contract
RP-533-1.

     The work described in this  paper was  not funded  by the U.  S.
Environmental Protection Agency  and  therefore the  contents  do  not  necessarily
reflect the views of  the Agency  and  no official endorsement should be
inferred.
                                     241

-------
                               REFERENCES

 1.    Leonard,  G.  L.,  Mitchner, M.,  and Self, S. A.  Particle Transport in
      Electrostatic Precipitators. Atmospheric Environment 14, pp 1289-
      1299,  1980.

 2.    White, H. J., Industrial Electrostatic Precipitators.  Addison-
      Wesley, 1963.

 3.    Leonard,  G.  L.,  Mitchner, M.,  and Self, S. A.  Experimental Study of
      the Effect of Turbulent Diffusion on Precipitator Efficiency.  Jour.
      Aerosol Science  13,  pp 271-284,  1981.

 4.    Leonard,  G.  L.,  Mitchner, M.,  and Self, S. A.  Experimental Study of
      the Electrohydrodynamic Flow in Electrostatic Precipitators.
      Accepted  for publication by Jour, of Fluid Mechanics.  See also
      Precipitation from Turbulent Flows.  Proceedings of the Int. Conf. on
      Electrostatic Precipitation, pp 208-256.  Monterey, Calif., October
      14-16* 1981.

 5.    Yamamoto, T., and Velkoff, H.  R.  Electrohydrodynamics in an Electro-
      static Precipitator.  Jour, of Fluid Mechnics 108, pp 1-18, 1981.

 6.    Busby, H.G.T., and Darby, K.  Characteristics of Electrostatic Pre-
      cipitators.   In:  Proceedings  of La Physique des Forces Electro-
      statiques, pp 229-253.  Grenoble, France.  September 27 - October 1,
      1960.

 7.    Dalmon, J.,  and  Lowe,  H. J.  Experimental Investigations into the
      Performance of Electrostatic Precipitators for P. F. Power
      Stations.  In:  Proceedings of La Physique des Forces Electro-
      statiques, pp 363-379.  Grenoble, France.  September 27 - October 1,
      I960.

 8.    Pejack, E. R. An Aerosol Tunnel Test Facility.  Jour. Environmental
      Sciences   May/June,  1981.

 9.    Leonard,  G.  L.  Effect of Turbulence on Electrostatic Precipitator
      Performance.   Ph.D.  Dissertation, Department of Mechanical Engine-
      ering, Stanford  University, 1982.

10.    Cooperman, P. A New Theory of Precipitator Efficiency.  Atmospheric^
      Environment  5,  pp 541-551, 1971.

11.    Allander, C., and Matts, S.  The effect of Particle Size Distribution
      on Efficiency in Electrical Precipitators.   Staub 52, p 738, 1957.

12.    Becker, H. A., Rosensweig, R.  E., and Gwodz , J. R.  Turbulent Dis-
      persion in a Pipe Flow.  A. I. Ch. E. Jour. 12, pp 964-972, 1966.
                                  242

-------
   VELOCITY  AND  TURBULENCE  FIELDS  IN NEGATIVE CORONA WIRE-PLATE PRECIPITATOR

          by:  H.P.Thomsen,  P.S.Larsen,  E.M.Christensen,  J.V.Christiansen
              Department of Fluid Mechanics
              Technical University of Denmark
              DK-2800  Lyngby
                                   ABSTRACT

     Back-scatter  laser  Doppler anemometry has  been used to obtain distribu-
tions of mean  and  rms values  of particle velocity in horizontal planes perpen-
dicular to  electrodes in a  0.3 m  wide by  0.6 m  high wire-plate electrosta-
tic precipitator for four types of  electrodes and a mean velocity of  1  m/s.
Particle motion reveals  complex three-dimensional flow patterns with rolls of
axial vorticity which are most  regular for barbed wire electrodes with axial
needles. The turbulence  level is shown to depend on current density and mean
velocity and to disappear when  the  inverse electrical Froude number is below
about 0.2  . However, rolls  persist. Results are discussed in terms of the
turbulent kinetic  energy production.

                                 INTRODUCTION

     Further improvement of the modeling of particle transport in electrosta-
tic precipitators  requires  detailed experimental studies of the ionized gas
flow and the motion of charged  particles under  the influence of electric
field and prevailing gas flow.  Laser Doppler anemometry (LDA) provides a means
for in-situ measurements of instantaneous velocity components of individual
particles.  Time averaging such  results provides mean and rms velocity compo-
nents of particles as well  as particle concentrations. The extend to which
such particle  data reflect  the  gas  motion depends on the governing parameters.

     In the absence of an electrical field (or  particle charge) uniformly
distributed micron-sized particles  closely follow the instantaneous motion of
the gas phase  to a sufficiently high frequency  that the LDA-technique provides
an adequate description  of  the  turbulent gas flow (Durst, Melling & Whitelaw
1981). The  governing parameter  is the Stokesian time constant
\= (2az/9V)(pg/p) , which  is less  than  10""* s , for example, for  2a = 5 ym
diameter drops,  V = y/p,   p  and  ps  denoting kinematic viscosity, fluid
density and particle material density, respectively.
                                     243

-------
     Net particle drift owing to non-uniform concentration distributions in
turbulent flows may be inferred from the changes  in concentration profiles
along a flow. Actual drift velocities of particles relative to gas are usually
too small to measure directly with any accuracy.  For  a  turbulent  diffusivity
of, say  D-r = 10"1* - 10~3 m2/s  in a channel of small dimension  L = 0.1 m ,
the drift velocity associated with turbulent diffusion  will be on the order of
DT/L -v/ 0.001 - 0.01 m/s .

     Drift velocities resulting from the Coulomb  force  on charged particles in
an electrical field, on the other hand, are typically on  the order of
WE '^ 0.05 - 0.5 m/s  and can be readily measured  in a laminar parallel flow.
The terminal drift velocity of small particles of radius   a  and  charge  q  in
a locally uniform electrical field of strength  E may  be calculated from
WE = qEC/(6irya) , where  C  denotes the Cunningham correction factor.

     In view of the smallness of the Stokesian time constant  T  and for weak
turbulence, terminal drift  WE  can be assumed to exist locally throughout
inhomogeneous fields. LDA-measurements therefore  are  expected to  reflect the
result of particle drift relative to the local gas motion.  In general,  when
turbulent diffusion can be ignored, particle velocity components  perpendicular
to the local Coulomb force equal the gas velocity components.  Specifically,
when the electrical drift velocity is small, particle velocities  approximately
describe the gas velocity. The foregoing remarks  are  useful when  interpreting
LDA-data.

     During the last -few years IDA has been used  in a number of experimental
precipitator studies. Employing pre-charged particles in  small parallel plate
precipitators, Jurewicz, Stock & Crowe (1977) and Ross  (1980)  measured axial
velocity and drift velocities at selected conditions, while Leonard, Mitchner
& Self (1982a) made extensive measurements of concentration distributions for
comparison with their mathematical model (1980) describing the effect of tur-
bulent diffusion on collection efficiency. Results of this study  infer the
magnitude of particle diffusivity. Masuda, et_ al  (1979) studied the velocity
and turbulence in a long 1:1 aspect ratio, negative corona, wire-plate preci-
pitator, finding turbulence to develop from the central part of the flow.
Grass  (1979), in a similar study of a 2:1 aspect  ratio  precipitator, found
wall turbulence to increase with the application  of the electrical field, and
evaluated the turbulent kinetic energy production due to  the near-wall shear
flow associated with the observed distortion of the axial velocity profile.
Data on drift velocity in a positive corona, wire-plate precipitator has been
presented by Lawless, et al_ (1981). Kawase, et^ £l (1980)  and Tedjojuwono, et
al (1981) observed increasing turbulence levels with  increasing electrical
field  in a small, axial flow, single-wire precipitator, studying  both positive
and negative corona, and presented velocity vector maps for particles at dif-
ferent current densities. Leonard, Mitchner & Self (1982b) carried out a com-
prehensive study in a small scale wire-plate precipitator for positive and
negative corona, combining LDA, hot-wire and visualization studies to deter-
mine flow patterns and turbulence levels to ascertain the role of turbulence
on.precipitator efficiency. Positive corona was found to  yield plane secondary
flows with wall-confined turbulence while non-uniformities of the negative
corona produced three-dimensional secondary flows and bulk turbulence. The
latter effect was shown to decrease with increasing mean  velocitv through the


                                     244

-------
precipitator, and to disappear for a mean velocity of  2-3 m/s at a mean
current density, which corresponds to an inverse electrical Froude number of
fg = (J'/b)/(pu*£) ^ 0.5 .  Here  b  denotes the ion mobility,  p  the gas densi-
ty and  U0  the mean gas  velocity. The inverse Froude number, being the ratio
of electrical force on the ionized gas to inertial force, was shown to be a
useful parameter for correlating the onset on turbulence at the centerline
for the case of positive  corona.

     It appears to be well established that increased turbulent mixing lowers
the parallel plate precipitator efficiency (Leonard, Mitchner & Self 1980,
and discussion by Cooperman 1982). For negative corona, wire-plate precipita-
tors, irregular corona discharge induces three-dimensional secondary flows
enhancing turbulence and  mixing. These effects are further enhanced by baffles
usually incorporated into industrial type wire-plate precipitators (Crowe,
Stock & Bernstein 1978).  Aside from the problems associated with deposition
and reentrainment, it appears that improvement of collector efficiency requi-
res further knowledge and search for possible means of control of secondary
flows and mixing. It is the purpose of the present study to provide new expe-
rimental data on distributions of mean and rms velocity of particles in a
full-width, negative corona, wire-plate precipitator.

                            EXPERIMENTAL FACILITY

     The present experimental study has been performed in a full-width, 2:1
acpect ratio, negative corona, wire-plate type laboratory precipitator facili-
ty (Figure 1). The  0.30  m  wide by  0.60 m  high by  5.0 m  long test section
has grounded aluminium side walls and perspex bottom and top. Vertical,  3 mm
diameter wire electrodes  are positioned along the centerline of the channel at
a  0.20 m  spacing, normally starting at  x = 0.35 m , where  x = 0  denotes
the inlet. The wire voltage can be adjusted over the range  - 35 kV to - 50 kV.
Electrode voltage and current to groups of discharge electrodes can be mea-
sured. In addition a portion (0.60 m long by 0.30 m high) of one side wall is
subdivided into 6 sections (each 0.10 m long by 0.30 m high) to which the
plate current can be measured individually, essentially eliminating inhomoge-
neous field effects from  bottom and top.

     Air is drawn into the test section through a contraction (1.2 by 1.2 m
inlet) equipped with two  screens (0.5 mm wire, 18 mesh per inch). The flow is
driven by the fan of a bag filter connected to the test section exit, yiel-
ding test section velocities in the range  0.1 - 2.5 m/s . Aerosol particles
(80 - 90 °C hot glycerin, atomized into 2 - 5 ym diameter drops) are feed
over the width on the inlet contraction in the midplane and serve as seeding
particles for LDA-measurements.

     Velocity and concentration measurements are made with a two-color, dual
beam, back-scatter laser  Doppler anemometer, equipped with frequency shift
(DISA 55X), amd employing a Spectra Physics 165-03 argon-ion laser, 2:1 beam
expander and a 600 mm focal length front lens to access the test section
through its perspex bottom (Figure 2). A rigid traversing bench supports the
LDA-optics and permits three-axes positioning of the measuring volume to
within  0.2 mm  accuracy. The measuring volume has a diameter of  0.16 mm
and a length of  0.17 mm  . Data is referred to x,y,z-coordinates, where axial

                                    245

-------
coordinate  x = 0  at the test section inlet, transverse coordinate  y  = 0
on the one collector plate, and vertical coordinate  z = 0   at  the bottom.

     Photomultipliers (DISA 55X08) incorporate preamplifiers and  low output
impedance for low sensitivity to external electrical noise.  Signal proces-
sing involves computer interfaced counters  (DISA 55L) operated  in combined
mode, allowing for bias-free residence time averaging of velocity and rela-
tive concentration. Averages were based on 500 samples in most  cases, after
establishing the level of accuracy based on 1000 samples. Since the LDA-
system has been only recently expanded to a full two-color operation, most
data were obtained with a one-color system. Consequently, axial and transver-
sal components of velocity were measured separately. To obtain  reliable tur-
bulence data requires careful operation of the LDA-system for optimum S/N-
ratio (greater than 15-30 dB). Comparison to hot-wire data in reference
flows and in the test section were carried out to ascertain  the accuracy of
the data. Such tests show a 2 - 5% accuracy on mean velocity and  5-10%
accuracy on rms-velocity without electrical field.

                            EXPERIMENTAL RESULTS

     The measured electrical characteristics of the precipitator  for four
types of discharge electrodes (Table 1) having axial spacing  0.20 m  is
shown in Figure 3. The low currents measured for electrode A  show an antici-
pated scatter, but are within  10%  of predictions based on numerical calcula-
tions of two-dimensional electrical fields. Electrode B  was not  found to
improve uniformity of corona discharge. The current on the rough  side (shown)
is much higher than on the smooth side (not shown), hence giving  an asymme-
trical electrical field. Electrodes  Ca  and  Ct  yield nearly  the same cur-
rent with little scatter in results, but  Ca  was found to give more stable
and uniform discharge as seen from the velocity distributions.  Because of the
low particle concentrations (about  1010  particles per m3) currents were
insensitive to variations experienced in concentrations, and Figure 3 applies
to all data presented.

                          TABLE 1.  ELECTRODE TYPES
     type
cross section
description
      A

      B
                       polished  3.0 mm  diameter brass wire.

                       M4/4 mm  threaded brass wire, ground to
                       leave two threaded edges. Flat side
                       facing collector plate.

                       Barbed wire electrode. Polished  3.0 mm
                       diameter wire fitted with  1 mm dia. by
                       12 mm  long pins spaced  100 mm  apart.
                       Pins point in axial direction.

                       Barbed wire electrode as  Ca. Pins point
                       in transversal direction towards plates.
                                     246

-------
     The uniformity of velocity profiles in the precipitator without and with
electrodes, for zero current, is shown in Figure 4. Except for the regions of
wall boundary layers and electrode wake (0.02 m downstream of the second elec-
trode) the turbulence intensity is uniform at about  2.5%  for a mean velocity
U0 = 1-1 ni/s .  This is typical of a wind tunnel of the present design. The
accuracy of this data is better than  2%  on  ux  and  5%  on  ux ms .

     With a  50 kV  negative corona voltage a comprehensive experimental pro-
gram was undertaken to determine the velocity field of particles in the up-
stream and central part of the test section where effects of bottom and top
should be small. The accuracy of the following data, because of the inherent
irregularity of negative corona discharge, is on the order of  10%  for mean
velocities of the more regular flows and about  20%  on the rms-values. The
accuracy is less on irregular flows.

     Figures 5.1 - 5.8 show the velocity vector and turbulence intensity at a
number of points in a horizontal plane between the second and third electrode
for a number of condictions. Figure captions give electrode type (Table 1),
z-coordinate for horizontal plane, mean velocity  Uo  (1 m/s for the first
six figures), and the inverse electrical Froude number  fg = (J'/b)/(pUo) ,
where  J' = JL ,  L  denoting the electrode spacing.

     Regarding mean velocities, as expected for negative corona, smooth or
rough electrodes (Types A and B) give irregularly distributed corona discharge
along wires, resulting in irregular ionic wind distortions of the axial flow
in a given plane (Figures 5.1 - 5.3). Although irregular, the recorded flow
appeared to remain unchanged during the period of measurements.

     Barbed wire electrodes with axial needles (Type Ca) give a fairly regular
flow distortion which is towards the collector plate in the horizontal plane
of needles (Figure 5.4) and away from it in the horizontal plane between
needles (Figure 5.5). This implies the existence of vertical velocity compo-
nents, which has been verified by a local check of data for the divergence of
the velocity in a horizontal plane. No measurements of vertical velocity
components has  been made so far. Yet, distributions of transverse velocity
components in vertical-planes (Figure 6.2) clearly show the existence of a
regular pattern of secondary flow, forming rolls of axial vorticity of alter-
nating sign. The precipitator height being  0.6 m  and the distance between
needles  0.1 m , suggest enough rolls that end effects from bottom and top
may be small.

     Barbed wire electrodes with transverse needles (Type Ct) give more
irregular secondary flows due to ionic wind which appears to vary in intensity
from needle to  needle (Figure 5.6). However, results indicate to secondary
flows as rolls  of axial vorticity (Figure 6.1).

     Two cases  of low mean velocity  UQ = 0.5 m/s (Figure 5.7)  and
UQ * 0.1  m/s (Figure 5.8)  are included to show pronounced effects of ionic
wind,  leading also to particle and gas flow reversals in a horizontal plane.
The results imply three-dimensional flow and only partial periodicity,
probably owing  to less regular corona discharge.
                                     247

-------
     In no case of a mean velocity of  UQ =  1 m/s  (Figures  5.1  to  5.6)  were
negative axial velocity components of particles measured. Assuming the  axial
component of the electrical field to be zero in  y,z-planes through electrodes,
the axial velocity component here equals the  gas velocity. It appears that gas
recirculation zones are suppressed at an axial velocity of   1 m/s  ,  but not at
0.1 - 0.5 m/s . Also apparent from several figures  is  the weakly pointed di-
stributions of axial gas velocity in  y,z-planes through electrodes  with re-
duced values near the wall and increased values near the center. For the data
as a whole, increasing values of  fE  yield  increasing transversal flow and
rolls of axial vorticity induced by ionic wind. In  view of  the  smallness of
the Stokesian time constant of particles, the velocity distributions of the
gas phase may be inferred by subtracting computed values of the electrical
drift velocity vector  WE  of particles from the measured particle velocity
vector. Near the collector plate  WE  is essentially perpendicular to the
plate and on the order of  0.2 m/s , for example, for  the data  of  Figure 5.6.

     Measured values of turbulence intensity  u^ rms/Uo  are 'indicated  in
Figures 5.1 - 5.6 by circles at each point according to the adjoining scale.
Of the several features, turbulence increases with  increasing   fE  ,  it  appears
to be high in areas of high ionic wind and strong shear layers, and  it  appears
to reach a maximum intensity in a region some distance from the wall. To
study further the development of turbulence, the flow  was traversed  in  axial
direction at three values of  y , starting well upstream of the first of the
type Ca  electrodes placed at  x = 750 mm . Figure  7 shows  the  increase of
"Xjrms  from the typical no-field level of  0.02 -  0.03 m/s , which  persists
to a position of about  150 mm  upstream of  the first  electrode. Although
irregular, the variation of data shows  ux rms  between electrodes to be
highest, on the average, at the intermediate plane  at  y =  220  mm  .  Note that
the rise in turbulence near the wall occurs well downstream of  the first
electrode.

     Figure 8 shows the rise in  ux ms  at  two axial  positions as the  current
density is increased from its lowest attainable value  (at - 35  kV).  Figure 9
shows  Ux rms/U0  versus mean velocity  Uo  without and withvfull  electrical
field at one location between the first two electrodes. The turbulence  inten-
sity without field is about  2%  except for velocities less than about  1 m/s,
where it is  3-5%. Note the suppression of field  induced  turbulence with
increasing mean velocity, as found also by Leonard, Mitchner &  Self  (1982b).
The suppression appears to be complete for  fE < 0.2 .

     Figure 10 shows the same data of turbulence intensity, now plotted ver-
sus  fE . Beyond a threshold, there is first a rapid increase in turbulence
with increasing  fE , followed by a nearly linear increase  which may level
off for  fE > 2
                         DISCUSSION AND CONCLUSIONS
     On essential points the present results obtained by  in-situ velocity
measurements corroborate the often indirect findings of recent  studies  of
negative- corona, wire-plate precipitator flows. In horizontal planes, the
ionic wind produces flow distortions of the type  suggested by Yamamoto  &
Velkoff (1981) for low mean velocity and smooth electrodes, but flow reversal

                                     248

-------
and recirculation was not found to exist at mean velocities at or above  1 m/s.
Here, the flow is strongly three-dimensional and most regular for barbed wire
electrodes with axial needles.  In vertical  y,z-planes, the three-dimensiona-
lity of flow is revealed as rolls of alternating axial vorticity, as suggested
by the study of Robinson (1976).  Turbulence appears to develop first in  the
bulk flow at electrodes, and between electrodes and wall (where it attains the
highest average levels), and later near the wall. While turbulence dissappears
for mean velocities above about  2 m/s (fg < 0.2)  rolls persist for type  Ca
electrodes. These results imply that turbulent mixing can be effectively
reduced by increasing the mean  velocity, but that mixing associated with
three-dimensionality of flow persists.

     For an explanation of the  influence of mean velocity on turbulence  level
we turn to the balance of turbulent kinetic energy  k = jujuj  of the gas
phase, which for steady mean flow is of the form

     u. 9k/9x. = P  + D  - e     ,                                         (1)
      1      1    iC    K.


expressing the change as contributions from production, diffusion and dissipa-
tion. Since any increase in  k   is due to increased production, we here con-
fine attention to this term. For  an ionized gas, including the electrical body
force  PeE4  into the momentum  balance, the usual derivation of Eq.(1) (see
for example Hinze 1975) then yields the production
     P.  = P.    + P.  ,  = - u!u!  3u./3x.  + p'u! E.     ,                     (2)
      k    k,s     k,b       i J    J    i     e j  j     '

where the electrical field  E-   has been assumed to be steady while velocity
u:  and charge  density per unit mass  pe  are expressed in the usual way as
the sum of mean and  fluctuating components,  uj = uj + u!  and  pe = pe + pe  .
Obvious additional electrical terms arise if fluctuations  E'  correlate with
those of  pe  and u!  .

The first term  in Eq.(2)  describes  shear production. In view of


     - uTuT = vm 3u./3x.    , hence    P,    = vm(9u./3x.)2      ,             (3)
        i j     T  j   i    '         k,s    T   j   i


where  v^  is the turbulent (eddy)  diffusivity of momentum, any distortion of
an initially uniform flow produces  velocity gradients, hence a turbulent
momentum flux in a velocity gradient, yielding production and increased turbu-
lence level. At low  mean  velocity,  the  ionic wind distortions of the mean
flow are large, giving a  significant shear production. Grass (1979) examined
his data for the increase in shear  production in the near-wall layer associa-
ted with the distortion of the  mean velocity distribution here. At high mean
velocity,  however, we  find that ionic wind distortions are reduced to ondula-
tions on the mean flow, yielding small  velocity gradients, hence small pro-
duction.
                                     249

-------
     The second term in Eq.(2) describes body force production,  originating
from a turbulent charge flux  - p^u!  in electrical field  E-  .  A  similar
term is known in thermally stratified flows, where buoyancy production, how-
ever, may be positive or negative depending on direction of heat flux. In
view of the large velocities of ions, it is not clear to what  extend body
force production is important. Assuming  Pg  to be associated  with the turbu-
lent motion, we estimate  P^ ^  by use of the Prandtl mixing length arguments
as

     P.  ,  = p'u! E. * u'£ (3p /Sx.)E. ^ V  ^p -E                           (4)
      k,b    e j  j          e   j  j    T   e


Employing the electrical field equations


     $•! = Pe/£Q   ,   J = PebE   ,   $-J = 0                              (5)


where  b  is assumed to be constant, Eq.(4) may be expressed,  for  example, as
     Present estimates of  P^ ^  and  P^ s  suggest that shear production
created by ionic wind distortion of the mean flow is principally responsible
for the increased turbulence production at large values of  fg . However,
further study is required to explain the phenomena near electrodes and diffe-
rences between positive and negative corona turbulence.

                              ACKNOWLEDGEMENTS

     The present study was supported by research grant 20138.M-654 from the
Danish Council of Scientific and Industrial Research (STVF) and by contribu-
tions to the precipitator facility from F.L.Smidth & Co A/S. The authors are
indepted to Messrs. L.Barlebo and P.Carlslund for their contributions to the
planning phase, preliminary hot-wire studies, and field calculations, and to
Dr. L.Lind for advice and many constructive discussions.

     The work described in this paper was not funded by the U.S. Environmen-
tal Protection Agency and therefore the contents do not necessarily reflect
the views of the Agency and no official endorsement should be inferred.
                                 REFERENCES

Cooperman,P.   1982  Particle Transport in Electrostatic Precipitators,
     Discussion, Atmos.Environ.,  16, 1568-1571.
                                    250

-------
Crowe,C.T., Stock,D.E. & Bernstein,S.  1978  Numerical simulation of the Fluid
     Mechanic, Electrodynamic and Particle Dynamic Interaction in Wire-and-
     Plate Electrostatic Precipitators, Model .Simul., Proc.(ISA), 9_, 149-153.

Durst,F., Melling,A.  & Whitelaw,J.H.  1981  Principles and Practice of Laser-
     Doppler Anemometry. 2nd ed., Academic Press Inc., London.

Grass,H.  1979  Zur Wirkung der Turbulenz in Elektroabscheidern,
     Staub-Reinhalt.Luft, 3g, 197-202.

Hinze,J.O.  1975  Turbulence, 2nd Ed.,  McGraw-Hill Book Comp., New York.

Jurewicz,J.T., Stock,D.E. & Crowe,C.T.   1977  Particle Velocity Measurements
     in an Electrostatic Precipitator with a Laser Velocimeter,
     AIChE Symp.Ser.,  73:165, 138-141.

Kawase.Y., Tedjojuwono,K. & Asakura,T.   1980  Application of Laser Doppler
     Velocimeter to Velocity Measurement of Charged Dust Particles in High
     Electric  Field,  Optik, 56, 283-292.

Lawless,P.A.,  Damle,A.S., Viner,A.S., Shaughnessy,E.J. & Sparks,L.E.  1981
     Laser Doppler Anemometry Measurements of Particle Velocity in a Labora-
     tory Precipitator,  3rd Symp.Transfer Util.Part.Control Techn., Orlando,
     Florida.

Leonard,G.L.,  Mitchner,M. & Self,S.A.  1980  Particle Transport in Electro-
     static Precipitators, Atmos.Environ., 14,  1289-1299.

Leonard,G.L.,  Mitchner,M. & Self,S.A.  1982a  Experimental Study of the Effect
     of Turbulent Diffusion on Precipitator Efficiency, J.Aerosol Sci., 13,
     271-284.

Leonard,G.L.,  Mitchner,M. & Self,S.A.  1982b  An Experimental Study of the
     Electrohydrodynamic Flow in Electrostatic Precipitators, (to appear in
     J.Fluid Mech.).

Masuda.S., Akutsu,K.,  Kanno,Y. & Ko,M.   1979  Motion of Small Charged Partic-
     les inside an Electrostatic Precipitator,  IEEE, IAS 14th Annu.Meet.,
     Cleveland,  Ohio,  139-145.

Robinson,M.   1976  Effects of Corona Discharge on Electrical Wind Convection
     and Eddy  Diffusion  in an Electrostatic Precipitator, Ph.D. Thesis,
     The Cooper Union  University.

Ross,J.N.   1980  Studies of an Electrostatic Precipitator using Laser Doppler
     Anemometry,  Optica  Acta, 27, 19-23.

Tedjojuwono,K.,  Kawase,Y. & Asakura,T.   1981  Studies of the Dynamic Behaviour
     of Charged Dust Particles in High  Electric Fields using Laser Doppler
     Velocimetry,  Opt.Laser Technol., 13, 187-192.

Yamamoto,T. &  Velkoff,H.R.   1981   Electrohydrodynamics in an Electrostatic
     Precipitator,  J.Fluid Mech., 108,  1-18.
                                    251

-------
u


U \" tf

i
Y
V
15
u
\
\«
u

0 1 2m
ar
=P
22


— H
23
[24]
1:  PARTICLE GENERATOR
2:  INJECTION FAN
3:  DIFFUSERS
4:  AIR INLET
5:  SCREENS
6:  INLET CONTRACTION
7:  TEST-SECTIONS
8:  OUTLET CONTRACTION
 9:  ORIFICE PLATE          17:
10:  BAG FILTER             18:
11:  FAN WITH POWER SUPPLY   19:
12:  AIR OUTLET             20:
13:  CABLE CONNECTION       21:
14:  HIGH-VOLTAGEvCABLE     22:
15:  H.VOLTAGE TRANSFORMER   23:
16:  MAIN KNIFE SWITCH      24:
                         12
CONTROL BOX FOR TRANSFORMER
GROUNDED PROTECTION FENCE
DOOR WITH PROTECTION SWITCH
WARNING LAMPS
TRAVERSING BENCH WITH LDA
LDA SIGNAL PROCESSOR
DATA TERMINAL
PRINTER
                                                                                   I
            Figure  1. Plan  view of precipitator  facility.
                                       1. BASE PLATFORM SUPPORTING
                                          3-AXES LDA TRAVERSING
                                          PLATFORM,
                                       2. AR.- LASER,
                                       3. OPTICAL UNIT (FIXED) WITH
                                          45° MIRROR (BEAM FROM X
                                          TO Y DIRECTION).
                                       H. FRONT LENS.
                                       5. MIRROR (BEAM FROM Y TO
                                          Z DIRECTION).
                                       6. MEASURING VOLUME FOR Uy
                                               VELOCITY COMPO-
      Figure 2.  Cross  section of  test section with  optical
                  traversing bench  for  positioning the x,y,z-
                  coordinates of the LDA measuring volume.
                                      252

-------
                      100
                                    40
     SO -V0 [kV]
                   Figure  3.  Mean  current  density  versus
                              negative corona voltage for
                              four  types of electrodes
                              (Table 1).
  300
V[mm]
  ISO
          NO ELECTRODES
          X = 550 MM
          (EMPTY DUCT)
                                                     0,1    u;  [m/«]
,  A - ELECTRODES
  Z = 300 MM
  X = 570 MM
  (20 MM DOWNSTREAM OF ELECTRODE A)
     0                          1,0       0              u^ [m/«]     1,0

      Figure 4. Distributions of  mean and rms  axial velocity,  J=0.
                                      253

-------
                              wall
300
 250
 200
 150
     550
                           650
                                                  750
                                                         [mm]
                                                                  NEftN VELOCITY:
                                                       TURBULENCE INTENSITY:



                                                            0-51



                                                         „   5 -  8 J



                                                         O   8 - 10 I



                                                         O  10 - 15 I



                                                         O  15 - 20 Z



                                                         O  20 - 25 I



                                                         O  25 - 30 Z



                                                        O  30 - 35 I



                                                        O  35 - HO I



                                                           HO - 15 1
  Figure
5.1 Velocity  vectors and turbulence intensity in horizontal

    plane between 2nd and  3rd electrode.

    Type A, f_ = 0.11 , 0   =1 m/s,  z =  300 mm.
               E           O
                              wall
300
250
200
150
    550
                           650
                                                  750   x [mm]
                                                                  MEAN VELOCITY:
                                                      TURBULENCE INTENSITY:


                                                           0-51


                                                        O  5-81


                                                        o  8 - 10 I


                                                        O  10 - 15 I


                                                        O  15 - 20 I


                                                        O  20 - 25 I


                                                        O  25 - 30 I


                                                        O  30 - 35 I


                                                        Q  35 - 10 I


                                                        Q  10 - 15 I



                                                        O"
  Figure 5.2  Velocity vectors and turbulence intensity in horizontal

               plane  between  2nd and 3rd electrode.
     Type  B, f  =Q436f
               E
                                       =1 m/s, z =  300 mm.
                                       254

-------
                              wall
 300 f""
 250
 I
 200
 150
                                                         MEAN VELOCITY:

                                                            1 m/»

                                                       TURBULENCE INTENSITY:

                                                             0-51

                                                             5-81

                                                         o   8 - 10 I

                                                         O  10 - 15 Z

                                                         O  15 - 20 I

                                                         O  20 - 25 I

                                                         O  25 - 30 I

                                                         O  30 - 35 J

                                                         Q  35 - 10 I

                                                            10 - 15 I
     550

 Figure
                           650
                                                  750   x [mm]
5.3 Velocity  vectors and turbulence intensity in  horizontal
    plane between 2nd and  3rd electrode.
    Type B, f
                        E
0..36,  U  =  1  m/s, z  = 270 mm.
        o
                              wall
300 iiiniut
250


?
200
150
    550
                           650
                                                  750
                                                        [mm]
                                                         MEAN VELOCITY:
                                                            1 m/»

                                                       TURBULENCE INTENSITY:

                                                            0-51

                                                         o  5-81

                                                         o  8 - 10 I

                                                         O  10 - 15 I

                                                         O  15 - 20 I

                                                         O  20 - 25 I

                                                         O  25 - 30 I

                                                         O  30 - 35 I

                                                         Q  35 - 10 I

                                                        Q  tO - 15 I

                                                        O5'-
 Figure  5.4 Velocity vectors  and turbulence intensity in horizontal
             plane between 2nd and  3rd electrode.
             Type C  , f  =0.54, U   =  1 m/s,  z = 300 mm.
                    a.    E          o
                                    255

-------
                              wall
300 <"""<
250
200
150
    550
                           650
                                                  750
                                    MEAN VELOCITY:


                                       1 m/»



                                  TURBULENCE INTENSITY:



                                       0-51



                                    ,  5- 81



                                    o  8 - 10 I



                                    O  10 - 15 I



                                    O  15 - 20 I



                                    O  20 - 25 I



                                    O  25 - 30 I



                                    O  30 - 35 I



                                    O  35 - 10 I



                                       tO - 15 I
  Figure 5.5 Velocity vectors and turbulence  intensity in horizontal

              plane between 2nd  and 3rd  electrode.
              Type C , £
                     a   £i
0.54,  U  = 1 m/s, z =  250 mm.
        o
                              wall
300
                                                                 MEAN VELOCITY:
250 +
200
 150
     550
                           650
                                                  750   * [mm]
                                  TURBULENCE INTENSITY:


                                        0 - 5 t


                                    o   5- 81


                                    o   8 - 10 I


                                    O  10 - 15 I


                                    O  15 - 20 I


                                    O  20 - 25 I


                                    O  25 - 30 I


                                    O  30 - 35 I


                                    O  35 - 1(0 I


                                    Q 10 - H5 I



                                    O5'-
   Figure  5.G Velocity vectors and turbulence intensity  in horizontal

               plane between 2nd and 3rd electrode.
               Type  Cfc,
 0..58, U   = 1 m/s,  z =  300 mm.
        o
                                       256

-------
       ^—f-
        t
              X
Figure 5.7 Velocity vectors and turbulence intensity in horizontal

           plane between 2nd and 3rd electrode.

           Type C , f  = 2.23, u  =0.5 m/s, z = 300 mm.
                 t   E          O
                                  V
                                                  X



                                                  X
                                             X



                                             X







                                             X
Figure 5.8 Velocity vectors and turbulence intensity in horizontal

           plane between 2nd and 3rd electrode.

           Type C , f  = 58 ,  U  =0.1 m/s, z = 300 mm.
                 t   E         O
                                257

-------
               Y=170 (MM)
              230
         300 -
         Z
        (MM)
         200
                            -i	1
               0
0.2
                        290
                                J	L.

0
0.2  UY (M/S)   0
0.2
Figure 6.1 Distribution  of transversal velocity component Uy in three
           vertical planes, y = 170, 230 and 290 mm  at x = 575 mm.
           U  =1.0 m/s,  Type A electrode.
            Y=160  (MM)
              220
                        290
     300
     Z
    (MM)
     200
                                    J	L.
                                    0
                UY  (M/S) 0
                                          0.4
Figure 6.2 Distribution of transversal velocity component uy in three
           vertical planes, y = 160, 220 and  290 mm at x = 575 mm.
           U   =1.0 m/s, Type C  electrode.
                                    258

-------
 0.8
 0.4
           600
                        600    X    WOO mm
Figure 7. UY „__ versus x at y =  190,
          A f ±.1I1O
                                  220  and 250 mm.
            .
         Type C  electrodes at x = 750 and 950 mm.
              3.
0.8
 0.4
                              i
i
  I      I      I
/=  190 fain
z =  300            X= 950 rnm^



                                650
            /--i"
           I	I	I	I
                             0.4    J     0.6  mA/m*
                       _L	i
                      0.2
                                   A   °-4
                                     at two points

         between first and second electrode, Type C  .
                                              3.
Figure 8. ux?rms versus  f
                     259

-------
    60
>OJ


40
                       +
                       o
                               •50 kV  (0.65
                                x* 850 mm

                                y- 220

                                z- 300
Figure 9.  Turbulence  intensity  u     /U  versus U  with
                               x.rms  o         o
          and without electrical  field at point Between

          first and second electrode, Type C  .
                                           Si
   60
   40
                                              o _
Figure 10. Turbulence intensity versus  f   =  (J/b)/(pU2)

          (data of Figure 9.)                        °
                          260

-------
    THE  EFFECT OF TURBULENCE ON  ELECTROSTATIC  PRECIPITATOR  PERFORMANCE

                                 D.  E.  Stock
                   Department  of Mechanical  Engineering
                        Washington  State  University
                          Pullman, WA 99164 - 2920
                                 ABSTRACT

     The  turbulent  character of the gas flow found in an  electrostatic  pre-
cipitator (ESP)  affects the motion of  the  particles (dust) and,  therefore,
the unit's performance  through two mechanisms.   The  mean velocity  profile
found in  an ESP  conveys  the  particles both  longitudinally  through  the preci-
pitator  and  in the transverse  direction.   Particle motion is also  strongly
affected  by  turbulent diffusion.  The  magnitude  of the turbulent diffusion
is expressed  through a  particle  turbulent  diffusivity  which depends  on  the
turbulent character of the gas, particle  size,  and the crossing  trajectories
effect.
     A particle  diffusion equation  is  developed  and the  coupling with  the
gas flow  field  and electric field  is  discussed.   Finally,  techniques  for
estimating the particle  diffusivity are presented.
       The work described  in this paper was  not  funded  by  the  U.  S.
       Environmental  Protection Agency and  therefore the  contents
       do  not  necessarily reflect  the views of  the  Agency and  no
       official endorsement should  be  inferred.
                                    261

-------
                                INTRODUCTION
     An  ESP  is  an  extremely   complex   mechanical   system.   Beyond  the
complexity of the physical unit with  its  requirements for correct plate and
wire spacing, for minimum gas  leakage  around the plates, and for controlled
rapping; the physical phenomena controlling the particle motion  is a coupled
system which  is difficult  to model.   The particle  motion  depends  on the
interaction between  particles, electric  field,  and  the  turbulent  gas flow
field.

     When  trying  to  understand  and   to  predict  the operation of  such  a
complex  system, one  would like  to isolate  the  various phenomena  and then
carry  out  some  form of  parametric study  on  each  variable separately.  The
success of this technique strongly depends on the insight of  the modeler.

     The Deutsch  Equation  [1], developed  in  1922 and widely used  to scale
ESP until  a  few years  ago, is an  example of a model  which  did not include
all  the  physical  phenomena.   It  neglects  the change  in space  charge and
assumes  infinite   mixing.    The  use  of  this  equation  has   limited  the
investigation of the importance of turbulent  mixing and wide plate spacing.
Only recently have these topics started to  attract interest.   In this paper
the equations governing particle motion is presented along with a discussion
of  coupling  between the  electric  field,  the  gas   flow field,  and  the
diffusion coefficient for the particles.

                                 BACKGROUND
     Considering  the flow  in  a wire  and  plate precipitator  to  be  two-
dimensional and treating the particles as a dilute, continuous phase, we can
write the differential equation for particle continuity. as
where both the particle velocity,  Up and V , and particle concentration are
instantaneous values which depend on both location and time. If the equation
is time-averaged, we obtain
Now, assuming  the  longitudinal  time-averaged particle  velocity is equal to
the gas velocity T_T,_and transverse particle  velocity  is equal  to the sum of
the gas  velocity,  V, and the particle  velocity relative to  the gas due to
the electrostatic forces on the particle, Y^ we can write
Using the gas phase continuity equation for incompressible flow
                                     262

-------
            77 3C  , - 8C  . - 8C  , T 8w _
            u9^+v9y  + w9y+cay-
For a  simple  velocity field,  the  two  velocity-concentration  correlation
terms  on the right hand  side can be modeled  with  the Boussinesq approxima-
tion,  Pick's Law gradient jiiffusimi model, to give _        _
               TT  3C  +ir^ ,-rr3C   T 9W _   _a- 8C    3.
               U   ~  +  V    + T   + C- +   °    +
Now, neglecting  changes  in the particle diffusion coefficient, D, neglecting
gradient in the migration velocity,  W,  and agreeing  all  symbols  stand for
average values,  we  get

                     ii 9C j. w aC  . ii 9C
                     U8x-+V9y-+W3y-=

Boundary  conditions  suitable  for  an electrostatic  precipitator  include
assuming no net  transverse mass flux at the center! ine,
                        •\r
                      D -g- - WC  = 0 at the centerline                  (8)

and assuming that very near the collecting wall turbulent dispersion effects
on the particle  vanish giving

                           |£ =  0  at the wall.                        (9)

The boundary condition at the wall assumes that in the viscous sublayer near
the wall,  the  particle  velocity due  to  electrostatic force  dominates all
other forces acting on the particle.   In  the  absence of the electric field,
this boundary condition could  no longer  be  used.  Under  such  conditions,
particle stopping distance in the sublayer must be considered.

     Two assumptions  used  in  developing equation  (7) need to be examined in
more detail.  Modeling  the  concentration-velocity  correlation  with  Pick's
Laws assumes that particle dispersion  is governed by a single characteristic
length scale and a single turbulent velocity  scale  [2].   These assumptions
are met for a smooth  wall precipitator with a well -controlled inlet.  Gradi-
ent diffusion is often used for  more  complex  geometries  since the  more cor-
rect model,  2nd  order turbulence  model [3], is much more complex to develop.
The validity of  the second assumption, neglecting the gradient in the migra-
tion velocity,  depends  on  the gradient in the  electric  field normal  to the
plate and on the gradient of  the  concentration  profile normal to the plate.
Large particle concentration gradients toward the wall would cause the elec-
tric field  to increase slightly and the migration velocity and particle con-
centration  to increase.  Under such conditions, this term could be important.
We have neglected  this  term  in past work, but  will  investigate  its import-
ance in future studies.

     Many of the past studies [4,5,6] have assumed uniform longitudinal vel-
ocity,  U, constant migration velocity, W, and constant  particle  diffusion
coefficient,  D.  Using these  assumptions,  insight into the operation of ESP
has been  obtained,  but also some  important aspects have been neglected.
                                    263

-------
                              MODELING COUPLING
     To use equation  (7),  the  gas velocity must be  known  and the motion of
the particle due  to  the electric  field  must also be  known.   We assume the
particle velocity  due to  the  electric  field is  equal  to  the equilibrium
velocity attained  by  a  particle  charged  to  saturation in a  field equal to
the local  field.  This is given by
                                    E Ende
                                u = _LJ1_°_ .                            (10)
                                       V
where E  is the charging field  and  E  is the local  field.   Since particle
historycis unknown E   is set equal   'to E .   The justification for assuming
local equilibrium ana for neglecting forces other than drag on the particles
is given by Eschbach [7].

     To use equation  (10)  the  local electric  field  must  be  determined. '' We
assume the field  is  uniform except for the influence  of  the particle space
charge.  The  details of  the modeling of the electric  field  are  given by
Eschbach and Stock [8].

                            MEAN GAS FLOW EFFECTS


     The diffusion equation (7), which governs  particle  concentration, re-
quires the gas  phase  mean  velocity, U and V,  to  be  specified.   Most of the
simpler numerical models [4,6,8] assume the axial velocity is uniform in the
transverse direction  and constant in  the  axial direction.  This simplifica-
tion allows major  trends to studies.   More complete, and therefore complex,
models [9] take into account variation  in the U and  V  velocity, but often
provide more information than can be easily assimilated.

     In the initial  region  of  the channel  formed by two plates the velocity
profile goes through  a  development stage  from a near uniform  profile  to a
rounded profile typical  of  low  Reynolds  number turbulent flow.  This devel-
opment occurs over the  initial  half of  the  stage with its  length and  flow
character depending  on  the initial velocity  profile and  turbulence charac-
teristic.  Given  the initial  conditions, numerical  codes  can  predict the
flow field development.  Measurements  need to be made at the inlet plane of
full scale precipitators to determine velocity profile, turbulence level and
length scale.   In  the initial  region,  the transverse gas velocity is toward
the centerline with typical velocities of 2% of the mean flow but 50% of the
particle migration velocity, W.   The effect  of  the  transverse gas velocity
must be evaluated for a particular  installation.

     Most  ESP  are built with   some form  of  baffles  or  stiffeners.   These
stiffeners can  greatly change the gas  flow field.   We have modeled one  par-
ticular geometry  [10].   Figures  1-4 show the concentration profiles for the
geometry given  in  Table  1.   Figure 1 shows how  the  smooth plate  configura-
tion is more efficient  than one with stiffeners.  It  is also interesting to
find that the stiffeners cause  the concentration to decay in an  exponential
manner which  explains why  many  commercial ESP's follow  the trend given by

                                      264

-------
the Deutsch  Equation [1] with a modified  particle  migration  velocity.   Fig-
ures 2-4  show  the  transverse  concentration profile.  The stiffeners make the
concentration  more nearly uniform.   The shape and magnitude of concentration
profile  would  probably  change with  changes  in  the turbulence  of the  gas
flow.

                         TURBULENT  GAS FLOW EFFECTS


     The  turbulent  characteristics  of the  gas  flow affect  particle  motion
through  its  influence  on mean flow  as  discussed  above and through the  affect
of turbulent dispersion.   Turbulent dispersion enters the diffusion equation
(7) through  the particle diffusion  coefficient,  D .  The diffusion coeffici-
ent enters  the equation when the gradient  diffusion  assumption  is made.   In
simple flow fields  where only one  length  scale  and velocity scale  exist,
this assumption is  known  to  work well.  Grid generated  turbulent  flow  in a
wind tunnel  is one  case  where gradient diffusion  can  be used.    In a  well-
conditioned  tunnel, the flow is isotropic  and decays in turbulent intensity
as the flow  moves  through the tunnel.   A smooth wall ESP with a  well-condi-
tioned inlet  as found in the  last stage  of a  multiple  stage  precipitator
comes close  to meeting these  requirements  except the passage  is  not square.

     Particle  dispersion is  determined by  the gas phase turbulence, the par-
ticle inertia, and by the crossing trajectory effect.   The crossing trajec-
tory effect  [11] refers  to a  particle which has a velocity  relative  to  the
fluid phase  (due to gravity  or  electric field)   and  moves from one eddy  to
another  at  a rate  faster than the average  eddy  decay rate.   Therefore,  par-
ticles with  a  drift velocity  lose  velocity correlation  faster than a  fluid
point and diffuse  less  than  a particle which follows  the  flow.   Wells  [12]
built a  wind tunnel  which, through  the use  of an electric  field,  could  con-
trol the drift velocity  of  charged  particles.    He  measured the  particle
drift velocity and mean  squared dispersion with a laser Doppler  anemometer
for two  particle sizes, fyim and 5^/m,  glass beads.   Figure 5 shows the  par-
ticle dispersion coeffients measured  in this study.   Dispersion  is seen  to
depend only  on the drift velocity,  not on  particle size.  With no drift vel-
ocity both  particles have the same  diffusion as a fluid point.   These  mea-
surements were taken in  a 35  x 35 cm tunnel with a  2.5  cm grid.  Measurements
were taken  between  .5  and 2.2 m from the grid  with a gas  velocity of  6.5
m/s.  The gas  turbulence intensity  fell from 4%  to 1.5% in the test section.
These results  support  the work by  Csanady  [13] where the effect  of crossing
trajectories is given  by

                      D
where  3 is  a parameter  describing the  relationship  between the  Lagrangian
and Eulerian  integral  scales and /IT7"is the gas RMS velocity  and V.  is the
particle drift velocity. This  relation  shows  that crossing trajectories will
become  important  when the particle drift  velocity  is comparable  to  the gas
RMS velocity.   For an ESP with  typical  drift velocities of 8  cm/s  and mean
gas velocities  of 2 m/s  at turbulence levels above 4%  crossing trajectories

                                     265

-------
effects can  probably be  neglected and  the particle  diffusion coefficient
assumed equal to the gas turbulent diffusivity. For lower turbulence levels,
as might be found in the last few stages of a smooth wall ESP, crossing tra-
jectories effects would reduce the particle diffusion coefficient below that
of the gas and result in higher collection efficiencies than expected.

                           SUMMARY AND DISCUSSION
     Numerical models can be used  to  help  understand ESP operation, but the
models must include the important physical  phenomena.  The turbulent charac-
ter of the gas flow is one such physical phenomena.  If we know the physical
geometry of the precipitator and the velocity and turbulence distribution of
the gas  entering  each stage of  a  precipitator, then the mean  gas flow and
gas turbulent  diffusivity  can  be  calculated  [14].  Using these  results  in
the particle  continuity  equation (7), the concentration  and efficiency can
be calculated.   If only small  particles  are  considered, then  the particle
diffusion coefficient can be set equal  to  the gas diffusion coefficient ex-
cept for modification to account for the crossing trajectories effect.  Con-
sidering  only  particle  collection  on the  plates, 'reducing  the  turbulence
level  and using smooth walls will improve collection efficiency.

                                 REFERENCES
1.   Deutsch, W., Bewegung und Ladung der Eteletrizitatstrager.  Anna!en der
     Physik, IV, Folge 68, pp.335-344, 1922.

2.   Tennekes, H. and Lumley, J.T., A First Course in Turbulence, MIT Press,
     1972.

3.   Lumley, J.T.,  Computational  modeling of turbulent  flows,   in Advances
     in Applied Mechanics. Vol.18, Academic Press, pp.124-176, 1978.

4.   Williams, J.C.  and Jackson,  R.,  The motion  of  solid particles  in  an
     electrostatic  precipitator.   Proceedings  of  the  Symposium  on Interac-
     tion  of  Fluids  and  Particles,  Institute   of  Chemical  Engineering,
     pp.282-288, 1962.

5.   Feldman, P.L.,  Kumar, K.S. and  Cooperman,  G.P.,  Turbulent diffusion in
     electrostatic  precipitators.   AIChE,  Symposium  Series,  Atmospheric
     Emissions  and  Energy Source Pollution,  Vol.73, No.1651,  pp.120-130,
     1977.

6.   Leonard,  G.L.  Mitchner,  M.  and  Self,  S.A.,   Particle  transport  in
     electrostic  precipitators,  Atmospheric  Environment,  Vol.14,  No.11,
     pp.1289-1299.

7.   Eschbach, E.J., Numerical prediction of electrostatic precipitator per-
     formance, M.S. Thesis, Washington State University, 1982.
                                      266

-------
8.    Eschbach,  E.J.  and Stock, D.E.,  Optimization  of collection efficiency
     by  varying  plate  spacing  within  an  electrostatic  precipitator.
     Environmental Inter., Vol.6, pp.177-180, 1981.
9.    Stock,  D.E.  and  Crowe,  C.T.,  The  effect of  electrodynamic secondary
     flow  on the  performance  of electrostatic  precipitators, Proceedings.
     1974  Heat  Transfer and Fluid Mechanics  Institute,  Stanford University
     Press,  1974.

10.   Stock,  D.E. and Eschbach, E.J., "Prediction of electrostatic precipita-
     tor performance,  Proceedings, Gas Bourne Particles, C76/81 Institute of
     Mechanical Engineering, London, 1981.

11.   Yudine,  M.I.,  Physical   considerations   on  heavy-particle  diffusion,
     Advances in Geology, Vol.6, Academic Press, 1959, pp.185-191.

12.   Wells,  M.R.,  The  effects  of crossing trajectories  on  the diffusion of
     particles  in  a  turbulent  fluid,  Ph.D.  dissertation,  Washington State
     University, 1982.

13.   Csanady,  G.T.,  Turbulent  diffusion  of heavy  particles  in  the  atmos-
     phere,  Journal of Atmospheric Science, Vo.20, 1963, pp.201-108.

14.   Barriga, A.N.,  Effect  of  protruding  normal  dividers  on  the turbulent
     flow  between  plane boundaries,  Ph.D.  Dissertation,  Washington State
     University, 1978.
                                    Table 1

                             Operating Conditions

              Wire to Plate Spacing                0.10 m

              Stiffener Height                      .028 m

              Distance between Stiffeners          0.40 m

              Inlet Gas Velocity                    .28 m/sec
                                                            _2
              Inlet Particle Concentration         4.57 x 10  kg/m3

              Applied Potential                    8 x 104 volts

              Particle Diameter                    2 x 10" m
                                    267

-------
                WITH STIFFENERS
   ICT8
      0   0.5   1.0   1,5  2.0
       AXIAL DISTANCE, m
Figure 1.  Net particle flux at various
           longitudinal locations
                    268

-------
c/c
 1.0

14
 0.5
     0.0
                 WITH STIFFENERS
                   ,*•

                   SMOOTH WALL
      0.0
           0.05
0.10
       DISTANCE  FROM WALL, m
Figure 2.
      Concentration at .375 m from
       inlet normalized by concen-
       tration at node 14
                                        2.0
     10

 C/C,4

    Q5
                                         1.5
              WITH
           STIFFENERS
                                C/C,4
             SMOOTH
              WALL
                                     1.0
                                    0.5
                       WITH  STIFFENERS
                                               SMOOTH WALL
      0        0.5       1.0
     DISTANCE FROM WALL.m

Figure 3.   Concentration  at 1.175 m
           from inlet normalized by
           concentration at node 14
                                     0.0
                                       0       0.05       OJO
                                     DISTANCE FROM WALL.m

                                Figure 4.  Concentration at 1.975 m
                                           from  inlet normalized by
                                           concentration at node 14
                                 269

-------
ro
«xl
o
     (cm2/s)
                            25
50             75

    vd(cm/s)
100
125
             Figure 5.  Particle dispersion coefficients from Y2 measurements
                            ,  57 ym particles;     , 5 ym particles.

-------
                   FACTORS LEADING  TO  ELECTRICAL BREAKDOWN
             OF RESISTIVE DUST LAYERS  AND  SUSTAINED BACK CORONA

           By:  Phil A.  Lawless
                Research Triangle Institute
                P. 0.  Box 12194
                Research Triangle Park,  North Carolina  27709

                Leslie E. Sparks
                Industrial Environmental Research Laboratory
                U. S.  Environmental Protection Agency
                Research Triangle Park,  North Carolina  27711
                                   ABSTRACT

     Recent theoretical work modeling  the  resistive dust layer has shown con-
siderable enhancement of  the electric  field at  the points of contact between
particles.  These theories  are  examined  with regard to evaluating the con-
ditions  leading to electrical breakdown  and sustained back corona discharge
via the  Townsend avalanche  mechanism.  Among other factors, the particle size
distribution is shown to  have a significant effect on the internal field in
the layer.  The resulting breakdown characteristics of the resistive layer
can be used to obtain the proper operating current density for the precipita-
tor and  to evaluate the effects on  collection of  excursions into the back
corona regime.

     This paper has been  reviewed in accordance with the U. S. Environmental
Protection Agency's peer  and administrative review policies and approved for
presentation and publication.
                                    271

-------
                                 INTRODUCTION

     Back corona is an electrical breakdown of a resistive dust layer which
results in the injection of ions (of the polarity opposite to that  from the
"charging" electrode) into the gas space of a precipitator.  The presence of
ions of both polarities leads to a reduction of charge on the suspended par-
ticles and to a lowering of the operating voltage of the precipitator.  Both
effects are deleterious to collection performance, with the severity of deg-
radation in proportion to the resistivity of the material.(1)

     The problem of determining the effects of back corona has two  basic
parts:  describing the conditions which lead to breakdown of the dust layer
and evaluation of the effects of back corona on particle charge and operat-
ing voltage.  An earlier paper (2) described the sustained breakdown, and
methods for calculating the effects on particle charge and operating voltage
are available.(3)  This work is aimed at improving the estimation of layer
breakdown conditions and extending the calculation of degradation effects to
more realistic conditions than previously described.

     The previous model of back corona discharge (2), called the "pore break-
down model," assumed that the electric field existing in a resistive dust
layer could sustain a continuous generation of ions in pores in the layer by
means of the Townsend avalanche mechanism.  The resistivity of the  layer was
assumed to be uniform and ohmic in character, and, based on those assumptions,
the values of the electric field in the dust required to sustain back corona
were several times higher than experienced in practice.  The equation relat-
ing electric field, E; current density, j; and resistivity, p; is:

                                 E = jp.                                (1)

     The interaction of the ions, the electric field in the layer,  and
the electric field in the pore is shown in Figure 1.

     In recent years, research on the nature of conduction paths in particle
layers (4,5,6,7) has emphasized the local enhancement of electric fields at
the points of contact between particles.  The results of these investigations
show that all phenomena which depend on electric field in nonlinear fashions
will have very different behavior in a particulate layer from a homogeneous
layer.  Since the Townsend avalanche mechanism is a highly nonlinear function
of the electric field, the particulate nature of fly ash layers is  expected
to play a very important role in the pore breakdown model.

                       TOWNSEND GAS DISCHARGE IN A PORE

     The Townsend avalanche ionization of a gas is a process caused by free
electrons.(8)  Because electrons are so much less massive than gas  molecules,
they lose very little energy in collisions with them.  In the presence of an
electric field, electrons may gain more energy between collisions than is
lost during collisions.  If the electrons gain sufficient energy, there is a
definite probability for ionizing a gas molecule during a collision.  When
ionization occurs, two electrons (the original one and the ionized  one)
                                     272

-------
                           f
            Figure 1.  Evaluation of electric field, Ep, in a pore.
are available for further ionizing collisions.  In this way, a single free
electron can produce a population of several million in a distance of a few
millimeters.  The number of new electrons created by a single electron tra-
versing a 1-cm path is called the first Townsend coefficient, a.

     A factor opposing the growth of the free electron population is electron
attachment,  a process by which a free electron and a molecule form a stable
negative ion, incapable of further ionization.  The attachment coefficient, T],
is the average number of free electrons forming negative ions in 1 cm of
travel .
     Because of the rapid growth of the number of free electrons, the total
gain of electrons,  G,  is an exponential function of the distance, x, over
which the avalanche occurs :
                               exp[(a - n)x] -
                                    a - n
(2)
     Both a  and  T]  depend on the value of electric field moving the electrons.
When the "effective  Townsend coefficient," a - r\, is negative, no multiplica-
tion can occur.  When a - n is positive, the avalanche will proceed.  For
air at standard  conditions, a and T\ are equal at an electric field of
24 kV/cm.
                                     273

-------
     As indicated by Equation 2 , a given value  for  the  gain can be achieved
over any distance, x, as long as the effective  Townsend coefficient is  large
enough.  This is true for distances larger than about  10 ym,  but below  that
length (about 100 mean free path lengths) , the  finite probability the electron
has for ionizing a given molecule reduces the effective Townsend coefficient
and requires a higher electric  field to produce the same gain.   This phenom-
enon is the basis for the minimum in the Paschen discharge  curve. (8) As a
result, the Townsend theory should only be applied  over macroscopic distances
of the order of 10 ym or larger.

     The Townsend mechanism can provide an enormous multiplication from a
single electron; but if no further electrons are supplied,  only a single
burst of ionization will occur.  Continuous ionization, such  as called  for in
the pore breakdown model, requires a steady supply  of initiating electrons.
These must originate from some  sort of feedback mechanism.

     The usual feedback mechanisms operational  in a corona  discharge are
photoionization and impact ionization.  Photoionization occurs  because  some
of the free electrons or negative ions may recombine. with positive ions, re-
leasing high energy photons in  the process.  The photons travel in all  di-
rections and may ionize neutral molecules to provide a  new  source of elec-
trons.  Only those photons which travel upstream of the electron flow direc-
tion will be useful in renewing the ionization  process, although all photons
will contribute to the total ionization.  Impact ionization occurs when the
positive ions gain enough kinetic energy from the electric  field to cause emis-
sion of secondary electrons from any surface they may strike.   Impact ioniza-
tion is only effective for renewing a negative  avalanche, one in which  the
electrons move away from the surface (e.g., a negative  corona wire).

     Masuda (9) has suggested that a third mechanism, negative  ion detachment,
may be responsible for providing the free electrons-, in  a dust layer. In this
mechanism, the presence of the  large electric fields near the dust pore de-
stabilizes approaching negative ions, liberating an electron  from each  one.
Within the confines of a dust pore, it is reasonable to expect  contributions
from all these sources .

     If the fraction of electrons (or ions) in  the  avalanche  which are  effec-
tive in renewing the process is denoted by y> the gain  equation is modified
to become:

                       G =  _ exp[(a - n)x]
                            1 - y {exp[(ot - n)x] -  1}

As the feedback factor, y> approaches 1, the value  of  (a -  r|)  needed  for a
given gain decreases.  This can lead to a significant  reduction in the value
of electric field required to maintain the ionization,  but  in  no case can the
field decrease below the value for which (a - n) =  0.   A plot  of the  effect
of feedback on the electric field required to produce  a gain of 108 is shown
in Figure 2.  The gain of 108 is chosen because of  estimates that there are
about 108 ions in a corona Trichel pulse. (8)
                                     274

-------
                140
                120
                100
                 80
                 60
                 40
                 20
                          0.9
                   0.1   0.2     0.5    1.0    2.0     5.0   10.0
                                Layer Thickness (mm)
         Figure  2.   Breakdown field  as  a function of layer thickness
                     for various feedback factors.


     When negative  corona  is  used in  a precipitator, the positive ions pro-
duced by back  corona  can emerge from  the dust layer at a high enough density
to form a streamer.   A streamer is a  self-propagating Townsend avalanche that
advances by photoionization.   Streamers  can lead to sparks by creating an
ionized channel between the corona electrode and ground plate.  Streamers
also can charge dust  particles positive to a significant degree, resulting
in their moving away  from  the collecting plate.(9)

     Because streamer formation is a  positive ion phenomenon, it cannot occur
as a manifestation  of back corona with positive corona.   Also, because there
is a certain density  of positive ions required to achieve the high local
electric fields,  low  current  density  back corona does not form streamers.
High resistivity  materials reach breakdown conditions at low current densi-
ties and fall  into  this category.   The remainder of this investigation is
concerned with only low current density back corona, without the formation of
positive streamers.

                      PARTICLE SIZE DISTRIBUTION EFFECTS

     Theories  of  the  conduction of electric currents through an array of
spherical particles all assume that there is a very strong constriction of
the current at the  points  of  contact  between spheres.(4,5,6,7)  In order to
                                    275

-------
maintain the same current through all parts of the  sphere,  the electric field
in the contact region must be much larger than the  average  field over the di-
ameter of the sphere.

     The presence of the intense field greatly enhances  the cohesiveness of
the particulate layer, modifies its field dependence,  and produces a resis-
tivity which is mildly field dependent.  Using the  theory developed by  •
Dietz (6) because of its simplicity, the expression for  the resistivity, p,
of a layer of spheres is:

                            p = Ps R (VE0)°'2>                        <*)

where ps is the surface resistivity of the material, R is the radius of the
sphere, % is the field at the contact point, and E0 is  the average applied
field.  Although the model is derived on the basis  of  surface resistivity
alone, monodisperse spheres, and an undetermined Eta, the trends indicated in
Equation 3 do hold fairly well for precipitated fly ash.  More complex'theo-
retical results (7) modify the dependences slightly and  distinguish between
volume and surface conduction but are not necessary for  this work.

     The principal conclusions from Equation 4 are  that  the effective resis-
tivity is proportional to particle radius and decreases  slightly with applied
field.  The former result indicates the potential for  size-dependent effects
in a polydisperse mixture of particles.  The latter result  should be used to
correct the effective resistivity to conditions other  than  those under which
it was measured.

     There is some experimental evidence for this size dependence.  Figure 3
shows the size-dependent resistivity of glass spheres  (10) , for several condi-
tions of temperature and humidity, and resistivity  measurements on fly ash
classified into narrow size ranges.(11)  Both plots are  consistent with the
linear dependence of resistivity on particle diameter  indicated by Equation 4.

     Evaluating the effects of size distribution on the  Townsend avalanche is
done by comparing a uniform average field producing a  certain gain with the
average field producing the same gain with a nonuniform  set of field values.
This situation is illustrated in Figure 4, showing  the effect of including a
single large particle in an otherwise monodisperse  array.

     The average field in both arrays is determined by the  current flowing
down each chain of particles.  For the monodisperse chain,  it is uniform from
particle to particle  (but not along the surface of  each  particle).  For the
second chain, the same current must flow through the large  particle as
through the smaller ones, leading to a higher field across  it because of its
higher effective resistivity  (Equation 4).  This region  of  higher electric
field enhances the gain for two reasons:   (a - r|) increases with field
strength, and the increased multiplication is effective  over a larger than
normal distance.  Because of the local field enhancement, the polydisperee
particle chain can produce the same total gain at a lower average field than
the monodisperse chain.
                                     276

-------
    1010
Figure 3.
         10            100          1,000
          Partfcto Dtemtw (ion)
Resistivity variation with particle diameter
(o - Reference  3;  ° - Reference 4).
  Figure 4.   Effect of polydispersity on  local field.

                         277

-------
     Two limiting factors have been assumed in evaluating  this model.  The
first is that only the average field across a particle  can be effective in
maintaining the Townsend discharge in a pore next to  the particle.   Even
though the electric fields in the contact region between particles may be
orders of magnitude larger than the average value, the  close separation of
the particles effectively encloses any electrons and  leaves room  for few ion-
izing collisions.  Thus, even though ionization may occur  in the  gap, it is
not capable of sustaining the discharges characteristic of back corona.
Some of the initial free electrons may come from microdischarges  in  the gaps,
but unless the average field is high enough to sustain  the avalanche, the
ionization will not produce back corona.

     The other limit is imposed on the maximum enhancement that can  be assoc-
iated with a large particle.  Because Equation 3 was  derived on the  assump-
tion of only one contact area in each hemisphere of the particle, it cannot
be expected to be correct if there are multiple contact points.  Multiple con-
tacts are the norm in a randomly packed polydisperse  layer.  Therefore, a
limiting function was invoked to keep the maximum electric field  to  some mul-
tiple of the average field, and the effects of this limitation were  tested.

     The previously used values for a and r| (2) were  fitted graphically, re-
sulting in the following power law expressions for a  -  ri:

                        (a - n)/P = B (E/P - 31.3)C,                       (5)

where:  B = C = 0, (E/P < 31.3);
        B = 7.47-10 *, C = 1.30, (31.3 < E/P < 39.6);
        B = 2.54'10~", C = 1.81, (39.6 < E/P);

E is the electric field (V/cm); and P is the pressure (torr).

     For the size distribution o.f particles in the layer,  lognormal  parame-
ters were used:  the mass median diameter, MMD; and the geometric standard
deviation, a.  With lognormal distributions, integrals  of  power law  expres-
sions over the size distribution can be evaluated analytically.  This is use-
ful for finding the most probable particle diameter from the MMD, but the
nature of Equations 2 and 5 prevents easy analysis.   Also, the limitation on
the maximum field cannot be expressed as a power in the particle  diameter.
For these reasons, numerical evaluation of the local'  electric fields and
equivalent gains was1, required.

     Four parameters were varied in the evaluation:   MMD,  a, E/P, and the
upper field limit, E^^.  Of these, E/P produced little change in the enhance-
ment of the electric field from values of 40 to 400 V/cm-torr; therefore, a
value of 90 V/cm-torr was chosen to represent the field that might te ex-
pected to break down a thin dust layer.(2)

     Since the upper field limit plays a very important role in determining
the enhancement, Figure 5 shows the results for an MMD  of  10 ym and  a 0
of 3.  It is evident that the enhancement grows less  rapidly than the maxi-
mum field.  This is due to the nature of the size distribution:   the higher


                                     278

-------
                                              10
20
          Figure 5.   Effect  of maximum local field on enhancement.

local fields  are associated  with larger particles, but the number of large
particles  decreases  rapidly  for sizes larger than the HMD.

     For examining the remaining relationships,  the maximum field was set to
a value 5  times  the  average  field,  yielding an enhancement of about 2 at
this HMD and  a.   This choice may seem large, but it yields observable varia-
tions in the  effects of the  other parameters.  From considerations of geom-
etry, it can  be  taken as an  upper limit on the local field.

     On the basis of the model assumptions, the enhancement should not depend
on the HMD of the dust layer (although the resistivity does, Equation 4).
The reason for this  expectation is  that the calculation of the average field
is based upon a  monodisperse layer  of particles  having the diameter most
likely to  be  found in the polydisperse layer; and the most probable diameter
is always  in  a fixed ratio to the MMD in a lognormal distribution.

     The numeric integration must use definite limits for its upper and lower
bounds;  as a  result, if the  most probable diameter approaches those bounda-
ries, significant amounts of the enhancement contribution may be ignored.
This is  borne out in Figure  6, where significant changes begin to occur for
MMDs greater  than 20 percent of the largest size in the integration, 50 ym.
Therefore, the original expectation of particle  size independence is
realistic.

     The variation of geometric standard deviation would be expected to have
effects  on the enhancement for the  same reason that MMD variations should
not:   the most probable diameter changes its relationship to the MMD when
                                    279

-------
                 3.0
                 2.0
                 1.0
                   1                   10                  100
                                    MMDI/mi)
           Figure 6.   Effect of mass median .diameter on enhancement.


a is varied.  When a is 1, the particles are monodisperse, with  the most
probable diameter and HMD coincident.  As cr increases, the most  probable di-
ameter decreases in relation to the HMD.  As in the case with  the MMD,  the
particle size distribution may spread so much that the finite  limits  of inte-
gration affect the enhancement.

     Figure 7 shows the trend of enhancement with CJ over  the range from 1
(monodisperse) to 5 (very polydisperse).  The relationship can be described
as approximately linear, with much of the curvature caused by  the numerical
methods used.

     The conclusions drawn from this application of particle-contact  theories
to the pore breakdown process are as follow.  Field inhomogeneities around
large particles may enhance the local electric field to the degree that
breakdown into sustained back corona is possible for average fields 2 to 5  ,
times smaller than for homogeneous layers.  The actual reduction depends very
strongly on the maximum value of local field that can be  attained in  relation
to the average field.  The reduction also depends strongly on  the geometric
standard deviation of the particle size distribution, but not  on its  MMD.

     In practice, if a fly ash with a = 3 is well distributed  in a precipita-
tor, a reduction in back corona sustaining field of about 2 is reasonable.
A reduction in sustaining electric field by a factor of 2 is a significant
reduction because it corresponds to a reduction in the current density re-
quired to cause the back corona by almost the same amount  (allowing for the
resistivity to change as well, according to Equation 4).
                                     280

-------
                       3.0
                       2.0
                       1.0
                          12345

                            Geometric Standard Deviation

                    Figure  7.   Effect of a on enhancement.
                     CURRENT DENSITY DISTRIBUTION EFFECTS

     Since  the  local  current density acts in concert with the resistivity to
produce the field  for breakdown,  it is important to know how much the current
density differs from  its  average  value at various locations in the ESP-  Nu-
merical techniques (3) are available for calculating the variation in the case
of uniform  corona, such as occurs in the positive glow mode.  The calcu-
lations, which  are in good agreement with experimental measurements, show
that geometrical effects  can be substantial.

     For a  geometry with  wire-wire spacing equal to the plate-plate spacing,
the maximum current density at the plate occurs directly opposite a corona
wire and is about  1.2 to  1.5 times the average value; the higher factors
occur at higher current densities.

     For wire-wire spacings less  than plate-plate spacings, the maximum ratio
is somewhat less,  1.1 to  1.2, and less dependent on current density.  In
effect, the corona wires  approximate a uniform sheet of ions, delivering an
almost uniform  corona current.

     For wire-wire spacings of twice the plate-plate spacing, the peak current
density ranges  from 2.2 to 2.4 times the average.  For wire-wire spacings of
4 or more times the plate-plate spacings, the peak current density is about
4.5 times the average. At the widest separations, each corona wire is essen-
tially uninfluenced by the presence of the other corona wires.

     In common  use the wire-wire spacing is between 1 and  2 times the plate-
plate spacing,  meaning that with uniform positive corona the maximum current
density would be 1.2  to 2.4 times the average value.  Since negative corona
is'most often used, the uniformity of corona assumed in the calculation is

                                     281

-------
not even valid.  Negative corona is found  to  emanate  from distinct spots  on
a corona wire, resulting in a very spotty  generation  of  current along the
wire.

     As a first approximation, each corona spot  emits current toward the
plate in a manner similar to a point-to-plane electrode.   The current density
distribution on the plate under a point  is given by the  Warburg distribution:

                        j(6) = j(0) cos56, 6  < 60°,                      (6)

wheire j(0) is the current density measured at an angle,  0, with respect to
the point, and j(0), the current density directly opposite the point.(12)  By
integrating j(6) over all angles, the calculated ratio of j (0) to the average
current density is  1.89.

     Since the corona wire is not a point  source, the actual  distribution
will be distorted from that of a point electrode.  Experimental measurements
of corona from a point embedded in a plate (13)  show  a current density distri-
bution with a sharper cutoff at a limiting angle, which is a function of the
applied field.  Under a corona wire tuft,  the distribution of current could
be considered a combination of the two types  of  distributions, with the War-
burg distribution dominating in a direction perpendicular to  the wire and the
embedded point distribution dominating in  the direction  parallel to the wire.

     It has been observed that negative  corona tufts  interfere with one
another, and the movement of one toward  another  is limited to some minimum
distance.  This is  interpreted as a mutual repulsion  of  the ion clouds emitted
from each tuft and  implies that current  density  distributions on the plate do
not overlap.  Because of the nonoverlapping,  the average current density  on
the plate will contain substantial areas of zero contribution; and this means
that the ratio of peak current density to  plate  average  current density will
be higher than indicated by Equation 6.

     Figure 8 shows an arrangement of Warburg tuft distributions along corona
wires, corresponding to the closest approach  of  areas without overlapping.
The wire-wire spacing is equal to the plate-plate spacing for this figure,
and the ratio of the total plate area to the  area covered by  current is 1.2.
This ratio reaches  a maximum of 1.27 when  the wire-wire  spacing is about
1.2 times the plate-plate spacing.  When the  spacing  ratio (wire-wire/plate-
plate) equals 1.73, the total area is 1.27 also  and increases with further
increases in the spacing ratio.  For the spacing ratios  commonly found in
precipitators, an area ratio of 1.25 is  a  reasonable  average  value.  If the
corona tufts were not densely packed, as at low  current  densities, then the
area ratio could be much larger.

     Multiplying the area ratio by the ratio  of  peak  to  average value of  the
Warburg distribution provides the factor relating the maximum negative cur-
rent density  (under the corona tufts) to the  average  plate current density:
2.36 for spacing ratios from 1 to 1.7.   This  ratio represents a conservative
estimate of the peak current density in  terms of the  average  current density.
                                     282

-------
              Figure 8.   Idealized arrangement of negative corona
                         tufts on wires.
                   COMBINING FACTORS TO CALCULATE BREAKDOWN

     It  is  now possible  to  relate the back corona sustaining field and cur-
rent density  for  a real  fly ash layer in a negative corona precipitator.  To
calculate the initial  homogeneous layer breakdown, the thickness of the dust
layer in Equation 2 is used to to find an (a - n) that produces a gain of
108.  For reasonable thicknesses, 1 to 10 mm, the field required will be large
enough that a »  n and Equation 2 can be rewritten:
This can  be  solved  for  (a -
                             G - exp [(a - n)x].
                                a - n = InG/x,
(7)
(8)
with x being  the dust  layer  thickness.   Equation 5 is the-.n used to determine
the E/P required to produce  that (a - r|).   Using the notation of Equation 5:
                           E/P
(9)
                                    283

-------
Recall that E/P is the standard variable used in gas ion measurements, with
the field, E (in V/cm) and the temperature-reduced pressure, P  (in  torr).
As an example, assume a thickness of 1 mm, a pressure of 760 torr,  and a
temperature of 150° C.  Then, a - n would be 184 and E/P would  be 87.7.  At
the reduced pressure of 490 torr, E is 43 kV/cm.  This value of electric
field is much higher than usually calculated from dust resistivity  and cur-
rent density values:  breakdown fields of 5 to  10 kV/cm are expected  in op-
erating precipitators.

     But assume that the particle size distribution has a cr of  3 and  that the
maximum local field is limited to 5 times the average.  Then, from  Figure 7,
the field enhancement in the pore is a factor of 2.25; this means the back
corona would be sustained by an average field of only 19.1 kV/cm.

     The average pore breakdown field is given  by Equation 1, relating current
density and resistivity-  Assume a resistivity  of S'1011 f2-cm,  as measured at
an electric field of 10 kV/cm; then the effective resistivity at 19.1 kV/cm,
calculated with the aid of Equation 4, would be 4.4'IQ11 ft-cm.   The current
density required to produce a field of 19.1 kV/cm in a resistive layer of
4.4-1011 ft-cm is 4.35-10 * A/m2.

     Presuming that this current density is produced directly under a negative
corona tuft, the_average plate current density  corresponding to that  peak
value is 1.84*10 "* A/m2 because of the factor of 2.36 between the peak and
average current densities.

     The apparent electric field in the layer at breakdown, given by  the
product of the average plate current density and the measured dust  resistiv-
ity, is 9.2 kV/cm, a value well within the range encountered in common
practice.(11)

     Until the back corona is fully established, there can be no feedback
mechanism operating to reduce the electric field.  However, once the  pore
channel has enough gain to initiate the Townsend avalanche, the intense
photoionization and impact ionization within the channel will provide very
high feedback factors, close to 1 in value.  There is evidence  for  this in the
surface potential measurements in Reference 14.  In that work,  Teflon® filter
paper was used as a resistive layer under a point-plane electrode.

     The layer was able to sustain an average E/P of 105 V/cm-torr  without
breakdown, with the breakdown field evaluated by Equations 8 and 9  as
112 V/cm-torr.  When the breakdown was firmly established by using  a  higher
corona point potential, the surface potential on the layer dropped  to the
point that the average E/P was only 44 V/cm-torr.  This corresponds to a
feedback factor of 0.8, using Equation 3.  Thus, these measurements support
both the initial breakdown calculation and the  use of heavy feedback  after
breakdown occurs.

     The feedback lessens the current density required to maintain  the dis-
charge, and, according to the pore breakdown model, the excess  current den-
sity flows directly into the pore.  The same factors may apply  to the
sustaining field as were previously calculated  for the breakdown field.

                                      284

-------
Therefore,  continuing the hypothetical layer example, at a feedback value of
0.8, the average enhancement corrected electric field is 8.8 kV/cm.  The cor-
responding  layer resistivity is 5.13'IQ11  ft-cm, and the local current den-
sity required is 1.7*10 "* A/m2.  The heavy ionization in the pore may upset
these calculations,  but the drop in E/P to 44 V/cm-torr will be reflected in
a drop of local current density.

     At present it is not possible to relate this local current density to
the plate average current density.   The back corona generates additional ionic
carriers which not only increase the total current but also change the dis-
tribution of current by neutralizing some  of the ionic space charge.

                                  CONCLUSIONS

     The pore breakdown model for back corona generation predicts average
layer electric fields which are much higher than usually accepted for operat-
ing precipitators, but which are in agreement with laboratory determined
values.  It is shown that the size distribution of particles in the dust
layer can account for about half of the discrepancy and that current density
distribution effects can account for the other half.

     The size distribution has been shown  to have an effect on the initiation
of back corona only  through the range of sizes encountered, characterized by
the geometric standard deviation.   For lognormal, or approximately lognormal,
distributions, the MMD does not have a noticeable effect.   Layers composed
of monodisperse particles should have the  same back corona initiation point
as homogeneous porous layers of the same average resistivity.

     The distribution of current density along the plate of a precipitator is
equally important because back corona will always be initiated at the areas
of maximum  current density.   The tuft structure of negative corona works
against achieving uniform current densities, but the spacing of the corona
wires can also be important in determining the inhomogeneity of the current
density.

     The ionic feedback that occurs within a pore after back corona begins
is so large that the field required to sustain the back corona may drop to
less than half the value required to initiate it.  This process results in a
considerable hysteresis in the conditions  required to turn on and turn off
back corona.   The Townsend avalanche conditions set an upper bound on the
amount of hysteresis.

     The pore breakdown model has  proven to be a useful and accurate model
for describing the back corona process.  It provides quantitative values for
layer breakdown,  for the degree of  hysteresis, and for the amount of back
corona current generated.   When applied with electric field models and bi-
polar particle charging models, it  should  be capable of describing adequately
the motion  of particles in precipitators operating in back corona.
                                    285

-------
                                  REFERENCES

 1.   White,  H.  J.   Industrial Electrostatic Precipitation.  Addison-Wesley,
     Reading, Massachusetts,  1963.

 2.   Lawless, P.  A.   Progress in modeling back corona.  In:  Third Symposium
     on the  Transfer and Utilization of Particulate Control Technology:
     Volume  II.   Electrostatic Precipitators.  EPA-600/9-82-005b.  U. S. En-
     vironmental  Protection Agency, Research Triangle Park, North Carolina,
     1982.   pp.  35-43.

 3.   Lawless, P-  A., and Sparks, L. E.   A Mathematical model for calculating
     effects of back corona in wire-duct electrostatic precipitators.
     Journal of Applied Physics.  51:242, 1980.

 4.   McLean, K.  J.   Electrical conduction in high resistivity particulate
     solids. Doctoral  dissertation, University of New South Wales, December
     1969.

 5.   McLean, K.  H.   Cohesion  of precipitated dust layer in electrostatic pre-
     cipitators.   JAPCA.  27:1100, 1977.

 6.   Dietz,  P-  W.   Cohesive force and resistivity between electrostatically-
     precipitated particles.   J. Electrostatics.   6:273, 1980.

 7.   Moslehi, G.  B., and Self, S. A.  Electromechanics of precipitated par-
     ticulate layers.  In:  Conference Record IEEE-IAS Annual Meeting, 1981.
     Library of Congress No.  80-640527.  p. 1102.

 8.   Loeb, L. B.   Basic Processes of Gaseous Electronics.  University of
     California Press,  Berkeley, California, 1961.

 9.   Masuda, S.   Back discharge phenomena in electrostatic precipitators. In:
     Symposium  on the Transfer and Utilization of Particulate Control Tech-
     nology: Volume I.  Electrostatic Precipitators.  EPA-600/7-79-044a
     (NTIS No.  PB295226).  U. S. Environmental Protection Agency, Research
     Triangle Park,  North Carolina, 1979.  pp. 321-333.

10.   Masuda, S.   The influence of temperature and moisture on the electrical
     conductivity of high-resistivity dusts.  Staub.  25:1, 1965.

11.   Spencer, H.  W., III.  Electrostatic precipitators:  relationship between
     resistivity, particle size, and sparkover.  EPA-600/2-76-144  (NTIS
     No. PB257130).   U. S.  Environmental Protection Agency., Research Triangle
     Park, North Carolina,  1976.

12.   Sigmond, R.  S.   Simple approximate treatment of unipolar space-charge-
     dominated  coronas:  the  Warburg law and the saturation current.
     J. Appl. Phys.   53:891,  1982.
                                    286

-------
13.  Collins, L. P., Self, S. A., and Shearer, D. D.  Point-source  corona
    current distribution in an external field.  IEEE Trans, on  Ind. Appl.
    Vol. IA-14:506, 1978.

14.  Thanh, L. C.  Back corona—Part I:  its formation.  J. Electrostatics,
    6:139, 1979.
                                   287

-------
                  ELECTRICAL BREAKDOWN OF PARTICULATE LAYERS

                      By:  G. B. Moslehi  and S.  A.  Self
                   High Temperature  Gasdynamics  Laboratory
                      Mechanical Engineering Department
                             Stanford  University
                          Stanford, CA 94305, U.S.A.
                                   ABSTRACT
     A  theoretical analysis  of  electrical breakdown  of  a current-carrying
particulate layer  modeled  as  a regular array  of  equi-sized resistive spheres
having combined  surface and volume conduction,  is presented.  This analytical
treatment  is  based on  an extension  of  a  theory of electromechanics  of the
layer and predicts the  onset  of  electrical breakdown of the layer in the form
of intermittent microsparks  in the gap between  the  contacting particles when
the  electric  field  at  the  contact  or  in the  surrounding gap  exceeds the
threshold breakdown value. The occurrence of breakdown is due to the existence
of a very  strong electric field  in and around the contact region as a result
of current  constriction at the  contact  area.  Two  possibilities  of breakdown
are examined:  i) breakdown in the gap where the gap height, d, is larger than
the gas mean  free  path, X, and ii) vacuum breakdown at  the contact where d <
\.

     The electrical behavior  of  the layer  after breakdown is also analyzed in
terms of a  simplified  equivalent  lumped  circuit using methods of conventional
transient circuit theory. In this analysis  the layer is modeled as a number of
capacitive spark gaps in series,  separated  by high resistances. In effect, the
discharge  propagates  through  the layer as a  cascade of  microsparks,  which
discharges  the layer  locally. The theory  predicts  increases of sparking fre-
quency  and average  current  as  the  applied  average field E.,  exceeds the
threshold average field for the onset of breakdown, E._.

     Experiments with  glass  beads  demonstrate the  existence of intermittent
microsparks  after  the  onset of breakdown,  whose  frequency  of  occurrence
increases with increasing E..

                                  INTRODUCTION
     Back discharge, which  seriously impairs the precipitation of  high  resis-
tivity  ash,  is  poorly  understood  and  not satisfactorily  incorporated into
                                      288

-------
computational codes or  design procedures  for precipitators.  It  is generally
attributed  to electrical breakdown of the  ash layer when the average electric
field
                                  E.
                                   (i)
exceeds  some  critical  breakdown value of the order of 10  V/m. Here, p. is the
apparent layer  resistivity  and J^  is  the average  current  density passing
through  the layer.
     This macroscopic  view of the phenomenon, while undoubtedly correct as far
as it  goes,   treats  the  layer  as  a  homogeneous medium  subject  to uniform
current  and  field.  However,  to understand back-discharge breakdown  in more
detail,  and  to explain the remarkably  low  value  of dielectric strength, it is
essential to  consider  the particulate structure of the layer.
     It  is generally  accepted  that the  onset  of  back discharge is  due  to
electrical  breakdown  in  the interstices of the ash  layer (as a  result  of a
large field  enhancement  in  the  contact region),  since it occurs at average
fields of the order  of a  few kV/cm which is many orders of magnitude less than
the inherent  dielectric strength of glassy materials such as fly ash.
       ONSET OF BACKDISCHARGE
     In a particulate  layer carrying
an  ion  current from  the  discharge
space to the  grounded  collector,  the
Current must  flow through  the  small
contact   areas   between   particles
(Figure la).  Therefore, the current
and  field lines  are  concentrated
near the  contacts  and  most of  the
resistance  and  voltage    drop   is
associated with these regions.  This
current constriction in  the contact,
results in a  strong field  enhance-
ment above the average  value of  the
applied field,  EA, in  and  around the
contact  region, which gives rise to
a  remarkably  large  electrical  com-
pressive stress,  PE,  in the layer.
Furthermore,   the  apparent resis-
tivity  of the  layer,  pA,  decreases
with increasing J^ or  E^ as a result
of    self-compression     (electro-
strictlon).

     The existence  of  a very strong
electric  field  in  and   around  the
contact  region, results  in the  onset
of electrical  breakdown  of  the  layer
(or  back-discharge)   in the form of
                                                                    Equi-
                                                                    Potentials
                   EOUIPOTENTIAL VQ
Figure 1.  Geometry for current  flow
through  resistive   spheres.     (a)
potential  and  field  distribution;
(b)  geometry in contact region.

intermittent microsparks  in the  gap
between  the   contacting  particles
when the (local) electric field  at
the  contact   or   surrounding   gap
                                     289

-------
exceeds  the  threshold  breakdown  value,  Eg.  At  the  onset  of  microspark
breakdown  the  average  field  across  the  layer, E^,  is equal  to the  average
(threshold) breakdown  onset field, E^g.  At higher  fields (E^ >  E^g),  micro-
sparks occur with  increasing  frequency so  that the average current  increases
(and hence  the  apparent resistivity decreases)  more rapidly than is  the case
prior to breakdown. However,  the  decrease in the apparent  resistivity,  pA, is
steady rather than catastrophic.

     Clearly, a homogeneous  leaky dielectric  model  for  the  layer fails to
represent  the  true structure of  the layer  and is  not able  to explain and
predict the behavior of the particulate layer.

     In  this  section  we  outline  an  analysis  of the  onset of  back-discharge
breakdown of a  current-carrying particulate layer  by  coupling the results of
our analysis for the field  distribution in the layer  reported  in References 1
and 2, with certain well-established results for electrical breakdown  in small
gaps. This  allows  one  to predict,  in terms  of  the  basic  parameters of the
system,   the average current density, J^g,  or  average electric field, E.g, at
which the onset of microspark breakdown occurs  and  its location in the gap.

ELECTRIC FIELD DISTRIBUTION IN THE GAP BETWEEN  TWO  CONTACTING SPHERES

     In  the past  we  have  developed  a  comprehensive  theory  of   the  electro-
mechanics of precipitated particulate layers (1,2). The  basic model  treats the
layer as a regular cubical array of equal size  resistive spheres  through which
the current is conducted to the collector either by volume  conduction, surface
conduction or, in general, a combination of both.

     In  References  1  and 2,  the  electric field in  the gap between two equal
spheres  in  contact  (Figure  1) having a  (finite)  contact angle 9Q is shown to
be
                           E <9>V ~= Emax G <9'V                        (2)
where G = 1 in the contact region itself and for 9>9Q>

                               _ [1  - F(a,9)/F(a,9o)]
Here, p,  = cosG,  u  = cos9. and  the field enhancement factor (FEF) is defined
as (1,2)
               FEF = E/E  = bD(a/4 - 1/2)  (E)(a/2 '  X)   .            (4)
                                E                A

     For small angles 9 and 9Q, the function F is approximately  given by (2)


                             F(a,0) « are)11                               (5)
                                       r
and

                             F(a,00) - aF(6o)n                             (6)
                                     290

-------
   TABLE 1 VALUES OF COEFFICIENTS AND EXPONENTS IN THE POWER LAW EQUATIONS
F = a

a
0
1
wn
10*
10
104
CD
n . Q .
oa/2
a: E
(«/4 -
- 1/2) ,«/;
>
*• cJ ' > o y ^ A' ' max E ^ A'
aw
1.5865
1.2057
4.3475
3.8322
1.0545
0.6970
0.6879
n
-0.1628
-0.1867
-0.2859
-0.5492
-0.8927
-0.9860
-0.9889
bfl
0.6767
0.7484
1.0391
1.5103
1.2112
0.9865
0.9766
bff
0.8226
0.8651
1.0194
1.2289
1,1005
0.9932
0.9882
a
0.4279
0.4314
0.4430
0.4502
0.4184
0.3996
0.3987
A(CT)
2.139
2.115
1.976
1.582
1.299
1.274
1.274
NOTE:     1.  In all  cases  0Q  is  in radians.  All values are in (MRS) SI Units.
          2.  Value of  F  for  a = 0 is actually Fg where Fs =  L^m (F/a).   For
             a £ 10    the  approximate expression F = oFg may be used.
          3.  For intermediate values of  a,  which are not given in the table,
             the functions F,  9Q and Emax may be evaluated using a logarithmic
             interpolation procedure.

(The exponent n used  in the  above expressions is identical  to  nF in  Ref.  2.)
Values of  ap,      n  and  a  for  different values of a are given in Table 1.

     For a  spherical  particle  of  radius a  the non-dimensional resistivity
ratio a is  defined as

                                   a = sa/p                                 (7)

where p is the (intrinsic) material volume  resistivity and s  is the surface
resistivity of the  particle.  Pure surface conduction corresponds to  o=0  and
for pure volume conduction a •* <=.

     In Eq. 2,  £__,.  is the maximum electric  field in the gap and is expressed
by (1,2)

                    _  Lim r ,    ,-> _    (a/4 - 1/2),
               Emax ~  e+e LEte>vJ - V            t
                         o
The contact angle  0  is  given  by (2)

                          90(a,EA)  =  bQ Da/2(EA)a  .                        (9)

     In Eqs. 4, 8  and  9  the material  constant D is defined as

                               D(a) = iteoA(o-)B                            (10)

where eQ is the vacuum permittivity, A(a) is  a  constant and B depends only on
the elastic properties of  the  particle  and is given by


                                     291

-------
                              B = 3(l-v2)/4Y                               (11)

Here, Y is the Young's modulus and v is Poisson's  ratio.

     Values of  bfi,  b™ and  A(a)  for different  values of  a are also given in
Table 1.
     Using Eqs. 4,  5  and 6 with Eq.  3 gives an  expression for G  in terms of
E., which together with Eqs. 2 and 8 results in
                                                2
                                                  1^-                   (12)
where  x E  9/9Q  is a  measure of  location in the  gap between two  contacting
spheres.
The gap height between spheres can be written  (for  small  angles)

             d = 2a(nQ - u) = a92(x2- 1) = a b2  (x2-  l)(Da)(EAl2a        (13)

where use is made of Eq. 9.
     Substitution for (x2-!)  from  Eq. 13 into  Eq. 12  gives
                                            2a E.
                         E(x,a,EA) = (l-xn)[—g-£]   .                     (14)

     Equations 12  and  13  give the values of the electric field  in the  gap and
gap  height  as a  function of  the  average electric  field, EA, the  normalized
location  x  and the resistivity  ratio a, and  are  necessary inputs  for  deter-
mining  the  breakdown condition from gas-discharge considerations, as discussed
below.

BACKDISCHARGE ONSET CONDITIONS
     The  condition for  back-discharge  onset  can  be estimated  by using the
 field  in  the contact area and  surrounding gap together with  results  from the
 literature  for  electrical  breakdown of  small  gaps between parallel  plane
 electrodes in air. There are  two mechanisms  to consider,  namely,  gas discharge
 breakdown for  gaps of height d ^  \ and vacuum breakdown for  d <\,  where X is
 the  gas  mean free path. Which  mechanism occurs first with  increasing E^, and
 where  it occurs,  will be determined by which  threshold field  for breakdown is
 first  exceeded.

     Figure  2  shows  the  composite breakdown  curve,  EB(d),  for  the threshold
 breakdown field  EB as a function of gap height d.  This curve  is  plotted using
 the  data taken from  literature and applies to metallic parallel plane elec-
 trodes in  air  at STP. The method  of construction  of this composite breakdown
 curve  is explained in Reference 1.  For  d ;> \,  in  the gas  discharge regime, the
 curve  of EB(d)  is a monotonically decreasing  function of  d. For  d £ \, in the
 vacuum breakdown regime, for short  gaps the local field  initiating breakdown
 is quoted  (3)  as Eg  =  EVB ~ (6.5 ± 1)  x 109 V/m. For the  purpose in hand we
 take EVB  = 6.5 x  109 V/m for gaps  less than d ~  1  urn where  the data for gas
 discharge  breakdown  yields  the same value. This  is a reasonable assumption
 because  at  this point d  is  close to the  gas  mean free path  \ ~ 0.06 \m (for
 air  at STP).                                                                  *

                                      292

-------
                     EB(gas)
10
                 1.0
         GAP HEIGHT.c
                             1O
Figure 2.   Graphical  procedure for
determining   breakdown  field   EAB
(Volume Conduction).

     From  the  expression  for  the
electric field  in the gap  between
contacting   spheres,   summarized  in
the previous  section  and  the  con-
ditions for electrical  breakdown in
uniform gaps  for  the two  distinct
mechanisms  discussed  above,  we  can
now  calculate  the  conditions  for
breakdown   in   the    gap    between
spheres.

     The  case  of vacuum breakdown is
relatively  straightforward.  Because
breakdown can  be  assumed  to  occur
when a certain field  EVB is  reached
anywhere  in the  gap,  and since  the
maximum field  in  the  gap occurs  at
the edge  of the contact area x  = 1,
then we have  from  Eq. 8 the vacuum
breakdown condition for the  average
layer  field
E  = (D)         FE   /          (15)

It may be  noted  that,  for a  given a,
this  condition   is   independent  of
particle  radius, and  that breakdown
occurs at  the  edge  of  the  contact
area  (or   in  the  contact   region
itself, where x
-------
breakdown is best illustrated by the
graphical procedure  shown in  Figure
2.  For a given  particle radius,  a,
and  material properties,  the  field
in the gap E(9,9 ) can be calculated
for given values of average  field E.
using    the    formulae   developed
earlier. Three such curves are shown
in  Figure  2 for the  case of  volume
conduction  (a-*"),  a  =  100  p,m, B  ~
10'11  mZ/.N  and for EA  = 1.0  x 10b,
1.5 x  106  and 2.0 x  10° V/m.  It  is
seen  that  the first  curve  does  not
intersect   the  EB(d)   curve,   the
second curve intersects Efi(d)  in the
vacuum  breakdown regime, while  the
third  curve additionally intersects
EB(d) in the gas discharge breakdown
region.  Interpolation   from   these
curves  shows  that  for  these  con-
ditions  breakdown  occurs  first  in
the  contact region  (d  ~ 0)  at  an
average  field for  breakdown  E»B  ~
1.3 x  10  V/m followed by breakdown
in  the  surrounding  gap  (at  a  po-
sition where  the  gap height is ~ 10
pm) at an average field of EAn ~ 1.8
  <« _n TT /                    Afi
x 10  V/m.

     To  avoid  the  tedium   and  in-
accuracy of  the  graphical procedure
illustrated    above,    a   computer
program has been developed to  calcu-
late   the   gas-discharge  breakdown
condition.  Using this method,  values
°f  EAB for  gas  discharge breakdown
have  been   obtained   for  the   two
limiting cases of surface and  volume
conduction as a function of particle
diameter D   =  2a (for atmospheric
pressure  air  at  25°C  and   150°C,
respectively) as shown in Figure  3.

     It is  found that  for  the  case
of volume conduction, with B ~ 10"11
m /N, and for air at 1 atm,  then  for
a < 140  p,m,  breakdown always  occurs
first  as vacuum breakdown   in  the
contact region (d ~  0)  at EAB ~  1.3
x  10   V/m, independent of particle
size (Figure  3).  For  larger  parti-
cles,,   however,   breakdown    occurs
first as gas discharge  breakdown in
   10'
   10
[KV/on]
   10
   .o»
    10'
10s
10s
Figure  3.   Average  electric  field
UAB   for   the  onset  of   breakdown
versus  particle diameter D. Volume
conduction  case.  For  D  g 280 \mt
vacuum breakdown occurs  first at EAR
~ 13 kV/cm.
the  surrounding  gap where  the gap
height  is ~  10  p,m  and  at  a  field
which  decreases   roughly inversely
with radius.

     Calculations of the  conditions
for  the   onset  of  back-discharge
breakdown  for  the case  of   surface
conduction lead to generally  similar
results.  However, because the  field
enhancement factor is lower  for the
case   of   surface  conduction,  an
appreciably larger value  of  average
layer  field  EAB is  required for
breakdown. For  instance for  a  = 150
Urn,  B  =  1011  m /N,  gas discharge
breakdown occurs  in  the  surrounding
gap  (at d ~  15  \u&) at  an  average
field  of  EAB ~  2.0 x  106  V/m,  a
factor  ~  1.7  higher  than  in the
volume conduction case.           ;
     While the  model on which  these
conclusions are  based  is idealized
and  the  breakdown  data  we  have
employed   are    only  approximate,
especially for  vacuum breakdown, we
expect these conclusions  to be  valid*
                                    294

-------
in  general   form  even  though  the  predicted  values  of  EAB  for  onset  of
microsparks  and  back-discharge may be only approximate.

     Because of  the form of  Eq.  15,  which, (for a ~  0.4,  from Table 1) shows
E.B « EyB  the  results are extremely sensitive to  the value  assumed  for the
local vacuum breakdown stress, EyB,  which is a quantity whose value for non-
metallic  electrodes  is difficult  to establish from the literature. What values
are available were  obtained  for 'small  gaps between (parallel  plane)  metal
surfaces,  and since  the  breakdown mechanism undoubtedly depends on the surface
state,  the appropriate value  of E^B (or Eg(d)  in general) for glassy materials
is difficult to  know with any certainty.

                        EXPERIMENTS WITH GLASS BEADS
     Using a  standard  resistivity cell under various conditions of temperature
and humidity, measurements  of  the apparent  resistivity p. of  the layer were
made as a  function  of the  (applied)  average field, E.. In these  tests glass
beads (450-500tim dia.) were  used.  In each test a new  layer of beads was used
from the same supply sample. The inechanical  pressure  due to  the weight of the
upper electrode  was  ~  10  g wt./cm  and the layer thicknes was about 5mm.
     The results of  these experiments are discussed in some  detail in Ref. 2.
Table 2 gives the values  of (dry bulb) temperature,  tj^,  and relative humid-
ity, Pr, for  each test. In  the high  temperature tests (t^^  30°C) the ambient
air was  heated   to  the  specified temperature.  Other  tests  were made under
ambient conditions.  Only  in experiment #1  was the humidifier  used, in order to
achieve a  relative humidity of  Pr ~ 100% at tdb ~ 28°C.

     Figures  4a  and b show a  few  examples of  the measured  apparent  resis-
tivity, pA, of  the  layer as a function of  the (applied) average field,  EA,
given in Table 2. Figures 4a and b are  identical,  with the exception that the
former has a  log-log scale and  the latter  a semi-log scale.
     Generally  in  all  of   these  experiments,  the  p^(EA^  characteristic  is
composed of  two distinct regimes:  pre- and post- microspark breakdown.
     The pre-microspark breakdown regime has two regions.  In  region I, for EA
in the range 104  £ EA & 2x10^  to 3x10^ V/m,  pA is approximately constant. This
is due to the fact that the compression  as  a result of the electric stress in
the layer is less  than that  due to  the  weight of  the upper  electrode.  In
region II,  for EA  between  ~ 4x10  V/m and ~2xlO  V/m,  the  PA(\)  character-
istic  follows  a  power  law (p.  <*  E.m) with  the  exponent  having  a  value
somewhere between -0.077 for the case  of  surface conduction  (t,,  = 28° C, P  =
100%)  and -0.4  for the  case  of volume  conduction  (hot and  dry conditions).
These  values are in  agreement  with  our  theoretical  predictions   which  are
reported in  Reference 2.
     In the   third  region (Figure  4a) , at  EA  = EAB  ~  (2-4)  x 105  V/m,  the
       curve  starts  to  deviate  from  the  power   law  dependence,  with  pA
decreasing more rapidly with increasing (applied) field EA- This is due to the
electrical  breakdown of  the layer  whi,ch occurs  in the form  of  intermittent
microsparks. For  EA between (4-5) x 10   V/m and the value for gross breakdown
of  the  layer,   the  P^(EA^  dependence is  exponential in form (Region III in


                                     295

-------
TABLE 2.  RESULTS OF RESISTIVITY CELL MEASUREMENTS ON GLASS BEADS AT
                       DIFFERENT VALUES OF TEMPERATURE
No.

1
2
3
4
5
6
7
8
9
10

11
12
13
14
15
16
Oven
Conditions
ffi]
28
23
22.5
23
22.5
22
22.5
29
46
46

61
119
133
151
162
185
P ,
100
~ 88.5
~ 82
~ 81
^ 80
~ 78
~ 81.6
* 70
23.2
22.9

12.1
1.2
0.8
0.5
~ 0.4
~ 0.2
Region II
« - v wm
PA ~ A
Region I pA in [S3 - cm] ;
p^ = const. E^ in [kV/cm]

9
4
5
9
3
9
4
5



1
7


4
4
PA>
[fi-cm]
.24xl07
.56xl010
.84xl010
.82xl09
.07xl010
.83xl010
.25xl010
.17xl010

«___
1 9
.27x10"
.12x10
_«___»
_«__
.54xlo}2
.60xl012
b
7.54xl07
3.53x10]-°
4.97xl010
6. 9 5x1 O9
2.07xl010
7.92xl010
3.37xl010
4.44xloJ°
6.28X1011
5.56xloJ-f
1 1
6.14x10,
3.55xlOJ-2

3.41xloJ2
2.71x10"
2.12xl012
m
-0.077
-0.172
-0.175
-0.21
-0.23
-0.23
-0.235
-0.259
-0.352
-0.337

-0.344
-0.408
-0.407
-0.382
-0.337
-0.420
Region III
PA = Po exP(aoEA>
p^ in [Q - cm] ; Gross
EA in [kV/cm) Breakdown


2
3
4
1
6
3
3
4
4

4
2
2
2
2

PO»
[Q-cm]
____
.60x10?-°
.46xl010
.15xl09
.36xl010
.ISxlO10
.16xl010
.46x!0j°
.71xlOU
.07x10,}
1 1
.43x10,
.50x10};
.72xlo}2
.46xioJ;
.09x10
	
[cm/'kV] [kV^cm]

-2.
-3.
-2.
-2.
-3.
-5.
-4.
-6.
-5.

-6.
-8.
-8,
-8.
-8.

—-._..
88xlO~2
53xlO~2
65xlO~2
16xlO~2
25xlO~2
52xlO"2
OlxlO"2
06xlO~2
45xlO~2
_9
74x10 ;
95xlO~,
50x10
92xlO~2
98xlO~

^~ ••_
15.8
16.0
13.4
16.0

15.8
16.0
16.0
16.0

15.4
13.8
12.6
13.4
12.0
— — —
(-) means: (i) "No Data Taken" or (ii) "No Data Available".

Figure 4b). The  gross  breakdown of the layer  occurs at E^ =  EGB ~ (12-16)  x
10^ V/m, with the lower values  corresponding to higher temperatures and drier
conditions. These  values  are  much  lower  than ~  4 x  10  V/m  required  for
breakdown across a 5 mm air  gap between two parallel electrodes  at STP. Table
2 gives the values of p.
                         and a.  for  the exponential  form
                                = Pr
                                          <~ aoV
in Region III which applies to  the  range  4 x 10  V/m
                                                             E
                                                               GB*
                                                                          (16)
                                                                         values
are calculated using a least squares fit  to  the experimental results. This form
of functional dependence of p^ on E^ has also  been observed by McLean and Huey
(4). In  these experiments, for the  cases close to volume conduction the micro-
spark  breakdown  of the layer occurs at  E^B ~  2  to 4 kV/cm which  is close to
the  theoretical  value of  ~ 7.2  kV/cm  for 500  p,m diameter beads  and volume
conduction,  taken  from Figure 3. The  difference is mainly due  to  the lack of
accurate breakdown data  for non-metallic electrodes  in  nonuniform fields and,
partly due to the  approximations of the  theory.
                                      296

-------
O EXP .!(T-23'C.f-BB.5Z!
« EXP .711-22.5'C.f-91.6X1
x EXP .nn-6rc.E-i2.ixi
D CUP •12I1-]19>C.{.1.2ZI

m EXP .isiT.ias't.^-o.jX!
I GROSS BRERHOOHN OF THE
 Linen
 (0- en,],
                    PA  '
                   to--4
V
      I

V
                                                                    O EXP
                                                                    « EXP
                                                                    X EXP
                                                                    d EXP
                                                                    A EXP
                                                                    X EXP
                          •29 C.f-B».52l
                          -22.5"C.£-81.6V.I
                          1-6TC.EJ.12.UI
                          I-IIS'C.f.l.2XI
                          1-133'C.^«0.6'X1
                          i-ie5'c.pr-o.jxi
                                                                      DflOSS BREflKOOHM

                                                                      LBTER
                                                   23 t iG7  8  9  10 11 f? 13  h 15 16
                                                             EA . [kV/fcm]


 Figure  4.   Resistivity cell  measurements  at different values  of  temperature;
 Glass  beads;  Dp ~  450 p,m.  Solid  (or dotted)  lines  are the best  fit to  the
 experimental  results in the corresponding  region.  Experimental conditions  and
 results  are  given in  Table  2.   (a)   log-log  scale  showing  pA  ~ const,  in
 Region  I and pA = b(EA)m in  Region II  (Power law  form).   (b)   Data points of
 Figure   4a  plotted  in  semi-log  scale showing the exponential dependence of

 PA " Po  exP (-  aoEA>'

                        POST-MICROSPARK BREAKDOWN REGIME
                                        eQ E>,  and for a
     Our  theory  of  the  electromechanics of  a  particulate  layer  (1,2)  shows
that there is a strong field  enhancement in and around the contacts due  to the
current  concentrations  there.  As  a  result  the  cohesive  electrical  stress  is
much larger  than one would calculate for  a  given  average  field, E.,  on the
basis of  a homogeneous dielectric having  a  permittivity of the  order of  the
layer material.  This is because the  cohesive stress is
highly inhomogeneous field   is much greater  than 

     Alternatively,  one  can  regard  the  layer  as  having a greatly enhanced
effective  permittivity  EA  such  that (1/2)  EA 2  = (1/2) e  .  Moreover,
because the stored energy density in the field  is also , the  energy
stored  for a given average  field  is much larger  than one would  calculate  for a
«n.i|orm dielectric.  (It may be  noted that energy density in  J/m3 and stress in
N/m  have the same dimensions).
                                       297

-------
     Both  the  enhanced cohesive  stress and  the energy  stored are associated
with  the  strong  fields local  to the  contact region.  In effect  the contact
region and surrounding gap  act as a capacitor.  There is  also a large resis-
tance associated  with  the  current concentration  in the  contact region which
makes the  average layer resistivity  p^ much larger  than the intrinsic resis-
tivity p of the material.
     Our previously  presented steady state  theory (1,2)  takes account of the
self-compression  of  the  layer  and predicts  a resistivity characteristic for
the layer  of the  form  (1,2)
                                     E

                                      A
                                                                           (17)
prior to breakdown. Also, using the calculated field enhancement together with
literature  results for  gas discharge  and  vacuum breakdown,  we  predict  the
onset  of electrical  breakdown in  the  form of  a microspark  in  the  contact
region.  These theoretical  results  can  be  used  to  explain the  form of  the
measured resistivity characteristic p^ (E^)  for layers of glass beads as shown
in Figure 4. On theoretical grounds one  expects that for E^ > E^g, microsparks
will  occur  with increasing frequency so  that the  average  current increases
more rapidly than  in the power law region.

     To analyze the layer's current-voltage  characteristic in the region after
the  onset  of  microsparking,  we model  the distributed  real  system of  two
spheres in  contact (Figure  5a)  by a simple  equivalent circuit shown in Figure
5b.
                                              The  contact   region  is   rep-
                                         resented by  a  capacitor C  in  par-
                                         allel  with a contact resistance  R
                                         together with  a  parallel  spark  gap
                                         having a sparking voltage VB-  These
                                         elements are in  series with a  par-
                                         allel  combination  of   C  and  R
                                         representing  the  capacity and  resis-
                                         tance  of  the  bulk of  the  sphere,
                                         away  from  the  contact  region.  The
                                         total  circuit is  connected between a
                                         supply voltage  Vg and  ground.
                                              This  division of  the  distri-
                                         buted  system  into  separate   parts
                                         representing  the contact  region and
                                         the  bulk  of  the  sphere  is,   of
                                         course,   somewhat   arbitrary.    In
                                         practice  the   distributed   system
        (°)                               consists of  infinitely many series-
                                         parallel R-C combinations.  However,
                                         it  does  represent  the  fact that the
Figure   5.     Two  current-carrying     bulk  of  the  resistance  and Capacity
particles   in  contact   and  their     *S  associated with  the  local  region
equivalent circuit,  (a)  Contacting     ln  and around  the  contact where the
particles    and    their    contact     spark  occurs-
region,  (b)  Equivalent circuit.
                           (b)
  ^.—— Equipoeintlal turfacaa
  ...... E - flild (and currtnt) lln.i

  _.—._ E - fl.ld liiui In ch. gap
                                      293

-------
ANALYSIS  OF THE EQUIVALENT CIRCUIT

     In  the  steady state,  for applied voltages Vg  such that sparking does not
occur,  the voltage across  the  contact  region and  spark gap as

                           V  =  V R /(R + R )   .                         (18)
                            c    s c    c     p'                            ^  '

For Vc < Vg  no  sparking  occurs.  As Vg is  slowly  raised  so  that V  =  Vg,  a
spark will  occur which  discharges the contact  capacitance  C  rapidly  and
reduces V to  essentially  zero.

     A transient  analysis of the  circuit gives  the voltage  V  across  the
parallel  combination  of R  , C  and spark gap as

                  Vc  (t) = AQ  exp <-t/T> +  Vs

where              R  R  (C +  C  )       R

                   r         c       -
                            R
and                     p     c         p    c          p    c
                                  T  = R C                                (21)
                                   P    P P
                                  T  = R C                                (22)
                                   c    c c

are the  time constants  of  the  component circuits.  Normally  we expect  the
condition tc »  TD to be satisfied. The coefficient  AQ  depends on the initial
condition assumed.

BEHAVIOR IN THE REPETITIVE  SPARKING REGIME

     Immediately  following  a spark,  the  initial condition is  Vc(t=0)=0.  Then
Eq.  19  gives for the  increase of  Vc:
                                   R
                   Vc (t) " Vs  (R  A )  t1 - «P(-t/T)]  .               (23)
                                 P     c
The total current  (through  the  parallel combination)  is  given by


              \  (t) = (R   +SR ) ^ + £r "  ^  exp (~ t/T)]   •           (24)
                         pc'
In Eq.   23 , when Vg > Vg  (R  + RC)/RC,  the voltage VQ  (t)  reaches Vg and a
spark occurs  after  a time given by         _._


where

                             ^
                                   B   p    c
is a normalized  supply  voltage such that  ^^™  1  corresponds  to  onset  of
sparking  and /^>  1 corresponds  to an overvoltage where  repetitive sparking
occurs.  With increasing c^~, the time tg decreases  and the  rate of sparking
increases.

    The  forms of the voltage  Vc(t)  and  of the total current Ig(t)  are shown
in Fig.  6a and 6b  respectively.  In Fig.  6a,  the voltage  across the  spark gap
is assumed  to  fall instantaneously to zero  at t  = tg  and then repeat its
previous   increase.  Figure  6b  is   drawn for   the   case  RCCC  > R0C_, when

                                     299

-------
         «.•»  /-
                      (b)
 Figure 6.  The forms  of the voltage
 Vc(t)   and  of  the   total   current
 Is(t).  (a)  Dependence  of  Vc(t)  on
 time,   (b)  Dependence  of  Is(t)  on
 time with  RCCC  > RpCp,  for TC/T  =>
 10 •
Is(t = 0) > I(t
       tB)
                         This  is  the
 condition  to  be  normally  expected
 and  shows  a current,  overshoot  im-
 mediately  following  the  spark.

 TIME-AVERAGE CURRENT IN  THE  SPARKING
 REGIME

     In    conventional   resistivity
 cell  measurements on  powders,  only
 the    time-average    current    is
 measured.  In the sparking regime we
 expect   the   average   current  to
 increase more rapidly than if sparks
 did not occur.

     The time-average  total current
 is by definition
B
                         dt  •  (27)

Using  Is  (t) from Eq.  24,  together
with Eq.  26, this can  be expressed
as
                                (28)
                             Note that  at the onset of  sparking
                             when  ^=  1,    =  (VB/Rc)  and
                             defining   a   normalized    current
                             ^=  R /V  we   can  write  the
                             normalized time-average^^- ^char-
                             acteristic  above the  point  (  ^"i
                             !> ^=  1)  for the onset of sparking
                             as

                                    -rle      (I/?*)   ,
     Above   the   point  of  sparking
onset,   the  characteristic ^(^\
departs  gradually  from  the  linear
form/'^'and  for^>>!  tends to an
asymptote (J^*^(Tc/T)   . Two cases
are distinguished according as TC/T
is  greater or less  than  unity.  For
the case  of interest, when Cc » C
and   Rc   > >    R     one  obtains
(TC/T) - (R /R )  >> 1  .   Thus,  the
current   rises1^  more  rapidly  than
linearly   for °^ >   1     tending
eventually   to    a   steeper   linear
asymptote for^>> 1.
                                         TIME-AVERAGE   RESISTANCE
                                         SPARKING REGIME
                                                         IN   THE
                                 Defining  a   normalized  resis-
                            tanceJ^^/jT such   that   j|?=  1
                            for  ^^ <  1,  below  the  onset  for
                            sparking, we have  from Eq. 29
                                         As^increases above unity .^
                                         smoothly from the  value ^0= 1 to a
                                         new    asymptotic    constant   value
                                         (T/TC).  For  the case of  interest,
                                         when  TC/-C ~ (Rc/Rp) » 1, the resis-
                                         tance falls  to  a   lower  asymptotic
                                         value for>> 1.
                            SPARK REPETITION RATE

                                 The  discharge  repetition  rate
                            fB is given by


                             fB= I= " T An (1 -
            In
                                     300

-------
        1  at  sparking onset fB = 0 and increases rapidly to an asymptotic rate
                           ,                     1  .                       (32)
                           D

DISCUSSION OF  THEORETICAL RESULTS

     From  the  foregoing  it  is  seen  that  our  simple  equivalent  circuit
predicts,  for  the reasonable  condition  RC » R  and GC  »  C ,  that the time-
average resistance should  decrease  for voltages  above the  onset  voltage for
sparking  and is qualitatively capable  of  explaining the  observed  decrease in
resistivity of the layer  of  glass beads below the power  law (E™ where m < 0)
form for  voltages  above  sparking onset.  Of course the equivalent circuit model
does not  take account of  self-compression of the  layer as  the steady state
theory does for  applied  voltages below that for sparking onset.

     The   transient  analysis   also  predicts  that   the  sparking  frequency
increases  quite rapidly  for  voltages  greater than  the  onset  voltage,  even-
tually becoming  linearly proportional to"for^>> 1.
     It  must be admitted, of  course,  that the model  is  a simplification and,
as with all equivalent lumped circuit representatives of distributed systems,
it is difficult to clearly  identify appropriate values  of  the circuit param-
eters.

                  EXPERIMENTAL OBSERVATIONS OF MICROSPARKS
     A direct observation  of  the occurrence  of  microsparks in  a  particulate
layer for  applied  fields  EA  >  E^B,  the  onset field,  would provide  strong
evidence  of the  correctness of the theoretical explanation derived in the last
section for the decrease  of  the  apparent  resistivity characteristic  p^
below the power  law  form for high fields.
     Using the resistivity cell, there are basically  two techniques available
for detecting sparks.  The  first  is  to detect  the  electrical pulses  (due  to
microsparks)  superimposed  on  the average  current  in  the resistivity  cell
circuit.  The second  is  to detect  the  light  pulses emitted  by  the individual
microsparks.  Both techniques  have been employed  in exploratory  experiments
described below.

     To   detect  the  electrical  pulses,  a  signal  resistor  (of  100  kQ)  is
inserted  in the  ground return from the cell to  the  high voltage power supply.
The light  pulses  are  detected  by a  photomultiplier tube  (UV enhanced  RCA
1P28B)  with a signal resistor (of 10  kQ)  in conjunction  with a  1  mm diameter
optical  fiber to collect  light from the  cell.  The voltages developed across
these  signal resistors associated  with  the electrical pulse detection system
and the  light  pulse emission detection unit  can be displayed on  a  storage
oscilloscope and photographed, independently or  simultaneously.
    Using this setup,  we  have performed experiments  with 500  urn diameter
glass  beads under  ambient  conditions in which both the  electrical pulses and
the light  pulses  emitted  by  the individual  microsparks  were observed  and
recorded.  The thickness of  the layer was about 5 mm.
                                     301

-------
     Figure 7 shows oscilloscope  traces  of  the electrical pulses for different
values of  the applied field E^ across the layer,  for 500 urn  beads  under  am-
bient air conditions.  In  Figure 7 (a) at EA ~ 10 kV/cm,  there are tens of  low
amplitude forward  pulses  with  a  smaller number  of reverse  pulses.  Figures 7
(b)  and  (c)  for  EA  ~  12  and  14  kV/cm,  respectively,  show  a  progressive
increase  in  the  number  of   microspark   pulses  and  their  amplitude with
increasing applied  field.  The  upward  or forward pulses  correspond to current
flow  in  the  same  sense  as  the  average  current through  the  layer,  and  are
current  pulses  due to microsparks.  They are  dominant and have  the  effect of
increasing the time-average current  through the layer. The downward or reverse
pulses have,  generally, a  much lower amplitude and  occur less  frequently than
the  upward  pulses. They  are  fast  and  similar  in  form  to  the  upward pulses
except for their sign. At  present, we do not  understand  the mechanism respon-
sible for this phenomenon.
     Figure  7(d) shows the waveform of  an individual  (forward)  pulse at E» ~
14  kV/cm which has a  very sharp  rise (~  15  to  20 us),  corresponding to  the
discharge  period,   followed  by  an  exponential  decay  (corresponding to  the
charging period) having a  time constant  of  ~ 160  us. The  fast rise time in  the
discharge  period,  supports  our assumption of instantaneous discharge in  our
theoretical analysis.
     Similar  experiments were  carried out using  a  single layer of glass beads
of  6mm   diameter,  under  ambient air conditions,  which, generally,  give  the
same form of  results.
     The  measured  spark repetition  rate,  fg,  as a function of  the (applied)
average  field,  E.,  for  a  single layer of  6mm glass  spheres  under  ambient
conditions is shown  in Figure 8, which  indicates that fg increases (linearly)
with increasing EA, in qualitative agreement  with the theoretical predictions.
Here, fg has  a value between ~5 to 25 Hz.

     In  addition, a few experiments  were carried  out in which the oscilloscope
traces for both light  and  electrical pulses were recorded for different values
of  the applied field,  EA,  across  the  layer.  Figure  9,  shows  an example of
these  tests  with  (one layer of)  6mm beads under  ambient conditions  at E^ ~
11.67  kV/cm.  Here,  the lower  trace  shows  the electrical pulses  and the upper
trace represents their corresponding light pulses.

     In  this  photograph  a single light  pulse is  shown which is clearly asso-
ciated with an electrial pulse. However, there are two other electrical pulses
for which no  corresponding light  pulse can be observed in the photograph. This
is  due  to the fact that  the 1mm optical fiber,   used for transmission of  the
emitted  light pulses  to the  PM tube, only  "sees"  a small portion of the sample
exposed  to  it and  cannot detect the  light pulses occurring at  a distant  lo-
cation.  In  Figure 9,  the light  pulse has a  sharp rise  corresponding to  the
discharge  period,  followed  by a  relatively  slow  exponential decay  having a
time constant of ~ 10  to  20  ms, as opposed to ~ 100 us for electrical  pulses.
     This  photograph  and the results  of other similar  experiments clearly
indicate  the existence of microsparks and pulsed  light  emission in a partic-
ulate  layer  for applied fields EA > EAB,  the breakdown  onset field.  Further-
more,  it  is  found  that the spark repetition rate increases with  increasing  E^«
                                     302

-------
uo
o
       Figure 7.

       10 kV/cm;
Electrical pulses.

(b)
                                                                               SOOmV
               500fjm diameter glass beads under ambient  conditions.   (a)   E
       at E ~ 14 kV/cm.
E ~ 12 kV/cm;
 A
(c)
E ~ 14 kV/cm;
 A
(d)   waveform of forward electrical pulses

-------
   [HI]
                        10
                                 IS
Figure 8.   Experimental increase in
the  average spark  repetition rate,
E
B'
A'
with  increasing   the   applied
The solid line is  the  best fit
to the data points.
Figure   9.     Upper   trace,  light
pulses;   lower   trace,   electrical
pulses.   6  mm beads,  ambient con-
ditions.  EA ~ 11.67 kV/cm.

       SUMMARY AND CONCLUSION
     The  first  part  of  this paper
presents   a   theory  of  electrical
breakdown  of  precipitated ash layer
which   appears   to  be  capable  of
explaining  quantitatively,  in terms
of fundamental parameters,  the onset
of  back   discharge  as  microspark
breakdown in and around  the contacts
between particles where  the field is
greatly  enhanced above  the  average
layer value. This theory enables one
to  predict the  conditions required
for  the  onset  of   the microspark
breakdown   of   the  layer  and  our
calculations   are    in   reasonable
agreement   with   the   experimental
results.   The   theory   treats   the
generalized case of combined  surface
and  volume  conduction  and  can  be
easily  extended   to  consider  the
effect  of   different  modes  of  com-
paction (2.).
     In an  extension of the  theory,
the  post-microspark  breakdown  be-
havior of the layer is qualitatively
explained,  in  terms  of a simplified
equivalent   lumped   circuit.   The
theory    predicts    increases    of
sparking    frequency   and    average
current  (and  hence  decrease  of p^)
as  the applied  average field,  EA,
exceeds the threshold average field,
                                          AB>
                                                     onset
                                                                breakdown.
                                              Experimental   observations  of
                                         microsparks confirm the existence of
                                         electrical  and  light  pulses  asso-
                                         ciated  with them.  Moreover,  in all
                                         cases   the   spark   repetition  rate
                                         increases (linearly) with increasing
                                         E^, as  predicted by our theory.

                                                   ACKNOWLEDGEMENTS
                                             This  work was supported  by  the
                                        Electric Power Research Institute.

                                             The   work  described   in  this
                                        paper  was  not  funded  by  the  U.S.
                                        Environmental  Protection Agency  and
                                        therefore   the  contents   do   not
                                        necessarily reflect the views of the
                                        Agency  and no official  endorsement
                                        should be  Inferred.
                                      304

-------
                                  REFERENCES

1.  Self,  S. A., Moslehi,  G.  B., Mitchner, M.  and Leach,  R.,  "Electromechanics
    and Reentrainment of Precipitated Ash," Proc. International Conference on
    Electrostatic Precipitation, Monterey, CA, October 14-16,  1981.

2.  Moslehi, G.  B.  and Self, S. A., "Electromechanics of Particulate Layers,"
    4th Symposium  on Transfer and  Utilization  of  Particulate  Control  Tech-
    nology,  Houston,  Texas,  October 11-15,  1982.  (Session  C-6,  this  con-
    ference)

3.  L. L.  Alston,  editor, High-Voltage Technology,  Chapter 4,  Oxford Univers-
    ity Press  (1968).

4.  McLean, K. J.  and Huey, R. M., "Influence of Electric field on the Resis-
    tivity of a Particulate Layer,"  Proc.  IEE,  Vol. 121, No.  1,  pp.  76-80
    (Jan.  1974).
                                   305

-------
                    ELECTROMECHANICS OF PARTICULATE LAYERS

                      By:  G. B. Moslehi and S. A.  Self
                   High Temperature Gasdynamics Laboratory
                      Mechanical Engineering Department
                             Stanford University
                     Stanford, California  94305, U.S.A.
                                   ABSTRACT
    In  previously  reported  work,  a  comprehensive  theory  of  the electro-
mechanics of a particulate layer was developed, treating it as a regular cubic
array of  equi-sized resistive spheres,  allowing  for both  volume and surface
conduction, and  taking  account of  self-compression of the  layer.  The theory
gives expressions for  various quantities of  interest including the contact
angle 9O,  the maximum electric field Emax, the average compressive stress PE
and  the  apparent resistivity  pA,  all in  terms  of the average layer current
density JA (or average field E^ = p^Jf^) anc* fc^e system parameters.
    In the present  paper  this  theory  is  extended in two ways. First, explicit
expressions for  the quantities 90, Emax,  PE  and  p^ are given  in the form of
power laws (in  EA) in which  the  coefficients  and  exponents depend  on the
system parameters,  in particular the  resistivity ratio a  =  sa/p,  where a is
the particle radius and s and  p are, respectively, the (intrinsic)  surface and
volume (material) resistivities  of  the particles. These expressions should be
operationally useful  for interpreting experimental data.  Second,  the theory
has  been  extended to  include all of the  six classical modes  of compaction for
equi-sized  spheres. The  results  are  qualitatively  similar  to  those for the
cubic array and are related to them by a set of multiplicative constants which
are of order unity.

    Measurements of the resistivity characteristics p.  (E.) of layers of glass
beads confirm the theoretical  results both qualitatively and  quantitatively.

                                  BACKGROUND
    In  this  section,  the  results  of  the  theory  of  electromechanics  of  a
current-carrying particulate  layer are  summarized. A detailed account  of  this
theory  is  given  in  References 1 and 2,  where the  particulate  layer is  modeled
as a  regular  cubic  array of uniform spheres through which a steady current is
                                      306

-------
flowing,  as illustrated in Figure 1, which shows  the distribution of electric
field  and  current lines.  Because  the current  must flow  through  the  small
contact  areas between  spheres,  the current  and field lines  are concentrated
near the contacts, and most  of  the resistance  and  voltage  drop is associated
with these regions.

    The  theory  for  the cubical  array was  based on  the distribution  of  po-
tential  V  for  a single sphere  of  radius  a with  point contacts  at opposite
poles  where current I  enters and  leaves  (Figure 2). For combined volume and
surface  conduction the  potential  is (1,2)
                                           «T — T?/ **•  ff  \\\ —
                                                                           (1)
Here z  =  (r/a),  |i  = cos  9,  p is  the volume  resistivity, s is  the surface
resistivity and  the nondimensional resistivity ratio  a = (sa/p);  a measure of
the relative contributions  of  surface and volume conduction. Figure 3  shows
computed results  for F(9),  for  points on the  surface with a as a parameter.

                                             This  solution  for  V(r,9)   was
                                         used to find an approximate solution
                                         for   the case  of  a finite  contact
                                         angle 9Q,  from  which  the  electric
                                         field distribution in and around the
                                         contact  between  spheres  was  calcu-
                                         lated.   From  this   field  the  elec-
                                         trical  attractive force  between two
Figure  1.  Model  for  current  flow
through array of  resistive spheres.
                                             o
                                                         B (degrees)
    SPHERICAL
      SHELL
Figure 2.  Model  for  current  flow
through  single   sphere  with  point
contacts  showing equipotentials.
Figure  3.  Function F(cr,9)  for po-
tential distribution at the surface
(r=a)  for  current  flow  across   a
sphere.   For
shown  as   a
F_ = Lim (F/a).
 s
comparison
is
broken  line,  where
                                    307

-------
spheres was calculated as a function
of contact  angle  9Q.  Then using the
Hertz  formula  from  elastic  defor-
mation theory,  the  contact angle  9Q
due  to self-compression  was  deter-
mined.

    In ' this  manner self-consistent
equations   (with  respect  to  self-
compression)  were  derived for  the
following   quantities  of  interest:
(i)  the  contact  angle   9Q  between
spheres;  (ii)  the  average  resis-
tivity  pA  of  the  layer;  (iii)  the
average compressive stress Pg in the
layer  and the attractive electrical
force  between two contacting parti-
cles   Fg   and   (iv)  the  maximum
electric  field  Efflax  in  the  gap
between contacting  spheres (2)
9  =D1/V2/5
 o        (o
     pF(a,UQ),forego
v

0
E   -
 max
        2/5
(2)

(3)

(4)

(5)
Here F   = Lim(F/cr),  and D and A are
           O+O
constants,  being functions  only of
the  sphere radius,  a,  and  its ma-
terial  properties.  The  character-
istic electric field £f is defined as

                                  (6)
where  F  and F1 are evaluated at z  =
1,  F'  = dF/d|j, and K = (F/F1)^2,. It
may be  noted  that  Eqs.  2  and   6
constitute  a pair of implicit  equa-
tions  determining 9Q andp7.

    The  field  enhancement  factor FEF
=  (Emax/^A^  as a  function  of contact
angle  9Q and resistivity  ratio a is
numerically   evaluated   and   the
results  are shown in Figure  4.
                                           o-
                                                       00(degrees)
Figure  4.  Field  enhancement factor
(FEF) dependence on a and  9Q.

    The value  of D in Eqs. 2-5, can
be evaluated  from D = it eQAB, where
e0 ~ 8.854 x 10~12 F/m is  the vacuum
permitivity,   A(o)   is  a  constant
given in Table 1,  and B is given by
B = 3(1 - v2)/4Y. Here, Y  is Young's
modulus and v  is Poisson's ratio.
    In  the  special  case  of  pure
volume  conduction  (o  *  °°),  which
applies for high temperature and low
humidity,  we  have  K •»  K^ *  1,  so
that  ^« EA  which when substituted
in  Eqs.  2-5  results  in  a  set  of
explicit  equations  (2).  Here,  the
values  of  A (and hence D) at a * "i
denoted  by  A^  and  Dy,  should  be
used.  In  the  case of  pure surface
conduction  (0=0),  which applies for
low  temperature and  high humidity,
the  quantity  K  •»• Kg  is  not a uni-
versal  constant  but is  weakly de-
pendent  on  E^  and on  the material
                                     308

-------
TABLE 1.  EXPLICIT FORM OF  THE SET OF SELF-CONSISTENT ELECTROMECHANICAL
                  EQUATIONS  OF  A PARTICULATE  LAYER
e0 = be D<«/:
r

lmF
bF(sa)
bp<*/4)
Z> 
-------
behavior  of the  layer  under these  limiting  conditions. Equations for  the
general  case  of  combined  volume  and  surface conduction were  also  given.
However,  these relationships  are not  explicitly  expressed  in  terms of  the
(applied)  average  field E^  (or  the average current  density  J^). They are in
the  form of a set of  transcendental equations  which must  be solved  numer-
ically.  This  complexity of  the  governing equations of the  layer  clearly
creates some inconvenience in practical  applications.
    In  this  section  we numerically solve  the  set of  general equations of  the
layer. It is shown that  in the range of  contact angles of practical  importance
all  the  quantities  of interest depend on  the  average field E^ according to a
power law. Applying a least squares curve  fit  to  the numerical  results the  set
of electromechanical  equations  of the layer are  derived  in the form of power
law  relationships  whose exponents  are  functions  (of  a and hence)  of temper-
ature and  humidity. These explicit expressions are useful from an operational
point of view  in interpreting ash resistivity  characteristics,  for example.

POWER LAW EQUATIONS

    A  close examination of  Figures 3 and 4,   where  log-log  scales  are used,
reveals  that,  to a close approximation,  in the range  of  the contact angles of
practical  interest (9Q  ~ 0.05°  to 1°)  the  functions  F(o-,9Q) and  FEF(a,90)
depend on  9Q according to a power law of the form
and                                    n
                   FEF  (ff.e) - a     (9EF)                               (8)
                           o     pEF


    It  follows  from the definitions of  FEF ( = F'/F) and K  (=  (F/F')/9^) that
the dependence of the functions F'(o",9o) and  K(a,9Q)  on 9Q  also have power law
forms

                      *'  (a,9o)=aFl(9"Fl)                                (9)

                       K  (a,9Q) -  ^(9^)                                 (10)

    In  the above relationships both  the exponents  n and the  numerical coef-
ficients a are  functions  of  or. A least squares curve fit procedure is applied
to the  numerical results for  F,F', FEF and K (for the  contact angles between
0.05° and  1°) and values of a  and  n are  determined for each case.

    The next step  is  to  relate the contact  angle 9Q and the  applied average
field EA in an explicit manner.  Using Eqs. 6 and 10, with Eq.  2,  after some
simplifications, gives

                        9  =  b  D(«/2)(E )«                               (11)
where                                   A
                          o  =  2/(5 + .21)                                   (13)
                                    310

-------
    Substitution of Eq.  11  into Eqs. 7-10  results in  a  set of  explicit ex-
pressions  for  the  functions  F,  F1,  FEF and K in terms  of EA
                  F (o-,EA) = bF D      (EA)                               (14)

and                               (m.,/2)    m..
                  K (a,EA) = bK D      (Ej K  .                          (15)


    The  expressions  for F'(
-------
    To evaluate  the parameters  of  interest  (9O>  pA,  PE,  E^^  etc.)  as  a
function of EA, for a given value of cr, one can find the values of A(a), a, mF
and coefficients bQ, bp,  bp,  and bg from Table  1.  Then one can calculate the
values  of   the  elastic  properties  constant  B  and  the  material  properties
constant D.  Finally,  for the  specified value  of  a,  the  functions 9Q(EA),
PA(EA)» PE(EA) and Emax (EA)  can be established, using Eqs. 11, 18, 16 and 17,
respectively. In general,  a  knowledge  of  particle radius,  a,  surface resis-
tivity, s,  and  (intrinsic) volume resistivity p  of the particle material and
their variation with temperature and humidity is necessary for evaluation of a
(and pA).
    From Table 1 one may note that the power law dependence of 0O, ?„ and E
on EA  are  rather  insensitive  to a, because a varies  very slowly with a (by a
maximum factor  of  ~ 1.13).  The coefficients b  vary rather  more  (by maximum
factors of  ~ 2.2 and  ~ 1.5  for  b«  and bg,  respectively) but  no  more than a
factor of  ~ 11 (in the case of  bp).  In the case  of  pA, the power law depenr
dence  on a is more pronounced. Here,  the exponent  mp decreases from ~ - 0.07
for surface conduction (0=0) to  ~ -  0.4 for pure volume conduction (CT •»• »);  a
factor of ~ 5.7 difference. The variation of bp with a is relatively slight (a
maximum factor of ~ 6).

        THE EFFECT  OF PACKING MODE ON THE ELECTROMECHANICS OF THE LAYER

    Our analysis so  far applies to the  case of  a regular cubic array of uni-
form  spheres. However,  in practice the particles  are usually more densely
packed which  suggests  the  necessity  of studying  the effect of packing mode of
spheres on the behavior of the layer.
    In  this  section,  the electromechanical equations  for  all  six classical
regular arrays of equi-sized spheres are derived in a unified form.
    Of the  many types  of packing of identical  spheres, we consider here only
regular  stable arrays.  Such  arrays are composed  of repeated  planar layers
which  are geometrically systematic (3).
    The parameter which characterizes the type of a (single)  layer is the base
angle,  oc^,  which is the  angle  between the two  sets  of rows having different
directions.  This angle  has  a  value between  60°  and  90°.   In  the following
analysis we  consider only  the  two particular cases of a  square base layer (otj,
=  90°) and a simple rhombic base layer (a^ = 60°).  It  may be noted that the
number of  contacts N^  between  any sphere and its neighbors  in the layer is N|j
=  4 for the square based layer and N^ =  6 for the rhombic  based layer  (Figures
5a and 6a).

    For each  of  these  cases  there are three simple geometric ways of  stacking
adjacent  layers,  thus producing  a  total  of  six  different types  of array.
Topologically, only  four  of  these are  distinct,  since two of the  square based
arrays  are equivalent  to two of  the rhombic  based arrays, but have different
spatial  orientations  (3,4).  However,  since we  are interested  in the appli^
cation  to current flow normal  to  the base  plane (resting for instance on an
electrode), we must  consider  the six cases as distinct.  In  this connection it
may also be noted  that the N^ contacts,  mentioned above,  do  not  carry  current,
since  they  lie on  the  equipotential mid-plane of  the  layer.
                                      312

-------
    These  -.six  classical  modes of  packing  for  spheres are  illustrated  in
Figures  5  and  6  for  the cases  of square  and  rhombic bases,  respectively
(3,4,5).  Each mode is  designated  by a  letter, S (= square)  or R (= rhombic),
followed  by  a  number  1,  2  or 3 which  refers to  one of  the three  ways  of
stacking  the  next  layer.  The two pairs of cases  (S-2  and R-l)  and (S-3 and R-
3) are the ones mentioned  earlier  which are topologically  identical but have
different  spatial  orientations.
    At this point  an additional complication should be mentioned;  that is the
concept  of twinned or  tripled  packings, as  opposed to the  basic  or standard
packing.
    Figures 7(a)  and 7(b)  illustrate two alternative ways of  stacking (I and
II) for orthorhombic  (S-2)  and  rhombohedral  (R-3)  arrays,  respectively. In the
standard  packing  either one of these ways is used exclusively,  and the whole
array  is  symmetric about  the mid-plane  of any layer.  However,  it  is possible
to create twinned arrays with  the  sequence 0 I II  or 0 II  I  where 0 denotes
the base  layer,  which do not   have  the basic  symmetry.  In the case  of the
tetragonal-sphenoidal array (Figure 7(c)) there are three alternative ways of
constructing  the  next layer which  gives  rise to two  types  of  twinned packing
and one type  of tripled packing, with the sequence 0 I II III for example (3).
                             (c)
Figure 5.  Regular arrays  of  iden-
tical  spheres   (3,4).  Square   base
layers (S).  (a)  simple cubic (S-l);
(b)  orthorhombic  (S-2);   (c)   pyra-
midal (S-3).  Top  view;  solid  lines
represent  the first layer and dotted
lines the  second layer.
   mm
   (a)
(b)
(c)
Figure  6.  Regular arrays of identi-
cal  spheres   (3,4).  Rhombic  base
layers  (R).  (a) cubical-tetrahedral
(R-l);  (b) tetragonal-sphenoidal  (R-
2);   (c)   rhombohedral  (R-3).   Top
view;   solid   lines  represent   the
first  layer  and  dotted  lines   the
second  layer.
Figure   7.   Alternative   ways  of
stacking  of  identical  spheres  (3).
(a)  orthorhombic  array  (S-2);   two
alternative  ways,  (b)  rhombohedral
array  (R-3);  two  alternative ways.
(c)   tetragonal-sphenoidal    (R-2);
three  alternative  ways.  Top view;
numbered spheres represent  the lower
layer, denoted by 0.
                                      313

-------
    In the analysis which follows, we consider  only  the standard packing for
all cases  except for  the rhombohedral array  (R-3),  where we  treat  only the
twinned packing.  The  reason  for this exception is that  from the electrical
point of view the twinned rhombohedral array has the desired form of symmetry
which is amenable to a simple analysis.
    For the set  of packings considered,  the 2N current-carrying contacts that
each sphere makes with the layer above and the layer below can be arranged in
pairs which form current  paths  lying along a diameter of the sphere. Further-
more, when N = 1 the  colatitude angle ac that this path makes with the z-axis
normal to the layers is zero,  while when N > 1, the angles occ for each current
path are identical but the paths have different azimuthal angles. This fact is
essential for the electrical  analysis, developed later,  and is the reason for
restricting the class  of arrays to those mentioned.
    In general each sphere has a total of N^  (= N^ + 2N)  contacts with other
spheres, which is sometimes called  the coordination number.  Values of Nt, N^,
N and cos ap for the six arrays considered are listed in Table 2.
       Table 2. CHARACTERISTICS OF REGULAR ARRAYS OF IDENTICAL SPHERES.
BASE
              SIMPLE SQUARE BASE  (S)
SIMPLE RHOMBIC BASE  (R)
                                                     Nv
                                                      6
Type
of — »»
Array

Nt
N
cos ac
P
Ac
AC
vc
e0/(e0)
Emax/CE

PA/ (PA>
Simple
Cubic Orthorhombic
(S-l) (S-2)
Standard Standard
6 8
1 2
1 /T/2
0.4764 0.3954
4a 4a2
2a /5a
8a3 4/5a3
c 1 0.9441
max>c l 0.9716
c 1 1.4575
c 1 0.5774

Pyramidal
(S-3)
Standard
12
4
/f/2
0.2595
4a2
4/2a^
0.8706
0.9330
1.8661
0.3536
Cubical-
Tetrahedral
(R-l)
Standard
8
1
1
0.3954
2/Ja2
4/3a3
1
1
1.1547
0.8660
Tetragonal-
Sphenoidal
(R-2)
Standard
10
2
7372
0.3019
2/3"a2
6a3
0.9441
0.9716
1.6829
0.5000

Rhombohedral
(R-3)
Twinned
12
3
/27/J
0.2595
2/Ja2
2/2a/^"
4/Ia3
0.9221
0.9603
2.2176
0.3536
    For  our purposes we  define  a unit cell  with an  effective  base area AC,
effective height Ac and volume Vc = ACAC-
    By geometry it can be shown that
and
                AC  =  [Nb  tan  (n/Nb)]a2

                Jl  = 2 a cos a
                               (19)

                               (20)
                                     314

-------
                                                                            o
    Since  the volume of a  single  sphere occupying the cell is V   =  (4/3)ua ,
the porosity P can be evaluated from P  = 1  - Vg/Vc = 1 - Vg/AcJ!.c. The  values
A , JL,  Vc and P are  listed in Table 2 for the  six simple  arrays considered.

ELECTROMECHANICAL EQUATIONS - GENERAL FORM
    For the basic cubic array each sphere has a  single  diametral current path
(N = 1) parallel to the z-axis  (normal  to  the layer) so that cos ac =  1,  and
the basic electromechanical equations were solved to give the derived  quanti-
ties 90,  pA,  PE  and Emax (Eqs. 2-5) for  this  case.
    For the  other  five regular  arrays  to be considered now,  the number  of
diametral  current paths N varies up  to  N = 4 and they make  different  colati-
tude angles «c to the z-axis with the N paths disposed  at different azimuthal
angles  $ around  the z-axis.  We  now have to  take these  facts into  account  to
determine  the modifications to the expressions 2-5 for  the derived  quantities
for the five  additional cases  considered.
    The basic solution for the potential due to  current flow between one pair
of diametrically opposed point contacts  (Eq.  1)  can  be  used  to  find the total
potential  due to a set of  N  diametrically opposed point  contacts (at different
angles  ac  and <(>) by  superposition. However, because  the potential only varies
rapidly in the  neighborhood of  each current-carrying  contact,  it  is  a  good
approximation for calculating the fields and stresses around each contact,  to
ignore  the contribution to the potential due  to the other current-paths. Thus,
the behavior in  the region of each  contact  can  be  treated as  decoupled  from
the effects of the presence  of the other contacts. With  this  approximation the
treatment  for the simple cubic array  may be generalized  as  follows.
    The potential drop across each half unit cell is obtained from  Eq.  1 with
z = 1 and  |i = \iQ as

                      p [(l/N)/(4a)]  F (o-.p.  )    ,   for a*0
                 V  =                      °
                  °    (sa) [(l/N)/(4a)]  F (u ]   ,   for o=0
                                         s  o

where (l/N)th of the current  I per cell  flows through each  contact.  ,Hence,  the
resistance per sphere is  R  = 2  VQ/I. Equating this  to  the average  resistance
per unit cell PA.^C/^C» we  find for the average layer  resistance

                      [(A /A  )/(2aN)]  p F (ff.ji )   ,    for o*0
                        C                     °                            (21)
                      [(A /I  )/(2aN)](sa) F f|i )   ,    for o=0
                        C   C              SO
    The average  compressive  stress is given by

                            PE =  [NFE (cos  OC)]/AC                        (22)

where Fg  is the  attractive  electrical force  at  one  contact, in the direction
of the  current path.
                                    315

-------
    In the earlier work for a cubic  array,  we defined a  characteristic field
(via Eq.  6)  in terms  of  E^ (which  in that case  was parallel to  the  current
path) to determine (via Eq. 2) the contact angle 9Q.  In the general case,  when
the current path makes an angle «c to EA, it is clear that we can use the same
formalism to calculate 9Q,  but that we  should use (EA cos «c) in  place of EA
in Eq.  6. Thus  in  general, Emax  and FE can  be  evaluated  in a similar  way.
Collecting together the results for the derived quantities of interest,  making
use of Eqs. 19  and 20, we have in general, in place  of Eqs.  2-5  for the basic
cubic case:

     9Q = D1/5  (cos aJ2/5g*2/5                                            (23)

          {Nbtan(it/Nb)]/[4N(cosac)]}pF(a,uo)    ,  for a*0


      A   {[Nbtan(7t/Nb)]/[4N(cosaJ]}(sa)Fs(M,o), for a=0

     PE = {N/[Nb tan  U/Nb)]} it eQ A  (a) D~2'5(cos c^)11'5 gf7             (25)

     E    = D~2/5  (cos a )1/5 Cfl/5                                        (26)
      max          *•      c'   0

In these equations cos 
-------
higher than for the simple  cuhic  array).  On the other hand the average layer
resistivity decreases  with increasing number  of  parallel current paths  N,  as
might be expected, being lowest for  the pyramidal and rhombohedral arrays  (a
factor ~ 3  lower  than  the cubic array).

    When equi-sized spheres  are packed under gravity,  for instance in  a  resis-
tivity cell,  it  is  probable  that  the packing mode  approximates the rhombo-
hedral or pyramidal types which have  the minimum  porosity, so  that the quanti-
tative results for these cases are most appropriate to use for comparison with
measurements.

                        EXPERIMENTS  WITH  GLASS BEADS
    The theory  described in  the  preceeding  sections  applies  to  equi-sized
spherical  particles which is,  admittedly,  rather  idealized.  However,  treatment
of the case  of  a randomly packed  bed of polydisperse  spheres  appears to  be
quite  intractable.  Therefore, it  is  important  to  devise  some  experimental
tests of  the theory to check how far  it describes  the behavior  of particulate
layers in  practice.
    One aspect  of  the theory  which  is amenable  to a  fairly straightforward
test is the prediction of the dependence of the apparent  layer  resistivity p^
on the applied  field  E^,  and how this  varies  with  the ambient  conditions
through the ratio,  a,  of  volume to  surface conduction.
    Using  a standard resistivity cell (ASME Power Test Code 28)  we  have  made
measurements  of  p^> as  a function of  E^. For these experiments,  glass beads
(450-500  p,m dia.) are  used rather than fly ash,  which is a glassy  material.
Since the  beads  are monodisperse, smooth spheres of  uniform properties, it is
to be  expected  that more reproducible and  interpretable  results should  be
obtained.  In  each  experiment  a  new  layer  of beads was used  from the  same
supply sample.  The procedure used was  standard.  In  particular the  weight
loading the top  electrode was  designed to compress  the layer to  the standard
compressive stress  of  ~ 10 g wt./cnr.  In addition,  the  layer was  ~ 5mm thick.
In each experiment  sufficient  time  (~  1 hour)  was  given  to the sample in order
to reach  equilibrium  with the conditions of  the  oven (i.e.  temperature and
humidity).  The electric current readings were  also  performed after a period of
~ 2  minutes  following  each step change made  in  the magnitude of  the  applied
voltage. This  period is found  to  be long enough to  let transients  die out.
    Figure 8a shows a  few examples of the measured  apparent resistivity, p^,
of the layer as a  function  of the (applied) average  field, EA, for  different
values of (dry  bulb)  temperature,  tdb, in  range of ~  20°  to 185°C.  Table 3
summarizes  the experimental conditions in these tests  and  a number of addi-
tional experiments which are  not  shown in Figure 8a. This  table gives the
values of  tdb and relative humidity Pr for  each test. For the cases  in which
t^ < 30°C the conditions of the oven were identical to those of  ambient  air,
except for Experiment  #1  in which the  humidifier was used in order to  achieve
a humidity of ~ 100% at  tdb ~ 28°C.  In all  the other cases where tdb  > 30°C,
the  ambient  air was heated  to  the  specified oven  temperature,  without any
change in  the specific  humidity.  However,  the  relative humidity  in these cases
is varied,  with  the lower values corresponding  to a  higher oven  temperature.
The general features of P^(E^)  characteristics are as follows.
                                   317

-------
                                              O
                                                      = 23°C.IJ=88.57.)
              A
           [a- cm]
00
00
                                              X EXP.-ll C.£-88.5X)
                                                                           « £XP..7n-2?.S*C.£«8i.621
                                                                           X EXP..11 (T-61* C.£-12.121
                                                                           D EXP..12IT-119*C.£-1.221
                                                                           * EXP..13IT-l33'C.£-0.821
                                                                           * EXP..16 (T-185'C.£-0.221
                                                                           ' CROSS BREAKDOWN OF THE
                                                                             LflTER
                                                                                                                               O
                                                                                                                               I
                                                                                                                               10
                                                                      1.0. ID"
                                                                                   1.0-105
                                                                                EA. (v/mj
                                                                                          (b)
          Figure 8.
          - 500'Vim.
             Resistivity cell measurements at  different values  of temperature;  Glass beadsj D   ~ 450
             Experimental conditions and results are given in Table  3.   (a) log^log  scale shoeing
p  =  b(E  )m in  Region II  (power  law form).   Solid (or dotted) lines are the best fit to the
experimental results.  (b)  Normalized  resistivity data.   Solid lines represent the theoretical
results of power law expressions (for  Region II).

-------
 TABLE 3.  RESULTS OF LEAST SQUARES CURVE  FIT TO  THE  EXPERIMENTAL DATA POINTS








Oven
Condition
No.

i
2
3.
4
5
6
7
8
9
10
11
12
13
14
15
16
'db.
[°C]
28
23
22.
23
22.
22
22.
29
46
46
61
119
133
151
162
185
pr«
[%]
100
~ 88.5
5-82
~ 81
5-80
~ 78
5 ~ 81.6
~ 70
23.2
22.9
12.1
1.2
0.8
0.5
~ 0.4
~ 0.2


Region I,
PA = const.
PA.
[Q - cm]
9.24 x 107
4.56 x 1010
5.84 x 1010
9.82 x 109
3.07 x 1010
9.83 x 10l°
4.25 x 1010
5.17 x 1010
__—
____
1.27 x 1012
7.12 x 1012

____
4.54 x 1012
4.60 x 1012
Region II
PA = b(EA;
PA "in [Q

)m
= cm]
EA in [kV/cm]

b
7.54 x 107
2.53 x 1010
4.97 x 1010
6.95 x 109
2.07 x 1010
7.92 x 1010
3.37 x 1010
4.44 x lo"
6.28 x 1011
5.56 x 1011
6.14 x 1011
3.55 x 1012
3.64 x 1012
3.41 x 1012
2.71 x 1012
2.12 x 1012

m
-0.077
-0.172
-0.175
-0.210
-0.230
-0.230
-0.235
-0.259
-0.352
-0.337
-0.344
-0.408
-0.407
-0.382
-0.337
-0.420

Region III,
Gross
Breakdown
EGB»
[kV/cm]
____
15.8
16.0
13.4
16.0
	
15.8
16.0
16.0
16.0
15.4
13.8
12.6
13.4
12.0

(--)  means: (i) "No Data Taken" or (ii) "No Data Available"

   We can divide the PA(EA) graph into three regions:

i.   For  an average  field  between  10  V/m and  (3-4)xlO  V/m,  the apparent
    resistivity in most  cases  is approximately constant  (Region  I  in Figure
    8a).  This is due  to the  fact  that the  compression resulting  from the
    electric  stress  is  less than that  due to the weight of  the  upper elec-
    trode. In fact,  from our theory the average  compressive stress of ~ 10 g
    wt./cm ,  corresponds to EA ~ 3.6 x  10  V/m  for the  case  of  pure volume
    conduction (a ->  »)  and EA ~ 1.73 x  10 V/m  for  the case of pure surface
    conduction  (0 -»• 0).  For   intermediate  values  of  0,  the  value of  EA
    required falls between  these two  limits;  for instance,  from Table 1, for
    a = 102 one obtains EA ~ 6.26 x 10* v/m, for PE ~ 10 g wt./cm2. Values of
    PA in Region I for each experiment are given in Table 3.

ii.  In the second region (Figure 8a), for 4 x 104 v/m £ EA £ 2 x 105 V/m, the
    PA(EA) characteristic  follows  a  power law form, pA  »  b  (EA)m>  with the
    exponent m (= rap)  having a value  somewhere  between - 0.077 for the case
    of surface conduction in Experiment  #1 (tdb  * 28°C, Pr = 100%) and - 0.4
    for  the   case  of  volume  conduction  (hot and   dry  conditions  with low
    relative humidity Pr). Values of  m and b,  for each test are evaluated by
                                    319

-------
      applying a least  squares curve fit  to the experimental  data points in
      the aforementioned range of EA and are given in Table 3. These values of
      m are in reasonable agreement with our theoretical predictions (for rap)
      which are summarized in Table 1.  In  may be noted, from Table 3 that the
      value of  m  decreases with temperature as  expected from  theory.  These
      results  clearly confirm  both qualitatively and  (approximately)  quanti-
      tatively the power  law dependence of  pA  on EA in  the  second region of
      the  PA(E^)  characteristic,  as  is explicitly  predicted by  our  theory.
      This non-ohmic  behavior is due to self-compression of the layer.

iii.  In the  third  region (Figure 8a), at  EA  = EAB ~ (2-4) x  105 V/m,  the
      PA(EA) curve starts  to deviate from the  power law dependence,  with p^
      decreasing more rapidly with increasing (applied)  field EA- This is due
      to the  electrical breakdown of  the layer  which  occurs in  the  form of
      intermittent microsparks,  as discussed in  the companion  paper  (6).  The
      gross breakdown of  the layer occurs at EA =  EGB ~ (12-16)  x  10  V/m,
      with  the  lower values  corresponding  to  higher   temperatures and  drier
      conditions.

    In Figure  8b  the results  of  Figure  8a are  compared with  each other by
plotting the normalized  resistivity (PA/PA   ) versus EA, where pA    is taken
as  the  (reference) resistivity  for E.  =   10   V/m.  Also  shown  are the theo-
retical lines  for  Region II of the PA(EA) characteristic  (the power law forms)
with a as a parameter, from Table 1.  In addition, the results of Experiment #1
(T ~ 28°C, Pr  ~ 100%) are also plotted.
    Generally,  it is seen that  in the pre-breakdown regime  the experimental
results fall within  the envelope of  the limiting theoretical lines for volume
and surface conduction and follow a power law form.  Particularly, the data for
the hot and dry cases lie close  to the theoretical curve  for volume conduction
(a  ->  »), as  is to  be  expected.  On  the other  hand  the  data  for the  ambient
condition  and  more  humid  cases fall between  the theoretical  curves  for the
volume  and surface  conduction  cases.  This  indicates   that  under the  given
ambient conditions,  both volume  and surface conduction are contributing to the
effects. Finally, the data for  Experiment #1,  where T ~  28°C  and Pr  ~ 100%,
lie close to the theoretical curve for surface conduction  (a = 0).

    An interesting phenomenon  observed  during Experiment  #1,  for the  case of
surface  conduction,  was that for  EA  £ 105 V/m,  pA  started  to increase. This
behavior  is attributed  to the  Joule heating  of the  contact  region. Calcu-
lations  show  that in the case of  surface  conduction at EA ~ 10^ V/m the true
current density in the  contact  area  is  Jc ~ 2.3  A/cm2, which is  a  factor of ~
1.8 x  10   higher  than the  measured apparent current density JA ~ 12.6 uA/cm .
Typically,  for  the case of volume conduction'with JA ~  5  x 10   uA/cm  an
-------
                         DISCUSSION AND CONCLUSIONS

    Tables  1  and  2  summarize the  set of  self-consistent  electromechanical
equations  governing the behavior of a  (monodisperse) particulate  layer  prior
to microspark  breakdown. A  close  examination of these tables  reveals  that the
dependence  of  9Q and Emax on  E^  and  packing mode is rather weak.  Also,  their
variation  with a at a specified  value  of  EA is slow. However,  the dependence
of PA  °n °" atl^ roode of packing is more pronounced. In this case,  the  value of
tj also  determines  the strength  of the  dependence  of  pA on  E^  through  its
influence   on  the  power  law  exponent  m^.  The  compress ive  stress  Pg  is
proportional  to EA  raised  to  a  power of  1.2  to  1.35,  instead  of EA  which
arises from a  homogeneous leaky dielectric model for  the layer.
    The  experimental  results  are  in agreement with the  theoretical prediction
regarding  the  P^(EA^ characteristic in  the power law  region (Region II),  where
p. a  (EA) , Here,  m (= mp)  lies in the predicted range and has  the correct
dependence  on  temperature and humidity.

                              ACKNOWLEDGEMENTS
    This work was supported by the Electric Power Research  Institute.
    The work described  in  this  paper was not funded by  the U.S.  Environmental
Protection  Agency  and therefore  the contents do not  necessarily reflect  the
views  of the Agency and no official  endorsement should be inferred.

                                 REFERENCES
1.   Moslehi, G. B. and Self, S. A.,  "Current  Flow  across  a  Sphere  with  Volume
     and Surface Conduction," J. Electrostatics  (in Press).

2.   Self,  S.  A.,  Moslehi,  G.  B.,  Mitchner,  M.   and  Leach,  R.,  "Electro-
     mechanics  and Reentrainament  of Precipitated  Ash,"  Proc.  International
     Conference on  Electrostatic Precipitation, Monterey, California, October
     14-16, 1981.

3.   Graton,  L.  C.  and Fraser,  H.  J. ,  "Systematic Packing  of Spheres,"  J.
     Geology, Vol. 43, No. 8, Part  I, pp. 785-909,  (Nov.-Dec.  1935).

4.   Deresiewicz, H.,  "Mechanics of  Granular  Matter,"  in Advances in Applied
     Mechanics, Vol. 5, Academic Press  (1958).

5.   McLean,  K.  J.,  "Electrical Conduction  in High  Resistivity  Particulate
     Solids," Ph.D.  Thesis,  E.  E.  Dept.,  Wollongong Univ.  College,  Univ.  of
     New South Wales, Australia, Chap.  7 (Dec. 1969).

6.   Moslehi,  G.  B.  and  Self,  S.  A.,  "Electrical Breakdown of  Particulate
     Layers," 4th Symposium on Transfer and Utilization of Particulate Control
     Technology,  Houston,  Texas,   October  11-15,   1982.  (Session  C-6,  this
     conference).
                                     321

-------
   LATERAL PROPAGATION OF BACK CORONA IN TWIN-ELECTRODE TYPE PRECIPITATORS

                by:   Senichi Masuda and Toshifumi Itagaki
                     Department of Electrical Engineering,
                     University of Tokyo
                     7-3-1 Kongo, Bunkyo-ku, Tokyo, Japan 113
                                  ABSTRACT

      The lateral propagation of back corona, discovered to occur in a tri-
electrode corona system, also occurs in a conventional electrostatic precipi-
tator of the twin-electrode system under certain circumstances.  The primary
factor in initiating this phenomenon is the mutual excitation of the wire
corona discharge and back corona.  At high resistivity of dust deposit and
with small discharge wire diameter, this mutual excitation becomes dominant
and the lateral propagation occurs from a single back corona on a plate
appearing at a local spot.  The dc base voltage in a pulse charging system
must be selected in a careful consideration of this phenomenon.  The detailed
conditions of its initiation and extinction in air at NTP are presented in
relation to various modes of corona discharge.

                               INTRODUCTION
      Investigations on the back corona occurring in a tri-electrode system
revealed that its lateral propagation takes place to cover the whole collec-
tion and third electrodes in the case when the magnitude of field intensity
in its collection field exceeds a certain threshold'(l).  This threshold is
a level to cause streamer-mode in the original back corona.  The wide-spread
secondary back corona, thus produced, is self-sustaining and cannot be elimi-
nated unless the field intensity be lowered well below its initiation thresh-
old.

      In the observation of back corona made at a large scale twin-electrode
precipitator for a sinter machine, one of the authors found with the aid of
a portable image intensifier that back corona occasionally occurs on the
dust-covered supporting frames of discharge wires.  This phenomenon results
from positive ionic current induced by the original back corona on the plate
and accumulating on the dust-covered frames similar to the lateral propaga-
tion in the tri-electrode system.  This finding motivated a careful observa-
tion of time-dependent development of back corona in a test precipitator of

                                     322

-------
twin-electrode system in the  authors'  laboratory.   As a result, the occurrence
of the lateral propagation of "back corona and the  concurrent hysteresis in
voltage-current (V-l)  curves  were confirmed in the twin-electrode system,
when dust resistivity was extremely high.  Again in this case, the lateral
propagation starts  to occur when the average field intensity in the collection
field exceeds  a certain threshold.  It was further confirmed in the laboratory
precipitator being  operated by pulse-energization  that the level of dc base
voltage,  Vb, is subjected to  the limitation posed  by the lateral propagation
of back corona possible to occur by chance somewhere on the collection plate.
In other  words, although it is possible to raise Vb by modification of corona
electrodes so  as to increase  corona initiation voltage, Vc, there is another
essential constraint of limitting the operating field intensity beyond which
the collection performance is drastically deteriorated by the lateral propa-
gation and the pulse-energization loses its control.  This paper is a pre-
liminary  report on  the lateral propagation of back corona specific to the
twin-electrode system occurring in air at NTP.

                           METHOD OF EXPERIMENT
     The experimental apparatus  consists of two parallel plates (25cm in
width, 15cm in length and spaced at 15cm) and a single discharge wire as
illustrated in Fig.  1.   One of the plates has a rectangular probe (lU.9cm in
width and 7-9cm in length)  at its center for measurement of corona current.
The plates are covered with paper towels (1.2mm in thickness and 0.03^-g/cirr
in volume density) to simulate the dust layer.   Resistivity of the paper
towel is adjusted by controlling the humidity of its ambient air at room
temperature and normal pressure, and its value  measured with ordinary parallel
plate electrodes.  Three wires with different diameters 0.5mm, 1.1mm and
2.0mm respectively,  and a stranded wire out of 0.35nnn wire elements with a
total diameter of 3.2mm are used.  The corona discharge on the wire and back
corona are observed with the aid of an image intensifier (EMI type 9912 with
maximum photon gain of 106) and  a portable image intensifier (HITACHI HS-690
with photon gain of 3.5x10tf). The experiment is made only for negative
corona discharge on the wire.

                           RESULTS OF EXPERIMENT
Corona Wire  with 0.5mm Diameter

     Figures 2 (a),  (b) and (c) show the voltage versus current density (V-J)
curves obtained with a 0.5mm discharge wire for three different resistivities
P = l.lxlO15, lAxlO13, and 2.?xl012 ohm-cm.
     When the  resistivity of the layer is very large at p= l.lxlO15 ohm-cm,
a distinct hysteresis  in the V-J curve occurs (Fig.  2 (a)).  As the voltage
is raised, negative  corona appears  on the wire at point B, and the corona
current begins to  flow.   The current gradually increases to reach point C
even though  the source voltage  is kept unaltered.  The corona points concur-
rently appear  one  after  another propagating in both directions along the wire,
and moving back and  forth (Fig.  3).   Glow of the back corona, on the other

                                    323

-------
hand, is so weak in this case that it does not allow a visual  confirmation of
its lateral propagation on the plate electrodes even with the  aid  of the EMI
image intensifier.  At point C, the time-dependent increase  in number of the
corona points stops.  At the initial stage when only a few corona  points
exist on the wire, no distinct back corona activity can be observed.  At the
final stage (point C), back corona as a dim light is observed  with the EMI
image intensifier, covering the entire surface of the plate  electrodes(Fig.It).

     As the voltage is reduced from point C, the current decreases slowly
with decrease in the number of the corona points.  The coronas do  not disap-
pear and back corona cannot be eliminated unless the voltage is reduced to
point D much lower than point B.  If the voltage is raised from a  point
between D and C, the current density J follows a curve on or close to the
original curve D-C.  In this case, the number of the corona points varies
with the variation in voltage.  These results provide a strong evidence that
the lateral propagation of back corona has occurred in this  case.

     Further raising the voltage beyond point C results in a current density
increase following the extrapolated curve of D-C.  The number  increase of
corona points remains small in this case, but each corona point becomes
increasingly brighter.

     The initiation point B of back corona and its lateral propagation can
vary by about IkV, with the concurrent fluctuation of point C.

     Fig- 2 (b) shows V-J curve measured at p = l.UxlO13 ohm-cm.   At this
lower resistivity level, back corona first occurs at point B'  with a substan-
tially larger corona current being required to produce the layer breakdown.
Once back corona starts, it again propagates laterally with the concurrent
rise in corona current up to point C1.  Raising or lowering the voltage from
point C1 causes the same effect as in the case of Fig. 2 (a).  When p is
about 10llt ohm-cm, almost the same V-J curve as shown in Fig.  2 (b) is
obtained.  Velosity of lateral propagation of back corona in these cases is
slow.   It takes 10 to 15 seconds from points B and B' to points C and C'.

     When p is smaller than 5x1012 ohm-cm, no hysteresis occurs in V-J curve
(Fig.- 2 (c)).  The number of corona points remains unchanged at constant
voltage, and neither increase in current nor propagation of the coronas is
observed unless the voltage raised.  The lateral propagation of back corona
does not take place at this resistivity in this particular case of the
experiment.

Corona Wire with 1,1mm Diameter

     In the experiment using l.lmm wire, neither hysteresis  in the V-J curve
nor propagation of the coronas along the wire occurs at constant voltage in
the resistivity range of 1011 to 6X101"* ohm-cm, indicating no  lateral propaga-
tion of back corona to occur.  An example of V-J curve 'is shown in Fig. 5.
     Instead, two remarkable phenomena-are observed in this  case.  One is
that the negative coronas on the wire take a form of diffused  glow in the
lower -resistivity range p < 1013 ohm-cm.  The other is that  the back corona
on the collection electrodes takes a form of separate domains  at higher

                                     324

-------
resistivity  range p >, 1013 ohm-cm.   The negative corona at onset manifests
itself as a few number of small motionless spots, or the diffused glow
surrounding a small portion of the wire.  In the range p < 1013 ohm-cm, the
increase in voltage produces the transition of these corona spots into the
diffused glow'so that it extends along the wire (Fig. 6 (a)).  A photograph
taken with a shorter exposure (Fig-  6 (b)) reveals this diffused glow to be
consisting of very small corona spots.  These spots are moving very fast
along the wire to produce the outlook of a diffused glow.   Fig. 7 shows back
corona in this case uniformly spreading over the collection plate.

     At high layer resistivity p ^lO13 ohm-cm, the diffused glow turns into
separate corona spots with voltage increased (Fig. 8).  The spots move around
back and forth along the wire.  As is clearly seen in Fig. 9» each  corona
spot on the wire produces its own domain of back corona on the plate, which
also moves concurrently with the corona spot.  The" back corona domains are
separated from each other, and behave as if they are repelling each other.
When the voltage is increased over 35kV with the current density beyond
2mA/m2, the number of corona points  becomes too large and makes it difficult
to recognize one-to-one correspondence between the corona points and the back
corona domains.  But the discrete nature of back corona domains still exists
although their contours become partly indistinct.

Corona Wire with 2.0mm Diameter

     When p is in the order of 10*  ohm-cm, the time-dependent rise in corona
current, much faster in speed, occurs at a constant voltage beyond back coro-
na initiation, and a slight hysteresis occurs in V-J curve (Fig. 10).  But
it differs from the hysteresis in Fig. 2 (a) and (b) for 0.5mm wire diameter
in that this time-dependent current  rise is not caused by the increase in
number of the corona points but by the mode change of corona in each original
spot.  At the onset, corona appears  in a form of tiny active lumps of glow
on the wire, but the current remains practically zero.  When the voltage is
slightly increased these active lumps of glow suddenly turn into a motionless
corona spot, raising the current at  a constant voltage, and back corona
appears at a small region opposit to this corona spot.  The corona spot
abruptly disappears when the voltage is lowered about IkV below corona initi-
ation threshold.  This presents a striking contrast to the case of Fig. 2 (a)
and (b) with 0.5mm wire where the corona extinguishing voltage lies far below
its initiation voltage.  The time required for the rise or drop of the cur-
rent in this case is less than 1 second, much shorter than that for 0.5mm
wire.  In spite of the hysteresis in. V-J curve, the lateral propagation of
back corona does not occur in-this case.  When the voltage is further in-
creased after the formation of one single corona spot, active lumps, of glow
appear renewed on both sides of this spot.  The spot and the lumps of glow
repel each other leaving dark portions in between.  Back corona in this case
also takes a form of separate domains, each corresponding to a spot or the
lumps of glow.

     At a lower resistivity in the order of 1013 ohm-cm, no corona point
appears unless voltage exceeds, say, U5kV.  Below this threshold, the corona
forms active lumps of glow (Fig. 11), each being accompanied by its own
domain of back corona (Fig.  12).  The lumps of glow move around rapidly along

                                     325

-------
the wire with the concurrent movement of the back corona  domains.  When the
voltage is raised beyond l*OkV, the current density exceeds  2mA/m2  at -which
value the lumps of glow turn into corona spots moving back  and  forth  along
the wire.  Concurrently, the boundaries of the back corona  domains are
blurred.

     When the layer resistivity is less than 1013 ohm-cm, corona after onset
takes a diffused form along the wire to have the same appearance as the dif-
fused corona with l.lmm wire at a low value of p.  In this  case, back corona
becomes also diffused losing its boundary of each region, and uniformly
distributed on the collection plates (Fig. 13).

Stranded Wire with 3.2mm Outer Diameter

     No lateral propagation of back corona occurs in this case  in  the entire
resistivity range between 1011 to SxlO14 ohm-cm.

     One-to-one correspondence between back corona domains  and  corona points
appears at a higher resistivity p >, 1013 ohm-cm  (Fig. lU).  Corona motion,
similar to those of the active lumps of glow appearing with 2.0mm  wire, oc-
curs, and it is enhanced with increasing resistivity.

     Fluctuation of current density at a constant voltage is very  large.  For
example, it fluctuates between 2.5mA/m2 and O.lmA/m2 at V = 30kV and p = 1.5X
1013 ohm-cm.  When a large domain  of back corona happens to cover a major
portion of the current probe, the current increases concurrently.  But the
domain rapidly moves out of the probe area following its corona point to
cause a sharp drop in current.

     Fig. 15 indicates the appearance of both negative coronas  and back
coronas at a lower resistivity p = S.SxlO11 ohm-cm.  Boundaries of the back
corona domains are blurred, and one-to-one correspondence between  the wire
corona and back corona disappears.  However, the flicker of a corona point is
accompanied by the concurrent change in the brightness of back  corona, causing
a slight fluctuation in the current density.  Both wire coronas and back
coronas at this lower resistivity are very static, so that  no large current
fluctuation is produced.

                                 DISCUSSION
     It must be emphasized that the present investigation is a preliminary
one made in air at NTP with clean negative corona wires, far from the realis-^
tic conditions.  Hence, the results obtained are primarily for understanding
the basic phenomena, and should never be extended to the practical precipi-
tators in operation.

     With .these factors born in mind, one can recognize from the results so
far obtained that the lateral propagation of back corona in this particular
case occurs only when the diameter of negative corona wire is very small (d =
0.5mm) and the layer resistivity very high (p ^lO13 ohm-cm).  The propaga-*
tion proceeds with a concurrent increase in number of discrete corona points

                                     326

-------
along a negative wire,  and not with  a continual extention of diffused negative
corona.   No  distinct  effect of the level of average  field in corona space is
observed, unlike in the case of  a tri-electrode system (l).   This  effect is
likely to be masked by  the fact  that,  in this  particular case when the lateral
propagation  can occur only at p >^ 1013  ohm-cm, the  initiation of  back corona
immediately  produces  its  lateral propagation.

     In general, it is  hypothesized  for  the lateral  propagation to occur that
both of the  following conditions must be met.   First,  lateral diffusion of
charge carriers (positive ions,  negative ions, electrons) should occur in the
original back  corona  either in gas space or on the surface of dust layer
and/or corona  wire.   Second, the neighbouring  region of both wire  and dust
layer must be  ready for initiation of new corona and back corona activities,
respectively.   It  can be  imagined that the first condition is easily met in
most of the  cases.

     In consideration of the experimental results described above, the second
condition is met in this  particular  case by a  very high resistivity p in the
layer easy to  cause its breakdown and a  very small wire diameter to produce
a very high  local  field enhanced by  ion  sheath of oncoming positive ions.

     In the  case of practical precipitators in operation, the following
factors are  to be  taken into account.  First,  the operating temperature is
much higher  to increase gas mean free path, thereby  enhancing the  lateral
diffusion of charge carriers, and also reducing the  breakdown threshold at
both dust layer and corona wire. Second, the  corona wires are not clean and
covered with high  resistivity dust deposit which will  easily experience break-
down by accumulation  of oncoming positive ions, similar to back corona
occurring on the collection electrodes.   As a  result,  the wire diameter may
have a much  less effect.   Under  these otherwise relaxed threshold  values the
effect of field intensity in gas space may again take  the most essential role
as the initiation  condition of streamer-mode back corona which is  actually
observed in  our laboratory precipitator, and a more  detailed investigation
is being attempted.   The  importance  of this field criterion in the twin-elec-
trode system is that  it poses a  strict constraint on the dc base voltage to
be applied to  the  corona electrodes- in pulse-energization.

                                CONCLUSIONS
     The preliminary investigations  are made on the lateral propagation of
back corona in a twin-electrode system in air at NTP with clean negative
corona wires.   The  results  obtained  lead to the following conclusions:

(l) The occurrence  of lateral propagation of back corona under a negative
    clean corona wire is  strongly dependent on the diameter of wire, d, and
    the layer  resistivity,  p.   In air at NTP, it appears only when the wire
    diameter is very small  (d = 0.5mm) and the layer resistivity very high
    (p >, 1013  ohm-cm).  No  distinct  effect of the field level in gas space is
    observed in air at NTP, unlike the case of the tri-electrode system.
    In other words, as soon as back  corona appears, it starts lateral propa-
    gation in  this  particular case.

                                      327

-------
(2)  At a higher operating temperature of an ESP with a larger gas mean free
    path, the lateral propagation is expected to occur at a larger wire
    diameter and/or lower dust resistivity.  The contamination of the corona
    wire with deposition of high resistivity dust also may result in the same
    tendency.
(3)  The lateral propagation of back corona produces at a constant voltage
    both a time-dependent increase in number of corona points extending along
    the wire and a concurrent gradual rise in corona current *  Once appeared,
    a negative corona point and back corona stabilize each other through their
    mutual action, so that they cannot be extinguished unless voltage be
    lowered to a level much lower than the back corona initiation threshold.
    As a result, a very distinct hysteresis in V-J curve occurs in the case
    when the lateral propagation of back corona takes place.
(h)  Irrespective of whether or not the lateral propagation occurs, a strong
    correlation exists between the appearance and motion of back corona and
    those of the negative corona points.  At a very high resistivity p >^lO1^
    ohm-cm and with a larger wire diameter d >^ l.lmm, back corona takes a
    form of separate domains, each corresponding to a discrete corona spot
    or a lump of glow on the wire.  At a lower resistivity p < 1013 ohm-cm
    and with a larger wire diameter, d = l.lmm and 2.0mm, the negative corona
    takes a diffused form surrounding the wire, and the boundary of the back
    corona domains becomes also blurred.  This diffused negative corona
    consists of very small corona spots moving around at high speed along
    the wire.

     A further investigation is attempted to clarify the initiation conditions
of the lateral propagation of back corona under the conditions prevailing in
a practical precipitator, with a special attention to pulse-energization
technology.

                                 REFERENCE


 1..   Masuda, S., Obata, S, and Ogura, Y.  Lateral propagation of back-
     discharge in a tri-electrode system.  Electrostatics 1979, Inst.
     Phys. Conf. Ser. No. hQ,  Institute of Physics (1979)


              The work described in this paper was not funded
              by the U.S. Environmental Protection Agency and
              therefore the contents do not necessarily reflect
              the views of the Agency and no official endorsement
              should be inferred.
                                     328

-------
                   Wire Electrode
                                Voltage Divider
     Paper
     Towel

Plate
Electrode
^
            ^
(^
                 in
                <15cm  >
                   Currem;
                   Probe
High
Voltage
Source
               1&
                                   :v;    £L.
          Fig. 1.  Experimental  Set Up.
                                  329

-------

~40
^20
c.
n
I 1 I
p=l .Ixio15 ohm-cm
Ci .
j/\
A i D j^"^ \B i
    o
                5         10
                     V  (kV)
                      (a)
15
^_^
^E 40
o
•* 20
0
I 1
- P=1.4><1013 ohm-cm

-
A'I iD'^-s
I
-

C1 -
jff

                        10
                   V (kV)
                     (b)
   80




<  40
C

°  20

   0
    _  p=2.7x!012 ohm-cm
                        I
    0
                                   15
                        10         15
                        V (kV)
       20
Fig.  2.   V-J curve with  0.5mm negative wire
                   330

-------
                                      (a) t = 0 second
                                       (b)  t = 2 seconds
                                       (c) t = 4 seconds
                                        (d) t = 10 seconds
Fig.  3.   Time-dependent  growth  in number of negative corona
         points  on  wire  observed with  a portable image intensifier.
                               331

-------
                                                  Dim light of
                                                  back corona
                                                Reflected image of
                                                negative corona on
                                                the wire
Fig.  4.  Back corona under 0.5mm negative wire at the final stage of
        growth in corona points (point C in Fig. 2 (a)) at
        p = 1.1 ><1015 ohm-cm observed with EMI image intensifies
        40
        30
        20
        10
           0     10    20     30    40
                     V(kV)

 Fig. 5.  V-J curve with l.lmm negative
          wire at p = l.lxio14 ohm-cm.
                                 332

-------
                  (a)  Exposure  time:  1/4  second
                  (b) Exposure time:  1/500  second
Fig.  6.   Negative coronas on l.lmm wire  at   p =  5.4xlOn  ohm-cm
         observed with a portable image  intensifier.
                              333

-------
                                                Back Corona
                                                    Wire
 Fig.  7.   Back  corona under 1.1mm negative wire
          at p  =  5.4xlOu ohm-cm observed with
          a  portable image intensifier.
          (Exposure time: 1/4 second)
Fig.  8.   Negative  coronas on  l.lmm
         wire  at p  =  5.5xl014 ohm-cm
         observed  with  a portable
         image intensifier.
         (Exposure time: 1/2  second)
Fig.  9.  Back corona under l.lmm
         negative wire at p = 1.1
         ohm-cm observed with a
         portable image intensifier.
         (Exposure time: 1/4 second)
                                    334

-------
           <

           'O
              40
              30
              20
              10
                      10     20
                          V(kV)
30
40
        Fig. 10.  V-J curve with 2.0mm negative
                 wire at P = 4.0X1014 ohm-cm.
Fig.  11.  Negative  coronas on  2.0mm wire
         at p = 4.1xl013 ohm-cm
         observed  with  a portable
         image intensifier.
         (Exposure time: 1/4  second)
       Fig. 12.  Back corona under 2.0mm
                 negative wire at p = 1.5*
                 1013 ohm-cm observed with
                 a portable image intensi-
                 fier.
                 (Exposure time: 1 second)
                                     335

-------
   Fig.  13.   Back  corona under 2.0mm negative wire
             at  p  =  4.2xl012 ohm-cm observed with
             a portable image intensifier.
             (Exposure time: 1 second)
Fig.  14.   Back  corona  under 3.2mm
          negative stranded wire at
          p  = 1.6xl013 ohm-cm observed
          with  a  portable image
          intensifier.
          (Exposure  time: 1/4 second)
Fig.  15.   Back corona under 3.2mm
         negative stranded wire at
         p = 8.8xlOn ohm-cm observed
         with a portable image
         intensifier.
         (Exposure time: 1/4 second)
                                     336

-------
                     FIRST MEASUREMENTS  OF  AEROSOL PARTICLE
                           CHARGING BY FREE ELECTRONS
                       James L. DuBard  and M.  G.  Faulkner
                Southern Research  Institute,  Birmingham AL 35255

                                Leslie  E. Sparks
                  Industrial Environmental Research Laboratory
                      U.S. Environmental Protection Agency
                        Research Triangle Park, NC  27711
                                    ABSTRACT

     The charging of fine aerosol particles by  free  electrons has been isolated
from negative ionic charging and measured  for the  first  time.  The charge and
size of individual particles were measured in a Millikan cell,  with charging
electric field from 0.82 to 8.2 kV/cm.   In the  particle  size range 0.5 to 3.0 Urn
diameter, the particle charge values are much larger,  and increase much faster
with particle size, than those predicted and observed  for negative ionic char-
ging.  The particle charge values show only slight dependence on the charging
electric field.
                                  INTRODUCTION

     A number of techniques for removing particles  from gas  streams  involve
electrically charging the particles.  Particles may be  charged  by impact of
positive or negative molecular ions produced  from a high voltage  corona dis-
charge in the carrier gas.  The physical mechanisms associated  with  ionic charg-
ing have been determined experimentally and described theoretically  (1,2,3).
Ionic charging theory has been applied in a mathematical model  which describes
the cleaning of industrial process gas streams by electrostatic precipitation
(4).  In a negative corona discharge, particles may be  charged  also  by free
(unattached) electrons.   The mechanisms associated  with free  electron charging
have not been determined.

     Ionic charging theory was tested by direct measurements  of the  charge
acquired by fine particles, 0.5 ym to 3.0 Urn  diameter,  using  a  Millikan cell
(5).  The charge and size of individual fine  particles  escaping laboratory
precipitators were measured at air temperatures from 38°C to  343°C.   The data
for positive ionic charging were in good agreement  with theory  at all tempera-
tures.   However, the measured particle charge with  negative  corona was signifi-
cantly higher than the predicted charge at temperatures above 200°C.  At 343°C,
the negative charge enhancement was about a factor  of two for 2.0 ym diameter
particles,  as shown in Figure 1.  The negative charge enhancement was attributed
to a contribution from free electrons which had escaped attachment to electro-
negative gas molecules because of the reduced gas density.  The top  curve in
Figure  2 shows the ratio of the negative-to-positive charges  in the  laboratory
                                      337

-------
data of Figure 1.  For  comparison,  the bottom curve in Figure 2  shows  the
numerator of that  ratio (that  is,  the measurements of negative charge)  replaced
by data from field tests  on  a  hot-side precipitator.  The charge  and  size of
individual fly ash particles escaping the precipitator were measured with a
Millikan cell.  In the  diameter  range 1 to 2 urn, 100 to 150 particles were
measured in each 0.1  Pm interval.   The data were obtained at a sampling port
downstream of the precipitator outlet.  Particle collisions with  duct walls may
account for the lower charge enhancement, and much greater scatter  in the data,
than that observed in the laboratory.  However, there remains a  strong  sugges-
tion of enhanced negative charging  of fine fly ash particles.  The  plausibility
of fly ash particle charging by  free  electrons can be shown by an estimate of
the surviving free electrons in  the flue gas.  Using the calculated electric
field distribution in the hot-side  precipitator and using electron  attachment
coefficients measured in  pure  oxygen  (6) (and adjusted to the flue  gas density
and composition) it can be estimated  that 2% to 4% of the electrons leaving the
negative corona discharge will survive halfway to the collecting  plate.

     In applications  to industrial  electrostatic precipitation,  it  is particu-
larly important to understand  the charging and collecting of fine particles
because they are the most hazardous to human health,  the most difficult to
precipitate from a gas  stream, and  the most responsible for visible emissions.
Although laboratory data  and field  data suggest that conventional negative ionic
charging of fine particles may be significantly enhanced by electronic charging,
there have been no data that isolate  electronic charging from ionic charging.
The preliminary data  on electronic  charging reported in this paper  show that
free electrons are much more effective than negative ions in charging fine
particles.  However,  the  electronic charging process cannot be aequately de-
scribed by adaptation of  the established ionic charging theory.
                                  EXPERIMENTS

     The experiments were performed  in  commercial ultrapure nitrogen gas having
1 ppm residual 02-  In electric  fields  the  order of 5 to 10 kV/cm, the electron
attachment coefficient in 1 ppm  02,  at  20°C,  is  4.2 to 6.5 x 10~5 cm"1 (6).
Within the dimensions of the aerosol charging apparatus, the probability of an
electron attachment was less than 4  x 10   .   A flux of free electrons was ob-
tained from a single needle point pressed  into a cold cathode plate and ener-
gized with negative dc high voltage.  The  electrical discharge from the needle
point was an intense streamer, pulsating at  6.0  Hz.  (The stability of a
conventional negative corona discharge  results from the negative ionic space
charge surrounding the emitting  point.  In a gas that is not electronegative,
that space charge control is missing.)  The  electrical parameters of the dis-
charge were independent of the power supply  and  were controlled entirely by the
electrode geometry.  The maximum cathode voltage between pulses was about 7.5
kV, and the maximum cathode current  during  a pulse was about 0.6 mA.  The
electrical discharge was extremely sensitive to  residual traces of ambient air
in the experimental chamber.  However,  once  the  chamber was cleared of
electronegative gases, the unusual character of  the electrical discharge was
entirely reproducible.
                                      338

-------
     The  aerosol charging apparatus was contained  in  a  standard cylindrical
vacuum chamber,  fitted with a viewport and with  gas and electrical feedthroughs.
The flux  of free electrons was obtained in a uniform  electric field between two
parallel  metal plates, 20 cm in diameter and separated  by 4 cm.  There was a
rectangular cutout in the center of the grounded plate  opposite the cathode
plate, to accommodate the aerosol charging assembly.  This cutout was covered
with wire cloth  having 49% open area.  The aerosol charging region was a rect-
angular void in  a teflon block set into the rectangular cutout in the grounded
plate. The grounded wire cloth formed the front boundary of the aerosol
charging  region.  The back boundary was a metal  anode plate set into the teflon
block. The separation between the grounded wire cloth  and the anode plate,
across which the charging electric field was developed,  was 1.2 cm.  The aerosol
generator was  a  Collison nebulizer operated with ultrapure nitrogen gas.  In a
preliminary series of three experiments, a dilute  water suspension of 0.8 Mm
diameter  polystyrene latex (PSL) particles was used in  the nebulizer.  The
nebulizer was  followed by a dessicator and a polonium-210 strip irradiating the
gas stream. The charge and size of individual particles  exiting the chamber
were measured  with a Millikan cell.  In every experiment, the Millikan cell was
operated  before  and after energization of the cathode needle point to be sure
that no frictionally charged particles were contained in  the gas stream.

     Three experiments, with different voltages on the  anode,  were performed in
order to  investigate the effect of the charging electric  field on the particle
charge.  The experimental parameters are listed in Table  1.   In order to inves-
tigate the variation of particle charge with particle size,  the driving gas
pressure  from  the gas cylinder was set such that the  gas  stream passed through
the dessicator quickly enough for the particles to be incompletely dried.  The
residual  water surface resulted in a particle size range,  measured in the
Millikan  cell, of 1.1 to 1.7 Mm diameter.  The dielectric constant of water was
used in calculations of particle charge.

     In order  to minimize precipitation of charged particles,  the high voltage
on the anode was applied in the form of a square wave,  alternately positive and
negative.   The square wave frequency was 100 Hz.  Thus, as shown in Table 1,  the
5 ms period of positive high voltage on the anode was just slightly longer than
the estimated  maximum aerosol exposure time of 4.1 ms.  Both times were short
compared  to 83 ms,  which was half the cathode streamer  pulsing period.   Thus,
some particles had the opportunity to be exposed to the maximum electron flux
and to experience a uniform charging electric field during the entire exposure.
Measurements were made on the most highly charged particles  observed in the
Millikan  cell.

     Each  data point for particle charge and particle size was calculated from
an average of  three up and three down velocity measurements  of the particle,
acquired within  30 seconds of charging the particle.  The stability of the water
surface on the particles was investigated for some particles by observing the up
and down motions in the Millikan cell for several minutes.   These data showed a
trend in  successive measurements of particle velocity;  the particles were
growing by condensation.   However, the data acquired  within 30 seconds gave no
evidence of a  changing particle size.
                                      339

-------
     In the theoretical  description of ionic charging of particles,  the critical
parameters are the particle  radius  (a),  the charging electric  field  (E),  and the
product of ion density  (NQ)  and  the aerosol exposure time  (t)  (2).   For example,
the charge acquired by  a large conducting particle as a result  of electric  field
charging is given by  (8)

                                 9       N0C
                    en  = 12ireo a2E  NQt + 46Q/eb » coulombs,               (1)

where b is the electrical mobility  of the ions.  In these  experiments,  the
analogous free electron  density  (N) was calculated from the measured  electron
current density which penetrated the grounded wire cloth and collected on the
anode, as

                    J =  e Ne v,  A/cm2,                                     (2)

where v is the drift velocity of a  thermal!zed electron swarm.   Several of  the
electronic charging parameters,  listed in Table 1, had to  be estimated in these
preliminary experiments.  The particle exposure time depended on the estimated
average aerosol velocity in  the  aerosol  charging region.   The electron flux from
a streamer discharge was  estimated  (from visual observations) to be  uniform over
one square centimeter of the aerosol charging region.

     Data points for  individual  particles in the three preliminary experiments
are shown in Figure 3,  along with dashed straight lines least-squares-fitted to
the data points to gui.de the eye.   The fits to the two data sets obtained with
8.2 and 4.1 kV/cm were  so nearly the same that only one dashed  line  is  shown.
Also shown in Figure  3,  for  comparison with the data, are  calculated curves for
negative ionic particle  charging under similar conditions.  Three interesting
features of Figure 3 may be  noted:

     a) The measured values  of particle  charge are very much larger  than  those
predicted for analogous  negative ionic charging.  The three curves in  Figure 3
were calculated using ion density NQ the same as the estimate of free  electron
density Ne, and using established ionic  charging theory (2).

     b) There is no dependence on electric field strength  shown in the data
obtained with 8.2 and 4.1 kV/cm  and only a slight dependence shown at  0.82
kV/cm.  For negative  ionic charging of particles in this size range, there  is a
strong field dependence,  as  shown in the three calculated  curves in  Figure  3.

     c) Accumulated electron charge increases more rapidly with increasing
particle diameter than  accumulated  negative ion charge.

If the same physical mechanisms  were assumed for electronic charging as for
ionic charging, the three curves of Figure 4 would result.  In  Figure  4,  the
experimental data are represented by the two fitted straight lines.  The  three
curves were calculated  in the same  manner as for ionic charging, but with the
ion mobility replaced with the much larger electron mobility obtained  from  the
electron drift velocity in Table 1.
                                      340

-------
     A second series of experiments  was  performed to obtain a direct experi-
mental comparison between electronic charging and negative ionic charging,
using a non-volatile aerosol.  Di  (2-ethylhexyl)  sebacate (DES) was used in the
Collison nebulizer.  The resulting aerosol  included fine particles spanning the
entire size range (0.5 to 3.0 Urn diameter)  that  could be measured in the
Millikan cell.

     The experimental objective was  to achieve the same value of Net for free
electrons in ultrapure nitrogen gas  as the  value  of NQt for negative ions in dry
air.  All of these experiments were  performed with 5 kV/cm electric field in the
aerosol charging region.  The high voltage  on the cathode needle point was
adjusted to control the discharge current so  that the measured anode current in
the aerosol charging region was different for electrons and ions by approxi-
mately a factor of 300 in order to compensate for the difference in mobility.
With ultrapure nitrogen in the chamber,  the cathode discharge was current-
limited by the external high voltage supply in order to change the pulsating
streamer into a low-intensity steady streamer giving a steady dc current (which
was nearly the same as the average current  with  the pulsating streamer.)  The
parameters for this second series of experiments  are listed in Table 2.  It
should be noted that the value Nt ~  10   cm  s achieved in all experiments is
lower by about an order of magnitude than the typical charging parameter in an
industrial precipitator.

     Individual data points for two  different pairs of experiments are shown by
the open and closed symbols in Figure 5.  Again,  only the most highly charged
particles found in the Millikan cell are shown on the graph.   (The values of
anode current for the closed symbols were lower  than those shown in Table 2, due
to a systematic measurement error, but the  ratio  remained approximately correct
at 300:1.)  The particle size range  shown in  Figure 3 is only a small slice of
that shown in Figure 5.  However, a  close comparison of measurements and calcu-
lations shows that the results of the two sets of experiments are consistent.
                                    SUMMARY

     These experiments have isolated the charging  of  fine  particles by free
electrons for the first time.  Although the  data are  of  a  preliminary nature,
they clearly cannot be interpreted with the  theory of negative ionic charging,
not even when the value of mobility is corrected to represent  free electrons.
This indicates that the physical mechanisms  involved  in  electronic charging
differ from those applying to ionic charging.  The estimated experimental values
of the charging parameter Net, shown in Table  1 and used as  NQt in the calcula-
tions of Figure 3, are believed to be a conservative  upper limit.   A more
accurate determination of that charging parameter  may result in an even larger
difference between ionic charging and electronic charging.  At least for the
rather low values of Nt achieved in these experiments, the difference between
ionic charging and electronic charging increases rapidly with  particle size.
This trend is consistent with the cases of negative charge enhancement described
in the INTRODUCTION.  In those cases, the density  of  unattached electrons aver-
aged over the wire-plate spacing would have  been a small fraction of the nega-
tive ion density.  Although the work presented here demonstrates that particle
charging by free electrons can be investigated in  a conclusive manner, further


                                      341

-------
experiments are needed  to  define  the physical parameters and to determine  the
dependence of electronic charging on Net,  a,  and E.  The data  from  these experi-
ments should lead  to an understanding of the  physical mechanisms, a capability
to develop a theoretical model, and  a basis for advantageous use of electronic
charging.


                                   REFERENCES

1.  Hewitt, G. W.  The  Charging of Small Particles for Electrostatic Precipi-
    tation.  Trans. AIEE 76:300 (1957).

2.  Smith, W. B.,  and J. R. McDonald.  Development of a Theory for  the Charging
    of Particles by Unipolar  Ions.   J. Aerosol Sci. 7:151 (1976).

3.  Smith, W. B.,  L. G. Felix, D. H.  Hussey,  D.  H. Pontius,  and L.  E. Sparks.
    Experimental Investigations of Fine  Particle Charging by Unipolar Ions - A
    Review.  J. Aerosol Sci.  9:101 (1978).

4.  McDonald, J. R.  A  Mathematical  Model  of  Electrostatic Precipitation
    (Revision 1).  EPA-600/7-78-llla,b.   NTIS PB284-614 and -615.   U.S.
    Environmental  Protection  Agency,  Research Triangle Park, NC (1978).

5.  McDonald, J. R., M. H. Anderson,  R.  B.  Mosley, and L. E. Sparks.  Charge
    Measurements on Individual Particles Exiting Laboratory Precipitators with
    Positive and Negative  Corona  at  Various Temperatures.  J. Appl.  Phys.
    51:3632 (1980).

6.  Grunberg, R.   Anlagerung  von  Elektronen in Luft und in Gemischen aus 02
    mit He, N2 und C02.  Z. Naturforschung 33A:1346 (1978).

7.  Dutton, J.  A  Survey of Electron Swarm Data.   J.  Phys.  Chem.  Ref. Data 4:577
    (1975).

8.  Pauthenier, M. M. and  M.  Moreau-Hanot.  La Charge des Particules Spheriquea
    dans un Champ  Ionise.  J. Phys.  et Radium 3:590 (1932).
                                       342

-------
TABLE 1.  APPROXIMATE  PARAMETERS  FOR MEASURED ELECTRONIC CHARGING AND CALCULATED
          IONIC CHARGING  OF  PSL-H20 PARTICLES
Anode positive voltage
Duration of anode positive
voltage, with 100 Hz
square wave on anode
Charge electric field E
Electron current through
screen into charging region
Electron drift velocity in
charging region, assuming
thermalized electrons (7)
Free electron density N in
kV
ms
kV/cm
MA
106 cm/s
108 cm"3
1
5
0.82
14
0.8
1.09
5
5
4.1
128
2.5
3.20
10
5
8.2
182
4.5
2.52
charging region, assuming
electron current distributed
         r\
over 1 cm

Aerosol velocity                        cm/s             268       268         268

Estimated maximum aerosol                ms              4.1       4.1         4.1
exposure time

Charging parameter N t               106  cm"3  s        0.44      1.31        1.03
TABLE 2.   APPROXIMATE PARAMETERS FOR AN EXPERIMENTAL COMPARISON OF ELECTRONIC
          CHARGING AND IONIC CHARGING  OF DBS PARTICLES
                               Negative  ions                 Free electrons  in
                                in  dry air                   ultrapure nitrogen


Electric  field                  5 kV/cm                         5 kV/cm

Ion mobility                    2 cm2/V's
Electron  E/N                                                    20 Td

Drift  velocity                  lO4 cm/s                        3 x  106 cm/s
Anode  current                    1 PA                            300  UA

Charge density                  6 x 108  cm"3                    6 x  108 cm"3

Charging  Nt                      106 cm"3 s                      106  cnT3  s

                                      343

-------
          10'
            15
       _o
       i
       <
       o
       O
       cc
          10'
             17
              0.1
                       D  NEGATIVE CORONA 27-28 kV
                       O  POSITIVE CORONA 26-27 kV
                      	 THEORY
                                          j    i
0.2            0.4

          PARTICLE RADIUS, jum
1.0
  2.0

700-204
Figure  1.   Measured  positive and  negative particle charge  and comparison with
            ionic charging theory,  at 343°C and  30 nA/cm* plate current density.*
                                         344

-------
    o

    £C
    111
    O
    ai
       2.0
       1.8
x
o
uj  1.6

P
M
2
       1.4
       1.0
                                     a LABORATORY PROTOTYPE ESP
                               b) FULL-SCALE HOT-SIDE ESP
         0.4       0.5        0.6       0.7        0.8

                                  PARTICLE RADIUS,
                                                    0.9
1.0
         1.1

      700-208
Figure 2.   Measured negative  particle charge, in proportion to the positive
           charge measured  in a  laboratory precipitator  at  343°C and 30 nA/cra2.
           (a)  Laboratory precipitator at 343°C and  30 nA/cm .  These data are
           the  ratio of average  values plotted in Figure 1.   (b) Hot-side
           precipitator (Lansing Smith Plant) at 360°C and  20 nA/cm2.
                                       345

-------
          1200
          1100
          1000
           900
           800
           700
           600
       IU
       O
       DC
           500
          400
           300
          200
          100 h-
(  8.2 kV/cm
•  4.1 kV/cm
4  0.82kV/cm
  • '"/*   • ' ~Z
   7   *    A
/  .     /   *
   .  -/•'     /
   V-            *
   ?  .      A *
             1.1
    1.2
                                        JL
                                                  _L
                               1.3       1.4        1.5
                               PARTICLE DIAMETER. Aim
                                          1.6
                        1.7
                     4100-HJ
Figure 3.  Measured particle  charge  acquired in laboratory  electric fields of
           8.2, 4.1, and 0.82 kV/cm.   The particles were charged  by free
           electrons.  Solid  curves  show the particle charge calculated for
           charging by negative  ions,  with an ion density the  same  as the
           electron density.
                                      346

-------
          1200
          1100
          1000
           900
           800
           700
        ta
        z
        D
        LIJ  600
        O
        cc
        u
           500
           400
           300
           200
          100
•  8.2 kV/cm
•  4.1 kV/cm
A  0.82 kV/cm
                                          I
                                                    I
             1.1       1.2        1.3        1.4       1.5

                                PARTICLE DIAMETER, fim
                                         1.6
    1.7
4100-189
Figure 4.   Calculated  particle charge acquired in laboratory  electric fields  of
           8.2, 4.1, and  0.82 kV/cm.  The  calculation uses  ionic charging
           theory, with the ionic mobility replaced by the  electronic mobility.
           Two dashed  straight lines, least-squares-fitted  to the data points  in
           Figure 3, are  shown for comparison.
                                       347

-------
               5000
               4000
               3000
             t
             z
             =>
             u
               2000
                1000
    CHARGING BY FREE ELECTRONS
    IN ULTRAPURE NITROGEN

DB CHARGING BY NEGATIVE IONS IN DRY AIR

 E = 5 kV/cm                     O
                         0.5     1.0    1.5    2.0    2.5

                                PARTICLE DIAMETER, jum
                                   3.0
3.5
Figure 5.  Measured  particle charge acquired  in a laboratory electric field of 5
           kV/cm.  The two data sets give  a direct comparison  of  ionic charging
           and  electronic charging.
                                       348

-------
                     GAS FLOW DISTRIBUTION MODEL TESTING

                         by:   D. R. Cook
                              J. M. Ebrey
                              D. Novogoratz
                              Lodge-Cottrell/Dresser
                              Houston, Texas  77002
                                  ABSTRACT

     Uniform gas distribution is critical to the operational efficiency of
both electrostatic precipitators and structural baghouses.

     The relationship of collection efficiency to gas distribution criteria
is discussed.   The Deutsch-Anderson equation modelled into a computer gas
distribution program indicates the effects of gas distribution deviations on
efficiency performance.

     Unique characteristics of the Lodge-Cottrell approach to gas flow
distribution,  pressure loss, and dust fallout testing are discussed, as are
the proprietary distribution devices used.  Test objectives and methods used
are described.

     These preceding discussions identify the need for gas flow correction
technology and its role in fulfilling todays technological requirements.

     The need  for gas flow correction in structural baghouses is discussed,
and some results presented.

     Finally,  model vs.  field test results will be compared for both the
electrostatic  precipitator and the structural baghouse.

ACKNOWLEDGEMENTS

     The authors wish to express their appreciation to Dresser Industries,
Inc.  for granting permission to publish this information and also our thanks
to our colleagues at Lodge-Cottrell U.K.  and U.S. for their assistance.
                                     349

-------
INTRODUCTION

     Todays Plant Manager has to live in an environment of highly regulated
emission control standards.  The requirements for efficiencies of 99.5% and
higher for particulate control devices on a continuing basis places a heavy
responsibility on both the plant owner and the original equipment supplier.

     In support of these demands, original equipment manufacturers must
respond by offering control devices where every facet of the design is
thoroughly scrutinized.  One such critical area is the need for controlled
gas distribution and minimized pressure losses.

     Consider now, two devices widely used in industrial gas cleaning; the
electrostatic precipitator and the fabric filter.  While these are very
different in their method of collecting dust, similar considerations hold for
the need to control gas flow within the filtration zone.  In our experience,
devices relying on resistance to flow only succeed in improving gas distri-
bution to the required standards with an unwelcome increase in pressure loss.
In contrast, when the gas distribution is controlled by use of guide vanes,
etc., then not only will the best distribution pattern be obtained, but also
the pressure loss for the system will be minimum.

     Our basic contention, therefore, is that a correctly designed gas
distribution system does not rely on excessive use of resistance devices to
control flow.

ELECTROSTATIC PRECIPITATORS

     The critical need for controlled gas distribution in precipitators is
self-evident.  Consider a fifty-duct precipitator and imagine that with the
flow control condition existing, one duct is completely devoid of gas flow.
This has the effect immediately equivalent to a loss of 2% of the effective
plate area.  If the precipitator has been designed for an efficiency of 99.5%
then the effect of the mal-distribution would be to increase dust emissions
by 10%.

     It is essential, therefore, that the gas distribution across the face of
the electrode system be reasonably uniform.  This can be illustrated by
consideration of the Deutsch-Anderson formula used to calculate the efficiency
for different gas velocities existing in different parts of the field as
determined by a survey.  A weighted mean efficiency is then produced which
shows how much the efficency falls below that attained when the distribution
is normally uniform.  Fig. 1 shows such a uniform velocity survey and the
associated mean efficiency.  Fig. 2 indicates the effects of gas mal-distri-
bution at this same position for the same flow rate.  Efficiency is reduced
from 99.03% to 98.6% while the RMS deviation increases from 15% to 35%.  To
achieve the same efficiency with the degree of mal-distribution present, the
collector plate area would need to be increased in the order of 10%.

     In addition to the uniformity of velocity profile, the gas must not by-
pass the field via the roof and hopper regions.  As a simple example, if 1%
of the gas completely bypasses the field, the efficiency can never exceed  99%


                                       350

-------
even when the dust removal from the remaining gases passing through the elec-
trical field approaches 100%, which is equivalent to an infinite migration
velocity.

     Another variable condition which distorts gas distribution is the
effect of dust buildup within the system.  As the gas flue approaches a
precipitator, the gas velocity must be reduced from the duct velocity of
15-23 meters/second to about 1.5 m/s.  This must be achieved with the knowledge
that as the gas velocity falls, so will dust tend to fallout.  Failure to
take this into consideration can result in partial blockage of the final gas
distribution system with the consequent serious effects on gas distribution
across the face of the precipitator.

     For example, with flat resistance plate distributors, the probability of
dust fallout is high behind the flat plate, effectively blocking a significant
area of gas flow at the lower end of the collecting electrode.  Another side
effect of such fallout was experienced with the chevron arrangement widely
used by some equipment suppliers a few years ago.  This was a very compact
gas distribution system used on large boiler installations.  However, due to
the low gas velocity, the inlet flares had substantial dust deposition which
in some cases resulted in structural failure of the duct system.

     The prerequisite for all correction devices installed in the system is
that their ultimate effect on flow distribution is accomplished with minimum
pressure loss and dust fallout.  Three typical examples which achieve these
criteria are; the proprietary top inlet mouthpiece (Fig. 3), with its triangular
splitters followed.by the tubular distributors which act as final smoothing
devices.  This system has the virtue of compactness and the ability to remain
free of buildup with relatively adhesive dusts such as high-sulphur fly ash
or cement dust, and has been widely used.

     Another example is the patented high-velocity wedge type inlet (Fig. 4),
which lends itself to a very compact mouthpiece so that the risk of fallout
is at a minimum.  This inlet is extremely valuable in retrofitting plants
with space limitations.

     Fig. 5 illustrates a typical horizontal straight inlet system.  This
uses honeycomb splitters and tubular distributors as the final smoothing
device.  The honeycomb splitters are of a spreading eggcrate configuration
installed in the upstream position of the flare where controlled spreading of
the gas takes place while still in a relatively high velocity area.  By
keeping the splitter in a high velocity area, the device is essentially self-
cleaning, thus eliminating the need for external rapping.   This is a particu-
larly low pressure loss system, and remains free of fallout problems to an
acceptable degree.

DECISION TO PERFORM A MODEL STUDY

     Having highlighted some of the relationships between controlled gas
distribution and particulate removal efficiency, it follows that flow correc-
tion must be considered an important aspect of equipment design and technology.
The question is now asked, how can we most economically produce a system
design to insure acceptable gas flow distribution?  It is not possible to


                                     351

-------
reliably calculate the total effects of gas flow correction devices,  and, in
view of todays large plant, on-site flow correction is not economically
feasible.

     The alternative is to accurately construct a scaled-down version of the
system and to study the effects that various distribution devices have on the
gas flow through the modelled system.

SELECTION OF MODEL SIZE

     If scale modelling techniques are to be used to identify and provide
solutions, then the first step is to decide the scale to which  the model will
be constructed.

     To obtain useful results which can be interpreted to predict full-scale
or prototype performance, modeling techniques have to attempt to satisfy
certain laws of similarity.

     Basically these are:  Geometric, Kinematic and Dynamic similarity.

     Geometric similarity means that the ratios of all corresponding  dimensions
on the model and prototype shall be equal.  This condition is easily  satisfied
irrespective of the model scale.

     Kinematic similarity exists between model and prototype when the ratios
of corresponding velocities and accelerations are the same throughout the
flow.

     Dynamic similarity between the two flows occurs when the ratios  of all
forces acting on corresponding fluid masses are similar.

     Satisfying the requirements of dynamic similarity of model and prototype
in typical gas cleaning systems would require that the force ratios,  as
represented by Reynolds, Euler and Froude numbers, be the same  in both
systems.  To simultaneously match these three force ratios is impractical in
this type of model study and only becomes attainable as the model to  proto-
type ratio approaches 1:1.

     In most industrial flue systems, the flow is said to be turbulent,
characterized by Reynolds numbers above the transition range of 2,000-4,000.
This transition range has been the topic of some controversy and  it has been
argued that due to internal geometric configuration, flow separation  probably
takes place at a much lower calculated Reynolds numbers than those  stated,
and when this has taken place, flow patterns will remain the same over a wide
range of Reynolds numbers.

     It has, therefore, been concluded by many that the velocity  profile
pattern within a precipitator model is independent of the calculated  Reynolds
number providing that number is above 4,000.  This has lead  to  the  widespread
use of small-scale models.  Our experience, however, has been  that  only when
operating high into the turbulent range does the pattern become less  depend-
ent of Reynolds number.

                                     352

-------
     Figure 6 illustrates the relationship between the model scale, the
treatment velocity within the precipitator model and the associated
calculated Reynolds number.  For this particular example a model con-
structed at 0.58:1 scale and operating at the prototype design treatment
velocity, will duplicate the Reynolds number of the prototype.  For a
precipitator of todays dimensions, a model scale of 0.58:1 is neither
practical nor economic.  A more acceptable compromise might be to con-
struct the model 1/8 full-size and operate at 2x the design velocity
with a resultant calculated Reynolds number of 12,000, i.e. high into
the turbulent range.  However, in order to test at design velocity the
Lodge-Cottrell approach would be to install every other collector plate,
thus increasing the cross section to wetted perimeter ratio,and maintain a
Reynolds number of  12,000.

     For the above reasons it is recommended that the largest practical
model scale size be utilized.

DISTRIBUTION CRITERIA AND MODEL STUDY OBJECTIVES

     Today, with the demand for higher operating efficiencies, some
customers and consulting engineers are specifying flow distribution
criteria to which they feel the supplier must conform in order to obtain
design efficiency.

     Typical of some recent specifications are the following extracts:

     1.    The precipitator study shall conform to the most recent IGCI
          recommendations.

     2.    RMS deviations of 15% or less shall be achieved.
     3.    Collecting plates to be modelled in all fields.

     4.    Tests shall be performed at equivalent velocity,  i.e.,  model
          gas velocity = site gas velocity.
     5.    Additional velocity surveys within the precipitator model will
          be performed at gas flow rates equivalent to 25,  50, 75,  100,
          125 and 150% design flow rate.

     Velocity surveys at significantly reduced velocities  are not justified
from a performance  viewpoint since the effective S.C.A.  would increase
proportionately.  However dust fallout in the ducting at low velocities is  of
concern  and should  be considered.

     In  addition to meeting these typical criteria, the model study
needs to satisfy other specific objectives.   Some of these are:
     1.    Obtain specified velocity distribution within the precipitator.

     2.    Minimize  systems overall total pressure loss.

     3.    Identify  and minimize dust fallout area.

     4.    Eliminate gas bypass of electrode  system.

     5.    Optimize  gas distribution to I.D.  fan inlets.


                                    353

-------
     Let's now address each of these points.

Velocity Distribution

     1.   Velocity surveys within the model are performed using hot-wire and
electronic direct-reading vane anemometers.  Measurement positions correspond
to 4 ft. centers of equal area on the prototype.  Signals generated by each
sensor are interfaced to a computer via a digital voltmeter and stored.  On
completion of a survey, the results are immediately printed out on the compute;
terminal for assessment.

Pressure Loss

     2.   Pressures and velocities within  the  ducting system  are  determined
with the use of a standard pitot tube and  inclined manometer.   Static
and velocity pressure measurements together with  ambient  conditions are
entered into the computer for rapid calculation of both velocity  and
total  pressure at each position in the system  of  significant  potential
pressure loss.

Dust Fallout
     3.   Cork dust is used as an evaluation tool for identifying areas
of potential dust fallout.  Dust is injected into the flue system at a
relatively low gas velocity sufficient to  allow fallout to occur. When
the pattern of dust deposit has become stable, the gas velocity is
incrementally increased and the clearing effect upon  the  dust  noted for
each velocity increment.  As a separate exercise, dust is introduced
into the system at a sufficiently high gas velocity to remain entrained.
By incrementally lowering the gas flow rate, areas can be identified
where  initial fallout is likely to occur.  Fallout can be controlled in
many cases by simple adjustment of existing turning vanes.  By allowing
the dust to be transported to the precipitator, an indication of  the
distribution across the face of a unit can be  also made by measurement
of the amount of dust fallout in each inlet hopper.   In a recent  study,
the dust mal-distribution across the face  of the precipitator  was re-
solved by extending the trailing edge of the turning  vanes at  a critical
position in the inlet ducting.

     During the dust test, no attempt is made  to either quantify  the
amount of fallout or predict the flow rate at  which this  would occur on
the prototype installation.  The dust is used  as a visual aid  to  identify
stagnant areas within the ducting system.

Gas Bypass of Electrode System

     4.   In practice,  less than ideal flow conditions  exist across the
collecting electrodes,  particularly at the upper and  lower extremities.  It
is very important that the full collector height be utilized while preventing
gas bypassing of the electrostatic field.   This is achieved by horizontal top
and bottom baffles across the full face of each field.  Gases receiving
partial treatment due to looping in and out of successive fields  can be
reduced to an acceptable level by this method.   Smoke or neutral  buoyancy

                                     354

-------
bubbles  are used  to  evaluate gas bypass,  particularly in the vicinity of the
last field.   To a large extent,  evaluation of this condition is by visual
means and  is  not  quantitative.   Relating  gas behavior observed in this area
on the model  to the  gas behavior predicted on the prototype should be made
with caution,  as  the extent  of  looping and disturbance will be a function of
time proportional to the model  scale.

I.D. Fan Gas  Distribution

     5.    The development of high efficiency I.D. fans has increased their
sensitivity to velocity distribution at the fan inlet.  Consequently, to
maintain efficiency,  fan manufacturers are imposing gas distribution standards
as additions  to the  model study.   Fortunately,  these criteria are not too
difficult  to  achieve because the measurement positions are normally in areas
of high  velocity.

MODEL CONSTRUCTION PROCEDURES

     Prior to the commencement  of a model study, the precipitator designer
has studied both  contract specifications  and plant site real estate constraints
to produce the most  economic ducting arrangement.  Economic, in this sense,
does not imply the cheapest, but the most efficient arrangement to transport
the gases  and remain within  the limitations placed upon him by site constraints.
It is extremely rare for ducting modifications  to result from model study
findings.  In fact,  it  becomes  the model  study  departments' responsibility to
"make the  layout  work".

     Lodge-Cottrell  started  model study work in 1947.  This practical experience
combined with theories  discussed in earlier sections of this paper led to the
adoption of a 1/8 scale minimum model size.

     Construction materials  are wood,  masonite, plexiglass and aluminum, with
the modelled  system  extending normally from the air pre-heater outlet to the
inlet side of the I.D.  fan.   The flue system from the outlet of the I.D. fan
to and including  a length of chimney are  also modelled when requested.  Where
necessary, turning vanes and internal bracings  are installed in the inlet and
outlet ducting.   Depending upon the configuration of the precipitator inlet
mouthpiece, a particular type of distribution device selected from various
proprietary designs  is  selected and installed in a starting pattern.  In the
precipitator  outlet  mouthpiece,  proprietary sawtooth distributors are
installed  in  a starting pattern.   The sawtooth  distributors ensure that gases
in the final  stages  of  cleaning remain within the confines of the electrode
system and are not prematurely  drawn towards the high velocity outlet ducting
in the vertical plane.   Within  the precipitator roof and hopper levels,
baffles  are installed and unless specified to the contrary, collecting
electrodes are simulated only in the first and  last fields of the precipitator
model.   As mentioned in the  previous section, some specifications require
that each  field of collectors be modelled.  This is based on the argument
that "the plates  act to  preserve existing horizontal distributions and that
the absence of several  groups of plates could make the modelled outlet
velocity distribution different  than it is in the actual unit."d)
Our experience shows  that in large models operating at relatively low treatment


                                     355

-------
velocities, collector plates downstream of the first field have a negligible
influence on preservation of the velocity profiles measured at the outlet of
the first field.  If a relatively uniform velocity distribution is achieved
at the outlet of the first field, then this degree of distribution will
progressively improve through the precipitator.

     Collector plates installed in the last field of the precipitator model,
however, dp_ play a role in controlling gases by minimizing the tendency of
the gas to horizontally converge towards the outlet ducting.  Avoiding
vertical convergence of the gases to the outlet ducting is achieved with
sawtooth distributors as discussed earlier.

     Following the installation of turning vanes, flow correction devices,
and distributors in the "starting pattern", the model is connected to one or
more of three test fans.  Each fan is rated at 9.4 m3/s at 406 mm w.g.,
offering the capability of testing to l/8th scale a prototype designed for
handling over 1,415 m^/s.

LIMITATIONS TO OPTIMUM DISTRIBUTION

     1.   In October 1981, the Industrial Gas Cleaning Institute (IGCI)
revised its .Publication No. EP-7, Gas Flow Model Studies.  This standard for
acceptable gas distribution states, "Within the treatment zone near the inlet
and outlet faces of the precipitator collection chamber the velocity pattern
shall have a minimum of 85% of the velocities not more than 1.15 times the
average velocity, and 99% of the velocities not more than 1.40 times the
average velocity."

     The revised standard does not impose any limitations on areas of low
velocity within the precipitator.  In an extreme case, Fig. 7, the survey
obviously does not follow any "normal" or Gaussian distribution but never-
theless conforms to the IGCI standard.  The effect of low velocity areas is
illustrated by the potential loss of efficiency from 99.15% to 98.6%.  This
would result in a 65% increase in dust emissions.  It is important to note
that the high efficiencies obtained in the low velocity areas are not
compensated by the lowered efficiencies at the higher velocities, due to the
exponential relationship between efficiency and velocity.

     Our recommendation is that the old IGCI standard "85% of the velocity
readings, within +25% of the average velocity, and all readings within +40%
of the average velocity"(2) together with the imposition of the 15% RMS
deviation at the end of the first field, is more than adequate for todays
needs.  We have used the above criteria for years and have many high efficiency
operational units to prove the veracity of our recommendation.

     2.   Unless specified differently, the model assumes that the gas
distribution at the inlet to the flue system—in the case of a boiler, the
air pre-heater outlet—is uniform.  There is no alternative to this since the
data is rarely available, but in practice not only is distribution at this
plane rarely uniform, it can also change as some degree of buildup occurs in
the air pre-heater.  Accepting that this condition exists, mal-distribution
is deliberately induced at the air pre-heater outlet to check effects upon

                                     356

-------
precipitator  distribution.   There is some smoothing due to turbulence as the
gas is  progressed  through the flue system,  but even so the very fact that
this situation  exists  implies that very fine correction of gas flow in the
precipitator  on the flow model is not justified.

     3.   None  of  the  flow distribution criteria  or work carried out takes
any account of  the turbulence induced by the ionic or electric wind.  It is
known that this produces gas movements of up to 3 m/s, that is higher than
the mean  gas  velocity  in the average commercial precipitator.   The dramatic
effect  of this  phenomena is demonstrated on Figs. 8 and 9.  Note the vertical
spread  of the gas  as the electric field is  applied.  This phenomena is
currently being studied  and we hope to have the opportunity to report the
results of this study  in the near future.

     The  cost of flow  correction devices on the prototype and  cost of carrying
out the flow  model work  escalates rapidly as distribution criteria becomes
more stringent.  In view of the pratical limits to optimum distribution
discussed above, the pre-1981 IGCI recommendations together with a 15% RMS
criteria  at the end of the first field constitute a reasonable practical
limit,  based  on our experience on 1/8 scale models.

GAS FLOW  CORRECTION FOR  FABRIC FILTERS

     Unless corrective work is performed on precipitators, mal-distribution
of gas  flow within the precipitator treatment zone.will remain there indefi-
nitely.   In the case of  a fabric filter, initial  mal-distribution to the
thimble plate will be  improved to some degree as  filter cake develops back
pressure.  However,  this infers that for some period of time,  higher than
designed  air  to cloth  ratios are experienced.  This is due to  the lack of
equal flow to filter bags.   These thoughts  prompted a recent model study of a
single  compartment within a multicompartment fabric filter.

     The  model  was constructed 1/4 full size, with 234 thimbles installed.
Fig. 10 shows the  extent of the model construction.  An additional simulated
second  compartment was constructed.   The purpose  of this addition was to
allow simulation of gas  flow take off,  as experienced in a multicompartment
baghouse.

     It should  be  mentioned that this exercise is part of an overall develop-
ment program  associated  with multicompartment fabric filters.   However,  this
discussion is limited  to the following objectives:

     1.   Determine the  velocity distribution of  inlet gases to the filter
         thimbles.

     2.   Determine the  distribution of "reverse  gas cleaning" gases to the
         filter thimbles.

     3.   Assess the disturbance and/or re-entrainment of hopper dust during
         filtering  and  reverse gas  cleaning.

     4.    Determine  system  pressure  loss.

                                     357

-------
Before Flow Correction

     Initial observations and measurements were made without flow  correction
devices, and the hoppers were partially filled with cork dust to a level
below the inlet entrance.

     Figure 11 shows the turbulent gas path with the resultant  scouring and
re-entrainment of hopper dust.  Negative flow during normal filtering mode
was observed over 54% of the thimble area, as shown in Figure 12.

     The total pressure differential across the model system was measured at
142 mm w.g.

After Flow Correction

     Continued tests led to the inclusion of flow distribution  devices
developed from our precipitator distribution technology and similar to those
used during the model study for the EPRI Fabric Filter Pilot Plant.

     The results of the final tests with the inclusion of flow  correction
devices showed that:

     1.   Positive gas flow through the thimbles gave an RMS deviation of
          14% across the cell plate.

     2.   A 23% RMS deviation during reverse gas cleaning.

     3.   The hopper dust remained undisturbed.

     4.   The pressure differential across the system was reduced to 91 mm
          w.g., a decrease of 51 mm or 36%.

     5.   Negative gas flow through the thimble plates was eliminated.

     With the addition of a few simple flow correction devices,  it is possible
to achieve acceptable flow distribution,  eliminate disturbance of previously
captured dust and reduce overall pressure loss.

CONCLUSIONS

     The approaches to flow model studies which have been discussed here have
consistently resulted in excellent correlation between onsite measurements,
and those predicted from the model.

     Figure 13  demonstrates the onsite measurements of two  recent precipitator
installations with those determined from their respective model  studies and
confirms our previous statement.

     Figure 14  (3)  shows the comparison between thimble plate velocity surveys
carried out on  the  Lodge-Cottrell designed EPRI Fabric Filter Pilot Plant at


                                        358

-------
Arapahoe  Station as  compared  to  the models  test  surveys.   These comparisons
are shown with  and without  gas distribution devices.

     The  improvement in velocity profile  using distribution devices  is evident,
and the close similarity  in velocity  contour between  model and prototype for
both conditions supports  our  confidence in  large scale modelling.

     Our  experience  in flow model studies has led to  the  following conclusions:

     1.   A  correctly designed gas distribution  system does not rely upon
         excessive  use of  resistance devices to control  flow.

     2.   The largest practical  size  model  should always  be utilized to
         ensure reasonable correlation between  the model and the full-size
         installations.

     3.   More  stringent  distribution criteria is not justified until existing
         practical  limitations  have  been addressed.

     4.   Flow  correction devices in  fabric filters can reduce re-entrainment
         of previously captured dust, and  reduce overall pressure loss.

     It is Lodge-Cottrell policy  to  conduct  gas  flow studies for large
 precipitator and fabric filter contracts.  Operating units have consistently
 met  or exceeded contract performance specifications.
     The work described in  this paper was not  funded by  the U.S. Environmental
Protection Agency and therefore the  contents do not necessarily reflect  the
views of the Agency and no  official  endorsement should be  inferred.
                                     359

-------
                                  REFERENCES
1.    Gas Flow Model Studies,  Publication EP-7, Industrial Gas Cleaning
     Instutute,  Revision 4

2.    Gas Flow Model Studies,  Publication EP-7, Industrial Gas Cleaning
     Institute,  Revision 3

3.    Cusing, Kenneth M., Wilson, Rufus Ray Jr., Smith, Wallace  B., Belkus,
     Paul and Carr, Robert.  Arapahoe Fabric Filter Pilot Plant, Air Load-
     Flow Distribution Test Results, EPRI RP 1129, November 1980

4.    Darby,  Ken.  Criteria for Designing Electrostatic Precipitators,  EPA
     Second  Symposium on the  Transfer and Utilization of Particulate
     Control Technology,  July 1979.
                                    360

-------
1.22
99.74Z
1.22
19.741
1.22
99.74X
1.13
99.84Z
1.13
99.84Z
1.07
99.B9Z
1.07
99.89Z
1.02
*9.92Z
1.02
99.92Z
1.02
»9.92Z
1.30
99.64Z
1.30
99.64Z
1.30
99.64Z
1.30
99.64Z
1.30
99.64Z
1.30
99.64Z
1.30
99.64X
1.22
99.74X
1.22
99.74Z
1.22
99.74Z
1.37
99.SOZ
1.37
99.50Z
1.37
99.50Z
1.37
99.50Z
1.37
99.SOZ
i.37
99. SOX
1.37
99.50Z
1.37
99. SOX
1.37
99. SOX
1.37
99.50X
1.45
99.34Z
1.45
99.34Z
1.45
99.34X
1.45
99.34X
1.45
99.34X
1.45
99.34Z
1.45
99.34X
1.45
99.34Z
1.45
99.34X
1.45
99.34X
1.S2
99.15X
1.52
99.15X
1.52
99.15Z
1.52
99.15Z
1.52
99.15X
1.32
99.15Z
1.52
99.15Z
1.52
99.15X
1.45
99.34X
1.45
99.34X
1.32
99.15X
1.52
99.15X
1.52
99.15X
1.52
99.15X
1.52
99.15X
1.52
99.1SX
1.52
99.15X
1.52
99.15X
1.60
98.94X
1.60
9B.94X
1.60
98.94Z
1.60
9S.94Z
1.60
98.94%
1.60
96. 941
1.60
98.94Z
1.60
98.94Z
1.60
98.94Z
1.60
98.94Z
1.60
98.941
1.60
98.94Z
1.68
98.69Z
1.68
98.69X
1.68
98.69Z
1.68
98.69Z
1.68
98.69X
1.68
98.69X
1.68
98.69Z
1.68
98.69Z
1.68
98.69X
1.68
98.69Z
1.T5
9B.42X
1.75
98.42X
1.75
98.42X
1.75
9B.42X
1.75
98.42X
1.75
98.42Z
1.75
98.42X
1.83
98.13Z
1.83
98.13Z
1.83
98.13X
1.13
9B.13Z
1.83
98.13X
1.91
97. BOX
1.92
97.73Z
1.92
97.73X
1.92
97.73Z
1.98
97.45X
1.98
97.4SZ
1.98
97.45X
1.98
97.45X
            AVERAGE VELOCITY -1.52 m/sec
            RMS DEVIATION   « 15%

             AVERAGE EFFICIENCY - 99.03Z

  FIGURE  1   NORMAL  DISTRIBUTION
0.51
100. OOZ
0.51
100, OOZ
0.51
100. OOZ
! 0.51
100.00Z
0.51
100. OOZ
0.51
100. OOZ
0.51
100. OOZ
, 0.51
100. OOZ
0.51
100. OOZ
0.51
100. OOZ
0.51
100. OOZ
0.51
100. OOZ
0.51
100. OOZ
0.51
100. OOZ
0.51
100, OOZ
0.51
100. OOZ
0.51
100. OOZ
0.51
100. OOZ
0.51
100.00X
0.51
100. OOZ
1.73
98.52Z
1.73
98.52Z
1.73
98.52Z
1.73
98.52Z
1.73
98.52Z
1.73
98.52Z
1.73
98.521
1.73
98.52X
1.73
98.52X
1.73
98.52X
1.73
98.52X
1.73
96.52Z
1.73
98.52X
1.73
98.52X
1.73
98.52X
1.73
98.52X
1.73
98.52X
1.73
98.52Z
1.73
98.52X
1.73
98.52X
1.73
98.52Z
1.73
98.52Z
1.73
98.52Z
1.73
98.32Z
1.73
98.52X
1.73
98.52X
1.73
9B.52Z
1.73
9B.52Z
1.73
98.52X
1.73
98.52X
1.52
99.15X
1.52
99.15Z
1.52
99.15X
1.52
99.1SZ
1.52
99.15Z
1.52
99.15Z
1.52
99.15X
1.52
99.15X
1.52
99.15Z
1.52
99.1SX
1.73
98.52Z
1.73
98.52X
1.73
98.52X
1.73
98.52Z
1.73
98.52Z
1.73
98.52X
1.73
98.52X
1.73
98.52X
1.73
98.S2X
1.73
98.52X
1.73
98.52X
1.73
98.52X
1.73
98.52X
1.73
98.52X
1.73
98.52X
1.73
98.52X
1.73
98.52Z
1.73
9B.52Z
1.73
9B.52X
1.73
96.52X
1.73
98.52X
1.73
98.52X
1.73
98.52Z
1.73
98.52Z
1.73
96.52X
2.49
94.62Z
2.11
96.82Z
2.11
96.82Z
2.11
96.82X
2.11
96.B2X
2.11
96.82X
2.11
96.82X
2.11
96.B2Z
2.11
96.82Z
2.11
96.82X
2.11
96.82X
2.11
96.82Z
2.11
96.82Z
2.11
96.82Z
2.11
96.B2Z
             AVERAGE VELOCITY - 1.52 B/SCC
             RMS DEVIATION    - 35%

             AVERAGE EFFICIENCY •= 98.60*

FIGURE  2   NON-UNIFORM  DISTRIBUTION
                                                       DESIGN MIGRATION VELOCITY
                                                                   • 8.0 cm/sec


                                                       EFFECTIVE CSA « 102m2


                                                       COLLECTOR PLATE AREA
                                                                   « 9290m2
                                                        DESIGN MIGRATION VELOCITY
                                                                    «= 8.0  cm/sec
                                                        EFFECTIVE CSA - 102m
                                                        COLLECTOR PLATE AREA
                                                                    «= 9290m2
                                            361

-------
             FIGURE 3
                      TOP ENTRY
                      NOZZLE
       PLAN
      AAAAAAA
      AAAAAA
      AAAAAAA
      AAAAAA
      AAAAAAA
      AAAAAA
             TRIANGULAR
             SPLITTERS
      FRONT
             TUBULAR
             DISTRIBUTORS
             FIGURE 4
                  SIDE
                      HIGH VELOCITY
                      WEDGE INLET
      I I I I I I I I I I I I I I I I I TR-SPL.TTERS
TUBULAR   -
DISTRIBUTORS
          PLAN

         FIGURE 5
£2^
              TUBULAR
              DISTRIBUTORS
                      HORIZONTAL
                      INLET
       PLAN
             ^HONEYCOMB
       FRONT
                         SIDE
                362

-------
, 14.6-
 hj 13.4-
 Q12-2-
 Ul
  11-0-
 JJ9*
 jjj 8-5-
 & 7.3-
 U
 U
   6.1
   4.9
                        FIGURE  6
   28,400
Illy
PRECIPITATOR
REYNOLDS NUMBER
                          CALCULATED  REYNOLDS
                          NUMBER FOR  PROTOTYPE
                             = 28,400
                          TREATMENT  VELOCITY
                             = 1.2 M/Sec.
. 2-4'
O 1.2-
U
JM
\ 6,000 \ ^ 	 ^***^^.,^
3,000 **--«._ "^- 	 ^_-^_ 	
e A
1
I
I 1
i
III
3
8
1
1
2
9
16
                      MODEL  SCALE
                         FIGURE  7
DESIGN:-
TREATMENT VELOCITY* 1.52 M/Sec.
MIGRATION  VE LOG I T Y= 8.00 Cm/Sec.
EFFICIENCY         : 99-15%
VE1
AVG.
0.51
M/Sec.
.OCITY
1-73
1
P
M
U)
•
TH
ROFI
1.73
LE
s
/
2.11
                        '2.49
                          ACTUAL:-
                          AVERAGE VELOCITY    =1.52 M/Sec.
                          VELOCITIES ABOVE  15% =15%
                          VELOCITIES ABOVE  40% = 1%
                          RMS %  DEVIATION     :35
                          AVERAGE  EFFICIENCY  =98-60%
EFF
100%
ICIENC1
98.52%
f
in
TH
•
0)
0)
PROF
98.52
ILI
%
•
/
96-82
                                                "94.62
                             363

-------
FIGURE 8
FIGURE 9
                          EFFECT OF IONIC WIND
 DEENERGIZED
                                                               ENERGIZED

-------
                    FIGURE  10
REVERSE
AIR DUCT
POPPET
VALVES
/ /  SN
OUTLET
DUCTING
INLET
DUCTING
          SIMULATED
          SECOND
          COMPARTMENT
        EXTENT  OF MODEL  CONSTRUCTION
                    FIGURE  11
                    ORIGINAL
                    DUST LEVEL
          DISTURBANCE  OF HOPPER DUST
     WITHOUT FLOW  DEVICES  INSTALLED
                        365

-------
        FIGURE 12
GAS  FLOW  PATH  THROUGH
COMPARTMENT  WITHOUT
FLOW  DEVICES  INSTALLED
           366

-------
                      FIGURE 13
490MW  BOILER:-
              MODEL
             SITE
  •40      AVG    +40%
       10%  RMS
              MODEL
-40      AVG    +40%
       12%  RMS
             SITE
 •40      AVG     +40%        -40      AVG    +40%

        "* "MS   END  OF 4th  FIELD6*  RMS
120MW  BOILER:-
              MODEL
             SITE
         AVG     +40%        -40      AVG    +40%
       11% RMS                   12.8% RMS
               END  OF 1st  FIELD
              MODEL
             SITE
 •40      AVG     +40%
       9% RMS
               END OF  5th  FIELD
-40      AVG     +40%
       5%  RMS
                          367

-------
            FIGURE  14
MODEL                  PROTOTYPE
 WITHOUT  DISTRIBUTORS
 MODEL                  PROTOTYPE
      WITH  DISTRIBUTORS
                368

-------
                           AIR FI.OW MODEL STUDIES

                              by:  L. H. Bradley
                                   United Engineers and Constructors
                                   Philadelphia, Pennsylvania 19101
                                  ABSTRACT

    This paper establishes the need for strict control of air flow model
studies and  the higher acceptance standards for test results which are required
to obtain the collection efficiencies needed to satisfy regulatory require-
ments.   The  importance of flow patterns and minimum pressure losses is reviewed
and discussed including the importance of dimensionless ratios such as Reynolds
Number, modified Fraude Number, and the momentum ratio.  The requirements for
design, scale,  model limits,  fabrications and test procedures as established
by I.G.C.I.  Standards and additional requirements are shown and discussed.
Data collection methods including limitations and evaluation methods for test
data are presented  and discussed.   Examples from recent air flow model studies
are presented and reviewed to illustrate items which should be considered.
Based on these results, a comparison of deviation from design collection
efficiency as a function of root mean square deviation is shown.

    The work described in this paper was not funded by the U.S. Environmental
Protection Agency and, therefore,  the contents do not necessarily reflect the
views of the Agency and no official endorsement should be inferred.
                                     369

-------
1.0  INTRODUCTION

         The use of scaled models for evaluation of  flow distribution patterns
     and to establish expected pressure losses has been  common practice for
     many years.  Because of the varied opinions regarding  the interpretation
     of the test data and the lack of field data to  verify  reproducibility in
     the prototype installations, the need for and the importance  of  model
     studies have been questioned for many years.  However,  over the  last few
     years, numerous existing installations have improved the  performance of
     the particulate collector by conducting model studies.  In addition, two
     factors have combined to emphasize the need for detailed  model studies.
     These factors are:

       a   The regulatory standards required a reduction in particulate
           emission level to 0.03 Ibs/mm BTU.  This, in  turn,  requires
           collection efficiencies ranging from 99.4% to 99.8% and

       b   The cost of energy has increased substantially.

         To obtain the efficiencies mentioned in (a) above,,  excellent flow
     distribution must be assured.  Item b above dictates that the system
     consisting of the collector and associated ducts be designed  for a
     minimum pressure loss.  For these reasons, United Engineers believes
     that detailed model studies are a necessity and therefore, includes
     this requirement for all particulate collection systems.

         The requirements for a model study often have been expressed in a
     single paragraph contained in the equipment specification.  This can
     and often does result in an inferior model study and misunderstanding
     between the owners, the Engineers, and the test personnel regarding the
     overall requirements.  It is believed that a separate  section of the
     specification should be devoted to the model study  and this section
     should include the scope, the design requirements,  the acceptance criteria
     for the test results, the preferred methods for data collection  and the
     acceptable evaluation methods.  United Engineers also  insists on the
     right of witness and the review of test data.   This paper will outline
     the requirements which are recommended for inclusion in the specifications
     for a model study and provides some example of  data collected during a
     model study.

2.0  MODEL STUDY REQUIREMENTS

         The requirements for a model study may be issued as a separate speci-
     fication or as a separate section of the equipment  specification. The
     following items are suggested as the minimum requirements to  be  included:

         A)  Purpose of the Study
         B)  Scope of the Study
         C)  Fabrication of the Model
         D)  Design and Acceptance Criteria
         E)  Data Acquisition
                                     370

-------
    F)   Data Evaluation
    G)   Special Requirements

    The inclusion of and the compliance with the details of the above
requirements will result in model study results which will dictate the
design and placement of flow distribution devices at the inlet and the
outlet of the precipitator for flow distribution as, well a® areas in the
associated ducts which require the installation of turning vanes to
correct flow disturbances.  The incorporation of these items in the
prototype unit will obtain minimum system pressure loss and will enable
the collector to operate at the designed collection efficiency.

2.1  PURPOSE

         The purpose of all model studies is simple:  optimize
     the flow distribution through the use of flow corrective
     devices and to obtain minimum pressure loss.  Both of the
     above must be consistent with acceptable engineering design
     standards.

2.2  SCOPE

         To obtain full evaluation of the gas flow distribution
     and the associated pressure loss, the scope of the equip-
     ment to be modelled should include all ducts from the air
     heater outlet to the inlet transition box of the collector,
     the collector including the inlet and outlet transition box
     and the connecting ducts to the inlet of the induced draft
     fan.  For installations with separate primary and secondary
     air heaters, both connecting ducts should be modelled and
     the flow should be adjusted to reflect the expected flows
     in the prototype unit.  In addition, the ducts should include
     internal struts, trusses, etc. to fully duplicate the flow
     distribution and the pressure losses.

         For units with more than one collector, both units should
     be modelled unless one is  a mirror image of the other.  If
     differences in ducts, turns, etc., exist, separate studies
     are required to determine both the flow distribution and the
     pressure loss.  In some cases, the same collector can be used
     and the connected ducts are modified.

2.3  FABRICATION

         To obtain similarity between the model and the prototype
     unit, the fabrication of the model and the associated ducts
     must be based on drawings of the prototype and be to the
     selected scale.
                                371

-------
         I.G.C.I.  standard E.P.,^7 specifies a scale of 1 to 16
     or larger and is commonly used in the United States,  A
     scaling factor of 1 to 12 is usually used in Canada, Great
     Britian, and  the European Countries.

         The fabrication of the system should include the inlet
     and the outlet transition boxes complete with flow distri -
     bution devices, the collector complete with plates in all
     electrical fields, ash collection hoppers and all associated
     ducts complete with internal struts, braces, etc.  Provision
     should be incorporated into the design of the model to per-
     mit easy installation of turning vanes, flow distribution
     plates, etc.   Some model studies have been conducted without
     installation  of all internal collector plates, but the
     inclusion of  all plates will provide a more accurate evalua-
     tion of internal drift or movement of gas flow within the
     collector between the inlet and the outlet.

         The material for fabrication of the model should permit
     visual observation of the flow patterns as an integral part
     of the test program.  Plywood or sheet metal may be allowed
     for fabrication of duct sections where visual observation
     of the flow pattern is not critical.

2.4  DESIGN CONSIDERATIONS

         General guide rules for the design of the model are
     found in I.G.C.I. standard E.P.-7, however, depending
     upon the intent of the test and the variables being studied
     consideration must be given to the following dimensionless
     ratios to insure continuity between the model study results
     and the performance of the Prototype unit.

          1)  Reynolds Number    = Inertia Forces
                                   Viscous Forces
          2)  Froude Number      _ Inertia Forces
                                 "" Buoyancy Forces
          3)  Momentum Ratio     _ Momentum - Stream 1
                                   Momentum - Stream 2

         The importance of these ratios to specific tests was
     discussed in a May 1974 article in Power Engineering1 and
     additional information was contained in a November 1976
     issue of Research-Development2.
                                372

-------
     2.4.1  REYNOLDS NUMBER

                The majority of model studies investigate only
            the gas flow distribution and the pressure loss and
            need only to consider the influence of the Reynolds
            Number.  Due to the difference in effective diameters
            between the model and the prototype, a match of the
            Reynolds Number is not feasible and the normal procedure
            is to match the flow velocity and then to verify that
            the Reynolds Number is in the turbulent region.  If this
            condition is met and dimensional similarity has been
            obtained, the flow patterns and the pressure losses
            existing in the model will be representative of the con-
            ditions to be expected in the prototype unit.

     2.4.2  FROUDE NUMBER AND MOMENTUM RATIO

                 For model studies to determine the thermal mixing
            of two gas streams entering at different temperatures
            and to evaluate plume effects for atmospheric discharges,
            both the density modified Froude Number and Momentum
            Ratios need to be considered.  For these studies,  it is
            essential that the values used for the density modified
            Froude Number and the Momentum Ratio match the calcula-
            ted values which are expected for the prototype installa-
            tion.  For normal cases,  the frictional component  will be
            small relative to the other forces and can be eliminated.

2.5  ACCEPTANCE CRITERIA

          The acceptance criteria outlined in the October 1981 edition
     of I.G.C.I standard EP-7 represents a minimum criteria which
     is required to ensure that the high design collection efficiencies
     will  be obtained.  The requirement that 85% of all data points
     be within + 15% of the average velocity is excellent and  repre-
     sents a level which has been specified by many consultants and
     some  manufacturers for several years.   Based upon a review of
     available test data and in consideration of the need to maintain
     high  collection efficiency during extended periods of operation,
     it would seem that the requirement for 99% of all data points
     to be within ± 40% of the average velocity should be modified
     to require that 95% of all data  points be within + 25% of the
     average and except for a few scattered points, the balance
     within the + 40% range*   Based on actual test, we believe this
     criteria to be obtainable.

        In addition to the recommended tolerance for flow distri-
     bution variations, the maximum root mean square (EMS) deviation
     may be specified and based upon  present criteria for the  high
     collection efficiencies, the RMS deviation should be in the 8 to
     14% range to assure compliance.
                               373

-------
2.6  DATA ACQUISITION

         The collection of the raw test data is a critical item
     in the performance of a model study since incomplete erroneous
     or non-reproducible data can lead to improper evaluation of the
     results and, therefore, incorrect conclusions may be reached
     regarding flow distribution which, could result in the installa-
     tion of wrong flow corrective devices and non-performance of
     the prototype unit.

         Calibration of all instruments used in the test is required
     and the calibration is to be verified before the start of each.
     series of tests.  Calibration records should be available for
     inspection and should be included in the final report of the
     model study.

         For most applications, the raw test data for flow deter-
     mination is secured using either a pitot tube or a hot wire
     anemometer with the selection being based on the existing
     velocity. A hot wire anemometers is normally used when the
     flows are below 40 ft. per second and are accurate to very
     low flows.  The resultant data is usually recorded on a
     continuous strip chart and must be converted to velocity data
     at predetermined locations to assure continuity between all
     tests.  The resultant data provides flow rate data but does
     not provide directional indications therefore other tests
     must be conducted to "verify direction of flow.

         Pitot tubes are used in areas where higher velocities
     exist and for static head measurements.  Care must be taken
     to ensure that all readings are taken at consistent locations
     and that the probe is inserted with care to assure correct
     alignment.

         To provide the directional component of the flow which
     is not obtained with a hot wire anemometer, it is common to
     use tufts of string or cotton attached to a small diameter
     wire which can be inserted through predrilled holes at
     selected locations to ascertain flow direction.  The presence
     of eddies, reverse flow, dead spots, etc., can be determined
     and proper flow corrective devices can be installed to correct
     the undesireable condition which will also assist in obtaining
     the minimum pressure drop.

         To assure proper evaluation of the velocity patterns, the
     minimum number of data points as prescribed by I.G.C.I, standard
     should include sampling in every thicd gas passage lane between
     collector plates and the vertical distance between data points
     should not exceed 10% of the plate height.  The pattern of sampling
     should be repeated for the inlet, at the midpoint of gas. passage
                                374

-------
     through the unit and at the outlet,   (The pattern is
     normally defined by assuming that the observer is in
     front of and facing the inlet.    Row 1 is located at the
     unit left side.   Vertical data  points are recorded from
     top  to bottom.)

         The final configuration should also be tested at flows
     of 75% and 50% of the design flow rate to verify that
     acceptable flow distribution is being obtained.   The unit
     is normally designed for the maximum expected flow based
     on the fan capability,  therefore, testing at flows higher
     than the design rate is normally not done.

2.7  DATA EVALUATION

         Careful and accurate evaluation of the raw test data
     obtained in the model study at  each sample point is essential
     to assure that the conclusions  reached will enable the  proto-
     type unit to operate in accordance with the specified require-
     ments.   As a minimum, the evaluation  should include the
     following methods:

         1)   Iso-Velocity Profile Plots
         2)   Velocity Histogram Plots
         3)   Root Mean Square Deviation - RMS
         4)   Visual Observations

         Examples of data collected  during recent model studies
     are  presented below with reference to each specific method
     of evaluation.  These are included to illustrate suggested
     methods.

     2.7.1  ISO-VELOCITY PROFILE PLOTS

                The Iso-Velocity plot is  an actual plot of the
            velocity existing at every sample point location and
            provides the best means  of understanding  the flow dis-
            tribution existing in the model.   By comparison  of the
            actual velocity existing in the model for the inlet
            the midpoint and the outlet,  determination of flow
            distribution can be made with respect to  the following
            criteria which are cf concern to ensure performance.
            Figures 1 through b are  presented to illustrate  fpeei*
            fically the following items which may be  evalua ted using
            the following:
                a)  Identification of areas within the model where
                    either high or low velocities exist.

                b)  The velocity profile existing at  the bottom o f
                    the model in the area just above  the ash hopper.
                                375

-------
    c$  A means to determine and to evaluate internal
        drift within the gas passage lanes as the flow
        proceeds from inlet to outlet.

    d)  The influence of internal obstructions,

    e)  The compliance with the specification require-
        ments as they pertain to flow distribution.

    Figure 1 represents a velocity profile plot for flow
distribution existing at the inlet of an electrostatic
precipitator model as it was recorded in a recent model
study.  In this instance, the velocities are shown as a
percentage of average velocity, however, actual velocities
as recorded may be used.  Note that except for a few data
points, the velocity points in the bottom rows are at or
below the average velocity.  This is desirable since it
reduces the possibility of re-entrainment.  For this pro-
file, the root mean square (RMS) deviation is 10,2% and
unless other problems are discovered in the evaluation,
the profile as obtained represents good flow distribution
and is acceptable.

    Figure 2 represents an unacceptable flow pattern which
was obtained from a different test series.  Even though the
RMS deviation is only 13.9%, the concentration of high flow
velocities in the center of the unit with low velocities on
the outer edges was not acceptable to the manufacturer or
to United Engineers and corrective action was taken to
modify the flow patterns.

    Figure 3 has been included to illustrate objectionable
condition existing at the bottom of the unit and represents
an unacceptable condition.

    Figure 4 has been included to illustrate the effects
of an internal truss on flow distribution which was located
upstream of the sample area and was observed in a recent
study.  The very low velocities as seen on the four sides
are a result of the gusset plates used to secure the truss.
In addition to the effects on flow distribution, each truss
is a source of additional pressure loss.  As an example, if
the obstruction caused by a truss reduces the total area by
12 to 15%, each truss can cause an additional loss of 0.10
IN-WG.

    Figures 5 and 6 have been included to illustrate internal
drift of gas flow within a unit.  Figures 4 represents the flow
distribution at the inlet and Figure 5 indicates the dis-
tribution at the outlet.  A comparison of Figures 5 and 6
                      376

-------
        indicate a definite drift of the ga.s flow, as it
        proceeds through the unit.  Because the majority
        of the velocities are within + 15% of the average.
        velocity and the RMS deviation are in the 8 to 12%
        range, this drift pattern can be accepted.  If a
        wide variation in velocities existed with high
        velocities at the bottom, the flow patterns should
        be corrected.

2.7.2   HISTOGRAM VELOCITY PLOTS

            A histogram is usually plotted as a bar graph
        and is plotted showing the number of data points
        existing as a function of the percent deviation
        from the average velocity.  The plots provide a visual
        means of verifying compliance with specification
        compliance regarding the distribution of all
        velocities and also provide a ready reference to
        indicate the number of data points outside the
        specification limits.

            Figures 7 and 8 are histogram plots taken from a
        recent study and represent conditions at the inlet
        and the outlet of the unit.  In addition to verifying
        compliance with specification requirements for flow
        distribution, these plots provide an indication of
        internal drifts although exact location cannot be
        verified.

2.7.3   ROOT MEAN SQUARE (RMS) DEVIATION

            This mathematical ratio provides a reliable
        indication of the capability of the measured flow
        distribution to obtain the specified collection
        efficiency.

            The data used in plotting Figure 9 has been
        compiled from a number of model studies with different
        collection efficiencies.   Using the actual velocity
        for each data point and the expected performance curves
        as furnished by the manufacturer, the collection
        efficiency for each point was calculated.   The collection
        efficiency was compared to the design value and the
        deviation from the design value was established.  This
        deviation from the design value has been plotted as a
        function of the root mean square deviation.   From this
        plot, it would appear that to obtain and maintain the
        high collection efficiencies now required, the FMS
        deviation should be in the 8 to 14% range depending upon
        the specific job requirements.
                             377

-------
        2.7.4  VISUAL EVALUATION

                   The use of  visual observations in the evaluation
               of flow patterns enables the test program to determine
               the direction of gas flow,  to correct objectionable
               flow patterns in ducts and  to evaluate drop out and
               re-entrainment  potential.

                   The determination of flow direction is accom-
               plished by using cotton tufts or stting attached to
               a small diameter wire and inserted into the flow stream.
               Areas of concern can be identified and corrective
               action taken.

                   The evaluation of dead  areas of drop out potential
               and hopper re-entrainment is performed using powdered
               cork which is deposited in  the duct area upstream of
               the model.  The unit is then operated at 50% flow,
               then at 75% flow and at 100% of design flow.  Results
               are noted for all cases to  determine the pickup poten-
               tial as the flow rate is increased, the presence of
               any areas where dropout may occur and the potential
               for hopper re-entrainment.   In a few cases, fly ash has
               been used and although good results were obtained, the
               procedure requires the addition of a small fabric filter
               to the outlet to avoid discharging fly ash into the
               laboratory area and also to protect the instrumentation.

   2.8  PRESSURE LOSS

            The pressure drop  raw data is  normally secured using a
        sensing type pitot tube and an inclined monometer to obtain
        static pressures at all locations.  The model data is converted
        to expected pressure loss for the  prototype unit by the square
        of the flow ratio: and  the ratio of the density.

   2.9  SPECIAL REQUIREMENTS

            This section should include the requirements for inspection
        of the model, witnessing of tests, review of data, and require-
        ments for submission of the test report.  The contents of this
        section may be varied  depending upon the requirements of the
        owner and/or Engineer.

3.0  CONCLUSION
         Operation of a number of installed units where the model studies
     were controlled to the standards as presented have demonstrated that
     the design collection efficiency and pressure loss requirements have
     been met.   Therefore, the importance of a detailed model study has
                                     378

-------
    been verified, provided  that  careful  attention has  been  given to the
    model specification, the collection of  data  and to  the  evaluation and
    interpretation of the results.
                              REFERENCES

1)   Experimental Flow Modeling for Power Plant Equipment,  Gerald  B.  Gilbert,
    Dynatech Corp. - Power Eng., May  1974.

2)   Scale Model Flow Testing, David J. Gibson, Research Cottrell,  Research/
    Development, November 1976.
                                   379

-------
 89  87
 92  95
104 104
 99  98
106  89
106  91
107 101
109 101
 93  92
 96  91
 87  90
 84  99
 71  87
             93 106
             91 109
             90  87
             96 101
             93 103
             95  96
             99  96
         88 106  98
         90 105 102
        105 102  96
        112  98  91
        110  93  96
        102 101  95
 83
 89
 99
 78
 81
 84
 87
    113 109
    117 112
    105 115
    111  78
    111
    109
    104
    115
    106
     99 106
     95 116
     95 115
    101 115
96
83
94
98
94
100 106
102 109
100 124
105 117
110 114
109 114
104 118
106 114
106 107
103 118
105 108
102  87
103  80
89
89
80
76
89
82
90
96
95
92
106
85
76
73 55
93 102
103
108
92
93
93
89
111 1
109
108
106
73
115
116
112
118
115
114
109"
109
108
107
94
74
108
116
110
111
112
110
109
112
108
116
113
86
75
111
116
111
119
111
108
115
113
112
115
123
72
105 58
111 107
117
119
118
111
112
119
115
109
113
122
116
117
111
97
98
94
98
111
116
115
88 96
80
112
113
104
102
85
86
88
82
102
105
96
82
91
90
79
76
82
85
90
107
110
94
98
97
74
       FIGURE NO. 1
   VELOCITY PROFILE PLOT
                                                 FIGURE  NO.  2
                                            VELOCITY  PROFILE PLOT
 90  75
 95  95
105 100
110 110
 95  95
 90 100
 80 105
 85 100
 85  90
   ^90
     95
105
125
115 105
140 115
 90  90
 90  90
105 100
 85  85
 95  75
 95  80
100 100
105  85
110  75
 95  70
110 110
 95  90  75  80  85  103  106   104/TI8
105  80  70  90  80
110  70  65  95  95  103
 85  70  75  90  90
 80  85  70  80  90  105.
 95  90  85  75  95
 90  90  90  95 105
 95  85 100 105 100
 85  80  95 .100  80
 90  75  80 100  95
                                                                    108  125
                                                                         120
120  70  75 110 [HO
        115 115 110 105  90  120  125
        125 105 100 110 125  130  145
       FIGURE NO. 3
  PROFILE PLOT-HIGH BOTTOM VEL.
  UNACCEPTABLE PROFILE
                                      99
                                      99   101
                                                                 114
                                                            110  137
                                                  FIGURE NO. 4
                                           PROFILE PLOT-TRUSS EFFECT
                                    380

-------
105 112 102
114 102 120
110 105
102  98
100  98
 98 100
 98 105
 94 110
110  98
100 104
 90 106
 98 100
 96  92
 90  88
 88  84
 98  94  96
115 108 100
100 105
 98 110
90
96
92
88
84
87
78
86
90
88'
80
"88
86
90
84
86
88
90
76
82
90
96
86 92
89 100
82
88
86
88
90
90
84
86
|94
100
105
100
94
jioo
78
86
92
88
80
86
80
95
105
110
102
100
86
82
82
84
90
88
80
82
96
100
90
110
104
126
120
1.14
122
112
110
106
104
100
95
90
98
90
105 110
105 100
124 116
116 120
108 106
100 100
106 100
100 106
 98 110
100  98
 96 105
108 104
100 102
 96  98
 98  96
106  90
110 100
118 102
116 100
110 100
 98 102
 96  96
100
108
110
100
                                 90
                                 88
                                 94
                                 96
                            100 100
                             98  96
                    100  90  96  96
                     FIGURE 5
                 PROFILE PLOT-INLET
106
108
104
100
100
98
96
88
94
100
102
100
96
96
90
98
112
110
98
106
110
114
114
110
98
106
96
98
90
88
96
110
104
106
98
106
120
110
112
104
98
100
96
96
100
100
110
98
110
94
100
126
100
108
112
110
100
100
98
102
106 100
110 108
115 110
120
128
132
130
124
116
116
120
110
90
LOO
LOO
"88"
86
86
90
78
80
84
94
96 92
96 88
100
104
112
100
180
84
90
88
70
82
80
78
90
90
82
98
98
102
92
78
80
84
86
84
80
82
90
96
90
80
96
100
100
88
76
78
82
90
86
82
78
80
96
96
88
96 100
96 98
94 100
96
84
82
80
90
92
86
80
fir
98
90
96
98
92
96
98
100
96
100
100
100 96
90 94
102
104
100
98
102
104
96
98
100
94
96
100
98
98
100
                    FIGURE NO. 6
                 PROFILE PLOT-OUTLET
                         381

-------
  1  I '  I '  I '  I '  I '  I '  I '  I
40 30 20 10 0  10 20 30 40

    FIGURE NO. 7  INLET
    HISTOGRAM PROFILE
       % DEVIATION
    FIGURE NO. 8 OUTLET
    HISTOGRAM PROFILE
       % DEVIATION
124

116   o

108   *
      v>
100 v>O
    oz
92 ZQ

 o. Q<
 84 < uj
    ii i O£
                                     68  tu
                                        UO
                                     60  Ouj
                                        sJ>
                                     52  >u-
                                     ,A
                                     36

                                     28

                                     20

                                     12
                                      8
                                      4
                                    124
             382

-------
                                      -* o
                                        o
                                        ro
                                        CM




                                           g


                                        o  ^
                                        CM  >

                                           8
                                       •O
    CM
                             eo
                      v.
      FIGURE NO. 9


AIR FLOW MODEL STUDIES


          383

-------
        COLLECTING ELECTRODE RAPPING DESIGNED FOR HIGH EFFICIENCY
           ELECTRIC UTILITY BOILER ELECTROSTATIC PRECIPITATORS

              by:   A. Russell-Jones and A.P. Baylis
                   Lodge-Cottrell Limited
                   Birmingham, B3 1QQ, United Kingdom.
                                 ABSTRACT

     The influence of collecting electrode rapping on the efficiency of an
electrostatic precipitator is examined.  Particular emphasis is placed upon
the limitations caused by dust re-entrainment.

     The way in which the effectiveness of the dislodgement blow is often
specified in terms of measured shock acceleration on the collecting
electrode, is shown to be unsatisfactory in view of plate and accelerometer
response variations.

     The means by which re-entrainment can be minimised is discussed and a
theory which can be used to explain observed performance is advanced.

     Long term performance of Electric Utility gas cleaning plant designed to
achieve the best compromise of rapping requirements is reviewed.  Examples
of 500 MW power plant precipitators, and larger, operating satisfactorily
over periods up to 17 years after commissioning, are cited.
                                    384

-------
        COLLECTING ELECTRODE RAPPING DESIGNED FOR HIGH EFFICIENCY
          ELECTRIC UTILITY BOILER ELECTROSTATIC PRECIPITATORS


                               INTRODUCTION

     The overall process of dust removal from the flue gas stream by
electrostatic precipitation can be considered to comprise 6 main stages:

     1.   Particle charging.

     2.   Migration through the gas to the collecting electrode.

     3.   Particle deposition on the collecting electrode.

     4.   Dislodgement of agglomerates.

     5.   Settlement into the hopper -

     6.   Withdrawal from the hopper -

     Only when the dust has been withdrawn from the hopper can it be
counted  as "caught".

     Dislodgement of  the accumulated  dust layer from the collecting
electrode must be effected at regular intervals.   It represents a series
resistance on the surface, which reduces the effective field voltage, and
hence the efficiency  of dust removal.

     In  order that the dust layer is  dislodged,  a shock impulse must be
applied  to the collecting electrode.   This shears the dust from the plate
as agglomerates of the initially deposited fine dust.  It is the means
of achieving this end which we will discuss in this paper-

     For some years,  it has been fashionable to define the effectiveness
of rapping of the electrode system in terms of a specified minimum
acceleration measured on the surface.  This specification is in multiples
of 'g1,  the acceleration due to gravity (9.81  ms~2).  Such a system of
specification is an over simplification of the requirements for dust
dislodgement, since it ignores the inter-relationship between acceleration,
displacement, and frequency.   We will endeavour to show that such is the
variation in these three factors that reliance upon acceleration alone as
an indication of satisfactory rapping effect can be misleading.
                                    385

-------
     The methods adopted by various manufacturers to effect dust dislodge-
ment are to some extent dependent upon the construction of their collecting
electrodes.  Broadly speaking they fall into three main classes:

     1 .  Top rapped.

     2.  Bottom rapped.

     3.  Multi-level rapped.

     These can be further sub-divided in terms of prime mover:

     a.  Drop rod rapping.

     b.  Tumble hammer rapping.

     c.  Vibratory rapping.

     These are illustrated in Figure 1.

     Bottom rapped systems are generally of the long, rolled form electrode
type and use internally mounted tumble hammers to produce the blow.
Similar tumble hammers are used for multi-level rapping.  Top rapped
systems on the other hand can be found using all three types of prime mover.
The difference in response brought about by these various constructions and
rapping blows are also to be considered in this paper.


                             THE SHOCK IMPULSE

     It has been generally assumed that the most important factor in
assessing a rapping system is the minimum acceleration measured on the
collecting electrode surface.  This is based on the view that a mass of
dust  'm1 on the surface accelerated by an acceleration  'a1 is subjected to
a force  'F1 according to the expression:

     F  =  m.a                         F  =  Force  (N)
                                       m  =  Mass (kg)
                                       a  =  Acceleration  (ms   )

When the shock impulse is applied to the collecting electrodes, a range of
forced vibrations is set up and propogated to the whole system.  This
impulse can be broken down into its constituent frequencies by Fourier
transform analysis.  Data have been variously reported showing how
electrodes of differing construction exhibit different frequency response
patterns.  Zarfoss (1) showed data which are consistent with our own
findings, viz. the frequency content of measurements made on a rolled form
electrode are significantly higher than those found on a channel and plate
type as used by Lodge-Cottrell.  These differences are extremely important
as we will demonstrate later.
                                     386

-------
                        EFFECT OF THE RAPPING BLOW

     Work was carried out some years ago in the U.K. by the Central
Electricity Generating Board (2) to try to determine the residual dust layer
thickness on a collecting electrode as a function of the local level of
acceleration when the electrode was rapped.  The data which were obtained
are shown in Figure 2.

     It would seem from the data in Figure 2, that an acceleration level of
at least 30'g' is required to ensure that only a small residual layer of
dust remains.  However, it is important to bear in mind that the data
refer to one dust, on one type of electrode, and measured with a particular
accelerometer.  All three of these factors have a significance which we hope
to demonstrate.

     The effect of a 30'g1 acceleration upon the dust layer is not in fact
a predictable one.  This was elegantly demonstrated by Juricic and Herrmann
(3).  In a small test cell, fly ash was deposited electrostatically onto a
collecting electrode.  Peak initial accelerations of 20 to 25'g1 were
applied to the deposited dust layers and observed by means of high speed
cinematography.  Two frequencies of vibration were used, viz. 100 and 500 Hz.

     The differences observed between 20'g1 acceleration for two frequencies
were quite startling and enlightening.  At 100 Hz. the dust layer was
largely broken up into re-entrainable debris, whereas at 500 Hz. the layer
slid off the plate in large agglomerates.

     We interpret these observations in the following manner-  The
acceleration in the dust layer subjected to sinusoidal vibration is
described by the equation:
                                            a  =  Acceleration (ms  )
     a  =  4Tffd                          f  =  Frequency (Hz)
                                            d  =  Deflection (m)

For an acceleration of 20'g' at 100 Hz. and 500 Hz., the linear displace-
ments are 5 x ,10-4 m and 2 x 10"-* m respectively.  The kinetic energy
imparted to the dust is given by the equation:
                                            E  =  Kinetic Energy (J)
                                            m  =  Mass of Dust (kg)
     E  =  %mv   =  4fT f d m               f  =  Frequency (Hz.)
                                            v  =  Velocity (ms^')
                                            d  =  Displacement (m)

Thus at a constant level of acceleration (e.g. 20'g') the kinetic energy
imparted to the dust at the two different frequencies is given by the ratio
E at 100 Hz.      4-fl 2(1002)(5 x 10"4)2m
E at 500 Hz.      4-tf2(500^)(2 x 1 Q-^m     1>e'   b

This significant difference in the kinetic energy terms can account for the
observed behaviour on dislodgement under the conditions of similar applied
peak acceleration, and hence similar applied force.
                                    387

-------
     It is reasonable to expect that similar effects should be  found  in the
dislodgement of dust in the full scale precipitator unit.  In the  light of
these observations it is clear that the frequency content  of the shock
impulse is important.

     Since it is well established that different forms of  collecting
electrode respond very differently to the impact blow in terms  of  the
propogated shock impulse frequency content, it follows that the dislodge-
ment of dust for a given acceleration level will be significantly  different.
Identical measured peak accelerations can therefore produce large
agglomerates, or debris, depending on the frequency content of  the local
shock impulse.

     For the reasons outlined above we maintain that it is misleading to
apply a specification minimum of a specific number of  'g1  and try  to  make
comparisons between different forms of collecting electrode on  that basis.
Indeed the method of measurement can account for a variation of more  than
3:1 in the 'g1 levels.
                     RAPPING ACCELERATION MEASUREMENT

     A major reason for disagreement between those supplying precipitators,
and those specifying them, and in particular laying down minimum
acceleration criteria, is that the measurement of acceleration  is  sensitive
to the type of accelerometer used.

     Accelerometers used for measurement of acceleration on collecting
electrodes have ranged in weight from 4x10~2 kg downwards.  Various
reports on, and studies of, the mass/acceleration effect have been made
(4, 5, 6).  It might seem surprising that when measuring shock  impulses on
a structure weighing over 1000 Kg, the introduction of  a 4 x 10~2  kg mass
could affect the measured result.  It must be recognised that the
accelerometer will nevertheless interfere with the vibration at the point of
attachment to the thin sheet material of the collecting electrode. The
degree of interference will be dependent upon the material thickness, and
the method of electrode construction.  Only an accelerometer with  zero mass
could measure the true acceleration level.  Figure 3 shows how  the mass and
acceleration are inter-related for a specific position  on a 13.72  m x 4.57 m
collecting electrode.  The observed acceleration levels for identical
rapping blows range from 18  'g1 to 85  'g1.

     Workers at Princetown University have attempted to explain this effect
in terms of refraction of the incident shock wave and developed correction
curves in terms of both frequency and accelerometer mass  (4).   The curves
permit the calculation of the theoretical acceleration  using an
accelerometer with zero mass.
                                     388

-------
     With the mass effect in mind we have used the lightest practicable
accelerometers currently available.  Both the major manufacturers of
accelerometers - Endevco of California, U.S.A. and Bruell and Kjaer of
Denmark - offer micro-miniature accelerometers with suitable characteristics,
weighing less than 5 x 10-4 kg.  For accelerometers of such a low mass no
significant correction need be made.

     Data obtained using a measuring system comprising an Endevco type 22
accelerometer and 'SMART' processing equipment, are shown in Figure 4 for
a complete survey of a 13.72 m x 4.57 m collecting electrode.  The data
show a spread of acceleration level from 62'g' up to 285'g', with the
point of lowest acceleration close to the bottom of the electrode.
Fourier transform analysis of the frequency content of the shock impulses
at the two extremes reveal much higher peak frequencies associated with
the individual plates where high 'g1 accelerations are measured.  This
serves to moderate the kinetic energy input to the dislodged dust, whilst
at the same time producing a sufficient acceleration to ensure dislodgement
of the thinner dust layer which tends to form at the top of each collecting
electrode as compared to the bottom (see later).
                        DISLODGED AGGLOMERATE SIZE

     Considering the six stages of precipitation listed earlier - it is
obvious that the agglomerates dislodged from the collecting electrode will
only reach the hoppers if their aerodynamic effective size is greater than
a critical minimum.   This minimum agglomerate size will vary according to
the position on the  collecting electrode and the field from which it is
dislodged.  The controlling factor is the residual contact time in the
precipitation fields.  This must exceed the time taken to fall into the
hopper.  The most critical field is therefore the outlet field.

     Figure 5 shows  the free fall terminal velocity as a function of
agglomerate size.  In order that a velocity of more than 7 ms~1 is achieved,
the equivalent sphere diameter needs to be at least 10~^ m.  The data
plotted in the graph are for an apparent particle density of 1000 kg m
based on a dust with a density of 2000 kg m~^ with an agglomerate voidage
of 50%.

     Another way of  looking at this effect is shown in Figure 6.  Here are
shown computer-calculated particle trajectories for various agglomerate
sizes.  The gas is typical flue gas at 150 C.  The assumption is made that
the dust is projected from the collecting electrode with no forward
velocity.  It is accelerated up to the gas velocity.  The trajectories
illustrate the important point that provided the agglomerate size can be
kept large, tall collecting electrodes with a height of upto 15.24 m can
be used with minimal re-entrainment risk.  They also emphasise that badly
designed rapping systems can result in excessive rapping re-entrainment
losses from the outlet field of a precipitator.
                                    389

-------
     In order that the size criterion can be met, the rapping  equipment
must be designed with attention paid to the interval between rapping blows -
the "periodicity" of the rapping mechanism.  It is essential that between
rapping blows a cake of dust is allowed to build up.  Since for  the majority
of dusts the adhesive strength of the dust to the plate exceeds  the cohesive
strength of the dust layers, dust will be sheared from dust.   This leaves
behind a tenacious layer referred to by Ruckelhausen as the "basic"
layer  (7).

     Ideally, the periodicity of the rapping blow should be optimised so as
to develop large dislodged agglomerates, but not so great as to  upset the
electrical operation of the plant.  Setting of the periodicity is generally
a compromise.  Its significance can be appreciated by reference  to Figure 7.
The plant performance for a precipitator field with an inlet dust burden of
approximately 1 gm~^ is shown by Darby (8) for a range of periodicity from
seconds to hours.  The performance factor (plotted as the effective
migration velocity) shows a steady rise through to a peak and  then a fall-of.
The latter is due to the dust layer being of a thickness whereby it has
started to affect the electrical performance of the precipitator.

     Variation in boiler load and coal ash content both have an  effect on
the rapping periodicity requirements.  To date, the settings of  periodicity
have been a fixed compromise.  However, as Bradburn and Darby  (9) point out,
microprocessor-based control equipment for electrostatic precipitators
open up the possibilities of programming in automatic compensation for
changes of this type.  The programming can also ensure that when an upstream
field  is de-energised the succeeding field rapping periodicities are
adjusted to compensate for the change in dust loading.  Only the last field
would  be left with periodicity unchanged in view of the increased risk of
re-entrainment of debris in the event of premature rapping.

     The work of Sproull (10) illustrated the advantages of allowing the
dust  'cake1 to build up on the receiving electrode surface.  His
experiment, although performed on a rigid "engine valve head"  type of
surface, rather far removed from a plate surface, showed 90% dislodgement
of fly ash with a deposition of 2.17 kg m~^ on the 6 x 10~3 m~2  collection
area,  with an acceleration of 30'g1.  With a deposition of only  3 x 10~1 kg
m~2 the removal was only 40%.  This ties up with the prediction  made by
Lowe and Lucas (11) who calculated that acceleration of the order of
105  'g1 would be required to get down to clean metal, free of  the  "basic
layer".
                       ELECTRICAL DUST DISLODGEMENT

     At the stage referred to above, and illustrated in Figure  7,  at
which the precipitator performance starts to deteriorate due  to excessive
dust build up, other mechanisms of dust dislodgement come  into  play.
Firstly, the mass of the dust can lead to local fracture of the cake.
This self cleaning mechanism leaves a "basic layer" which  can itself be
quite thick.
                                     390

-------
     The second mechanism which can dislodge the dust is arcing, i.e.
electrical  flash-over from the discharge electrode to the collecting
electrode.   We have carried out experiments to quantify the magnitude of
the acceleration induced in the collecting electrode when flash-over takes
place.   We  have found it to be a significant quantity when compared to
the magnitude of the conventional mechanical rapping blow.

     In Figure 8 are shown the shock impulse traces obtained using an
Endevco "SMART" system.   The accelerations were measured on a bottom row
centre  plate of a 13.72  m x 4.57 m collecting electrode of the standard
Lodge-Cottrell catch space type.  Values of acceleration, and general
form of the impulse for  the mechanical and electrical rapping can be seen
to be of a  similar order.

     The level of acceleration induced in the sheet by the flash-over is
adequate to effect dust  dislodgement, suggesting that provided the plant
is run  with a high degree of arcing it would keep itself clean without
mechanical  rapping.  This would be inadvisable, since frequent arcing
results in  decreased time-averaged field voltage, and hence reduced dust
deposition  efficiency.  In addition, there is the possibility that a
sudden  surge of current  to, and through, the dust layer may disrupt the
'cake'  and  produce excessive amounts of debris which are readily
re-entrained.
                             DUST SEGREGATION

     When each collecting electrode is rapped, most of the dust should fall
as agglomerates.   Some will be re-entrained as ultra-fine debris, and the
remainder will be of  an intermediate size which will settle only slowly.
In all but the final  field of the precipitator this dust will be recharged
and reprecipitated.   Hence, as we proceed down the length of the
precipitator  the  lower halves of the collecting electrodes will be covered
with progressively more and more reprecipitated intermediate size debris.

     The agglomerate  dislodged from an upstream plate has porosity, and
thus when redeposited produces a layer of higher porosity still.  The
layer has consequently a reduced cohesive strength.  This renders it more
susceptible to break-up on being dislodged from the electrode surface.
For this reason,  it is best for the rapping blow to be applied at the top
of the collecting electrode producing the higher accelerations where they
are needed to dislodge the thinner, more compact dust layers, and the
lower accelerations in the areas where the thicker layers with lower
cohesive strength are located i
                                    391

-------
                        DISCHARGE ELECTRODE RAPPING

     An electrostatic precipitator with the best design of collecting
electrode rapping, perfectly tuned to the operating conditions of the
boiler, would fail to perform well quite rapidly if equal attention were
not paid to the discharge electrode rapping.

     Accumulations of dust on the discharge electrode wires cause
suppression of the corona and hence prevent proper charging of the dust
particles in the gas stream.  Thus, effective rapping blows must be
transmitted to the discharge elements.  At the same time the elements must
not sway or bow, for any such movement adversely affects the plant alignment
and results in reduction in field voltage due to premature flash-over.
In addition, electrical arcing is detrimental to the integrity of long,
small diameter electrode wires since each flash-over results in local spark
machining.  For these reasons, the U.S. Electric Utility Industry moved
away from the use of weighted wires to the rigid frame design of discharge
electrode.

     Figure 9 shows the basic elements of a rigid design, and how the
electrodes are incorporated into the precipitator in a rigid interconnected
framework.  This gives the best possible plant alignment, even with field
heights upto 15.24 m.  Maximum unsupported wire length is short, ensuring
excellent transmission of the rapping blow to the wire elements and
minimised risk of fatigue failure.
                              FATIGUE FAILURE

     It has been observed on competitors plant that the use of over-weight
hammers to produce high 'g1 rapping has on more than one occasion led to
fatigue failure of electrode components, leading to unscheduled shut-down
of the power plant to permit rectification.

     Any rapping system must introduce a risk of fatigue failure over years
of continuous operation.  The typical energy input level at the anvil
position of a large collecting electrode is 30 J.  By calculation and good
design one can minimise the risk of fatigue failure, but the only really
reliable method of assessment is an accelerated life test.  For this
reason we carry out an accelerated life test on any development on a full
scale unit in our rapping test bay.  There is provision for hanging pairs
of collecting electrodes upto 15.24 m in height with a field length of
4.57 m.

     High  'g' rapping is sometimes put forward as a solution to the
problems resulting from high resistivity fly ash.  Despite the high
induced accelerations, there remains a "basic layer" of dust which in the
case of high resistivity fly ash is still sufficient to affect the
electrical performance of the plant adversely.  In addition, since the
dust layer retains charge unable to drain to earth due to the insulating
nature of this "basic layer", the cohesive strength of the dust deposited
                                    392

-------
onto the "basic layer" is reduced and high 'g1 rapping can therefore result
in increased production of re-entrainable debris and hence significantly
increased emissions.

     Lodge-Cottrell's approach to rapping systems design ensures that the
risk of fatigue failure is small.  This in turn means that provided regular
maintenance is carried out, the precipitator performance is maintained over
the lifetime of the plant.
                   LONG TERM HIGH EFFICIENCY PERFORMANCE

     A newly constructed plant will always perform well when first on-line.
After a period of 'running-in1 of about 800 hours an initial assessment
of the efficiency can be sensibly carried out.  Even this test may give
misleading data.   The only meaningful acceptance test is that carried out
after several months operation, since any poor design features will have
made themselves apparent by then.

     The performance of the precipitator could deteriorate over this period
for a whole variety of reasons.  One of the reasons would be failure of the
electrode rapping system to prevent long term build up of dust on the
collecting electrodes.  If no such deterioration is observerable over this
period of time it is almost certain that the same will be true after many
years of operation,  provided that regular plant maintenance is carried out.

     By far the most reliable indicator of the adequacy of any manufacturer's
rapping system is the long term high efficiency performance of the plant.
Lodge-Cottrell is able to cite plant operating on a wide range of coal
types throughout  the world, and in particular plant in the U.S.A. in which
low sulphur coal  gives rise to the so called "difficult" dusts.  Figure 10
shows the long term performance data for some typical precipitator
installations.
                                   393

-------
                                CONCLUSIONS

     Our experience in the design of electrostatic precipitator rapping
systems for high efficiency operation has lead us to the following
conclusions:

     1.  Rapping system must maintain adeguate electrical operation.

     2.  Rapping must remove dust whilst minimising re-entrainment.

     3.  Mechanisms must be accessible and easy to maintain.

     4.  Rapping periodicity must allow adeguate  'cake' build up.

     5.  Acceleration is a misleading measure of rapping adequacy.

     6.  High 'g' rapping does not solve high resistivity dust problems.

     7.  High 'g1 rapping can increase rapping re-entrainment.

     8.  High 'g' rapping can induce major fatigue problems.

     The Lodge-Cottrell precipitator has been scientifically designed on
the basis of the data presented.  The rapping systems for the collecting
and discharge electrodes have been proven in the field over many years
to be effective and reliable and to provide long term high efficiency
performance of the dust collecting plant.


                             ACKNOWLEDGEMENTS

     The authors wish to thank their colleagues for their help and
assistance in the preparation of this paper, and the Management
of Lodge-Cottrell Limited and Dresser Industries for granting permission
for this paper to be published.

     The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the view of the Agency and no official endorsement
should be inferred.
                                    394

-------
                                REFERENCES

1 .    Zarfoss,  J.R.
     New precipitator technology for particulate control.
     Third Symposium on the Transfer of Particulate Technology, March 1981,
2.    Madden, M.  and Wrigley, W.
     The relationship between dust accumulation on precipitator collector
     electrodes  and hammer rapping forces.
     Central Electricity Generating Board Report EPP.113,  May 1968.
3.    Juricic,  D.  and Herrmann, G.
     Modelling and  simulation of dust dislodgement on collecting plates in
     electrostatic  precipitators.
     Modelling and  Simulation 9  (Published:  Instrument Society of America)
     Proceedings  of Ninth Annual Pittsburgh  Conference (1) 161-166,
     April 1978.
4.    Chang,  N.,  Billington D.P.  and Nagy, D.A.
     Effect  of accelerometer mass on the flexural vibration of plates.
     International  Journal of Solids and Structures 14 (10) 851-860 (1978).
5.    Hardie, J.  and Kelsell P.H.
     Precipitator manufacturers  methods of measuring acceleration levels
     induced by  rapping precipitator plates  and wires.
     Central Electricity Generating Board Report EPP.93, December 1967.
6.    Heiber; M.
     Plane plate  rapping experiment.
     EPRI Plate  rapping and reliability study.
     Heiber  Engineering, Wachung N.J.  unpublished report   June 1977.
7.    Ruckelshausen  K.
     Uber entwicklungsstand der  elektrostatischen staubabschieder unter
     besonderer berucksichtigung der abreinigungsprobleme.
     Doctorate thesis,  Technischen Hochschule, Stuttgart,  No. DK-68511
     Geise-Druck  Kg.  Offenbach/M 1957.
8.    Darby,  K.
     Criteria  for designing electrostatic precipitators.
     Second  Symposium on the Transfer of Particulate Technology, July 1979.

9.    Bradburn, K.M.  and Darby, K.
     Electrostatic  precipitator  energization and control systems.
     Fourth  Symposium on the Transfer and Utilization of Particulate
     Technology,  October 1982.

10.   Sproull,  W.T.
     Fundamentals of electrode rapping.
     Journal of the Air Pollution Control Association 15(2) 50-55, 1965.

11.   Lowe, H.J. and Lucas,  D.H.
     The  physics  of  electrostatic precipitation.
     British Journal of Applied  Physics Supplement No.2 S40-S47, 1953.
                                    395

-------
                   DROP ROD
                                                     TUMBLE HAMMER
                                                                                     VIBRATOR
CO
IO
                                                                                                   TOP
                                                                                                   RAPPED
                                                                                                   BOTTOM
                                                                                                   RAPPED
                                                                                                   MULTI-LEVEL
                                                                                                   RAPPED
                                                      FIGURE 1
                                      ALTERNATIVE RAPPING METHODS AND POSITIONS

                                                                                                                   3   3
                                                                                                                                                 X  SINGLE MEASUREMENT

                                                                                                                                                 O  MEAN OF 5 MEASUREMENTS
    10           20           30
        ACCELERATION  ('g')

              FIGURE 2

       RESIDUAL DUST THICKNESS
          AS A FUNCTION OF
     MEASURED ACCELERATION LEVEL
(DATA SOURCE  C.E.G.B. GREAT BRITAIN)
                                                                                                                                                                           40

-------
90.
80.
70.
60.
50.
40.
30.
20.
10.
 0.1
                          1
                                                  10
                                                    -3
                                                      kg)
 MOUNTED ACCELEROMETER MASS   (10

               FIGURE 3

MEASURED ACCELERATION AS A FUNCTION OF
      MOUNTED ACCELEROMETER MASS
                                                                         100
ft


262
202
156
123
100
126

279
120
113
98
74
76

285
154
91
83
62
68

279
120
113
98
74
76

262
202
156
123
100
126
                                                                                                                                      DATA IN
                                                                                                                                      MULTIPLES
                                                                                                                                      OP  'g'
                                                                                                                  FIGURE 4

                                                                                                      RAPPING ACCELERATION SURVEY FOR A
                                                                                                  13.72 x 4.57 m  COLLECTING ELECTRODE PAIR
                                                                                                    USING ENDEVCO TYPE 22  ACCELEROMETERS

-------
                                                                                                              FIELD LENGTH  (m)
                                                                                                              1      2
CO
vo
00
           100
         0.001
                                                      STOKES LAW CURVE
                               EFFECTIVE PARTICLE DENSITY
                                               -3
                          EQUIVALENT SPHERE DIAMTER (10
                                       FIGURE 5

                            TERMINAL FREE-FALLING VELOCITY
                                   AS A FUNCTION OF
                              EQUIVALENT SPHERE DIAMETER
                                                                    10,000
                                                                                                                                                           OUTLET
                                                                                                                                                           FLARE
                                                                                                                                                       GAS VELOCITY
                                                                                                                                                                -1
                                                                                                                               FIGURE 6
                                                                                                       PARTICLE AGGLOMERATE TRAJECTORIES AS A FUNCTION OF THEIR
                                                                                                                  EQUIVALENT SPHERE DIAMETERS C10~6 m)

-------
11.

10,
1 Win.
1
0 100
10 Min. 1 Hour
W W
1000 10,000
1 Day
W
100,0
                                  PERIODICITY OF RAPPING BLOW (a)

                                             FIGURE 7

                             PRECIPITATOR PERFORMANCE AS A FUNCTION OP
                                        RAPPING PERIODICITY
                                    399

-------
     DISCHARGE  ELECTRODE  ASSEMBLY
            FIGURE 10


LONG TERM PLANT PERFORMANCE DATA

           U.S. PLANT
                                                                              CENTRAL IA
                                                                              Pacific Power & Light

                                                                              RUSH ISLAND
                                                                              Union Electric

                                                                             GEORGE NEAL
                                                                              Iowa Public Service

                                                                              IATAN
                                                                              Kansas City Power &
                                                                                         Light
                                                                             LA CYGNE
                                                                              Kansas City Power &
                                                                                         Light
                                                                             •WEST BURTON


                                                                             FIDDLERS FERRY


                                                                             EGGBOROUGH
                                                                                                     RATING
                                                                                                     (M.W.)
   700


   550


   500


   650


   650
   500


   500


   500
COMMISSIONING




    1974


    1976


    1975


    1979


    1977
                                                                                                              U.K. PLANT
                                                                                                              (C.E.G.B.)
                     1965
    1969
                                                                                                                        1965
                                          CURRENT
                                       STACK OPACITY
                                        (PER CENT)
3.0


3.0


8.0


2.5


2.5
                                                                                                                                         CURRENT EMISSION
                                                                                                                                          (10-3 kg Mm"3)
                                            110
                           115
                                                                                                                                                96
WMHEN TNUSS SMCINC AMHl

-------
        ELECTROSTATIC PRECIPITATOR AND FABRIC FILTER OPERATING
                      AND MAINTENANCE EXPERIENCE

          by:   P.  Goldbrunner
               Burns and Roe
               Oradell,  NJ  07649

               W.  Piulle
               Electric  Power Research Institute
               Palo Alto, CA  94303
                               ABSTRACT

     This paper summarizes the "Reliability Assessment of Particulate
Control Systems" (RP 1401) performed by Burns and Roe, Inc.  for the
Electric Power Research Institute.   Its purpose was to determine the
performance and availability of electrostatic precipitators  and fabric
filters used to control flue gas particulate emissions from coal fired
electric utility power plants.  Representative precipitator  and fabric
filter installations,  with associated ash handling systems,  were studied.
Operation and maintenance histories were compared.  The results are
presented in graphic form.
                                   401

-------
                             INTRODUCTION
     The utility industry has installed several hundred electrostatic
precipitators (ESPs) for flue gas particulate removal in the past two
decades to comply with the Clean Air Act of 1970 and its more stringent
amendments of 1978.  More recently, fabric filters have been installed
to achieve required emission levels.  Some of these ESP and fabric
filter installations are retrofits to existing generating units, and
others are.original equipment on new plants.  A large number of early
ESPs were of "hot-side" design, particularly those in which fly ash
resistivity was deemed to be a problem.  Recently, there has been a
trend towards rigid frame ESP designs instead of the weighted wire
designs which served the industry well in the past.  Fabric filter
designs currently operating include reverse air, shake-deflate and pulse
jet cleaning, with reverse air predominant in the utility industry today.

     Many problems in achieving required outlet emissions are reported
with ESPs of both hot-side and cold-side design and of rigid frame or
weighted wire construction.  These problems are often reported in a
general way lacking specific details.  The application of fabric filter
technology to utility boilers for particulate removal has indicated
problems as well.  These problems are generally due to excessive pressure
drop at startup or, as in the case of Clay Boswell and Monticello, are
of specific nature and are discussed later in the text.

     EPRI has just completed a program to better define and identify the
specific details of these reported problems.  Armed with this information,
the utility will be better prepared to avoid or correct problems through
proper engineering, design, operation and maintenance.

                               APPROACH
     Operating and maintenance histories of ESPs and fabric filters at
22 generating stations were surveyed as part of this program.  These
stations had a total of 38 ESPs and 7 fabric filters installed on 13,233
MW of generating capacity.  Selection criteria for ESP examination were
that they be of modern designs, more than 99.5% efficient, no less than
100 MW in size, and having a minimum of four electrical fields.  Fabric
filters were selected primarily on the basis of cleaning method, unit
size, and bag material.  Representation of a variety of manufacturers,
coals and lengths of operation were additional considerations in the
selection process.

     Table 1 lists the installations surveyed.  They include five eastern
hot-side, six eastern cold-side, four western hot-side, and four western
cold-side installations.  The intent was to- reveal differences in operation,
maintenance, and performance of the various applications.  In-service
dates for the ESPs range from 1971 through 1979.  The four generating
stations with fabric filters included in the study used western coals
exclusively and began operation between 1977 and 1980.

                                    402

-------
TABLE 1.   PRECIPITATOR AND FABRIC FILTER INSTALLATIONS
Utility
Potomac
TECO
TECO
Penelec
TVA
TVA
TVA
CP4L
CP4L
AEC
Buckeye
NIPSCO
NIPSCO
KCP4L
TUGCO
Utah P4L
P.S. of
Colorado
Colorado
Ute
Oklahoma
Gas & Elec.
K.C. Board
Pub. Util.
Minn. P4L.
TUGCQ
Col. Springs
Oept. Pub.
Utils.
F. = Eastern Fuel
W = Western Fuel
Plant Unit No.
Potomac
Big Bend
Big Send
Homer City
Kingston
Widows Creek
Bull Run
Sutton
Sutton
Tombigbee
Cardinal
Shahfer
Shahfer
LaCygne
Monticello
runter
Coiranche
Hayden
faskogae
Kaw
Clay Boa-ell
Monties 11 o
R.O. Nixon
1-5
2
3
3
1-9
1-4
1
1
Z
2,3
3
14
15
2
3
142
142
1
445
142
142
142
1
Unit
Capacity Hot Cold
MX Side Side Design Mfr.
5x100 E
420 £
420 E
640. E
4x1 50/ E
5x200
4 x 130 E
950 E
105 E
110 E
2x233 E
618 ' E
520 H
511 W
674 W
750 W
2x420 X
2x350 '(*
192 W
2x555 »
2x44 Fabric Filter
2x69 Fabric Filter
2x575 Fabric Filter
200 Faoric Filter
MM Joy-'lestern
ww Joy-Western
WW R-C
'illt R-C
RF AAF
RF C-E
RF Carborundum
MM R-C
WM Buell
W R-C
WW R-C
RF C-E
WW Sue 11
RF L-C
RF C-E
WW Buell
WW R-C
WW Buell
WW R-C
ICA
Joy Western
W-F
Joy Western
Operating
Date
1978
1974
1976
1978
1976
1977
1978
1975
1975
1979
1977
1977
1979
1977
1977
1977
1971
1974
1977
1979
1978
1977
1980
WH * weighted wire
SF * rigid frame
                        4Q3

-------
     Since performance of the ash removal system directly affects ESP
and fabric filter operation, it was included as part of the assessment.

     To assess the problems with ESPs and fabric filters, meetings were
held on-site at the selected generating stations with plant operating
and maintenance personnel.  The results presented here are primarily
based on information received during these on-site meetings, augmented
by data taken from specifications and performance tests.  The on-site
meetings concentrated on two areas:

     o    Review of maintenance records to obtain data on component failures
          frequency of failures, and repair times.  Discussions with
          plant personnel to learn first-hand their problems with the
          ESP or fabric filter, its ash removal equipment, and auxiliaries,
          in order to supplement frequently incomplete data from user's
          files.

     o    Determination of the unit outages and load reductions resulting
          from ESP or fabric filter ash removal or related system malfunctions.

     In this paper, the results are presented separately for ESPs and
fabric filters.

                      ELECTROSTATIC PRECIPITATORS
     A goal of the study was to find the causes of ESP operating and
maintenance problems affecting reliability and availability.  The on-
site meetings concentrated on obtaining historical data on  failures of
ESP and ash handling components.  Detailed records were not maintained
at all plants, and variations existed between recordkeeping methods at
those which did.  Therefore comparisons of component failures between
ESPs of different designs at different plants are not always definitive.

     The ESPs surveyed were installed on units ranging in size  from 100
to 950 MW.  Specified and actual operating parameters are presented in
Table 2.  Specific collecting areas varied from 275 to 600  ft /1000 ACFM.
Fuel ash contents ranged from 5 to 25%, sulfur contents from" 0.3 to 3%.

     Of the 38 precipitators reviewed, 26 meet or exceed their  design
efficiency, the remaining twelve either meet required emissions or use
remedial methods to enhance ESP performance.  At the Kingston plant,
initial performance problems required use of a higher sulfur coal to
maintain compliance outlet emissions.  (This has since been corrected
through modifications to the rapper systems.)  The Comanche and Hayden
stations use flue gas conditioning on their hot-side units  to maintain
emission requirements.  The Hunter station has experimented with "off-
line" rapping to'minimize ESP performance degradation.

     Failure frequency rates for the ESP components and associated  fly
ash handling system components are summarized graphically in Figure  1.  It
is apparent from this exhibit that the ESP has significantly lower
failure rates than the fly ash handling system.

                                   404

-------
                                                                  TABLE 2  Preclpltator Specified veraua Operating Conditions
Plant
Potoanc Rival

Big Dend

llomer City
Kinuntoo

MldoMS Creak
Bull Rim
l.V. Suitor.
Tontilghee
Cardinal
Schahfer
Schalifar
L. Cygna
Hontlcello
Hunter
Comnche
Harden
Mtnkogee
UnlKa!
1-Z
3-5
2
3
3
1-4
5-9
1-4
1
1
2
2»3
3
14
15
2
3
1.2
112
1
4*5
Heating value
of Coal
(Btu/lb)
Spec'd Actual
11,000-
12,700
11,000-
12,700
9.700-
11, BOO
11,400
11,500
12,208
12,288
11,030
10,700
12.000
12,000
12,009



8.125
6.161
11,500
7,883-
9,108
10,500
8,700
13,000
13,000
11,200
11,800
11,700
11,600
11,600
11,600
11,500
12,400
12,4cm;
11,550
11,900
10,400
10,400
8,100
5,461
11,400
8,500
9,600
8,793
Ash
Content
of Coal
(S)
Spec'd Actual
10-25
10-25
10.4-
17.7
11.0
20
15.5
15.5
30
16.3
19.0
19.0
12.90
25
11.0
11.0
19.7
14.31
5-11
3-8.5
9.54
5.0
8
a
10.9
11.0
18.0
16
16
13.8
15.
12
12
14.7
14.8
10.0
9.8
6.0
24.51
11.5
.16-. 9
13.5
5.54
Sulfur Content
of Coal
 465,000
1,140,000
3,100,000
NA
NA
3,000,000
MA
1,865,700
3,100,000
1,666.000
2.180,000
Inlet
lenperature( *F ]
Spec'd Actual
550-651
550-651
301
291
220-301
325
325
320
270
625
640
775
750
660
750
302
NA
262
828
775
264
N/A
N/A
320
N/A
293
318
342
337
266
622
594
730
641
HA
NA
314
NA
258
NA
706
294
Inlet Duat
Loading
(gr/acf)
Spec'd Actual
4.5
4.5
9.9
4.7
4.5
2.3
2.3
6.8
5.5
2.6
2.8
1.98
2.0
.66-2. t
NA
3.6
6.0
2.65
1.89
2.45
3.84
1.42
1.42
4.59
4.5
2.8
2.0
1,37
3.59
4.2
3.6
2.9
2.1
2.30
HA
NA
1.01
NA
2.47
NA
1.52
.775
SCA
Haling
(ft /tOOO ecfm
Spec'd Actual
413
413
331
322
440
475
450
561
570
275
335
326
492
326
367
603
420
440
296
377
635
430
501
356
398
510
427
445
604
586
377
480
385
619
NA
NA
576
NA
443
240
373
542
Efficiency"
%
Spec'd Actual
99.5
99.5
99.8
99.78
99.54
99.2
99.2
99.6
99.6
99.5
99.40
99.5
99.8
99.5
99.5
99.4
NA
99.6
NA
99.6
99.66
99.81
99.72
99.78
99.73
99.65
99.47
98.95
99.84
99.94
99.79
99.2
99.2
99.84
NA
NA
99.7
NA
99.78
NA
1
98.85
99.24
o
en
                                Mote:   Actual conditions taken  fr«n review of operating and test  data.

-------
     The highest failure category in the ESP is  the "high voltage
system."  Here the primary  causes of failure are support insulators  and
discharge electrodes.  The  second highest failure category is  "TR  set
trip out," the cause of which  is  normally external to the TR set  itself.
Very few actual TR failures were encountered.  The TR trips were caused
by high hopper levels, electrode failures, other ground-related  occurrences,
and malfunction of the protective relay designed into the TR control for
low voltage  protection.
  33
  27-
  Ib
  »2
        Figure 1.  SUMMARY OF FAILURE ANALYSIS PRECIPITATOR vs FLYASH HANDLING SYSTEMS
Kay;

DE Rap
CE Rap
Rap Cont
Pmp/Blwr
Cont
                                                                       LESEHD
Discharge Electrode Rapper
Collecting Electrode Rappers
Rapper Controls
Pump and Blower
Controls
                                                                       ASH REM7
     Pursue  HIVolt TR Trip  CE (tap HE R«p RtpCont  Mlic
                                            Hoppers Pup/Blur  Cont  Piping  Mlic
      Fly ash handling system problems  are  the cause of approximately 75%
 of the total particulate removal system failures.  The largest failure
 category is "hopper  equipment" which includes hopper pluggages, the
 primary contributor  to this category's  high failure rate.   On-site
 discussions indicated that continuous  ash  removal and auxiliary hopper
 equipment such as heaters, vibrators,  and  fluidizing air  (when properly
 installed) reduce hopper pluggage problems.

      During the on-site meetings, load reduction and outage data related to
 the ESP was requested of plant personnel.   Records were not usually
 available, but in some cases oral information was uaed.   Of 19 stations
 visited, only 6 (Potomac 1-5, Big Bend 2&3, Sutton 1&2, and Hayden 1)
 reported availability* reductions.  These  reductions ranged from a high of
 * Availability %
              Boiler Avail. Hours  - ESP Outage Hours
                        Boiler Avail.  Hours
                                            x  100
                                     406

-------
2.6% to less than  1%.   A number of plants  indicated that sometimes ESP
problems are corrected during outages which are charged to other  pieces
of equipment.

     Basic ESP and ash handling system design differences were  examined
to determine their possible reliability variations.  The following
subsections present these comparisons.

RIGID FRAME VS. WEIGHTED WIRE DESIGNS

     Figure 2 compares component failure rates between weighted wire and
rigid frame ESPs surveyed in this study.   The principal difference is in
the category of TR set trips.  The major contributor to higher  TR set
trips in weighted  wire design ESPs was malfunction of the TR  control
unit, not wire failures.  Most of the weighted wire designs investigated
had""spark rate" controllers which contained printed circuit  boards
having high failure rates.

     Figure 3 compares O&M costs of weighted wire and rigid frame designs.
Electrical costs and maintenance labor are equal in both designs.
Maintenance material and operating labor costs are somewhat higher for
the weighted wire  designs, but there is basically little difference in
overall O&M costs  between the two types.

HOT SIDE VS. COLD  SIDE DESIGNS

     Of the stations visited, those with hot-side ESPs burning  western
coal require remedial  measures (usually flue gas conditioning)  in order
to meet emission standards.  The stations  burning eastern coal  meet
compliance requirements without gas conditioning.  Figure 4 compares the
average number of  ESP  component failures for the cold- and hot-side
units.  No conclusions can be reached although hot-side units have a
higher failure rate in four of the six categories.  Table 3 shows the
average O&M costs  for  the hot- and cold-side ESPs.  Total O&M costs are
shown to be almost equal for both designs;  these values may be  biased by
exclusion of gas conditioning from operating costs.  No allowance has
been made for the  thermal losses inherent  with hot-side designs.
               TABLE 3.  AVERAGE O&H COST (HILLS/KUH) FOR COLD-SIDE 4 HOT-SIDE ESP
                          Prec1p1tator         Cold-Side    Hot-Side
                   Maint. material              0.011       0.012
                   Maint. labor                0.029       0.023
                   Subtotal                  0.040       0.035

                          Fly Ash System
                   Maint. material              0.043       0.038
                   Maint. labor                0.040       0.050
                   Subtotal                  0.083       0.088
                          Operating Cost
                   Labor                    0.044       0.054
                   Electricity                0.070       0.072
                   Subtotal                   0.114       0.126
                   Total                     0.237       0.249
                                    407

-------
 C 20 -I
 i
 a
 z
mo
li"-1
 a,
u. uj
 X
    10
    9
         UJ
         5

            i	
                                                               LEGEND


                                                          RIGID FRAME


                                                          WEIGHTED WIRE
MAIN POWER  HIGH VOLTAGE
 SUPPLY      SYSTEM
                              T/R SET
                                TRIP
                                          PLATE
                                         RAPPERS
 WIRE      RAPPER    MISCELLANEOUS
RAPPERS    CONTROLS
         Figure 2.   ESP FAILURE RATES. RIGID FRAME vs WEIGHTED WIRE
    0.10
 =!
 2
 u>
 o
 u
    0.0«
    0.04-
    0.02
    O.OO
                                                             LEGEND

                                                      	RIGID FRAME

                                                      ——	WEIGHTED WIRE

                                                            	 OVERALL AVERAGE
           ESP       ESP      OPERATING  ELECTRICAL
        MAINTENANCE MAINTENANCE    LABOR       COST
         MATERIALS     LABOR

     Figure 3.  ESP OSM COST. RIGID FRAME vs WEIGHTED WIRE
   FLY ASH REMOVAL SYSTEM

        Figure 5 depicts,  in  bar chart format, the various types of fly ash
   system failures for vacuum and pressurized systems  and their frequency
   of occurrence.  Pressurized system hopper equipment failures are high,
   presumably because of the  amount and complexity of  the equipment.  The
   pressurized system also has a higher control system failure rate.  Vacuum
   hydraulic systems have  a high frequency of piping system failures because
   the fly ash and water slurry causes system pluggage and erosion or
   corrosion.  Both pressure  and vacuum producing equipment also have
   relatively high failure frequencies.  Failures are  slightly higher with vacuum
                                       408

-------
     IO
  •fl
  at

  I-
  U.
  O   S
  ci
  UJ   *
  O

                                                                 L£
-------
systems probably because  in the plants studied they were usually of water jet
eductor design, the pumps  for which had a high failure rate.  Blower
failures in both types of  systems were normally due to fly ash seepage
into the lubricating oil.   Vacuum system blowers also failed because of
upstream filter pluggage and ash carryover.

     Figure 6 compares the average  number of fly ash removal system
failures for hot- and cold-side units.   The  ash system components affected
by this ESP location are the hopper equipment and the piping.  For hot-
side units,, the hopper equipment failure rates are lower, presumably
because there is less condensation  in the ash, and the piping system
failure rates are higher because of expansion problems encountered at
higher operating temperatures.   Figure 7 presents the breakdown  of
labor and material costs for O&M of pressurized and hydraulic vacuum
type systems.  It shows very little difference between labor and material
costs for the two systems  but a significant  difference in the costs for
electrical energy.
        4S
        40
    5
   <0
    o
    Ul
    ec
    b.
    O
    UJ
    o
    K.
    Ul
        10
          7
                                LEGEND


                                   COLD UNITS


                                   HOT UNITS
             HOPPER
            EQUIPMENT
  PRESSURE
VACUUM SOURCE)
CONTROL
SYSTEM
            PIPING
                                                         772
                                                          MISCELLANEOUS
             Figure 6.  FLY ASH SYSTEM FAILURES. HOT VS COLD
                                   410

-------
x
9
•x.
^
in

i
   0.10-
o.oe
  0.06
  0.04-
  0.02
  0.00
       Ul
       3
       J
       I
       I
                                                       LEGEND
                                                VACUUM
                                                  PRESSURIZED
         FLYASH MAINTENANCE
            MATERIALS
                           FLYASH MAINTENANCE
                               LABOR
 FLYASH SYSTEM
ELECTRICAL COSTS
            Figure 7.  o a M COST OF FLY ASH REMOVAL SYSTEM
 OPERATION AND MAINTENANCE  COSTS

     The O&M costs for the ESP installations studied are summarized in
 Figure 8.  The largest portion of  particulate removal system O&M costs,
 ESP electrical cost, is  estimated  at 70% of the installed TR set capacity.
 (Coal type, unit load, and other variables may cause actual power requirements
 to differ.)  Fly ash system maintenance was considerably more costly
 than ESP maintenance.  ESP maintenance is labor-intensive, while fly ash
 removal system maintenance is fairly evenly split between labor and
 materials.

                            FABRIC FILTERS
     Relatively few utility boilers  currently use fabric filter systems
when compared with the number using  ESPs;  however, this situation is
rapidly changing.  In August  1982 there were 17,500 MW of fabric filter
systems in operation, on order,  or under construction.   To establish a
data base from which future design and  operating improvements can be
identified and from which changes to existing installations can be
recommended, the operating and maintenance histories of four utility
fabric filter installations were investigated.

     On-site meetings were held  at the  plants to obtain first-hand
operating and maintenance information.   The plants were selected to be
representative of typical design and operating conditions covering as
wide a variety of parameters as  possible.   The four plants included
fabric filter installations by three different manufacturers and cover
extremes of climatic conditions.   Table 4 presents the general design
features of the four fabric filter installations and their ash handling
systems.
                                   411

-------
      Figure 8.  PARTICULATE REMOVAL SYSTEM OPERATING AND MAINTENANCE COSTS
          A

          V

          E

          R H

          A

          E
            K
           C U
            H  .e*
           0 II
                                                    LEGEND
                    EBP Halnt.
                             Fl-rtsh Mtint.
     The Kaw, Clay Boswell,  and  R.D.  Nixon fabric filter installations
are cleaned by the reverse air cleaning method,  Monticello by the shake-
deflate method.  All four plants currently use fiberglass bags with a
10% by weight coating of Teflon  B.

     For the units using reverse-air  cleaning, the normal design gas-to-
cloth ratio is between 2.02  and  2.26  to 1;  for the unit that uses the
shake-deflate cleaning method, it  is  3.1 to 1.  These ratios are within
the typical range for the two respective bag cleaning methods.  Normal
gas-to-cloth ratio is based  upon one  or more fabric filter compartments
out of service for cleaning  or maintenance.   The number out of service
is dependent on the conservatism of the design.   Higher gas-to-cloth
ratios generally lead to greater pressure drops  across the fabric filter,
thus requiring more fan power to move the gas through the system and
reducing bag life.

PERFORMANCE

     All four installations  visited are in compliance with the specified
particulate emissions levels, although the systems at Kaw plant have not
yet achieved their guaranteed collection efficiency.   The collector at
Clay Boswell station met guaranteed efficiency after some mechanical
modifications which stopped  particulate leakage  around the bag thimbles.
No performance tests have been run at Monticello.   The R.D. Nixon station
has an outlet dust concentration significantly lower than 0.03 Ib/MBTU
and a clear stack (opacity less  than  5%).
                                    412

-------
                                Table 4.  PARTICULATE REMOVAL SYSTEM DESIGN  DATA
Station Name
Units

Manufacturer
Coal Analysis
Type
Moisture Content (% by wt. )
Ash Content (X by wt.)
Sulfur Content (% by wt. )
Heating Value (Btu/lb)
Normal Gas-to-Cloth Ratio
Gas Flow to Baghouse (ACFM)
Flue Gas Temperature (°F)
Inlet Oust Loading (qr/acf )
Outlet Loading (gr/acf)
Outlet Loading (lb/1068tu)
Removal Efficiency
F to F Pressure Drop (in. water)
Bag Size (L x D)
Bag Material
Bag Finish
Antlcollapse Ring/Bag
Inlet Thimble Length
Bag Cleaning Method
Number of Compartments
Bags Per Compartment
Gross Cloth Area (ft2)
Kaw
1 and 2
2 x 44 MW
Fabric
ICA

Bituminous
6-12
15
5
11,000
2.02:1
200,000
350
-
0.005
-
99.86
6
36' x 8"
Fiberglass
10% Teflon B
3
8"
Reverse air
9
248
111,600
Clay Boswell
1 and 2
2 x 109 MW
Filter Design Data
Western
Preci pi tator

Sub-bituminous
25
10
1.0
8,500
2.26:1
348,000
240-390
1.0-3.5
0.01
-
99.7
6
32' x 12"
Fiberglass
10t Teflon B
7
4"
Reverse air
8
240
176,275
Monti cello*
1 and 2
2 x 595 MW

Wheel abrator
Frye

Texas lignite
26-37
23
0.5
5,500
3.1:1
1.840,000
400
9.0
0.009
-
-
6.5
30' -6" x 11"
Fiberglass
Silicon graphite**
0
12"
Shake-deflate
36
204
671,936
R.O. Nixon
1
200 MW

Western
Preci pi tator

Western Sub-bit.
15
6
0.4
10,500
2.03:1
1,071,000
295
5.25
0.005
0.022
99.86
5
31 '-9" x 12"
Fiberglass
10% Teflon B
7
12"
Reverse air
36
156
527,825
Ash Handling System Design Data
Manufacturer
Type
Hoppers per Baghouse
Level Detector
Warm A1r Flu1d1zer
Hopper Heater
Insulation
Vibrators
A-S-H
Pneumatic vacuum
18
Capacitance
No
No
3-in. Fiberglass
Yes
United
Pneumatic vacuum
8
Nuclear
Yes
Yes
3-in. Fiberglass
No
United
Pneumatic pressurized
12
Capacitance
No
No
2-1n. Fiberglass
Yes
A-S-H
Pneumatic vacuum
18
Capaci tance
Yes
Yes
4-in. Fiberglass
No
Fabric filters  Installed  in  parallel with ESPs.
Being replaced  by 10% Teflon-  B.
                                                   413

-------
             Table 5.  OPERATING AND MAINTENANCE DATA

                        Kaw      Clay Boswell
FABRIC FILTER STUDY

  Monti cello
R.D. Nixon
Pressure Drop History
Design W.G.
Operating W.G.
Bag Failure Data
Bag Failures
Points of Failure
Location in Baghouse
Replacement Bag
Material
Component Failure Data
Inlet Valves /Opera tors
6
10-12
Note 1
Lower portion
Random
10% Teflon B/
fiberglass
None
Outlet Valves /Operators None
Reverse Air Fans
Flue Gas Dist. Problems
Cleaning Effectiveness
Baghouse Service Hours
Boiler Service Hours
Baghouse Restricted
Load Hours
Baghouse Outages
Motor bearing
None Reported
Some problems
during low
loads
N/A
N/A
N/A
None reported
6
6
100 (1980)
300 (1979)
Lower portion
Random
10% Teflon B/
fiberglass
Sticky damper
limit switches
Viton boot
failures
Motor bearing'
None Reported
Some problems
in winter and
at low loads
7931 (1980)
8094 (1979)
7931 (1980)
8094 (1979)
230 (1980)
200 (1979)
None reported
6.5
10-12
Note 1
Random
Random
10% Teflon B/
fiberglass
None
None
Motor bearing
Extensive
Many
problems
N/A
N/A
N/A
None reported
5
4
9 (1980)
Thimble/lower half
Random
10% Teflon B/
fiberglass
None
Seal replacement
None
None
None
6650 (1980)
6650 (1980)
None
None reported
 NA - Not Available
 Note 1:  Definitive data not available; on-site reports indicate major problems.

OPERATION AND MAINTENANCE

     Operating  and maintenance data for the four fabric filter installations
are presented in Table 5.  Aside  from pressure drop  and bag life problems,
there were reports of minor component failures, including failure  of
ductwork, valves, valve operators,  and reverse air fan motor bearings.
Some failures of ash handling equipment were also reported (not shown in
table); however, ash removal system problems have not been great on
fabric filters.   Perhaps this is  because the flue gas entrance to  the
bags is through the hoppers, which  keeps them naturally warm and fluidized.
                                     414

-------
PRESSURE  DROP

     The  reported flange-to-flange pressure drops range from 4 in. to 12
in. of water.  The pressure drop of the system at the R.D. Nixon plant
is 3.5 in.  of water and within the design specification.  The pressure
drop at the Clay Boswell plant is normally within the specification (6
in. of water) but condensation during extremely cold weather has occasionally
caused a  hard-to-remove cake to form on the bags.  This results in
increased pressure drop, at times leading to a reduction in load due to
insufficient fan capacity.  High differential pressures, up to 12 in. of
water, have been experienced at Kaw and Monticello, the reasons for
which have not been completely determined.  At Kaw, load is occasionally
reduced.   At Monticello, flue gas flow rate through the fabric filter
system has been reduced to minimize pressure drop; to compensate, the
flue gas  flow rate through the parallel ESPs has been increased.  Monticello
also reports a unique problem of ash particles seeping through the bag
cloth; as yet, this problem is unresolved.

BAG LIFE

     The  R.D. Nixon plant had been on line over 1-1/2 years with only a
few bag failures, whereas Monticello reported a bag life of 6 months to
a year, and of this writing, even two years.  The Kaw plant has had some
bag life  problems.  It operates at higher pressure differentials, is
used for  cyclic operation, and uses a high sulfur coal, all of which may
affect bag life adversely.  Clay Boswell1s records show poor bag life
during the first year of operation; meetings on-site revealed that
considerably fewer bag failures now occur.

GENERAL

     It is recognized that this is a sampling of a very limited number
of installations.  It is not representative of the overall performance
of fabric filters as reported in other more broad based EPRI reports
such as RP 1401 "Operating History and Current Status of Fabric Filters
in the Utility Industry" dated July 1981 and currently being updated.

          The work described in this paper was not funded by the  LS.
          Environmental Protection Agency and therefore the contents do
          not necessarily reflect the views of the Agency and no official
          endorsement should be inferred.
                                  415

-------
          ECONOMICAL FLY ASH COLLECTION BY FLUE GAS CONDITIONING

                  by:   E. L. Coe,  Jr.
                  Wahlco, Inc.
                  Santa Ana, CA 92704
                                 ABSTRACT

      This  paper  discusses  the application of sulfur trioxide  flue  gas
conditioning  to tne collection of high resistivity fly ash  in  coal-fired
power plants.    A large portion of the world's coals produce ash falling in
this  category.    Performance figures for a 600-Mw U.S.  plant having  this
type  of  equipment in operation are given.   This plant  had  conditioning
installed  as   a  part of the original plant equipment.   A  comparison  of
emissions to those of baghouses is included, along with data on maintenance
and operating  costs.

      The  work  described  in  this  paper was  not  funded  by  the  U.S.
Environmental    Protection  Agency  and  therefore  the  contents  do   not
necessarily  reflect  the views of the Agency and no  official  endorsement
should be inferred.
                                    416

-------
                               INTRODUCTION
    Sulfur trioxide conditioning  of  flue  gas  to  improve  the  dust collection
performance of  an electrostatic precipitator  was first demonstrated  in  1912
at a smelter in Garfield,   Utah (1).   Its application to fly ash collection
dates  back to  1950 (2).    Increasing  use of  low-sulfur  coal,   reduction of
permissible  emission  levels  and the  availability  of practical   operating
equipment have  combined to cause  the installation of  this type  of  system on
approximately  130 boiler units world-wide in the past decade.   There  has
been  no  instance  of   a  failure  of  these units  to  control  fly  ash
resistivity.

     It  is the purpose of this paper  to  show from  actual experience   that
these  conditioning  systems   can be advantageously applied   to new plant
installations  and  that  their   use should not   be  confined  to  retrofit
situations.

                            SYSTEM DESCRIPTION
     Systems   for  catalytic   generation of  sulfur trioxide  for  flue  gas
conditioning   have been described  in a number of  publications   (3,  4,  5).
This description will therefore be brief.  The basic  process is the same as
that  used  and proven in the manufacture of sulfuric acid.    In  the  most
widely  used   process,   sulfur trioxide (SOo)  is  made from  sulfur  dioxide
(SOj).   S02   can be purchased as  a liquid,   or can be generated by burning
elemental  sulfur.   Except for test work or for  installations  of  limited
life expectancy,  the use of  liquid S02 is uneconomical  because its cost as
a  feedstock  is three to four times as great as that  of  sulfur to produce a
given amount  of SO.,.   This paper  will therefore  consider only  the  sulfur-
burning system, a typical flow diagram for which  is shown in figure 1.

     Filtered  ambient  air  is pressurized by a centrifugal  blower  and
electrically   heated  to combine with sulfur in a refractory-lined  burner
compartment.    The electric heaters are controlled so that their heat  plus
that  from the combustion of the  sulfur results  in the  temperature of  the
gas  stream  at the burner exit being about  800°F, suitable for  catalytic
conversion.    The  gas stream then passes through the converter compartment
which  is  usually constructed as  a part of  the same  enclosure  in which the
burner is located.  Excess air in  the combustion  gas  stream provides oxygen
for  the conversion of S02 to SOo  over a vanadium pentoxide type  catalyst.
An  additional rise in temperature occurs due  to  this oxidation,  and  the
converter  outlet temperature becomes 800 to 1200°F,   depending on the rate
at which sulfur is fed to the system.   The gas conditioning stream contains
air,  S03  and  a small amount of  unconverted  S02«   Water  vapor  is  also
present,  and  this makes it necessary to keep the  temperature above the acid
dew point until the conditioning gas is combined  with the boiler flue gas.
For  this  reason  the   connecting  piping and the   injection probes  are
carefully insulated.

     In  the  U.S. ,  sulfur is conveniently available  in  molten form and  is

                                    417

-------
handled  with steam-heated  piping, vessels and pumps.  A variable-rate
metering pump controlled by a plant signal proportional to coal feed or
boiler load delivers  sulfur to the burner.   Insertion of a ratio station in
the  feed rate signal circuit permits setting  the injection rate as a
percentage of maximum capability,  and it is thereafter maintained at an
essentially constant volume  percentage of the flue  gas as  boiler load
varies.

     Commercial systems  are  instrumented and  interlocked  to  be  self-
policing and  highly automatic  and normally operate  unattended.   As
previously stated the basic system  technology  is taken from acid  plant
practice, but items not useful or essential  in power plant operation,  such
as  air dryers  and multi-stage  converters, have  been omitted  to gain
reliability and simplicity of operation.   Systems are pre-packaged as
integral skid-mounted units appearing generally as shown in figures 2 and 3
for  burner-converter and pump assemblies.

                            PLANT DESCRIPTION
     One  U.S.   power  plant  to date has  incorporated  a  cold-side
 electrostatic precipitator with flue gas conditioning  in its  original
 design.  This is the Pleasant Prairie Station of the Wisconsin Electric
 Power Company, located nearKenosha, Wisconsin (6, 7).  This 616 Mw unit
 (580 Mw net) was brought on line in  June of 1980 and was  extensively tested
 for acceptance and environmental  compliance in the summer of 1981 without
 any unusual maintenance or adjustments preceding  the  tests.  The
 precipitator is a wi re-and-weight type of U.S. design constructed in a
 double-decked configuration of four precipitator units having two chambers
 each with  eight electrical fields per chamber for a total  of 64 electrical
 fields with 64 power supplies.  The  general arrangement of  the precipitator
 is shown by figure 4.   Design parameters  are listed in table  1.

     The coal  supplied  to  this unit is  Western  sub-bituminous  which
 originates in the Powder River Basin  of Wyoming and,  as shown  in table 2,
 is relatively low in sulfur  and ash content and somewhat high in  moisture.
 Analysis of the ash shows high calcium and  magnesium percentages and fairly
 low sodium content.  It would be expected that this ash  would exhibit a
moderately  high  resistivity,  and in-situ  measurements, as shown  in table 3
determined  this to be the case (8).  It was  necessary to make measurements
without conditioning at reduced  boiler load which caused a decrease in gas
temperature.  It is expected  that  the  unconditioned  resistivity would  have
been somewhat  higher  if 300°F could  have  been  maintained  for all
measurements.  Nevertheless the capability  of  SO-j  injection  for adjusting
ash resistivity  to a value best  suited to efficient precipitator operation
is clearly shown.   The quantity of   SO,  in the flue gas  entering the
precipitator was measured simultaneously with resistivity.  At 15 ppm
injection 1.7 ppm SOo was observed in the  gas stream.  At all other points,
the quantities were too low to be experimentally significant;  that is,  they
were less than one ppm.  The  difference between the observed  concentration
and the injected quantity indicates deposition on the fly ash.


                                   418

-------
                               TEST RESULTS
     Acceptance  tests   conducted on June  9,   10  and  11  in  1981   gave   the
results  shown  in Table 4.   These  tests  were conducted by  a  third  party
using   EPA  method  17   and   were  inlet/outlet   tests   of   the    complete
precipitator installation.    It  will be  noted  that the design efficiency of
99.72%  was exceeded by  a wide margin,   and that  the  precipitator migration
velocities  averaged 8.0  cm/s.     Since   it   appeared   that  the   design
requirement  could  have  been met by a  smaller  precipitator,  it was   of
interest to determine the effect of reducing the  effective collection area.
Therefore  the tests dated June  16 were  conducted on  a single chamber (one-
eighth of the precipitator) with two fields nearest the  outlet  deenergized,
leaving  six  in line of gas  flow in service.   As would  be  expected   the
efficiency  decreased,   but it continued to exceed the design  value.    The
migration  velocity  increased  to about  10  cm/s,  presumably  reflecting
increased loss of ultra-fine  particles.

     In  October  of  1981,   an  additional single-chamber test  run by   the
Southern Research Institute under EPRI sponsorship verified  that  the  single
chamber  with  all  fields in service produced a  collection   efficiency   of
99.975%,  corresponding   closely to the  acceptance test  performance of   the
full installation.   It  was thus determined that  performance of the  tested
chamber  fairly represents that  of all four precipitator units.    Inlet  and
outlet cascade impactor measurements on  the chamber produced the  fractional
efficiency data illustrated in figure 5.   Included for  comparison on this
figure  is previously published  data (9,  10)  from EPRI  sponsored tests   of
two  baghouse  installations.    It  is obvious that a properly  sized   and
operated  precipitator   can match or exceed the fine  particle emission   and
"clean stack" capabilities of baghouses.

     All  performance data for the Pleasant Prairie Station   reported here
was  taken  with the flue gas conditioning system  injecting  approximately
8.5 ppmv  SO, into the  flue gas  stream.    Without conditioning  it has been
determined  that  approximately   75%  boiler load  produces   stack opacity
exceeding  20%,  the  applicable limit at  this station.    It was   therefore
unacceptable to the owner and the surrounding  community  to operate for test
purposes  without conditioning.    With conditioning,   the stack opacity   at
full  load is normally  in the range of two to  five percent.   Visually   the
stack appears clear.

                     UNCONDITIONED COMPARISON PLANT
     Published  information  on the Wyodak plant of Pacific Power  &   Light
appears   to  present  a  reasonable comparison of  the  performance  to  be
expected   from  an unconditioned precipitator system (11).    This plant  is
located   near  Gilette,   Wyoming,  in the Powder River Basin and  its  coal
analysis   corresponds closely to that of  the Pleasant  Prairie  fuel.   The
plant  is  rated  330  Mw  and  has  a  rigid-frame  two-unit  precipitator
installation  with  40  bus sections,  40 power supplies  and  five  fields
handling   1.954  million acfm.   It has been tested with four of  the  five

                                   419

-------
fields  in operation which yields an SCA of 600.   Performance tests during
each of the past three years gave outlet emissions levels of 0.0044, 0.0057
and  0.0047  gr/acf  respectively  with an inlet  loading of  1.34  gr/acf.
Corresponding  migration velocities range from 4.6 to 4.8 cm/s.   When  all
five  fields  of  the  precipitator are in  operation,  giving  an  SCA  of
approximately 750, the outlet loading is stated to be 0.003 gr/acf which at
the  same  inlet  concentration indicates an efficiency of   99.78%  and  a
migration  velocity of 4.1 cm/s.   It may be worth noting that at least one
authority  on  the application of rigid-frame cold-side  precipitaters  has
indicated  that migration velocities in the 4 cm/s range should be used on
low-sulfur  coal  fly ash in high-efficiency  situations  (12).   It  might
thereby be inferred that Wyodak. performance is in the expected range.

     Clearly,  one  should  not  carry comparisons between the  Wyodak  and
Pleasant Prairie precipitators to a point of excessive refinement,  because
this is not an exact "apples to apples" situation.  The Wyodak boiler is by
Babcock and Wilcox,  the Pleasant Prairie unit is a Riley Turbo type.    The
inlet  loading to the Pleasant Prairie precipitator is slightly  higher  at
approximately  1.5 gr/acf.   Comparative particle size data is not at  hand.
Nevertheless,  at  two-thirds the size (in SCA) the Pleasant  Prairie   unit
produces  one-fourth to one-eighth the emission concentration of Wyodak and
operates at a migration velocity twice as great in spite of having a higher
efficiency.   The  two plants represent up-to-date practice and  burn   very
similar coal.  General comparison is not unwarranted.

                            ECONOMIC COMPARISON
     For  convenience,  a  600 Mw plant is assumed,  requiring cleaning  of
particulate matter from 2.6 million acfm of flue gas at 280 F-  The fuel is
low-sulfur  coal,  and the required collection efficiency  is  99.9%.   The
precipitator  may be of any design not having excessive sneakage,  rapping,
reentrainment or hopper  losses;  either  rigid  frame  or  wire-and-weight
style.   From the actual plant data in preceding sections of this paper  it
is determined that the precipitator design migration velocities (unmodified
Deutsch  equation) will be 4.0 cm/s for the unconditioned case and 8.0 cm/s
for the conditioned case.  Note that this is conservative in that it favors
the unconditioned case, since the unconditioned plant did not achieve 99.9%
efficiency  at  4.0  cm/s  migration velocity  but  the  conditioned  plant
exceeded it at 8.0 cm/s.

CAPITAL COST

     The cost of precipitators is assumed to vary as the 0.74 power of  the
size  ratio  (13).   For  99.9%  efficiency with a  migration  velocity  of
8.0 cm/s  the  conditioned precipitator has an SCA  of  440  ft2/1000 acfm,
giving a collecting area of 1.144 million ft2.  Assuming the purchaser will
conservatively wish to add a ten percent safety factor,  1.258 million ft2.
will be purchased.   At $20.00 per square foot installed, this precipitator
will cost $25.2 million.   The conditioning equipment adds $2.5 million for
a total of $27.7 million.
                                   420

-------
     By  the  ratio of migration velocities  the unconditioned   precipitator
with the same safety factor will have twice  the collecting  area and  by   the
0.74 power rule its cost will be 1.67 times  as  great,   or $42.0 million.  It
is assumed that ash handling connection costs,  to the  extent they  vary  with
precipitator size,  are covered in the assumed collecting  area  unit cost.

     The  conditioned  installation shows  a  capital  cost  advantage in  this
example of $14.3 million.

ANNUAL COST

     At 20 ppmv injection with an annual capacity factor  of 70%, the example
system  will  use 670 tons of sulfur.   At $200/ton,   the sulfur will  cost
$134,000  per year.  This figure is relatively   conservative,   inasmuch  as
Pleasant Prairie uses less than half of this injection rate.

     In calendar year 1981,   the Pleasant  Prairie plant used 323  man-hours
in  maintaining its conditioning system.   A  portion  of this was applied  to
correction of a minor installation design  problem,   but taking into  account
the possibility of  somewhat increased maintenance requirements as  equipment
ages,  an average 350 man-hours per year is  projected. At  $30.00  per hour,
the  annual  maintenance  labor  expense is  $10,500.00.  Annual   cost   for
maintenance material is estimated at $20,000.00.   This amount is  somewhat
speculative inasmuch as it exceeds experience to date  by  a  wide margin,  but
looks  toward  the  possibility of major overhauls which have not   yet  been
encountered.

     Operator  expense  is estimated at one  man,   one  shift,   40 hours   per
week.   At  $30  per hour an annual cost of  $62,400.00 is  determined.    No
plant having this type of equipment,  to the writer's  knowledge,   assigns a
man to this duty full time,  nor has added  to the normal operating  staff  for
this  purpose.   Nevertheless,  some  time  is  consumed  in operating   the
equipment, and this amount is decidedly conservative.

     Power is required to operate the blowers,  heaters, and heat tracing  of
the conditioning system.   The example plant will represent an average  load
of 165 kw.  Year-round operation at $0.025/kwh  will  cost  $36,135.  Steam for
heat  tracing is included in this figure as  though it  were  obtained  from an
electric boiler.   This power consumption  must  be offset  against the excess
precipitator  power  required by the unconditioned  unit.   The relatively
large  fields  iu  the Wyodak precipitator required  the  use   of   40 power
supplies  to  treat  1.954 million acfm.   By gas volume  proportion,   the
unconditioned  unit of this example would  have  about 52 supplies,  and   the
conditioned unit 26.   Each supply will use about 60  kw and  the reduction of
26  supplies  saves 13.6 Mwh per year or  $340,000.00. The difference  of
$304,000   is   the  net  reduction  in cost of  power  from   use  of   the
conditioned system.

     The reduction  in the number of precipitator power supplies reduces its
load by approximately 1560 kw.   The net reduction in  load  is  thus 1395 kw,
not  including  hopper heating.    At a capacity  charge  of $1000/kw,   the
reduction  in investment cost is  $1.395 million.   Table 5 is a summary of


                                   421

-------
the economic effect of using the conditioned system.

     This  economic comparison is conservative.   No credit has been  taken
for reduced maintenance,  hopper heating power or simplification of the ash
handling system due to reduced precipitater size with conditioning.  Sulfur
costs have been taken at about twice the usage rate experienced in  service
on  Power River coal,  and the tonnage price of sulfur is on the high  side
for present deliveries.  Installed precipitator cost at $20 per square foot
of  collecting  surface is conservative compared to some recent figures  in
the $30 range.   The amount of money allocated to operation and maintenance
is  greater  than  that experienced by any FGC installation  known  to  the
writer.   Refinement  of  these factors would be expected to  increase  the
advantage shown for flue gas conditioned precipitaters.

                                CONCLUSIONS
o    The  viability of sulfur trioxide flue gas conditioning in conjunction
     with  cold-side electrostatic precipitators for collection of fly  ash
     from  low-sulfur coal has been demonstrated in full  scale  continuous
     new plant operation.

o    Economics favor use of the conditioned system,  based on comparison of
     full-size existing operating U.S. plants.

o    Emissions   less  than  those  measured  on  certain   baghouses   are
     obtainable, economically and reliably.

Precipitator   performance   for  moderate  dust  resistivity   is   highly
predictable  and  reliable.   For  high dust  resistivity  predictions  can
be,  and  are  being  made  but the degree  of  uncertainty  is  inherently
greater.   It  is  difficult  to understand why one should  accept   random
fluctuations  at  Mother  Nature's whim when  such  variations  affect  the
performance  of  pollution control equipment,  and when it has been  proven
that they can be eliminated with net savings in money,  space,  energy  and
headaches.
                                   422

-------
                                REFERENCES

1.   White,   H.J.    Industrial Electrostatic  Precipitation.  Addison-Wesley
     Publishing Co., Inc.  Reading,  MA.   1963   p.  12

2.   White,  H.J.  Electrical Resistivity of Fly Ash.   Air Repair 3: 2, 79
     Nov.  1953

3.   Archer, W.E.   Fly Ash Conditioning Update Power Engineering
     81:  6,  76  June, 1977

A.   Brennan,  J.H.   and Reveley,   R.L.   Flue Gas Conditioning With SO., To
     Improve  Precipitater Performance   In;   Proceedings of American  Power
      Conference  Chicago, IL  1977  p. 569.

5.   Brines,  H.G.   and Reveley, R.L.   Flue Gas Conditioning to Reduce Size
     and   Costs of  a New Precipitator at Public Service  Company of Colorado
     - Arapahoe  Station Unit No.   1   In;   Proceedings  of  American  Power
     Conference  Chicago,  IL.  1978  p. 742.

6.   Eskra,  B.J and  McKiney, B.C.   One  Year's Operating  Experience With SO,
     Conditioning  on a Large Coal-Fired Unit's Electrostatic Precipitator.
     Paper  82  - 49.3  presented at  the 1982 Annual   Meeting   of  the  Air
     Pollution Contol Association.  New Orleans,  LA   June 1982

7.   Coe,   E.L.  Jr.   Conditioning  Power Plant Flue  Gas       Pollution
     Engineering April, 1982  p. 36.                              ^~"

8.   Altman,  R.F.;  Gooch, J.P.; Bickelhaupt, R.E. and Dismukes, E.B.  Flue
     Gas  Conditioning Studies In;   Proceedings International Conference on
     Electrostatic    Precipitation.  Air  Pollution   Control   Association.
     Pittsburgh,  PA.  1982  p.  131

9.   Ensor,   D.S.;   Hooper,  R.G.   and  Scheck, R.W.   Determination of the
     Fractional  Efficiency,  Opacity   Characteristics  of  the Fractional
     Efficiency  Aspects of a Fabric  Filter Operating on a Utility  Boiler.
     EPRI-297  Electrical  Power Research Institute, Palo Alto,  CA  1976

10.  Ensor,  D.S.  et  al. Kramer Station Fabric Filter Evaluation.  EPRI-CS-
     1669  Electric  Power  Research  Institute, Palo Alto,  CA 1981

11.  (Unsigned)  Cold-Side Precipitator Handles  Low-SulfurCoal.   Electric
     Light and Power  August 1982   p. 21

12.  Brandt, J.  Exhaust Gas Dust Collection  From Boiler-Furnaces for Fuels
     with  a  Low Sulphur -  And High  Ash  Content In; Proceedings, Second
     Indo-Geman Power Plant Symposium   vl, p22/IV-7   New Delhi, India
     January, 1982

13.  Stenby,  E.W.;   Grimm,  R.P.   and  Mora,   R.R.  Update on Economics for
     Particulate  Control    In;   Proceedings  of the Coal  Technology  '80
     Conference  Houston,  TX  November  1980  p. 559

                                   423

-------
                                                                                                            PHECIPITATOR
to
                                                                            MOLTEN SULFUR
                                                                              AT 175-310F

                                                                           SATURATED STEAM
                                                                            AT 45-50 PSIQ TO
                                                                             STEAM JACKET
                                                                            CONTROLLER
                                                                              TOBOOF
                                                                                                   SO3-t-H2O:H2SO4 •
                                                                                          CONVERTER
00
o
o
O    I
A  FLUE
U  GAS
O
O
                                                                                                        800-1200F
                                                                                                                    PROBES
                                                                                                       OOOO'
                                   Figure  1.   Sulfur-burning  flue gas conditioning system

-------
Figure 2.  Sulfur Burner and Converter
Figure 3.   Sulfur Pumping Unit - Eight
           Pumps for  Four Boilers.
                            425

-------
      TEST PORTS
OMR.OW
              Figure 4. Pleasant Prairie Precipitator Arrangement
            Table 1.  Pleasant Prairie Precipitator Specifications
            Design Gas Volume
            Design Collection Efficiency
            Number of Precipitators
            Number of" Fields in Series
            Collecting -Plate Height
            Treatment Length
            Total Collecting Plate Area
            SCA
            Drift Velocity, w
                                  426
2,600,000 ACFM @ 290°F
99.72
4
8 @ 4 1/2 Ft.
36 Ft.
36 Ft.
1,223,424 Sq. Ft.
468 Sq. Ft./lOOO ACFM
6.38 cm/sec.

-------
  10 T
         0 - KRAMER BAGHOUSE

             (ENSOR.ET AL. 1981)
                                                           T90.00
z
o
a.
h-
UJ
z
ui
0,
   10 "
      B - NUCLA  BAGHOUSE

          (ENSOR, ET AL. 1976)


      A - PLEASANT  PRAIRIE  ESP

          w/ SO3  CONDITIONING
u
o

*
0.
10'+
                                                           99.00  >

                                                                 Z
                                                                 UI

                                                                 o

                                                                 UL
                                                                 u.
                                                                 UJ
                                                        • -99.90
                                                                 UJ
  102
    10
      •I
            PARTICLE   DIAMETER  (MICROMETERS)
                                                             99.99
                                                        10'
          Figure 5.   Particle  size dependency of collection.
                                  427

-------
                    Table 2.  Coal and Ash Analysis

                              Average                    Range

       Moisture,%              29.2                  20.9-32.1

       Ash                     6.12                  5.31-7.26

       Sulfur                  0.47                  0.19-0.58

       HHV,  BTU/Lb.           8213                  8052-8376
                                  ASH
         Si02,%       30.6                   MgO         8.1

         A1203        12.6                   K20        0.24

         Fe203         7.2                   Ti02        1.1

         CaO          22.2                   Na20       0.90
           Table 3.  Effect of SO, Injection on Resistivity


   S03                              Gas                    In-situ
.Injection                       Temperature              Resistivity
  ppmv                              °F                      ohm-cm


    0                               260                   3 x 1010
                                    280                   1 x 1011

    5                               300                   5 x 109

    7                               300                   1 x 109

   15                               300                   7 x 107
                                   428

-------
              Table  4.   Pleasant  Prairie  Precipitator Tests
                         with  S03  Flue  Gas Conditioning


Date               SCA          Efficiency     w                    Outlet
1981    MW   Ft2/1000  ACFM        %         cm/sec    gr/DSCF    Lb/106BTU
6/9
6/10
6/11
6/16*

587
580
583
572
572
506
505
502
354
348
99.97
99.945
99.97
99.93
99.89
8.15
7.55
8.22
10.42
9.94
0.0008
0.0017
0.0012
0.0021
0.0034
0.0016
0.0034
0.0024
0.0042
0.0068
*Tests of  6/16  conducted  with two fields  (1/4)  of  precipitator de-energized.


                 Table  5.   Economic  Effect  of Conditioning
                 :========s==s=:s=s=========:==========s3ss==

                                  Unconditioned                Conditioned
Investment Cost                      $l,000's                    $l,000's

Installed  cost  -  precipitator        42,000                      25,200

             gas conditioner            -                         2,500

Capacity charge                       Base                        (1,395)
        TOTAL                        42,000                       26,305

        Difference                                 15,695

Annual  Cost
Fixed charges @ 20%
Energy
Sulfur
0 & M
Base
Base
-
Base
(3,139)
( 304)
134
93
        TOTAL                       Base                          (3.126)


                                    429

-------
    EXPERIENCES AT DETROIT EDISON COMPANY WITH DECLINING PERFORMANCE OF
              SULFUR TRIOXIDE FLUE GAS  CONDITIONING EQUIPMENT

          by:  L. A. Kasik
               W. A. Rugenstein
               J. L. Gibbs
               Detroit Edison Co.
               Detroit, Michigan  48226
                                 ABSTRACT
     The generation of sulfur trioxide  in  some  of our Company's sulfur-
burning flue gas conditioning systems has  been  found to decrease signifi-
cantly over a period of a few years.  The  deteriorating performance went
undetected because SC>2 to 863 conversion efficiency was measured with
what later proved to be an  inadequate procedure.   A fall-off in 863 output
can seriously affect precipitator  collection  efficiency resulting in opacity
and/or mass emission excesses.  The vanadium  pentoxide catalyst bed was sus-
pected to be at fault because in some systems very little heat was being
generated in the bed.  A new test  procedure,  a  modified EPA Reference Method
8, was used to overcome the inadequacies of  the old procedure.  Using the
new procedure, conversion efficiency on one  particular boiler was found to
have dropped from about 80  percent to about 30  percent after five years of
service.
                                    430

-------
                                BACKGROUND
COMPANY AND PLANTS
     The Detroit Edison Company  serves  the  Southeastern Michigan area with
a system generation capability of  8450 MW.   Pulverized coal firing accounts
for approximately 72 percent of  this  generation capability.  The Company
employs cold-side electrostatic  precipitators  exclusively for particulate
emission control.  A number of different  coal  types are used system-wide,
ranging from low sulfur western  sub-bituminous to medium sulfur eastern
bituminous.
     To comply with the State of Michigan  regulations on sulfur dioxide
emissions, the amount of  sulfur in  the  coal  is  limited.   Of interest in this
paper are those units which control  S02 emissions  by burning a variety of
eastern coals having a sulfur content of 1.0 percent or  less.   This type of
coal can be supplied from upwards of 10 different  mines.  All  of the boilers
were designed to burn coal of much  higher  sulfur content.  The installed
precipitators were, as a  result, incapable of meeting particulate emission
and opacity regulations,  due primarily  to  the highly resistive ash (greater
than approximately 10*1 ohm-cm) produced by  the low sulfur coals.  The
particulate emission problem has been solved by using manufactured sulfur
trioxide to condition the flue gas  and  particles.   A summary of those plants
using 803 flue gas conditioning (FGC) is provided  in Table I.


THE FGC SYSTEM
     Most of the FGC systems used  by  Detroit  Edison are the sulfur burning
type supplied by Wahlco, Inc.  Figure 1  shows in schematic the typical FGC
system.  Molten sulfur is delivered by truck and  stored in a heated tank at
ground level.  Sulfur is transferred  to  an adjacent pump house which delivers
a metered quantity of sulfur, by means of  a reciprocating piston pump with
check valves, to the burner.  The  sulfur burner  is part of a skid-mounted
package  of components.  The major  components  are:   1)  a constant speed fan
to deliver filtered air to the burner; 2)   an air  heater to maintain optimum
thermal  conditions in the burner at low  sulfur delivery rates; 3) the burner,
to oxidize sulfur to sulfur dioxide;  and 4) the  catalyst bed, which is van-
adium pentoxide on a silica based  substrate in pellet  form, to further
oxidize  the sulfur dioxide to sulfur  trioxide.  The air/sulfur trioxide gas
mixture,  of approximately 4-5 percent concentration, is then piped to a grid

                                     431

-------
of nozzles located  in  the  flue gas duct just downstream of  the  air  heater.
All systems were  sized to  provide a maximum of 40 ppm sulfur  trioxide  at
100 percent boiler  load.


     It  is important  to note that considerable heat is generated when  sul-
fur is burned, 3982 Btu/lb sulfur, and when sulfur dioxide  is catalytically
converted to  sulfur trioxide,  1957 Btu/lb sulfur.  The transport air serves
two additional purposes:   1) to cool the gases and 2) furnish sufficient
oxygen for burning  the sulfur and further converting the SC>2  to 803.   The
air, in  conjunction with the air heater, regulates thermal  conditions  in
the burner.   Burner thermal conditions are important and the  temperature
there should  normally  be between 800F and 825F for proper  functioning  of the
catalyst.  The catalyst manufacturer recommends keeping gas  temperatures in-
to the catalyst bed below 850F to avoid impairing catalyst  performance.
                    PROBLEM IDENTIFICATION AND ANALYSIS


     The Marysville Power Plant,  as early as May, 1980, reported  difficulty
in  controlling  stack  opacity to the low levels once achievable  when  the FGC
system was  new.   To stay below the legal opacity limits it was  now necessary
to  lower load on the  unit.  By the summer of 1981, plant  personnel felt suf-
ficient operating experience had  been gained to conclude  that opacity  levels
fluctuated  depending.on the source of the coal.  The plant also reported
that with the more troublesome coals they were unable  to  effect reductions
in  opacity  by manipulating output from the FGC system.


     An investigation was begun to determine the reason behind  the opacity
fluctuations  and to  find a method of control less costly  than reducing load.
Several factors have  a bearing on stack opacity.  In very general terms
these are:   the precipitator, boiler operation, the flue  gas conditioning
system and  the  coal,  or more correctly, the ash.  Study in the  first two
areas was not  considered worthwhile, in the early stages  at  least, because
very  little correlation could be  made between  them and the observation that
opacity changed when  the coal changed.  Boiler operation, especially as  it
relates  to  coal pulverizing however, was not completely forgotten as a
factor.  In addition, plant personnel reported both the precipitator and
boiler equipment were in good working condition, having received  special
attention during the  period that  opacity had been a problem.  The next two
plant areas were to  be the prime  study targets.  The coal, known  to  produce
highly resistive ash, and for the reasons stated, could be cited  as  a .source
of  the problem.  However, flue gas conditioning should be able  to compensate
for the high ash resistivities.  Ash resistivities were calculated for each
coal  source,  using the techniques of Bickelhaupt  (1),  and found to range  from
1.1 x 1011  to  1.5 x  1012 ohm-cm.   The FGC system was  identified as the primary
study area.


     Operating data on several FGC systems and precipitators  were taken in


                                     432

-------
an effort to isolate an  area  for  further  study.   On some units precipitator
current densities were found  to be  low,  indicating high resistivity ash
conditions. Furthermore, by increasing  the  sulfur rate of the FGC system on
these units to slightly  higher  levels,  current  densities and stack opacity
improved slightly.  Following  this,  the  sulfur  burning rate was raised to
the highest level possible without  sounding the  burner over-temperature
alarm (1000F).  This procedure  of overconditipning should produce two very
noticeable effects.  First, a  bluish-white  plume should be seen emanating
from the tip of the stack.  Second,  opacity degrades rapidly, particularly
when the plates are rapped, due to  the  very low resistivity ash in the
precipitator.  Neither of these effects were noticed in our tests.  Finally,
it was noted that on some FGC  systems  there was  very little increase in
gas temperature through  the catalytic  converter.  Recognizing that the re-
action of sulfur dioxide to sulfur  trioxide is  exothermic, and by making
adjustments for heat loss to  the  surroundings,  a temperature increase pro-
portional to the sulfur  rate  is expected.   Based on the above observations
it was concluded that there was sufficient  sulfur as evidenced by high burner
outlet temperatures and  that catalytic  conversion efficiency was likely very
low on some of the systems.  It was  decided that tests would be performed to
determine the efficiency.
                         CATALYST  EFFICIENCY TEST


     The FGC equipment manual  contains  a suggested procedure for testing
for conversion efficiency.  Several  tests  were  performed using the proce-
dure with disappointing results.   Test  efficiencies were consistently in
the range of 93 to 96 percent.   Some (2)  (3)   suggest conversion efficien-
cies for catalyst in new condition to be  approximately 80 percent.  Not
only were our test results high  by comparison but also inconsistent with
our observation of very low heat release  in  some of the converters.


     Consultation with technical experts  from EPA, Southern Research
Institute, and Monsanto Corp.  resulted  in  our using a modified version of
EPA Reference Method 8 to determine  conversion  efficiency of SC>2 to 863
in the FGC system.  A representative sample  of  system outlet gas was taken
through a sample line heated to  at least  600F to prevent premature acid con-
densation.  The EPA Method 8 was modified  such  that two impingers (bubblers)
of 80 percent isopropyl alcohol  (IPA) in water  were used to remove 863,
followed by two impingers of 5 percent  hydrogen peroxide (l^C^) solution
in water to remove SC>2.  Both  gaseous components are converted to sulfuric
acid via the following reactions:


                         In the IPA/water  solution


                              S03  +  H20 —
                                    433

-------
that used in FGC  systems  is  not  known with confidence.  When the FGC equip-
ment was first bought,  the  catalyst was believed to be good for the life of
the equipment unless  it was  damaged in one of three ways:  1) exposure to
temperatures in excess  of 1200F,  2) plugging the bed of pellets with dust
or ash, or 3) contamination  with excessive moisture.  The effects of such
damage are usually  immediate and quite noticeable.  However, it is believed
the catalyst can  be damaged  in a gradual, less noticeable manner.  The
catalyst manufacturer claims their product can function anywhere between
800-1200F, but at temperatures above approximately 850F, catalyst life and
performance level is  significantly diminished.  It is preferable to have the
gases (S02 and air) enter the catalyst at as low a temperature as possible
to sustain catalytic  action.


     In addition, the manufacturer reported that catalysts can develop a
temperature memory.   If a catalyst bed is operated at a higher temperature,
its performance remains approximately the same as it was at the lower temp-
erature.  However,  after  operating for a significant time at high tempera-
ture there is a reduction in conversion efficiency after lowering the oper-
ating temperature back  to the original temperature..  The long-term effect
of such temperature cycling  is not known but shortened life and reduced
performance is suspected, especially when the temperature excursions are
several hundred degrees and  the  temperature cycling is frequent.


     The test results suggest that cycling of the FGC system between "run"
and "standby" modes and over a range of sulfur burning rates may also con-
tribute to poor catalyst  performance.  Trenton Channel Unit 9 is base-loaded
and has one of the  oldest FGC systems.  Its tested conversion efficiency is
also one of the highest.  The other units, while not considered peaking
type, are operated  at several levels of load and undergo more frequent
start-ups and shutdowns.


     Indications  are  that stack opacity control has been substantially im-
proved at our Marysvilie  Plant since the catalyst was replaced.  It has been
reported that higher  unit load at lower average opacity is now possible while
burning the same  coal once  considered troublesome.  Unfortunately, there has
not been sufficient opportunity  to perform any definitive testing to quantify
the extent of improvement.   Further work is planned along these lines.


                            CATALYST REPLACEMENT


     The decision to  replace catalyst material is dependent on several
factors and not  solely  on an arbitrary level of efficiency.  Acceptable
stack opacity (and  low  particulate emissions) is possible with a catalyst
bed having very  low efficiency.   Simply by burning more sulfur, needed
levels of 803 can be  generated to lower ash resistivity to the proper
range.  A limitation  exists, however, because only an amount of  sulfur  can
be burned that does not create excessive temperatures in the FGC system.

                                      434

-------
                       and in the peroxide  solution


                            S02 + H202 -*• H2S04


     Instead of the in-train filter between the  IPA and the H202 imping-
ers,  quartz wool was used at the top of  the IPA  impingers to prevent cross-
over  of acid mist into the H202 solution.


     As a further modification of EPA Method 8,  the respective samples were
titrated with a standardized sodium hydroxide  (NaOH)  solution using a
phenolphthalein indicator to simplify analysis for  field technicians.


     Catalyst bed conversion efficiency  was determined as the ratio of the
amount of NaOH necessary to neutralize the  803 solution to that required
for both the S02 and 803 solutions, times 100 percent.


                               TEST RESULTS
     Test results to date have confirmed  that  several  of our FGC systems
have very low conversion efficiency.  Based  on test  results for systems
using new catalyst, a conversion efficiency  of 80  percent seems reasonable
using fresh catalyst.  Tested efficiencies for systems having catalyst from
3-1/2 to 6 years old range from 30 to 85  percent.  Table II gives the
efficiency test results for the old catalyst and  the new catalyst plus the
catalyst manufacturer's test results on samples of the old catalyst.


     The catalyst manufacturer uses a test procedure slightly different
than the test for efficiency but both are measures of  conversion perfor-
mance and give equivalent results.  Their test results are reported as
"activity".   The test for activity consists  of splitting a stream of  S02
gas of a given concentration and sending  one part  through a sample of the
old catalyst and the other through a fresh batch  of  catalyst of the same
type.  The amount of 803 in each outlet gas  stream is  then measured and
used to form a ratio which relates the "activity"  of the two catalysts.  The
conversion efficiency test consists of extracting  a  gas sample downstream
of the catalyst bed of an operating unit  and determining the relative
amounts of S02 and 803.  "Activity" and "efficiency" are related by a
factor that  derives from the fact that fresh catalyst  is not 100 percent
efficient in converting the S02 to 803.   The equivalency factor between
activity and efficiency should be approximately 0.8  based on the test re-
sults for new catalyst beds.  This general relationship appears to be good
for the limited number of tests performed to date.


     Catalyst bed material has a finite life expectancy but the value for


                                    435

-------
The coal too, may be of  a  type  that  does not require much 863 for
resistivity attenuation.   Flue  gas  temperatures, which have a significant
effect on ash resistivity, may  be  adjustable over a narrow range and by
operating at a slightly  lower  temperature,  less resistivity attenuation via
FGC may be needed.  This technique  can have application on older boilers
that typically operate at  flue  gas  temperatures of 300F and higher.  Unfor-
tunately, these same boilers do not  incorporate sufficient flexibility in
gas temperature adjustability.   However, should the coal change to one with
significantly higher ash resistivity,  the flue gas temperature be high with
no provision for control,  and  the  sulfur burning rate be near maximum level,
then a marginal to poor  catalytic  converter will create opacity problems
and need to be replaced.


     Instructions on catalyst  replacement is not covered in very great
detail in the FGC equipment manual.   For those contemplating catalyst
replacement the following  may  be helpful:
     .In the  standby mode,  purge the system of 802 an<* ^3 gases
      for at  least 30 minutes.
     .Remove  insulation  from the top of the burner/converter box but only
      for that  portion above the converter.
      .Break  the  seal  weld  and  remove the converter lid.
      .Scoop  the V2C>5  pellets  out  of converter compartment and place
      them in  an  appropriate  disposal container.
      .Wire brush  the  compartment and lid brick surfaces of dust and debris,
      then vacuum.
      .Inspect  the  internal  brickwork for damage.  In two units  at Detroit
      Edison,  cracks  and  compartment baffle wall separations  from the main
      side wall  were  found  and had to be repaired.  In one case where
      baffle wall  damage  was severe, reconstruction resulted  in an  improved
      design by  interlocking the baffle wall with the side walls.
      .Screen  new V2C>5  pellets of dust and refill converter  compartment
       following  instructions in the equipment manual.
      .Replace  lid,  seal weld,  replace insulation and purge  with  hot  air  as
      soon  as  possible to prevent moisture contamination of the  pellets.
                                     436

-------
     As a guide to planning, catalyst bed replacement  using the  above
procedures spanned four days, requiring a total  of  approximately 200
man-hours.  In most instances, catalyst replacement  is best done during a
regularly scheduled boiler outage.
     The catalyst pellets are fragile, dust easily,  with  the  dust
irritating to the eyes, skin and respiratory  tract.   Approved protective
equipment including goggles, masks, gloves and  other clothing were used
during the operation.  In addition, dust exposure  testing was performed by
a Company Hygienist that showed levels were well below the permissible
exposure limits in the "Material Safety Data  Sheet"  (4).
                                CONCLUSIONS
     Based on our test results, consultation with  the  catalyst  manufacturer and
observations of FGC unit operation,  the  following  conclusions  can be drawn:


     .Catalyst conversion efficiency can decrease  gradually to  unacceptably
      low levels in as little as 3 to 4 years.  Probable  explanations for
      this are:  1) operating the catalyst bed  at  temperatures  above 850F,
      2) temperature cycling of the  catalyst bed over  several  hundred
      degrees and 3) start-up/shutdown and sulfur  burning rate  cycling of
      the FGC system.


     .Low converter efficiency alone is  not justification for  replacing the
      catalyst.  The decision to replace the catalyst  should be based on
      the ability of the FGC system  to adequately  condition the fly ash.
      However, to deal effectively with  problem situations,  conversion
      efficiency should be measured  periodically.


     .The amount of heat released in the catalytic  converter,  as evidenced
      by the difference between inlet and outlet temperatures,  is a good
      indication of conversion efficiency.  Each FGC  system has slightly
      different physical characteristics and as a  result, displays slightly
      different temperature profiles through the catalyst bed.   As a result
      no general guidelines can be provided relating  temperature increases
      to efficiency.


     .The basic EPA Reference Method 8 using the modifications  mentioned
      earlier is a satisfactory procedure for determining conversion
      efficiency.


     .The ability of a particular FGC systm to  overcondition the flue gas/ash
                                    437

-------
      i.e., produce a bluish-white  plume  at  the  tip of the stack and/or
      spikes on the opacity monitor,  is an  indication that 1) adequate
      sulfur is being pumped and 2) converter  efficiency is sufficiently
      high for the particular coal  being  burned.   If overconditioning is
      demonstrated, yet opacity is  unacceptably  high, then other problem
      areas should be explored.  For  example,  there may be poor 803 gas
      distribution as a result of plugged injection ports.
                            ADDITIONAL  STUDIES
Despite feeling that we have dealt effectively with  the problem of poor
catalyst performance, the work  is not  finished.   Considerably more can be
learned about the overall gas conditioning  process.   Additional activities
are planned that will result in a better understanding of how F6C systems
perform and how performance affects opacity.   Following are steps we feel
should be taken to achieve the  above objective:

     .Obtain more information on the theory and  practical limitations of
      vanadium catalysts.
     .Prepare one or more FGC systems  for  short-term testing by,


          1)   calibrating flow and  temperature measuring apparatus,
          2)   performing current efficiency  tests  at  several levels of
               sulfur burning, and
          3)   cleaning SO^ injection nozzles.
     .Measure both equilibrium 803 levels  and  in  situ ash resistivity
      for one or more
          1)   coal types,


          2)   flue gas temperatures,  and


          3)   803 generation rates  on the  FGC  system.


     .Evaluate FGC system performance  over  the  long-term by




                                     438

-------
          1)    performing periodic conversion efficiency  tests,
          2)    obtaining historical information on  temperature  and  duty
               cycle,
          3)    recording opacity data on systems before  and  after  catalyst
               replacement to quantify improvement  increment.
     Information obtained from the outlined steps will help  formulate
better operating practices and problem solution guidelines resulting  in
more reliable  plant operation.
     The  work described in this paper was not funded by  the U.  S.  Environ-
mental Protection Agency and therefore the contents do not necessarily
reflect the views of the Agency and no official endorsement should be
inferred.
                                     439

-------
          Table I.  SULFUR TRIOXIDE FLUE GAS CONDITIONING SYSTEMS
                             AT DETROIT EDISON

Plant
Name
Conners Creek
Conner s Creek
Conners Creek
Conners Creek
Marys vi lie
Marysville
Marysville
Marysville
Trenton Channel
Trenton Channel
Trenton Channel
Trenton Channel
Trenton Channel
Harbor Beach
Monroe
Monroe
Fennsalt
Port Huron
Unit
No.
15
16
17
18
9
10
11
12
9
16
17
18
19
1
1
2
-
—
Unit
Size.MW
62
62
62
62
37
37
37
37
500
44
44
44
44
103
750
750
14
5
System
Type*
SB
SB
SB
SB
SB
SB
SB
SB
SB
SB
SB
SB
SB
SB
SB
SB
SA
LSD
Start-up
Date
May,
May,
May,
May,
Dec.
Jan.
Jan.
Nov.
June
Oct,
Oct,
Oct,
Oct,
Sept




1978
1978
1978
1978
1976
1977
1977
1976
, 1976
1978
1978
1978
1978
, 1979
**
**
-
_
*SB - Sulfur Burner, SA - Sulfuric Acid, LSD - Liquid Sulfur Dioxide.


**Systems were started and subjected to brief acceptance/trial  operation in
November 1980; have not been operated since.
                                   440

-------
FLUE - yr-K
GAS f — *
FLOW ' N!->
AIR
HEATER
AIR
i f . \ AIR
BLOWER HEATER
^__MniTCKI Gill CI ID
PRECIPITATOR
u / \
If 1
\ /
vv
S03,S02
AIR T(
INJECTK
i r
S02
•I-
s+o^so, ^B
S03
SULFUR f
BURNER



AND 	 ^
)
DN
:s
i

EFFICIENCY
TEST
SAMPLING
LOCATION
  AT 275-3IOF
          CATALYTIC
          CONVERTER
     PUMP
     HOUSE
 SULFUR
STORAGE
oo
   TRUCK
  DELIVERY
FIGURE 1  SULFUR  BURNING  FLUE GAS
          CONDITIONING SYSTEM
                 441

-------
             Table II.  SUMMARY OF FGC CATALYST TEST RESULTS*

Plant
Name
Conners Creek
Conners Creek
Conners Creek
Conners Creek
Marysville
Marysville
Marysville
Marysville
Trenton Channel
Trenton Channel
Trenton Channel
Trenton Channel
Trenton Channel
Conversion efficiency
Unit
No.
15
16
17
18
9
10
11
12
9
16
17
18
19
as
Conversion
Efficiency,
Old Catalyst,
%
48
49
69
73
74
Not Tested
Not Tested
30
83
70
56
85
68
SOi x 100
"Activity",
Old Catalyst,
%
60
66
**
**
28
68
38
45
**
**
84
**
**

Conversion
Efficiency,
New Catalyst,
%
84
82
**
**
Not Tested
81
75
76
**
**
Not Tested
**
**

                        S02 + S03
*Tests were performed over the period September, 1981 to October, 1982.
**Catalyst bed not replaced.
                                    442

-------
                           REFERENCES
1.    Bickelhaupt,  R.  E.,  A Technique for Predicting Fly Ash Resist-
     ivity,  EPA-600/7-79-204.   U.  S. Environmental Protection Agency,
     Research Triangle Park,  N. C.  1979, 115 pages.


2.    Klipstein,  D. H., Atkins, R.,  Sulfur-Burning Recommended for
     Precipitator  Gas Conditioning.   Electric Light and Power, April,
     1976.
3.   Proposal to Detroit Edison Company,  803 Conditioning System
     for Pennsalt Power Plant, Wyandotte, Michigan, Joy Manufacturing
     Co. July 12, 1971.
4.   Material Safety Data Sheet, equivalent to OSHA Form 20 and
     supplied by Monsanto Company, 800 N. Lindbergh, St. Louis,
     Missouri.
                               443

-------
           ESP CONDITIONING WITH AMMONIA AT THE MONROE POWER PLANT
                          OF DETROIT EDISON COMPANY

                by:  E. B. Dismukes
                     J. P. Gooch
                     G. H. Marchant, Jr.
                     Southern Research Institute
                     Birmingham, Alabama  35255
                                  ABSTRACT

      An investigation of ammonia conditioning was conducted recently at
Monroe Unit 1, a 720-MW unit burning coal with about 2% sulfur and collecting
fly ash in cold-side ESPs around 270°F (132°C) .  Experience has shown that
8 ppm of ammonia substantially lowers stack opacity.  Tests confirmed the
effect on opacity and showed that ammonia increased the ESP efficiency, from
99.72 to 99.83% on an overall mass basis and from 99.20 to 99.68% for
particle sizes below 6 ym.  Particle-size data and ESP electrical data indi-
cate that conditioning mechanisms include space-charge enhancement of the
electric field by means of a fume of ammonium sulfate or bisulfate.  The
resistivity of the ash without conditioning (<1 x 1010 ohm-cm) was not
altered by ammonia addition.  Data on outlet particle sizes show evidence of
reduced rapping reentrainment through increased ash cohesiveness.
                                     444

-------
           ESP  CONDITIONING WITH AMMONIA AT THE MONROE POWER PLANT
                          OF  DETROIT EDISON COMPANY
     This  paper  discusses  one of  the most recent field studies of flue gas
conditioning  that have  been performed by Southern Research Institute under
sponsorship of the  Electric Power Research Institute (RP724-2).   This speci-
fic study  was cosponsored  by EPRI and the utility company concerned, Detroit
Edison Company.

     Ammonia  is  used at Monroe at an added concentration of 8 to 9 ppm to
improve the efficiency  of  cold-side ESP collection of fly ash from a coal of
moderate sulfur  content (about 2%)  and low alkalinity (less than 10% total of
the oxides of alkali and alkaline earth metals in the fly ash).   Ammonia is
thus used  at  Monroe under  quite different circumstances from those in which
the classical conditioning agent, sulfur trioxide, is applicable.  Details
about the  design of the ESPs at Unit 1 are given in Table 1;  the scope of the
test plan  executed  at one  of the  four ESPs (in November 1981)  is described in
Table 2.

     The specific mechanisms of ammonia conditioning anticipated at Monroe,
on the basis  of  prior research at other plants (1), were (a)  space-charge
enhancement of the  electric field in the interelectrode space of the ESP and
(b) suppression  of  rapping reentrainment from the ESP plates.   Operation of
both of these mechanisms was confirmed; neither a resistivity effect on the
ash nor an agglomeration of ash particles prior to precipitation was indi-
cated by the  experimental  data.

     A major  objective  of  the research was to quantify the degree of improve-
ment in ESP performance achieved  with ammonia addition.  Data on mass effi-
ciencies and  stack  opacities are  presented in Table 3.  The level of
performance was  unexpectedly high even without ammonia addition, but evidence
of a significant improvement was  nevertheless obtained with ammonia addition.
The improvement  in  collection efficiency as a function of particle size
observed with impactors is shown  in Figure 1; the improvement was most
pronounced for particle diameters smaller than 1 urn.  During the time the ESP
performance data were being acquired, the discovery was made that some of the
outlet gas was probably bypassing the available sampling ports.   Whether this
was actually  so  and what its impact was on the reliability of the data cannot
be ascertained until the next outage occurs.

     The second  major objective of the investigation was to identify the
mechanisms by which ammonia is beneficial.  Prior work (1) has shown that
ammonia conditioning in a  cold-side precipitator involves interaction with
normally occurring  sulfuric acid  to produce a mist of ammonium bisulfate
        or, if the  mole ratio is  appropriate, the normal sulfate

                                     445

-------
                              NH3(gas)                 fume)
                   (vapor)
                          \.-L i\ni. f r,nn\
                                                         fume)
X
              *0ccurs as a liquid above 300°F (150°C).
     Data from Monroe bearing testimony to the occurrence of one or both of
these chemical reactions are given in Tables 4 and 5.  The sampling strategy
for flue gases was such that one technique was designed to determine free,
uncombined ammonia and sulfuric acid vapors whereas another technique was
designed to determine the total of the free vapors and the combined forms as
particulate substances.  The data in Table 4 reveal that when ammonia was
added little of either substance occurred in the free forms whereas roughly
the expected amount of ammonia and sulfuric acid occurred in the combined
forms.  The analysis of fly ash from the ESP hoppers (Table 5) indicates a
recovery of about 50% of the added ammonia in the ash.  The data in this
table for inlet and outlet hoppers were.combined proportionally to obtain an
estimate of 4 ppm accounted for out of the total of 8 to 9 ppm added; the cal-
culated recovery in hoppers would have been higher, as it was in ash
collected in the mass train, if allowance had been made for disproportionally
higher collection of small, ammonia-containing particles in the outlet hopper.

     The first mechanism of conditioning considered was resistivity modifica-
tion.  The in situ resistivity data obtained with and without ammonia addition
fell within the two areas plotted in Figure 2.  Although some of the resis-
tivity values recorded with ammonia addition were considerably higher than
any result obtained without ammonia addition, there was virtual coincidence
of the points within the two areas corresponding to the "centers of mass"
(i.e., the points representing the average values of the logarithm of resis-
tivity and the reciprocal of absolute temperature).  Hence, no effect of
ammonia on resistivity was evident above the scatter in the data, which is
attributed to several factors:  (a) the customary lack of precision in field
resistivity data, (b) the lack of constancy in the background concentration
of sulfuric acid (also see Table 4), and (c) the high sensitivity of resis-
tivity changes to even minor changes in the sulfuric acid concentration.

     Documentation of three^factors pointing to a space-charge effect is
given in Table 6.  This effect arises when the fume of NHuHSOi* or (NHiJaSOi,
is charged in the ESP and the electric field is thereby.increased, with
accompanying increases in the charge level of fly ash particles and the
collecting field near the plates.  The voltage effect, as the table indicates,
was more prominent in the first field than in any of the following three
fields in succession.  The voltage effect is displayed differently in
Figure 3, which was obtained with the inlet TR under manual rather than auto-
matic control.

     Data pertaining to the effect of ammonia on rapping reentrainment were
obtained with a photoelectric device (Large Particle Sizing System) at the


                                     446

-------
ESP outlet.   Figure 4 gives a graphic illustration of the suppression of the
counting rate at a selected range of particle sizes (11.0 to 14.5 ym) with
ammonia addition.  Figure 5 compares relative counting rates as a function of
particle size.  The effect of ammonia is most logically attributed to a sup-
pression in rapping reentrainment, for existing theories provide no basis for
attributing effects of the observed magnitude to the alternative mechanism—
a change in the effectiveness of the primary collection process.

     A fourth possible mechanism of conditioning—agglomeration of fly ash
particles in the gas stream before they are collected (2)—was considered,
although there is no known evidence of ammonia as a specific agent causing
this effect,  The data in Figure 6 do not exclude this effect, but obviously
they give no suggestion of it.

     The third and final objective of this research was to compare the degree
of improvement in ESP performance achieved with that theoretically predicted,
using the principles incorporated in a computer model of ESP performance (3).
This could only be done for the space-charge mechanism, in view of the
limited ability of the model to predict any effects associated with rapping
reentrainment.  Accordingly, attention was focused on experimental fractional
efficiency data from impactors for small particles and theoretical predictions
for these particles in the absence of rapping reentrainment.  Figure 7 com-
pares the ratios of migration velocities based on experiment and theory.

     Obviously, Figure 7 shows a serious discrepancy for particle diameters
less than 2 ym.  This discrepancy seems to be most logically attributed to an
anomaly in the impactor data, which if operative would give a bias in the
direction of an unfavorable collection efficiency when ammonia was absent.
The operation of the impactors is associated with adiabatic cooling of the
sample gas stream, which can be as great as 30°F (17°C)-  Such cooling could
cause complete condensation of all of the sulfuric acid vapor present in the
absence of ammonia and add materially to the apparent concentration of fine
particulate matter present at the ESP outlet.  The effect of cooling in the
impactors in this respect would be minimal with ammonia present, on the other
hand, because of the almost complete conversion of the acid vapor to a
particulate form upstream from the ESP and its removal from the gas stream in
the ESP.

     In summary, ammonia was found to cause an improvement in ESP performance
at Monroe even in the face of high performance even without ammonia treatment.
The improvement is most graphically seen in terms of the reduction im mass
penetration (about 40%) in the ESP tested.  Data on stack opacity confirmed
that a marked improvement in ESP performance occurred.  The primary mechanisms
identified were space-charge enhancement of the electric fields responsible
fot charging and collecting fly ash particles and suppression of the loss of
collected ash by the process of rapping reentrainment.  Alternative condition-
ing mechanisms—modification of fly ash resistivity and particle agglomeration
prior to collection—seem to be inoperative or without consequence.

     The work described in this paper was not funded by the U. S. Environ-
mental Protection Agency and therefore ,the contents do not necessarily reflect
the views of the Agency and no official endorsement should be inferred.

                                     447

-------
                                 REFERENCES
1.   Dismukes, E.B.  Conditioning of fly ash with ammonia.  J, Air Pollut.
     Contr. Assoc.  25; 152, 1975.

2.   Potter, E.G. and Paulson, C.A.J.  Improvement of electrostatic precipi-
     tator performance by carrier-gas additives and its graphical assessment
     using an extended Deutsch equation.  Chem. Ind. 532, 1974.

3.   McDonald, J.R.  A mathematical model of electrostatic precipitation
     (Revision 1):  Vol. I.  Modeling, and programming, the.electrostatic pre-
     cipitation process.  EPA-600/7-78-llla.  U. S. Environmental Protection
     Agency, Research Triangle Park, North Carolina, June 1978.
                                    .4-48

-------
          TABLE 1.  DESIGN FEATURES OF MONROE UNIT 1


     Boiler      Babcock and Wilcox
                 5,700,000 Ib steam/hr
                 750 MW

     ESPs        Research Cottrell
                 4 parallel units
                 4 electric fields each unit
                 2,360,000 acfm  (66,800 m3/min)
                 240°F (115°C)
                 SCA:  190 ft2/1000 acfm  (37 m2/[m3/sec])
           TABLE 2.  TEST PLAN FOR ONE OF FOUR ESPs
Week   NHs           Temperature           Measurements/Samples


 1     ON    "High"—ca. 280°F (138°C)      Schedule A

 2     ON    "Normal"—ca. 250°F (121°C)    Schedules A and B

 3     OFF   "Normal"—ca. 270°F (132°C)    Schedules A and B
NH3 ON = 8-9 ppm.
A = Ultrafine particle sizing, large particle sizing, resis-
    tivity, flue gas analysis, fly ash and coal for analysis,
    ESP electrical data.
B = Mass train and impactors.
                              449

-------
                        TABLE 3.  ESP PERFORMANCE
                                             With NH3
                  Without NH3
Test conditions*
  Mass loading, gr/scf (g/m3)
  Temperature, °F (°C)
  SCA, ft2/1000 acfm (m2/[m3/sec])
  3.0 (6.9)
252 (122)
156 (30.7)
  3.2 (7.3)
269 (132)
154 (30.3)
Test results
Mass efficiency, %
Stack opacity, %t
Emission rate, lb/106
Btu (ng/joule)
99.83
8.5
0.009 (3.9)
99.72
15.6
0.015
(6.4)
*Coal:  % S = 1.9, % ash = 9.8, Btu/lb = 12,600 (joule/kg = 29,300).
tValues of stack opacity were related to the performance of the one ESP
unit tested in a complex fashion.  The value with ammonia on was
achieved with ammonia addition in all four ESPs preceding the stack.
The value with ammonia off was reached with ammonia off only in the one
ESP tested and one other ESP but with ammonia addition continued in the
other two ESPs.
                 TABLE 4.  GAS ANALYSIS AT THE ESP INLET
                                    With NH3     Without NH3
             Ammonia, ppm
               Vapor                0.2-0.4          0.1
               Total                6.8-9.5       Not det'd

             Sulfuric acid, ppm
               Vapor                0.1-0.9        0.3-1.8
               Total                3.2-7.6        1.6-5.4
                                   450

-------
            TABLE 5.   ANALYSIS .OF FLY ASH FROM ESP HOPPERS


Ammonia , wt %
Inlet hopper
Outlet hopper
Sulfate, wt %
Inlet hopper
Outlet hopper
With NH3
0.033*
0.095*
0.67
1.4
Without NH3
0.007
0.033
0.55
1.1
            *Equivalent to 4 ppm as vapor (mass train sam-
            ple,  equivalent to 7 ppm as vapor).
          TABLE 6.   DOCUMENTATION OF THE SPACE CHARGE EFFECT
                                      With NH3          Without NH3
Ultrafine sizing,  0.018-2.0 ym     3.1 x 1013/DNm3     2.6 x 1013/DNm3

Impactors, <0.5 ym                 0.0138 gr/acf*      0.0041 gr/acf*
                                  (0.0316 g/m3)       (0.0094 g/m3)

Voltaget, 1st field               44.2 ,kV             38.5 kV
*Compare difference = 0.0097 versus
   calculated value = 0.0078 as
                 or = 0.0134 as
tAt current limit (j = 38 nA/cm2).
                                  451

-------
icr-
101-
Z"
•
o
1— 1
h-
,_
LU

£ 10°:
;
^_
Z
LU
0
C£
LU
Q_
10'1-
m-2
—
-
-
•
-
-
•
-
-
Q 0 D ;
[ NH3OFF 0
. •
. «
Q
A *"*
' 9

a
9 m ~
: D '
® * o
NH3ON
* n
; »« •••:
9 A

9
*•.
	 1 	 1 	 1— I — i i i i I i 	 1 	 1 — i i i i i
-u. u
-90.0

>-
o
LU
H— 1
O
1— 1
LU
-99.0 ft

•
z
LU
0
Q£
LU
Q_
:99-9
.QQ QQ
10"1                10°                 101
  PARTICLE  DIAMETER  (MICROMETERS)
    Figure 1. Fractional efficiencies measured with impactons.
4X0-120
                     452

-------
     1x1012
E
-C
o

H
>
u>
HI
cc
     Ix1010
      1x109
                                                    MONROE ASH
                       0 ppm
                                                    PREDICTED FOR
                                                    INDICATED SO3 CONCN
                                      MEASURED WITH 4 ppm SO3
                                      (LABORATORY)
                MONROE ASH

                MEASURED
                WITH NH3
                MEASURED
                WITHOUT NH3
      1000/T(°K)
            °F
                   2.8
                   183
2.6
233
2.4
291
2.2
359
2.0
441
1.8
541
   1.6
   666
4260-112
Figure 2.  Comparison of observed and predicted resistivities of fly ash from Monroe Unit 1
         and another similar fly ash.
                                   453

-------
£0
MS
N
Z
U MB
N
EE
Z 3S
X
>-
f- 30
2 -
u
^ 2B
H
z IS
LJ
o:
o: IB
^
s

B
1


_ NH3ON
NH3OFF +
• a


• a +


'• a +
-
,
L n +
•
•
-
; a +
-

B IS 2B ZS 30 3S MB MS SB
                 5ECDNI>RRY  VDLTREE/KV
       +  LSI  -  II/IH/BI  -  PIMMDNIR ON
       D  UBI  -  I I/IB/HI  -  HMMDNIR OFF
/£ure 3. Current-versus-voltage In Inlet field with and without ammonia addition.
                            454

-------
oc
o
8
ui
                                                                            NH3ON
               j_jj	j  ii Mm   i       i     ,  i  mill  i
.   i    ul
          NH3OFF
                                      TIME
                                                                             4260-136
          Figure 4. LPSS Data for particles 11.0 to 14.5 urn in diameter at the ESP outlet.
                                       455

-------
 CO

z
 CO


z

CO


1    5
01
o

o
o
OQ
O

P
OC

2
u.
o

o
      1.0
                                                 I      I      1
2.0
4.0       6.0


DIAMETER, Mm
8.0  10.0
   20.0



4260-124
        Figure 5. Relative number concentrations detected by the LPSS with ammonia

                 on and off.
                                     456

-------
  3.5
c
o

I  3.0
o  2.5
c
o

I 2.0
€
'i
ai
§1.5
I
  0.5
    O1—
    0.1
                                     I
                                I
0.2     0.3        0.5     0.7     1.0

                    PARTICLE DIAMETER,
                                             2.0
                                                     3.0
                                                               5.0
7.0    10.0

   4260-114
              Figure 6. Ratios of inlet mass concentrations of fly ash with and without
                       ammonia addition.
                                               457

-------
    2.0
    1.8
    1.6
    1.4
 CO

Z
S
    1.2
    1.0
                I
                      COMPUTER MODEL
I
I
I
                                                       I        T
                                                       I
                                                                IMPACTOR DATA
J_
      0.01     0.02        0.05      0.1
                0.2        0.5

               DIAMETER, jum
                           1.0     2.0
                                      5.0     10

                                          4620-118
             Figure 7. Enhancement of migration velocities with ammonia conditioning
                      predicted and observed.
                                             458

-------
           FLY ASH CHEMISTRY INPICES FOR .RESISTIVITY-• AND EFFECTS
           ON ELECTROSTATIC PRECIPITATOR DESIGN AND PERFORMANCE
                by:   Herbert J.  Hall
                     H.  J.  Hall  Associates,  Inc.
                     Princeton,  New Jersey 08540
                                  ABSTRACT
     Field data  on coal-ash properties and in situ fly ash resistivity meas-
urements  for  24  cases of cold side precipitators are correlated with various
ash chemistry indices for resistivity prediction.  Coals studied,  principally
low sulfur, are  from eastern and western U.S.A., western Canada, and several
other countries;  boiler sizes 50-800 MW.  Most universally applicable general
indices for low  S coals seem to be those based upon sodium content,  per se,
and on a  silicate type index long used by the author.  Under special condi-
tions,  other  indices may also be useful.  Ash chemistry indices proposed by
investigators such as Selle, et al, Dunston,  Matts, Soviet ESP designers,
Bickelhaupt,  Hall,  and others are reviewed.   Comparisons of useful indices
with resistivity data calculated from a computer program based on Bickel-
haupt 's work  are illustrated for an additional 16 cases which are also com-
pared with the in situ field data.  Some quantitative aspects of coal-ash
properties as affecting precipitator design and performance are discussed.
                                     459

-------
                        INTRODUCTION AND BACKGROUND
     In the last decade, high resistivity ash problems and effective precipi-
tator design for very high, reliable performance  (99.5-99.9+%) have received
much attention in the ESP community worldwide.  Earlier research by White and
Hall (1946-58) explored and developed many fundamental aspects of  this field
- including ash chemistry factors, various physical and chemical conditioning
systems, correlation of precipitator electrical operating parameters with ash
resistivity factors and process conditions, and new equipments and automatic
controls.  Later work by the author included studies on high temperature/high
pressure corona and resistivity effects, electrical waveform optimization for
high p dusts, electron beam possibilities for high particle charging and in
situ 302^803 conversion, and the role of ash chemistry factors in  resistivity
problems.  Over the past 13 years, I have been closely associated  with pre-
cipitator design, operations and upgrading under a wide variety of high re-
sistivity ash conditions in many different applications.

     Earlier investigations have been continued and significantly  advanced;
new ones have been added, and some have reached a degree of maturity.  An
early recognition of the fact that western U.S.A. coals required a different
approach to viable ESP design was that of Selle, et al (1) in the  early
1960's.  This work showed the critical effect of Na2O ash content, per se,
and some correlations of resistivity with other ash chemistry indices.  Suc-
cess with the hot precipitator at -700F (370C) for coping with resistivity
problems in low S eastern coals, beginning circa 1968, was transferred as an
expected solution to western coal problems by a number of vendors.  Although
some installations have worked well, many did not.  The fine work  of Bickel-
haupt (2,3) beginning 1973-74, has indeed clarified the critical role of al-
kali metal ions, principally sodium, in determining low S coal, fly ash re-
sistivity and in the phenomena of sodium depletion and deteriorating ESP per-
formance over time.  Some of the most difficult ashes for electrostatic pre-
cipitation include the highly lignitic types with low S, low Na, low Fe and
very high CaO+MgO contents, as well as the high silica+high alumina content
(80-90+%) bituminous types.

     In addition to conditioning with various chemical additives,  other ap-
proaches to high resistivity ash problems have surfaced recently - e.g., mas-
sive SCA, new pulse energization systems, hybrid two-stage ESP's,  relooks at
wet wall and cooled charger electrodes, wide duct technology; rigid frame
type, high intensity rapping designs, and others.  A recent paper  (4) gives
several references to results in progress on these techniques.

     Regardless of precipitator design approach, it is first essential to
assess ash resistivity conditions to ensure ultimate long-term, reliable,
high performance at minimum size and cost.  Recent advances in sampling tech-
niques, particle size determinations, and specific chemical constituent anal-
yses of industrial dusts  (5), as well as new resistivity measurement stand-
ards (6),"should be helpful.  Coal and ash chemistry indices to assist in
boiler-ESP operations analyses, in resistivity prediction, and in  coal selec-
tions can be useful.  Although over the years many investigators have devel-
oped such indices and techniques, no coordinated summary of these  indices and

                                     460

-------
their application exists.  Many have not previously been published.  Bickel-
haupt*s technique (2), now being widely used, has had only limited correlation
with field data.  It is our purpose here to try and shed some more light in
these areas.

                SUMMARY AND DEFINITIONS OF COAL-ASH INDICES
     Various coal and ash chemistry indices for predicting fly ash resistivity
and electrostatic precipitator operations may be summarized as follows:

1.  COAL SULFUR CONTENT   •< 1% wgt = low S coal; > 2.5% wgt = high S coal.
Coal sulfur content alone is not generally useful for predicting resistivity
or ESP performance.
2.  COAL S-ASH INDEX, a   a = — '- - - -- long used by the author for eastern
                                % asn
bituminous type coals.  For a. -<0.1, generally high resistivity ash at 300+F;
a > 0.15, generally good precipitation capability.  Gas temperature, ash spe-
cific surface, and relative conversion of S02->SO3 in the boiler are important
concomitant influences .

3.  COAL SODIUM- ASH YIELD RATIO  (.7).   * N\ > 0.035, (atomic sodium and ash
contents by weight in coal) ,  found boiler fouling problems in certain small
Australian boilers.  Loss of any available sodium in the coal due to fouling
in boiler back passes can strongly influence ultimate sodium content of ash
actually reaching the precipitator - hence bulk resistivity and ESP perform-
ance.

4.  ASH TYPES   Lignitic , where CaO+MgO > Fe2O3 (% wgt ash contents)., is typ-
ical of many western U.S.A.,  western Canada, and some other country coals.
Bituminous ,  where Fe2O3 > CaO+MgO, is typical of eastern and midwest U.S.A.
coals - also some western coals.  Bickelhaupt (2)  defines these types on
atomic number basis as Ca+Mg > 3.5% and K < 1% for western ash; Ca+Mg. <3.5%
or K - 1% for eastern ash, respectively.

5.  SLAGGING AND FOULING FACTORS

     It is desirable to have a handle on slagging and fouling properties of
ash related to specific coal sources and boiler conditions.  These are com-
plex subjects now receiving much attention.  Of interest are such questions
as the use of high excess air to help control slagging conditions and effects
on gas temperature, gas flow rate, SO2-»-SO3 conversion; possible loss of al-
kali metals due to fouling;  and overall effects on particle sizing, ash re-
sistivity at the ESP, and performance vs boiler load.  For eastern bituminous
coals, ash base/acid ratios have long been used with coal sulfur and Na20 ash
contents for evaluation indices.  Ash viscosity temperatures and alkali metal
indices are now being more widely used for lignitic type coal ash.  In
assessing suitable coals by source and local areas of mines for good boiler
and ESP performance, we recommend evaluation not only of probable ash resis-
tiyity properties, but also of .potential slagging and fouling aspects.  These
latter interesting subjects are beyond the scope of this paper; henoe, we

                                     461

-------
fold their tents with brief words and steal away to more ash chemistry indices
and resistivity matters.

6.  ATOMIC PER CENT LITHIUM + SODIUM

     Basis for Bickelhaupt (2,3) computer model for predicting resistivity
from coal bore samples.  It is applied to both western and eastern U.S.A.
coals in combination with other ash components - Ca+Mg, K and Fe.  Note that
this refers to atomic per cent by number for the various chemical species
present in the ash from coal fired in the laboratory at 1050C.  Computer pro-
grams produce p as a function of temperature (with and without SO3 effects) ,
and include corrections for selected moisture content of gas, electric field
(usually lOKV/cm) , and surface conductivity based upon a known 803 content or
on SO3 calculated as 0.4% of SO2 as determined from stoichiometric coal data
with 30% excess air.  Final resistivity prediction is the net result of chem-
ical, moisture, and S03 surface resistivity operating in parallel with volume
resistivity for gas temperatures. « -350-400F (177-204C) .  The computer pro-
grams were derived from careful laboratory resistivity measurements in simu-
lated flue gas environment.  Some further correlation of this method with in
situ resistivity field data are presented later.

7.  SILICA + ALUMINA CONTENT, y   y = SiO2 + A12O3 wgt % in ash.  For y typi-
.cally - =80% of ash, high resistivity problems can be expected at 300F+ gas
temperature .

8.  x/y INDEX =    e  a    expressed as per cent, where Fe, Na, K are mono-
atomic wgt % of iron, sodium and potassium in the ash.  This is a type of sil-
icate index long used by the author, particularly for eastern bituminous
coals.  Useful correlations with resistivity and precipitator performance
have been found.
                              Na-pO
9.  SILICATE INDEX   HSiO = - »••  . expressed as per cent, where ash con-
                            y+CaO+MgO   *          *
stituents are % wgts.  We have used this index with low S, lignitic and bitu-
minous ashes with fairly good results on resistivity prediction and precipita-
tor operations - general applicability.
10.  ALKALI -SILICATE INDEX   ASi = „ .      „  , %, where K and Na are mono-
                                   S1O2+A12U3
atomic wgt % in ash, gives fairly good general correlation within  limits.
Some anomolies have been found with  certain bituminous type  ashes.

11.  Na20 ASH CONTENT   Shown as a primary index for ion conduction mechanism
developed by Bickelhaupt.  White  (8) illustrates good in situ resistivity
correlation for a group of western coal ashes.

12.  SILICA-SODIUM INDEX   We have found that cold  side precipitators  gener-
ally have good performance capability on low S, subbituminous , western type
coals at sGOOF when  the ratio Na20/Si02, as per cent wgts in ash, is  equal
to or greater than 0.05-0.07.  Frisch (9) has indicated similar results in a
plot of p vs SiO2/Na2O for western ashes from seven states.

13.  ALKALI SULFATE INDEX  (ASI)  Matts  (10) has described this index long

                                     462

-------
used by Flakt, Inc.  The ASI is the sum of ash components, % wgt, of K20,
Na20 and P205 plus portions of the S03 content used in sulfating the potas-
sium and sodium oxides.  Any free 803 left after remaining chemical conver-
sion of CaO to CaSC>4 is also added.  The ASI was developed for assessing re-
sistivity and ash precipitability from coal samples.
14.  SOVIET INDEX (11)   Ks =                where g^ and ^^ are we±gnt
% in coal ash, and A, M, H2, S are ash, moisture, hydrogen and sulfur con-
tents % wgt in the coal as received .  Some low moisture , low S , very high ash
content coals are used in the U.S.S.R.

15.  MODIFIED SOVIET INDEX   KH = Ks/Na2O.  We have sometimes used this varia-
tion to give weight to the sodium content of ESP ash expressed as Na2O, %
weight .

16.  BUREAU OF MINES OXIDE INDEX (1)   BMO = Ca.°+Mg°   where items are % wgt
                                             Na2O+SO-j
in ash.  Used for western low S coals by Selle, et al.

17.  DUNSTON OXIDE FACTOR (12)   DOF =  a2°+Fe2O3+CaO wgt % ratio is corre-
                                           A1203
lated with ash softening temperatures.  For DOF > 0.4, improved performance
of small size precipitators is reported on industrial boilers burning low S
eastern coals.  Dunston gives an interesting discussion of mineral components
of fly ash related to conditions under which various coals were formed.

18.  SO4 CONTENT OF WATER SOLUBLE ASH

     For eastern U.S.A. coal precipitators, good performance at =i300F+ with
modest SCA typically requires a minimum of =0.6% SO^ by weight of total ash.
This reflects the amount of H2SO4 adsorbed on fly ash particle surfaces.  It
can be achieved either by combinations of adequate coal S-ash factors with
suitable gas temperatures and normal S03 production in the boiler system, or
by injecting SO3/H2SO4 conditioning agent.  Let us reiterate that sulfur con-
tent of coal, per se, is not very useful in assessing reliable resistivity
and precipitator performance conditions.  SO3/H2SO4 adsorption processes are
enhanced as temperature is reduced.

                                DATA SUMMARY
     Table 1 summarizes basic coal and ash chemistry data together with in
situ precipitator ash resistivity measurements for 24 cases among various
coals and conditions.  Ash chemistry is based upon samples of ESP inlet ash.
The in situ data were carefully measured at ESP inlet with well-known point-
plane apparatus.  Most of the base data are from our files and private com-
munications.  Items 6-9, 16, 16a are from Gooch (13)  and Bickelhaupt (.2,14);
item 15 is from Carr and Ensor (15); and lignite data are from Gronhovd, et
al (16) .

     Table 2 summarizes coal-ash data for another 16 cases for which indices
were calculated and compared with resistivities calculated as shown on the


                                    463

-------
TABLE 1.  SUMMARY COAL-ASH PROPERTIES AND IN SITU RESISTIVITY DATA
% by wgt. coal %
Plant Coal
1.
2.
2a.
3.
4.
5.
6.
7.
8.
9.
10.

11.
12.
12a.
13.
14.
15.
16.
16a.
17.
18.
19.
20.
21.
E. Bit.
E. Bit.
E. Bit.
E. Ky.
E. Ky.
E. Bit.
E. HiS
E. HiS
E. Bit.
E. Bit.
70% S.Afr.
30% WC
W-SB
W-SB
W-SB
W-SB
W-SB
W-SB
West
West
Austr.
W. Can.
W. Can.
Lignite ND
Lignite ND
M
6.4
6.0
6.0
4.0
4.0
7.5
2.1
10.84
11.7
8.4
10.3

27.6
25.3
25.3
23.4
28
12.7

13.94
=3
20
20


Ash
13
17
11
11.7
7.4
18
11
11.2
10.9
25.8
15.7

5.4
7.8
7.8
6.0
5.6
10.6

5.21
32.8
14
15.8


S
0.9
0.99
1.8
0.7
0.76
•0.5
3.28
2.05
0.81
0.79
0.56

0.45
0.40
0.40
0.48
0.40
0.61

0.41
0.42
0.20
0.19


H2
4.6
4.4
5.0
5.1
5.0
=5.0
5.0
3.95
3.87
3.74
4.92

3.45
3.6
3.6
2.86
3.5
4.93

4.24
4.8
4.8
4.8


K BTU Si02 +
lb A1203
12.3 84.6
10.6 81.0
12.5 67.0
12.6 85.5
13.1 84.0
11.2 80.5
=12.5 64
11.05 72
12.01 82.8
=10.6 84.6
10 78

8.83 57
8,5 66.6
8.5 64.3
8.77 68.5
8.25 48.2
10 62
75.2
10.56 75.6
9.72 90
77.4
77.3
36.2
34.1
A1203
29.1
31.0
22.0
30.2
27.5
. 27
18. S
21.7
29.3
27.2
30

16
23.1
22.3
21.3
15.4
17.4
24.6
23.7
25
22.6
26.4
14.5
13.0
Fe203
7.2
6.7
14.5
7.0
6.3
6.5
22.7
13.1
7.1
7.0
4.74

7.5
6.4
7.3
4.0
5.8
5.72
5.5
5.9
7.24
4.5
4.8
11.8
9.5
wgt. in ash
CaO
2.3
1.5
2.0
1.5
1.4
7.7
4.73
4.95
3.7
0.8
7.9

30
17.4
18.8
16.2
26.1
16.1
8.7
8.6
0.33
11.7
13.2
29.3
25.4
CaO +
MgO
3.3
2.2
3.0
2.2
2.9
9.1
5.7
6.0
4.6
2.1
9.9

35.5
22.2
22.6
19.7
30.8
18.9
12.3
12.2
0.81
13.5
14.2
43.1
34.3
Na2O
0.3
0.61
2.6
1.0
1.5
0.45
0.54
0.67
0.27
0.45
0.38

1.15
0.91
1.4
0.75
2.10
0.33
0.50
0.29
0.07
0.36
2.26
1.40
6.9
K20
1.1
2.7
4.3
2.5
2.8
1.5
2.5
2.1
2.1
3.4
0.85

0.4
0.86
0.80
0.53
0.30
1.13
1.72
1.8
1.01
1.3
1.0


so3
0.5
0.31
0.42
0.30

0.3
2.77
2.29
0.70
0.10
3.3

2.8
1.3
1.5
4.5
11.6
8.7
0.55
1.2
0.20
1.1
0,42
5.1
9.9
°F
300
257
262
330
275
300
311
315
325
290
266

320
300
300
300
300
300
306
293
300
350
350
310
335
H20
Vol.
=7.0
=7.0
=7.0
=7.0
=7.0
7.0
7.2
8.2
9.0
8.1
10.0

8-9
11
11
10-11
11-12
10
8.2
8.3
6.5
=10
=10


Resistivity
ohm-cm
1012
SxlO11
3-5xl09
SxlO11
=3. SxlO1 °
1012
1.7xl010
2xl010
2.7xlOU
SxlO11
2.4xl012

l.lxlO11
l.SxlO11
7xl010
5x10 n
4x10 10
4x10 12
1.4xlOn
3x10 u
3-5x10 13+
2xl012
4x10 10
2.3X1010
SxlO9
Size
MW
375
200
200
80
Small
250
160
122
350
250
285

350
800
800
480
650
525
128
204

150
150



-------
                                                           TABLE 2.
                                                                     SUMMARY COAL-ASH PROPERTIES AND RESISTIVITY DATA

                                                                     CALCULATED ON BICKELHAUPT MODEL COMPUTER PROGRAM
Ul
Coal
1.




2.

3.
4.
5.
a.
b.
c.
d.
e.
a.
b.

*
**
W. Can.
W. Can.
W. Can.
W. Can.
W. Can.
S. Am.
S. Am.
Spain
E.Bit.
E.Bit.
M
27.5
27.5
27.5
27.5
27.5
6.93
6.93
10.0
=4
4.3
Ash
18.6
23.4
23.4
23.4
23.4
6.54
11.0
13.0
=10
5.64
S
0.49
0.45
0.45
0.45
0.45
1.0
0.6
0.6
=0.8
1.1
H.2
2.93
2.67
2.67
2.67
2.67
4.85
4.85
4.50
=5.0
5.0
BTU/lb
6800
6192
6192
6192
6192
13000
11700
11160
=12.5
12600
Si02 +
A1203
78
84
64
74
56
74
81
76
83
80
.6
.0
.3
.2
.5
.1

.0

.4
A1203
22
22
21
22
20
15
20
30
28
28
.71
.53
.12
.26
.33


.4

.76
Fe
3
3
9
5
14
10
7
6
5
8
2°3
.86
.46
.04
.06
.43
.51
.0
.28

.85
CaO
7.57
5.93
11.80
9.38
14.03
3.88
3.80
8.61
0.7
2.32
CaO +
MgO
8.5
7.14
13.64
10.5
15.3
7.2
7.1
12.0
1.4
3.84
Resistivity, No
Na20
0.47
0.86
1.34
1.82
2.49
0.84
0.40
0.41
0.30
0.35
K20
1.28
1.49
0.81
1.20
0.74
1.10
0.80
0.71
1.60
2.47
so3
6.06
2.19
9.88
6.37
9.65
4.29
2.60
3.65
0.6
2.59

5
1
4
2
1.

1
1

1
300F
.4xlOn
.3xlOU
.6x10"
.SxlO10
08x10 10
2xlOn
.IxlO12
.SxlO12
2x10 12
.7xl012
Max @
9.3xlOn
3xlOn
9x10 10
SxlO10
2. IxlO10
2. SxlO11
1.4xl012
2x10 12
2xl012
1.7xl012
S03
p with SO3
F %H2O
Vnl.
360
370
383
392
374
340
340
340
300+
300
12.7
13.1
13.1
13.1
13.1
10.8
10.8
8.1
=7.0
7.8
300F
'gxio10




4. SxlO1 °
1.2X1011
l.SxlO11

4x10 n
Max
6x10 U




1.7xlOu
7.4X1011
1012

1.3xl012
F
400




360
380
400

350
Hot Case
6.


7.


a.
b.
c.
a.
b.
c.
E.Bit.
E.Bit.
E.Bit.
S. Afr.
S. Afr.
S. Afr.
5.0
5.0
5.0
8
a
8
12.5
12.5
-12.5
16
11
14
0.6
0.6
0.6
0.5
0.4
0.3
4.5
4.5
4.5
3.5
4.2
4.0
12500
12500
12500
=10000
=10000
=10000
88
88
88
75
93
96
.2
.47
.67
.9
.4
.7
35.2
35
35
34
24
34
.34
.42
.5

.7
4
4
4
6
2
1
.66
.66
.66
.0
.7
.1
1.02
1.02
1.02
9.0
1.0
0.1
1.37
1.37
1.37
11.0
1.6
0.2
0.69
0.40
0.20
0.15
0.5
0.04
0.77
0.77
0.77
0.3
1.1
0.2
1.57
1.57
1.57
4.0
0.3
0.2
3
1
4
1

1
.9xlOn
.IxlO12
.9xl012
.SxlO13
9x10 n
.OxlO11*












7.8
7.8
7.8
=7.0
=7.5
=7.5
3.2xlOn-
6.5X1C11
l.lxlO12



i . sxid9 @
4.2xl09
l.SxlO10



600
600
600



           ESP after mechanical collector.
          **
            Washed coal.
                                                         Note:  Ash  chemistry analyses  based upon laboratory coal samples..

-------
Bickelhaupt (.2). model.  In this group, we gratefully acknowledge permission by
F. L. Smidth and Co,, Inc., to use data for items 1-3, 5»

     For Table 1 an attempt was made to select cases as close as possible to
-300F gas temperature, not only for practical ESP operations, but also fpr
reasonable comparisons with the calculated resistivity data on Table 2,  For
the in situ data, we have a range of 260-350F, overall average 304F, and
standard deviation -8%.  For the 18 closest cases, the average is 3Q4F and
the standard deviation is 4.3%.  The range of data in Tables J. and 2 trayerse
important practical applications for modern fly ash precipitators with rea-
sonably good diversity of coal sources and ash properties.

                           RESULTS AND DISCUSSION
     For low S coals, our silicate index HSiO is fairly generally applicable.
Figure 1 shows the in situ fly ash resistivity data  CTable 11 and the resis-
tivities calculated on Bickelhaupt's model  (30QF, no S031 as a function of
this index.  Curve A can be represented by the equation, In p = 25.85-2.9 In
(HSiO).  The bulk of the standard computer calculated data from Table 2
agrees quite well with the in situ data, most of which are for low sulfur
coals.  Curve B, having essentially the same slope, might indicate nominal
minimum expected resistivity for medium to high S eastern type coals.  It
would appear from our results, however, that some caution should be exercised
in using the Bickelhaupt method with 503 correction for low S coals, particu-
larly those having lignitic type ash.  Note that using the standard 803 cor-
rection in Table 2 would drop indicated Bickelhaupt resistivities by about
one-half to one order of magnitude below the in situ data curve.  With the
high CaO ashes the beneficial effects of SOo may indeed be largely lost in
affecting the resistivity of the bulk ash.

     The importance of alkali content in controlling the bulk resistivity of
low S coal ashes seems to be well established.  Figure 2 illustrates the
strong effect of sodium from our data.  Another variation of such effects re-
sides in the alkali silicate index, ASi, which is plotted against resistivity
in Figure 3.  This shows generally pretty good correlation for both bituminous
and lignitic ashes, in situ and Bickelhaupt derived resistivities.  The
alkali-silicate index was derived from long experience with our other silicate
type index x/y, where it was noted that some spread among points occurred be-
tween eastern and western in situ data.  Deviations also tended to occur in
bituminous ashes for very low and very high iron content.  The x/y index is
overly sensitive to wide variations in Fe content, most consistent results
being obtained for typically =*4-10% as Fe2C>3,  Fe content apparently has some
effect on resistivity, but to a much lesser extent than sodium.  According to
Bickelhaupt, the iron may in some way have a catalytic effect on the useful-
ness of the potassium ion.  Eliminating the Fe from the x/y index pulls the
diverse data much closer together as represented in Figure 3,  There are,
however, a 'few points still to the right of the average curve in the region
3xl011-4xl012ohm-cm.  The reasons at this 'time are obscure.

     Long standing in resistivity prediction is the Alkali Sulfate Index  CASI)
used by Flakt  (101 -  Considering the temperature variations and diversity of

                                     466

-------
                                  In p = 25.S5-2.9 In (HSiO)
                                  AVG. IN SITU DATA
                                    260-350F
Hin p expt'd
E. Bit. coals
Ned-high S
    IN SITU DATA (Table 1)
       • E. Bit.
       is. Bit, HiS, (SRI 6,
       • E. Bit, (SRI 8, 9)
       OW. Subbit.
         So. Afr. £ H. Can.
       4 Australia
       AW. Can.
       • Lignite
    Computer calc. p on Bickelhaupt
    8 300F, no SO3 effect (Table 2)
        ?CaOtMgO/Fe203  S i.n
        HCaO+MgO/Fe2O3 >  1.0

                    Na-,0
             SiO2+Al2O3+CaOfHgO'
                                                                         10
                                                                           13
                                                                         101
10'
                                                         In situ western   —m

                                                                 (8)
                 Legend as in Figure 1.
                                        Bickelhaupt"
                                        calc. low S
                                        no SO3, 300F
                                 ash content,  % wgt.
                                    i         I
                  0.1                      1.0
  In situ measured fly ash resistivity and resistivity calculated
  by computer on Bickelhaupt model (300F, no SO3 effect) as a
  function of ash silicate index.  Chemical constituents % wgt.
  in ash and HSiO expressed as per cent.
    .01                     .1                       1                      10

   Figure 2.   Fly ash resistivity vs Na2O ash content for low S coals, =300F.

-------
      10" _
Figure 3.  Fly ash resistivity vs
          alkali silicate index.
          K, Na are monoatomic wgt.
          % in ash.
Figure 4. Fly ash resistivity
          vs alkali sulfate
          index.
our data, it can be considered a useful index as shown in Figure 4,  In this
case, the in situ data show up more or less about 0.2-0.5 order of magnitude
below the Bickelhaupt model calculated data a,t 300F, no S03.  Examination of
alkali index alone (lO^OHS^OH^C^) , without any 863 effect as in the ASI,
correlates quite well with in situ resistivity data on low S western coals.
albeit the negative slope of the curve is quite high  (-10.4) in a log-log
plot.  This again points up our conclusion that the S03 effect in low S, lig-
nitic ash appears not very significant in controlling effective ash resistiv-
ity in precipitators.  This is not to say that some 803 does not sulfate the
sodium and potassium oxides, but rather it is the alkali metal ions, per se,
that seem to have the greatest influence on ultimate ash resistivity.

     A plot of p vs silica-sodium index SiC>2/Na20 shows the same general re-
sults indicated by Frisch (9) for low S western coals  (250-300F gas).  Our
data extend his curve to about 101 "* ohm-cm including low S bituminous type
ashes, both in situ and Bickelhaupt calculated resistivities  (300F, no 803).
Maximum -spread of a few points above and below the average curve is typically
about 0.5 order magnitude on resistivity.
                                     468

-------
     Data plotted for indices Ks, KH, BMO and DOF show less effective general
correlations with resistivity under the diverse conditions in our data bank.
However,  within limits and possibly for certain specialized applications as
previously indicated, they can be useful.  High values of Ks, for example,
definitely would be associated with low moisture, low S, high ash coals having
bituminous type ash exhibiting high resistivity properties.  Our plot of in
situ eastern bituminous data shows a reasonable correlation with p - e.g.,
Ks =10 at 1010, 18-20 at 1011, 35-45 at 1012, and 150 at 4-5xl013ohm-cm, re-
spectively.  With few exceptions our in situ data for DOF  <0.4 shows high
resistivity expectations for low S eastern bituminous coals as suggested by
Dunston.   Our data for the BMO index fall into widely separated groups with
different slopes.

SUMMARY OF USEFUL INDICES
     Table 3 summarizes useful indices with equations for typical average
resistivity predictions.

    TABLE 3.  SUMMARY MOST USEFUL ASH CHEMISTRY INDICES FOR RESISTIVITY
         Index
1. HSiO =
2.
             Na2O
          y+CaO+MgO'
       Average Resistivity
         Formula (-300F)

%   In p = 25.85-2.9 In (HSiO)
                                  Application

                                 Low S general
    In p = 25.777-2.67 In (Na20)
    In p = 26.487-2.572 ln(Na20)
3. ASi =
            K+Na
                                In situ western
                                In situ western
                                   White (8)
In p = 25.45-2.53 In (Na2O)  Low S coals Bickelhaupt
                             model calc.,  no SO3
In p = 26.00-3.56 In (ASi)   General, some anomolies
                             in bituminous ash
4. ASI = alkali sulfate  In p = 32.49-5.06 In (ASI)       In situ data
                         In p = 33.50-5.16 In (ASI)   Bickelhaupt model calc.,
                                                      no 503
5. AI = K20+Na20+P205    In p = 33.71-10.4 In (AI)
                                    In situ western
6, Si02/Na2O

          Na+K+Fe
    In p = 18.22+1.652 In  .,T1 %.  Low S western and
                          (Na20)
7. x/y =
    In p = 34.64-5.19 In x/y
    In p = 34.55-5.89 In x/y
                             eastern
                               In situ E. bit.
                               In situ western
COMPARISONS AMONG RESISTIVITY INDICES AND OTHER FACTORS


     An example illustrated for the Bickelhaupt method (3) has the following
properties used for comparative analyses:  coal as received % wgts, C = 57.2,
H2 = 3.74, 02 = 3.03, N2 =1.02, S = 0.79, M = 8.41, ash = 25.8, =10,300
BTU/lb; ash analyses % wgts, Li2O = 0.04, Na20 = 0.45, K20 =3.7, MgO =1.4,
                                     469

-------
CaO = 0.7, Fe203= 6,7, A1203 =  27.6,  SiO = 58.2,  Ti.02 - -1,7 ,
= 0.2; gas analyses, S02 = 680  ppm vol,  H20 = 8,228% vol, SO3
                                                                   = Q-.-lf SQ3
                                                                 2.72-ppmvol.
     The ash is bituminous with very low slagging,  fouling potential and high
silica + alumina = 85.8%.  Most of  the chemical indices predict resistivity
-1012ohm-cm in situ at =300?, which corresponds with the base Bickelhaupt
model with no SO? effect - the net  p with all corrections, including S03,
being 4.4-7.7X10-1-1ohm-cm.  This figure compares best with predictions using
the x/y, ASi and ASI indices.  Conservative estimates would place nominal ex-
pected ash resistivity in the range 7x10-^ to
SOME PRECIPITATOR DESIGN AND PERFORMANCE EFFECTS
     Important factors affecting bulk  ash resistivity and precipitator per-
formance capability include ash chemistry, gas temperature, pressure and mois-
ture content, ash specific surface,  and degree of H2S04 adsorption..  In addi-
tion, other major factors include  SCA,  electric field strength as limited by
sparking or other effects, allowable average current density and quality of
its distribution, duct width, degree of TR match to -load, and effectiveness of
automatic voltage controls and rapping systems.  One must also be concerned
with boiler operations, excess air  slagging and fouling conditions - partic-
ularly loss of valuable sodium in  critical cases  hydrophobic ash properties,
and the amount and particle size distribution of any carbon present in the
ash.  Further, it is important to  remember that the dynamic ash resistivity as
seen by the operating precipitator may vary significantly due to temperature
variations, or due to changes in ash composition, field to field, as well as
between coal samples and ash actually  reaching the ESP.  Subtle matters of
experience and judgment are often  necessary in assessing overall effects of
any predicted resistivity condition  from coal samples, per se.  We have often
investigated correlations of dynamic ash resistivity as calculated from ESP
electrical data with various ash chemistry indices.
     Precipitator performance  fac-
tors, e.g., power density,  current
density, effective migration veloc-
ities, etc., can be correlated
with coal-ash chemistry  indices
and resistivity  (4) under various
operating conditions.  Figure  5
illustrates some typical effects
in three small boilers for  average
ESP current density as a function
of the coal S-ash index  a,  on
eastern and midwest bituminous
coals.  Figure 6 shows ESP  collec-
tion effeciency as a  function  of
ash chemistry index x/y  for three
different plants firing  various
eastern bituminous coals.   Ash
chemistry index correlations at
any given plant are expected to
                                          Hln-vgt. ESP'*
                                           0 80 MM, I. Bit. coal, lon-Md. S, 32ST gai
                                           D 100 MM, MldMIt Bit. 00.1, UP «ft.r MC, 1-1.5* S, 305F
                                         — A 100 MM, HldMK Bit. - cycl. ballm 3.5-4.71 S,3lSOf



                                          I
                                              495F
                                               X
                                      Figure 5. Average precipitator current
                                                density vs coal S-ash index, ou  w
                                     470

-------
                                   Figure 6.
produce reasonably consistent re-
sults for known conditions.  Ex-
perience shows that well designed
ESP's with good, uniform, current
density distribution can deal
with ash resistivities as much as
an order of nagnitude higher than
that possible in poorer designs.

     Our experience with high
silica + alumina ashes - =80% of
total ash content shows that hy-
drophobic properties may be pres-
ent to inhibit full usefulness of
moisture and often 803 condition-
ing.  These fly ashes are gener-
ally susceptible to effective NH3
conditioning, including high tem-
perature (700-750F) cases.  We
have also used NH3 conditioning
successfully on cat cracker,
aluminum silicate dust at similar
high gas temperatures; optimum mixtures of NH3 and water vapor content exist.
Many western coal ashes can be treated with SO3 conditioning, albeit at levels
often requiring two or three times that required for typical high resistivity
bituminous ashes.  In troublesome hot precipitators at relatively low gas
densities, we have found the critical bulk ash resistivity to be as much as
an order of magnitude lower  (-109ohm-cm) than the customary figure experi-
enced with low temperature ESP's.  Precipitators with difficult ashes have
also been treated successfully with sodium additives to achieve -1.5 to 2.5%
Na20 in the ESP ash.

                                CONCLUSIONS
Precipitator performance as a
function of x/y ash chemistry
index for three plants, eastern
bit. coals.  Na, K, Fe are mono-
atomic % wgts. in ash.
     Several coal-ash factors and/or ash chemistry indices related to assess-
ing resistivity conditions and operations in cold side fly ash precipitators
(=300F)  have been investigated for a wide variety of coal-ash properties using
in situ resistivity measurements in comparison with resistivities calculated
from Bickelhaupt's computer model (2).  Ash chemistry indices can be useful
not only for precipitator design and performance analyses, but also for selec-
ting preferred coal sources to achieve optimum results.

     The strong influence of alkali content, principally sodium, in control-
ling the bulk resistivity of low sulfur coal ashes from diverse sources is
confirmed and well established.  For both low S lignitic and bituminous type
ashes,  the most generally applicable indices seem to be those based upon so-
dium, per se, and on a silicate type index HSiO = Na20/Si02+Al203+CaO+MgO,
where items are % wgt in ash and HSiO expressed as per cent.  Fairly good
correlation between these indices and resistivities, both measured in situ and
calculated on the Bickelhaupt model (300F, no SO3) has been found.  For high
CaO lignitic ashes,  and indeed for many bituminous low S coal ashes at gas
                                     471

-------
temperatures 300+F, the potential benefits of 503 may be largely lost or
greatly reduced.  Our results suggest that caution be exercised in using the
Bickelhaupt method with 863 correction in such cases.  Our investigations on
the Alkali Sulfate Index (ASI) and its variations for lignitic western ash
also suggest that, while some 803 may sulfate the potassium and sodium oxides,
it is alkali metal ions, per se, that seem to have the greatest influence on
ultimate ash resistivity.  Other indices can also be useful under special con-
ditions.  Equations based on our data for average resistivity predictions at
-300+F from seven indices are presented.

     Low sulfur bituminous type ashes having very high silica + alumina con-
tent (80-90+%) are often deficient in alkali and exhibit very high resistivi-
ties.  Our experience shows that such ashes are susceptible to NH3 condition-
ing even at high gas temperature (-700-750F).  These are also good candidates
for pulse energization, as are the difficult lignitic ashes too.

     Many factors are involved in the net dynamic ash resistivity conditions
as seen by a precipitator and some discussion is included.  Subtle applica-
tions of experience and judgment are often necessary in assessing overall ef-
fects of any predicted resistivity condition from coal samples, per se.  Uni-
form corona current density distribution delivers a premium in precipitator
performance; such well designed systems can handle ash resistivities up to an
order of magnitude above those possible in poorly designed equipment.

     Accurate and careful chemical analyses of dusts are necessary for reli-
able resistivity prediction.  We look forward to future progress in analyses
of actual chemical species present in fly ash, and to developments for coping
with persistent ash resistivity problems.
                                 REFERENCES
1.  Selle, S. J., Tufte, P. H. & Gronhovd, G, H.  A study of the electrical
    resistivity of fly ashes from low-sulfur western coals using various
    methods.  Paper 72-107, 65th Annual Meeting of the Air Pollution Control
    Ass'n, Miami Beach, Fla., June 1972.

2.  Bickelhaupt, R. E.  A technique for predicting fly ash resistivity.
    EPA-600/7-79-204.  Symposium on the Transfer and Utilization of particu-
    late Control Technology, U.S. EPA, Research Triangle Park, N,C., Aug.1979.

3.  A Manual for the Use of Electrostatic Precipitators to Collect Fly Ash
    Particles.  EPA 600/8-80-025,  U.S. EPA IERL, Research Triangle Park,
    N.C., May 1980.  pp. 340-361,

4.  Hall, H. J.  High voltage power supplies and microprocessor controls for
    electrostatic precipitators.  Paper presented at the International Con-
    ference on Electrostatic precipitation, Monterey, Calif., 14-16 Oct. 1981.
    Proceedings published by Air Pollution Control Ass*n, 1982.
                                     472

-------
 5.  Farthing, W.E.  Particle sampling and measurement.   Envir.  Sci.  & Tech-
     nology,  Vol.  16, No.  4,  1982.   Review of Third Symposium on Advances in
     Particle Sampling and Measurement,  Datona Beach,  Pla.,  Oct. 1981.

 6.  IEEE Standard criteria and guidelines for the laboratory measurement and
     reporting of  fly ash  resistivity.  #548-1981.  Institute of Electrical
     Engineers, Inc.   New  York, N.  Y., 24 April 1981.

 7.  Bosio, M. and Durie,  R.  A.  The control of boiler fouling during the com-
     bustion  of Leigh Creek coal (Elect.  Trust of S. Australia). Paper 8,
     Austr. Inst.  of  Fuel  Conference on the Changing Technology  of Fuel,
     Adelaide, 5-7 Nov.  1974.

 8.  White, H. J.   Resistivity problems in electrostatic-precipitation.
     JAPCA, vol. 24,  No. 4, April 1974.   p. 334.

 9.  Frisch,  N. W. and Dorchak, T.  P.   Impact of fuel  on  precipitator per-
     formance.  Pollution  Engineering, May 1978.

10.  Matts, Sigvard.   Coal-ash compositon and its effects on precipitator per-
     formance.  Flakt, Inc.  Technical Bulletin,  Vol.  No. 1,  No.  3, Oct.  1977.

11.  Kropp, L. I., Shmigol, I. N.,  Chekanov, G.  S.,  et al.   Joint US/USSR test
     program  for reducing  fly ash resistivity,  JAPCA, Vol.  29,  No. 6,  June
     1979.                            '

12.  Dunston, James B.,  Jr.  Effects of ash chemistry  on  precipitator perform-
     ance. Paper  81-17.3, 74th Annual Meeting of the  Air Pollution Control
     Ass'n, Phila., PA., 21-16 June, 1981.

13.  Gooch, John P.  Electrostatic  precipitator performance.   EPA report
     600/7-79-044a, Feb. 1979.

14.  Bickelhaupt,  Roy E. and Sparks, L.  E.  Predicting fly ash resistivity -
     an evaluation.  EPA report 600/9-80-039a, Sept. 1980.

15.  Carr, Robert  C.  and Ensor, D.  S.   Evaluation of the  George  Neal  electro-
     static precipitator.   EPA report 600/9-80-039a, Sept. 1980.

16.  Gronhovd, G.  H., et al.   Some  studies on stack emissions from lignite
     fired power plants.   Lignite Symposium, Grand Forks  Energy  Res.  Lab.,
     BuMines.  9-10 May 1973.


     The work described in this paper was not funded by the  U.S. Environ-
mental Protection  Agency and therefore the contents do not necessarily reflect
the views of  the Agency and no official endorsement should be inferred.
                                    473

-------
     A NEW ENERGIZATION METHOD FOR ELECTROSTATIC PRECIPITATORS

               MITSUBISHI INTERMITTENT ENERGIZATION SYSTEM
              by:  Takashi Ando
                   Mitsubishi Heavy Industries, Ltd.
                   Kobe Shipyard & Engine Works
                   Kobe, Japan

                   Naoji Tachibana
                   Mitsubishi Heavy Industries, Ltd.
                   Kobe Shipyard & Engine Works
                   Kobe, Japan

                   Dr. Yoichi Matsumoto
                   Mitsubishi Heavy Industries, Ltd.
                   Takasago Technical Institute
                   Takasago, Hyoogo, Japan


                                   ABSTRACT

     Mitsubishi Heavy Industries, Ltd. developed  a new energization  system for ESPs
and has proved its favorable performance on many operating full-scale ESPs with high
resistivity  ash.   This  paper is to introduce the  new equipment named  Mitsubishi
Intermittent Energization System (  abbreviated to MIE ) which supplies high voltage to
ESP intermittently with adjustable supplying and suppressing times by means of power
control  thyristors with electronic circuit.  MIE mitigates the degradation of collecting
efficiency  due to back corona and its effect can be estimated by voltage-current
characteristics at gas load  operation.  The best improvement is expected for the case
of low voltage and large current.  A long term actual operation on an iron ore sintering
machine has shown that the modified migration velocity W|< of 100% at conventional
energization is improved to  147%  by MIE and at the same time the power consumption
of 100% is reduced to 20%.
     The advantages of MIE can be summarized as follows:
    (1) improvement in collecting efficiency for high resistivity ash
    (2) energy  saving
    (3) small additional cost
This paper includes explanation of  equipment, result  of laboratory test and measure-
ment of full-scale ESPs.
                                INTRODUCTION

     In Japan, the change in the energy situation which started from the energy crisis
in 1973 gave impetus to a review of fuels.  As a  result, coal has been again highlighted
as a fuel  for thermal power plants, and the planning and construction  of coal-fired
thermal power plants are positively progressing.  However, coal has many environment-
al problems in contrast  to oil.  In particular, amount of dust (fly-ash) in the- flue gas

                                       474

-------
produced by pulverized coal combustion is more than 100 times that of oil combustion.
Therefore, coal-fired thermal power plants require high performance of ESP.
     However, the electric resistivity of fly-ash varies widely according  to the kind of
coals and has a significant effect on the performance of ESP.  The  value of resistivity
has a peak  in the temperature range (130°C to 150°C) of the flue gas at the outlet of
the air preheater where a cold-side ESP is installed, and such conditions are unfavorable
for stable  and efficient dust collection.  This  trend increases as the  sulfur content of
coal decreases.   Japan with few coal resources is necessarily dependent on overseas
imports. However, most of overseas coals are  low sulfur content less  than 1% and  the
fly-ash produced from such coal has extremely  high resistivity and significantly  impairs
ESP performance.
     Under the  circumstances,  our current  task  for  ESP was to  develope a  new
technology to meet the following requirements;
        (1) to mitigate degradation of ESP performance  due to high  resistivity dust
        (2) energy saving
        (3) low initial cost
     Our  studies have resulted in the development and commercialization of a  new
technology of an "Intermittent Energization System" (marketed under the trade name of
the Mitsubishi Intermittent Energization System: MIE). A new high voltage power supply
equipment  has been developed for MIE incorporating modern electronic technology.


                           CONCEPT AND EQUIPMENT

     The  fundamental  concept of MIE  was found during a study of the relationship
between the voltage-current waveforms and the collecting efficiency . At that time, a
momentary high ESP voltage was encountered just after  switching on the power supply
to an ESP with a heavy back corona.

PRINCIPLE

     When high  resistivity dust  deposits on the collecting electrode,   it acts as an
insulation layer and, therefore, can be expressed by the equivalent circuit consisting of
a resistor  R and a  capacitor C as shown  in Fig.   la.   The transient  change of  the
voltage across the dust layer Vd is expressed by the following Eq. (1) when a constant
corona current (i=Io) has been applied to the dust layer from time t=0:

     Vd=R.I0.(l-e-t/to)                                                      (1)

where to is the time constant (t0=C x R ) and is calculated (1) by Eq. (2).

     t0=8.85-Er.Pd-10-14 (Sec)                                                (2)
where
     Er : specific dielectric constant of the dust layer
     Pd : specific resistivity of the dust layer

     For instance, to = 0.27sec  when   Er=3 and Pd=1012ohnvcm.  Fig. lb  shows a
graphical representation of the Eq. (1).  The  voltage Vd rises  gradually and  a  back
corona occurs when the electric field strength Ed = Vd/d (d: thickness of the dust layer)
exceeds its break-down electric field strength Edfc>.   That is;

                 ^ Edb                                                     (3)

                                       475

-------
     It takes time t[ from the start of current Io to the occurrence of a back corona as
shown in Fig. Ib .    The back corona starting time  ti is calculated by  Eq. (it) which is
obtained from Eqs. (1) and (3).

         t1=t0-ln(l/(l-Edb/(Pd-io)))                                           (4)

where  io is the current  density.   For instance,  t0=0.27 sec and ti=  0.109 sec when
Edb=10kV/cm, i0=0.3mA/m2, Er=3 and Pd=1012 ohm-cm.
     Therefore, by  deenergizing before  the  back corona starting  time tj and  re-
energizing  when the voltage Vd returns to the original low level,  ESP can be operated
suppressing back corona.  The  method of repeating energization and  deenergization at
given intervals is the basis of the MIE.
     Theoretically,  the dust migration velocity w is expressed by:

         w=k-E0-Ep                                                          (5)

where
     k   :    constant
     E0 :    charging field  strength
     Ep :    precipitating field strength
As Eo  and  Ep are  roughly proportional to  the peak  value Vp and the mean value Vm of
the ESP voltage respectively, the value of w is expressed by:

         wOCVp-Vm                                                          (6)

     Once a back corona occurs,  positive ions are  discharged from the dust layer into
the gas space  to  increase  the ESP  current  and reduce Vp.  However, Vp does  not
decrease by use of  MIE because it tries to interrupt  the ESP current before a back
corona  starts.   Furthermore,  even  if there are  periodical interruptions of the ESP
current, Vm does not decrease so much because of the equivalent capacitance across
the electrodes.  In consequence, w in Eq.(6) will increase by MIE.
     The effect of  MIE was  experimentally  demonstrated by a "visual model ESP"
which was  made with one set of collecting plates and discharge electrodes housed in a
transparent plastic case.   The photographies in Fig. 2   show the flying trajectories of
the dust.  In each photo, the collecting electrode is located at the left hand side and the
discharge electrode  at the right hand  side.  The collecting electrode was coated by a
mica sheet  to produce a simulated back corona discharge.  Pearlite particles were used
as dust  and, being fed from the top, were left to drift downward by gravity.
     Fig. 2a is for  a conventional energization  and indicates that the dust particles
come near to the collecting plate but are repelled, and, finally, go  down through the
ESP without capture.   Fig. 2b is for  MIE and clearly shows that most of the particles
migrate toward the  collecting  plate without repulsion.  These  two photographs simply
illustrate the increased efficiency of MIE over the conventional energization.

POWER SUPPLY EQUIPMENT

     As mentioned  above, the  method of  repeating energization and deenergization at
given intervals is the basis  of  the MIE.    This is graphically represented  regarding the
ESP current by Fig. 3.   Since the current flow time TI  ranges from several to tens of
milisecond and the current pause time T2 ranges from several to hundreds of  milisecond
, a conventional thyristor controlled power supply equipment fortunately can be utilized
with modification of the control circuit.

                                       476

-------
     Fig. 4 shows a block diagram of a power supply equipment for the MIE. The main
power  lines consisting  of  thyristors,  a high  voltage  transformer  and  high  voltage
rectifiers are the same as for conventional energization system.  The only difference is
the control circuit: it should be modified.   A  periodical blocking unit is added on the
signal line from the conventional control circuit to the thyristors . This unit blocks the
signal to the thyristors, periodically, so that the current to ESP  may be  shaped as shown
in Fig.  3.   TI and T2 are adjusted  by manual operating knobs or  automatic control
circuit.  The current flow level Ij is automatically adjusted.  For example, by using a
spark rate control or a current limiting control.   This addition, however, necessitates
the comlete rearrangement of the printed board.
     Typical  waveforms of  the  output  voltage  and  current photographed  on an
operating MIE power supply equipment are shown in Fig. 5.  Fig. 5a is  for conventional
and Fig. 5b for MIE when the duty ratio Re is 1/3 .   The duty ratio Re is introduced as
ratio of Ti/(Ti+T2) assuming Ti and T2 are expressed by numbers of half-cycle.   For
instance, the duty ratio Re is 1/3 when Tj=two half-cycles and T2= four half-cycles.


                                   TEST PLAN

     Laboratory tests and gathering the operating data from full-scale ESPs were made
so as to  confirm the effectiveness and  obtain concrete  design guides for MIE  with
respect to its two primary aims, i.e., performance improvement and energy saving.
     The tests were carried out under the following conditions for MIE  :-
         a) Current-flow time TI  : one to four half-cycles
         b) Duty ratio Re        : 1 to 1/20
         c) Peak current  value    : nearly rated current level

TEST BUG WITH PILOT ESP

     Since adjustable  ranges of test conditions are  limited in the actual operating
plants, a coal  firing  test rig with a pilot ESP was used.   Fig. 6 shows the schematic
diagram of the test rig.  Primarily, the rig consists of a pulverizier, a  furnace, coolers
and an ESP.  The test rig is designed to fire pulverized coal, reproducing the equivalent
exhaust gas and ash that are generated from  commercial boilers.  The associated pilot
ESP consists of two fields and employs the same discharge and  collecting electrodes as
those of full-scale ESPs.
     Fourteen (1*) sets of tests were carried out firing each of seven  Australian coals
and three Canadian coals.

FULL-SCALE ESPS

     Eight (8) ESPs are now in commercial operation with MIE: five coal-fired thermal
power plants , two iron ore sintering plants and one oil-fired thermal power plant.  23
sets of the operating data were gathered in total.   The kinds of coal fired in the coal-
fired  thermal power plants were  two  American  coals,  two  Australian  coals,   two
Japanese coals, one Canadian coal, one Chinese coal, one South African coal.


                         TEST RESULT AND DISCUSSION

IMPROVEMENT OF COLLECTING EFFICIENCY
                                       477

-------
     The improvement  of collecting  efficiency  was evaluated  by the enhancement
factor  H that was introduced by Feldman (2).   The H value is defined as the ratio of the
modified migration velocity w^ for MIE to conventional and the wk is calculated by the
following equation:-
where
         7 : collecting efficiency
         A : collecting area
         V : volumetric gas flow rate
         m : exponent depending on inlet particle size distribution
            m=0.5 in this paper

      According to observation of variation of the  H values and V/I characteristics, it
was found that test results were roughly classified into three types as follows:-
      Type I  :  V/I characteristics show a steep increase; no voltage rising with
               increasing current due to heavy back corona.  The maximum efficiency
               stands at Re of nearly 1/3 to 1/5.
      Type II :  V/I characteristics situate between types I and III.   Back corona takes
               place to some extent. The  maximum efficiency stands at Re of nearly
               1/2 to 1/3.
      Type III:  V/I characteristics look normal ; no back corona is suggested from V/I
               characteristics. No improvement of efficiency is expected by MIE.
      Regarding the pilot ESP, one representative result of each type is listed in Table.
1.  Regarding full-scale ESPs, one representative result of each type is listed in Table.
2 for the coal-fired thermal power plants.   One test result each of an oil-fired thermal
power plant as well as an iron ore sintering machine are listed in Table. 3.  Their V/I
characteristics are shown in Fig. 7-   The enhancement factors are  graphically shown in
Fig. 10.

      The remarkable facts derived from the test results were as follows:-

(1)    The largest  H value of Type I was 1.47 for case H on a full-scale ESP and the
      same was 1.72 for case A in laboratory tests.  The largest H value of  Type II was
      1.09 for case F on a full-scale ESP  and 1.31 for case C in laboratory tests.  The H
      values of full-scale ESP seem to be nearly 20% less than those of laboratory test.
(2)    The Type classification  can be made primarily  by the dust resistivity, but not
      always.   It rather corresponds to the steepness of V/I characteristics.
(3)    The lower the voltage of V/I characteristics is, the larger the enhancement factor
      is.  For instance, referring to Tables  2 and 3 and Fig.  7, case H with a voltage of
      24 kv had the H value of  1.47 whereas case E with 30 kv had 1.16.
(4)    The classification is interconnected with the properties of coal as  stated below:-
      Type I  :  Low sulfur, low alkali overseas coals
      Type II :  General overseas coals except Type I
      Type HI:  High sulfur, high alkali domestic coals
      Therefore, performance for MIE can also be estimated more or less by the kind of
      coal and its properties.
(5)    Type III  is  anticipated  in  case of  the hot  side  ESP  because the resistivity
      decreases by increasing the operating temperature.   Certain data, however, also
      indicate the  Type II performance and  suggest occurrence of a back corona even in
      hot-side ESP.

                                        478

-------
ENERGY SAVING

     It is shown by Fig. 8 that the power consumption is nearly proportional to the duty
ratio.  Therefore, energy saving is easily estimated by the duty ratio with which MIE is
operated.   Type I has the most effective energy saving.   The saving amount is 70 to
90% .  Type II is 50 to 70%.
     It is  natural  that  the  power  consumption can  also  be reduced by decreasing
current in  conventional  energization.    Comparison of power  consumption  between
conventional energization and MIE are shown in Fig. 9.  Fig. 9a is for Type I and Fig. 9b
is for Type III.  It should be  noted that regarding energy saving, MIE in some cases is
effective even for Type III.  As is seen in Fig. 9b, very small degradation of w|< takes
place.   That is, energy  saving of 50 % is achieved with  a little  degradation of the
efficiency for Type  III.

VOLTAGE AND CURRENT WAVEFORMS

     Fig. 5 shows the voltage and current waveforms obtained from a full-scale ESP of
type I.   The value of Vp x Vm referred by Eq. (6) was calculated from  the waveforms
and are shown in Table 5 as well as the corresponding migration velocity w, which was
calculated from the measured collecting efficiency.  It is noted that the value of Vp x
Vm was remarkably increased by MIE with a corresponding increase of the w value.

     The effect of MIE is summarized in Table 5 as classified by Types.


                                  CONCLUSION

     It can be said that the MIE is recommendable for ESPs where a back corona takes
place more or less.   The modified migration velocity W|< is improved to 110-150% with
the power  consumption  reduced to 10-30%  in the favourable cases.    The  power
consumption can be reduced by 50% with a small degree of degradation in the collecting
efficiency even in cases where no improvement in efficiency is expected.  Moreover,
the initial cost for  MIE is rather small  since  the main power lines of the conventional
energization systems need no modification ; only the control circuit must be modified .
     The MIE was commercialized due  to the social needs for mitigating the degrada-
tion of ESPs in the coal-fired thermal power plants  and also for  the power saving . It
should  be  mentioned additionally that  this  was  also  made possible  by  the recent
advanced  electronic  techniques  utilizing  power  control  thyristors  and electronic
circuits, which are  capable of turning on and off a  large  amount of current at a high
speed with sophisticated control.
     At present, the developement of Mitsubishi Intermittent Energization System is
still in  the progress of refinement.  The main target of our current  research for MIE is
development of full  automatic control system utilizing microcomputer.
                                       479

-------
                                 REFERENCES

1.    Harry 3. White, Industrial electrostatic precipitation, Addison-Werley Publishing
     Co., inc. (1963)

2.    Paul  L.  Feldman  and  Helmut  I. Milde,  Pulsed  Energization for  Enhanced
     Electrostatic Precipitation in High-Resistivity Applications, Symposium on Trans-
     fer and Utilization of Particulate Control Technology, July 1978
                    TABLE 1. TEST RESULT OF PILOT ESP*
Cases
Kind of coal
Coal characteristics
Ash wt%
Combustible %
sulfur wt%
Gas temp. °C
of dust ohm -cm
Collecting efficiency
Conventional
MIE at Rc=l/2
MIE at Rc=l/3
MIE at Rc=l/5
MIE at Rc=l/10
Enhancement factor
Conventional
MIEatRc=l/2
MIE at Rc=l/3
MIE at Rc=l/5
MIE at Rc=l/10
Type for MIE
A
Canadian I

13.8
0.3
150
7x1012

88.45
88.75
92.01
94.13
89.55

1.00
1.02
1.37
1.72
1.09
I
B
Australian I

11.2
0.8
150
4.5xlOH

99.76
99.65
99.54
99.27
-

1.00
0.88
0.80
0.66
-
Ill
C
Australian II

11.6
0.6
150
1.1x1012

96.10
97.43
97.55
96.70
95.10

1.00
1.27
1.31
1.11
0.86
II
 Gas flow rate
( designed )
                  ,.nrv
                  500
                                       480

-------
 TABLE 2. MEASUREMENT OF FULL-SCALE ESP FOR COAL-FIRED BOILERS
Cases
Unit capacity
Designed conditions
Gas flow rate m-*N/H
Inlet dust «/m3\i
content g/nrPN
Outlet dust «/rr,3M
content g/nrPN
Kind of coal
Coal characteristics
Ash wt%
Combustible .<,,
sulfur wt%
Gas temp.
ofedustV1 y ohm -cm
(D5uos%8dt) '«*»•
Collecting efficiency (%)
Conventional
MIE at Rc=l/2
MIE at Rc=l/3
MIEatRc=l/5
MIE at Rc=l/10
Enhancement factor
Conventional
MIE at Rc=l/2
MIE at Rc=l/3
MIE at Rc=l/5
MIE at Rc=l/10
Type for MIE
D
185 T/H
190,000
20.0
0.08
Domestic
22.6
0.3
110
1.4x1010
14.0
99.97
99.95
99.90
1.00
0.83
0.72
III
E
185 T/H
190,000
20.0
0.08
Canadian II
10.2
0.2
113
4.5x1012
17.0
99.59
99.66
99.72
99.73
1.00
1.07
1.1*
1.16
-
I
F*
75 MW
280,000
0.44
0.2
Australian III
18.4
0.4
198
7.0xlOH
3.4
86.02
87.16
84.21
83.97
1.00
1.09
0.88
0.87
II
having other MC and ESP ahead
                              481

-------
TABLE 3. MEASUREMENT OF FULL-SCALE ESP FOR OTHER PLANTS
Cases
Unit capacity
Designed conditions
Gas flow rate rn^N/H
Inlet dust rt/r«3M
content g/nrPN
Outlet dust «/m3M
content g/nv*N
Plants
Gas temp. °C
Resistivity . 	 ^m
of dust ohm -cm
(D5uost%grr, »*•»
Collecting efficiency (%)
Conventional
MIEat Rc=l/2
MIEat Rc=l/3
MIEat Rc= 1/5
MIEat Rc=l/10
Enhancement factor
Conventional
MIEat Rc= 1/2
MIE at Rc=l/3
MIE at Rc=l/5
MIE at Rc=l/10
Type for MIE
G
150 T/H
165,000
0.25
0.05
Oil-fired
thermal
power plant
154
10*
0.25
97.45
96.97
96.55
94.14
1.00
0.91
0.84
0.60
III
H
262m2
512,000
1.00
0.03
Iron ore
sintering
plant
95 to 112
1012 to 1Q13
20.0
88.28
89.24
91.78
92.58
1.00
1.08
1.36
1.47
-
I
                        482

-------
                     TABLE it. SUMMARY: EFFECT OF MIE
Types
       Type I
     Type II
    Type III
Dust resistivity
 ( ohm-cm )
  High  resistivity
        to 1013)
Medium resistivity
  (1010 to 1012)
  Medium-Low
(less than
V/I
characteristics

Optimum „
duty ratio c
Enhancement factor
(at optimum duty)
Power consumption %
Coal
characteristics
\ / J

1/5 to 1/3 1/3 to ly
1.1 to 1.5 1.1 to 1
10 to 30 30 to 5
overseas
low sulfur general ove
low alkali
I
T ^ T

'2 1/2 to 1
.3 0.8 to 1.0
0 50 to 100
domestic
rseas high sulfur
high alkali
 Plants
•Coal-fired thermal  -Coal-fired thermal
   power plant        power plant
 •Iron ore sintering
      plant
                   •Coal-fired thermal
                      power plant
                   •Oil-fired thermal
                      power plant
                     •Cement plant
                   •Refuse incinerator
    TABLE 5. COMPARISON* OF VOLTAGE WAVEFORMS AND PERFORMANCE
Items
Re
Vp kV
Vm kV
VpxVm
w %
Conventional
energization
1
30
24
720
100
MIE
1/3
46
21
966
117
MIE/Conventional



1.34
1.17
  referring to Fig. 5
                                    483

-------
               Vdb
    I I i=lo

     R >  vd

 v1-
a. Equivalent circuit
    Figure 1.  Electrical characteristics of dust layer
a. Conventional                  b- MIE

   Figure 2.  Flying dust trajectories
                484

-------
                      Tl
        H

        0
         C
         0)
        u
                   Current-flow time   T2 :  Current-pause time
                        Figure 3.   Current  to ESP
Thyristors
High voltage
;transformer
High voltage
rectifier
      ESP
       Figure 4.   Block  diagram of MIE power supply equipment
                                485

-------
                      kV
                      60

                      40

                      20

                      0
                      mA
                      3000
                      2000

                      1000

                      0
             t (10msec/div)
                       kV
                      60

                      40

                      20

                      0
                       mA
                      3000
                      2000

                      1000

                      0
             t (20msec/div)
   a. Conventional
b.  MIE  (Re = 1/3;
       Figure 5.   Voltage  and current  waveforms
       Furnace
    Coal
Figure  6.   Schematic diagram of coal  firing test rig
                           486

-------

1.5





1.0



0.5





0
0.4
-





_


1
/


i
i
i "~
A |C ^03
! £
1 •=
1
1 &
1 "A
, . c 0.2
' / OJ
1 / «
' / B S
' / £
| / M
; / R o-i
/ I/
\ Si
) />
^-''-'.
















J
D /
E


H


/
/
/
/
/
/ /
;
/ F /
/ /
/ /
/ / G
/
(*
/
/
/ /' /''
/y
^.->Z''
10 20 30 40 10 20 30 40
Voltage (kV) Voltage (kV)
a. Pilot ESP b. Full-scale ESP
               Figure  7.   Gas load V/I characteristics
     100
5
ft    50
          1/10 1/5   1/3     1/2
                  Duty ratio Re

        Figure 8.  Power consumption
                        487

-------
-p.
00
00
                        1.0
                       0.5
                       1.0
                                         Conventional
                                                 50



                                       Power consumption ratio (%)


                                           a. Type I  (case E)
                                                                      100
                                         HIE
                                              ' Conventional
                                                 50                    100


                                       Power consumption ratio (%)


                                           b. Type m  (case D)
                       Figure 9.   Comparison of  energy  saving
                                                                                               1.5
                                                                                               1.0
                                                                                               0.5
                                                                                                        /

                                                                                                       /      B
      1/10 1/5   1/3     1/2


               Duty ratio  Re


               a.  Pilot ESP
                                                                                               1.5
                                                                                               1.0
                                                                                               0.5
     1/10 1/5   1/3     1/2                     1



               Duty ratio  Re


               b. Pull-scale ESP



Figure 10.   Performance  improvement of MIE

-------
            SOME  MEASURED  CHARACTERISTICS  OF AN ELECTROSTATIC
         PRECIPITATOR OBTAINED  USING A MICROCOMPUTER CONTROLLER

              by:   M.J.  Duffy,  T.S.  Ng,  Z.  Herceg  and K.J. McLean
                   University of  Wollongong
                   Wollongong,  N.S.W. Australia. 2519.
                                    ABSTRACT
     This paper presents a new microcomputer control  system for  electrostatic
precipitators.   The system, currently in its second stage  of development,  is
capable of actively controlling,  monitoring and recording  the characteristics
of an electrostatic precipitator.

     The results of tests carried out on the precipitator  along  with  the out-
put of a dust monitor mounted on  the precipitator  output are also presented
in an attempt to determine the optimum operating conditions.

     The work described in this paper was not funded  by the U.S.
Environmental Protection Agency and therefore the  contents do not necessarily
reflect the view of the Agency and no official endorsement should be  inferred.
                                    489

-------
                                INTRODUCTION
     Electrostatic precipitators are one of the most efficient and economical
means of removing particulates from effluent gasses in large coal-fired power
stations.  The basic principles of operation, design and application of
electrostatic precipitators have been well documented (1, 2 and 3) and so
will not be discussed here.

     Since Australian coals are low in sulphur content the resultant fly ash
has a high resistivity.  It is this high resistivity which adversely affects
the performance of the precipitator.  In particular the  'back corona' formed
on the collecting plates has the effect of reducing the spark-over voltage,
increasing re-entrainment and reverse charging the suspended particles by
positive ions injected into the gas stream.  The consequence of these effects
is to reduce the precipitator's performance.

     The most efficient operating conditions within the precipitator have in
the past been largely a matter of 'ballpark' approximation relying upon
occasional adjustment of the automatic analogue control system.  This method
of operation is adequate provided that the conditions within the precipitator
remain static.  However as internal conditions vary with unit load, the
degree of contamination and the type of coal used, this method becomes
inefficient.

     With the development of relatively cheap C.P.U.  and memory chips the
use of microcomputers is becoming more pronounced in precipitator control
as pointed out by Hall (4) for a variety of control and diagnostic purposes.
The computer is able to react to changing operating conditions as they occur
resulting in the theoretical maximum possible efficiency being obtained (5) .

                             PROJECT DEVELOPMENT
     The development of the Computer Controlled Electrostatic Precipitator
System (C.C.E.P.) was proposed in three phases. The initial phase required
that the computer replace the existing automatic analogue control system (6) .
By discretely sampling the precipitator zone voltage, current and spark rate
the computer is able to supply an appropriate control input to the zone
based upon a simple algorithm.

     In the next phase the precipitator conditions are recorded and incorpor-
ated with the output of a dust monitor, mounted in the precipitator output,
and stored on cassette tape.  This information is later used to evaluate the
relative performance of the computer and analogue control systems, to quantify
the voltage-current characteristics of the precipitator in normal operation
and in static tests as well as in comparing the results obtained using
different control algorithms.  It is the purpose of this paper to report some
of the results achieved to date, explain the capabilities of the system and
detail the projected development.
1.  C.P.U. Central Processing Unit, e.g., 6502.

                                     490

-------
     Having characterised the effects of different control strategies on
precipitator performance the final phase will be to incorporate the, input
from the dust monitor as well as the past history of the precipitator in an
adaptive control algorithm using a high level language like BASIC for the
manipulation of the information and primitive ASSEMBLY routines for input-
output purposes.

   The system to be described in this paper is currently installed in one
pass of a four pass electrostatic precipitator at the Electricity Commission
of New South Wales Munmorah Power Station, New South Wales, Australia.
Subject to satisfactory performance it will be extended to include the
remaining passes in a complete computer network system which will be capable
of informing an operator of the performance of his precipitator using
pictorial and numerical displays.

                        DESCRIPTION OF THE SYSTEM
   Figure 1 shows the block diagram of the power supply and existing control
hardware in one zone of the electrostatic precipitator.  The saturable
reactor is used as a control element to vary the primary voltage to the high
voltage transformer.  The reactance of the reactor is controlled by the d.c.
current in its control winding.  It is this d.c. control current that is
indirectly varied by the existing analogue control board using feedback
supplied from the precipitator voltage, current and spark rate.

   The control system (C.C.E.P.), shown within the dashed box of Figure 1,
replaces the existing analogue inputs on the control boards of the five
zones by changing over the contacts of a relay.  These contacts are control-
led by a count down timer which is refreshed periodically, while the computer
is in control of the precipitator, so that it never times out.  In the event
of a power failure, the hardware ensures that control is automatically passed
back to the analogue control system.

   The microcomputer itself uses a 6502 C.P.U. chip together with twenty
kilobytes of memory and six, eight bit input output ports which are
connected to interfacing hardware as shown in Figure 2.  The hardware
comprises buffer amplifiers for each zone voltage and current together with
an analogue multiplexer and a high speed analogue to digital converter (A/D).
The spark rate is buffered and recorded separately using digital counters
which are read periodically by the computer.

   Control of the precipitator is achieved by using eight bit latches and
digital to analogue converters (D/A) for each zone.  The D/A's are connected
to the zone through buffer amplifiers which scale the input so as to vary the
precipitator secondary voltage in two hundred volt steps.

                             CONTROL ALGORITHM
   Until now, all of the control on the precipitator has relied upon
emulating the present analogue control system.  That is, to control the

                                    491

-------
                                               PRECIPITATOR
         TO ZONES
         I, 3, 1
         AND 5
COMPUTER
CONTROL
FROM ZONES
2, 3, H
AND 5
Figure  1:   Power Supply and Existing  Control Hardware
            with Microcomputer System  Attached.
1 ZONE 1 1 ZONE
D/A D/A




LATCHES LATCHES







5 n 5




2







IZONE
D/A



LATCHES
i-




6


JF^DUST
V I
BUFFERS
b

MULTIPLEXER
111
A/D
LI








MONITOR




3
L

IZONE
D/A




LATCHES




hi
.

8



BUFFER


COUNTERS



BUFFER


„]


4







IZONE
D/A



LATCHES
•

8







5


-r
V
•5
ZONE
CURR
10


CPU

MEMORY

I/O PORTS

\


60
T
I
5
F
s ,
VOLTAGES
5
}


Figure  2:   Microcomputer with Interface Circuitry.
                        492

-------
precipitator using the spark rate as the reference variable with specific
limits being imposed upon voltage and current.

   The current control algorithm, shown in Figure 3, requires the computer
to increase its control voltage by one step until the precipitator spark
rate reaches a reference value.  If the reference is exceeded the computer
will back off the voltage.  If, however, the reference is exceeded by a fixed
amount, the computer will assume that the zone is sparking heavily and so
will back off the voltage more vigorously and then set a flag.  On the next
pass through the zone, approximately one minute later, the computer will
examine the flag.  If it is set and the excess sparking has ceased the
computer will restore the control voltage to a few steps below its former
value.

   It was hoped that this strategy would maintain zone voltages close to
their maximum vlaues for more efficient dust collection while, at the same
time, avoiding long time delays while the computer recovered from excessive
sparking.  Field test results tend to support the validity of this
argument.
               Figure 3:  Flow Chart for Control Algorithm (1 Zone)
                                    493

-------
                             RESULTS  OF  TESTS
     A typical plot of one zone's performance is shown in Figure 4 with the
spark rate plot shown in Figure 5.  From these results it is easy to see that
the computer is indeed capable of regulating zone spark rates although
this is not the case for spark rates in excess of twenty sparks per minute.
This appears to be due to the instability caused within the zone due to the
excess spark rates.

     Having plotted the performance of a control algorithm the next step is
to determine the optimum spark rate.  To enable this the computer can be
programmed to dynamically increase the reference spark rate by a fixed amount
at fixed intervals of time.   Again the performance is evaluated by using the
output of the dust monitor connected into the output of the pass.

     This test was carried out again by taking the precipitator voltage, on all
zones,below the spark-over voltage and manually increasing the voltage to a
point of heavy sparking.  The computer printout for this case is shown in
Figures 6 and 7 for one zone.

     In both cases the results, although at this stage preliminary, showed
that the dust output is fairly independent of spark rate over a wide range
of values.  This may, however, be due to excessive dust contamination as
earlier tests indicated a much greater sensitivity.  Peaks in dust output
occur for voltages below the spark-over potential and at spark rates in
excess of one hundred sparks per minute.  This may be observed in the dust
monitor output plot for the above case as shown in Figure 8.

                                 CONCLUSIONS
     While at this stage no definite conclusions regarding precipitator
optimum performance may be inferred it is reasonable to say that the
microcomputer control system has demonstrated its versatility in a wide range
of operating environments and proved superior to the existing analogue control
system in several instances.  That is not to say that the system has proved
to be the ultimate in precipitator control either, but over the test period,
it has provided much valuable information which was previously unobtainable.
This information must ultimately benefit the research and development of
electrostatic precipitators and hopefully lead to a better understanding of
their internal operation under many diverse operating environments.

                              ACKNOWLEDGEMENTS
     The author gratefully acknowledges the financial assistance provided
by the National Energy Research, Development and Demonstration Programme
administered by the Commonwealth Department of National Development and
Energy as well as the assistance, both financial and material, provided by
the Electricity Commission of New South Wales and its officers.
                                    494

-------
ISTflRT TIME:  01:56: 1/9/1982    HONE 3 PBGE  05
  0.   60.  120.   180.  240.  300.
                 TIME  (niln.)
                    J
         3G0.  420.  480.
                                                 Key
                                                     Average  Current
                                                     Peak Current
                                                     Average  Voltage
                                                     Peak Voltage
                                                     Computer Control
                                                     Signal	
 Figure 4:  Voltage, Current  and Control Signal of Precipitator
            with Microcomputer  in Normal Operation (1 Zone).
 Key

 	 Spark
     Rate
 	 Reference
	Spark Rate
a
a.
in
                                 60.  120.  180.  240.  300.  3G0.  420.  488.
                                             TIME  (rain.)
 Figure  5:   Spark Rate (sparks/minute)  During the Same Period.
                               495

-------
Key
    Average Current
    Peak Current
    Average Voltage
    Peak Voltage
    Computer Control
    Signal
     260
     240
     220
     200.-
    -180.-
    tlBB.J
    o
                              STRRT  TIME: 13:2
                                                2/B/1982
                ONE ,2 PRCE  03
        I!    i  i .,
••}-•£•	Ji..,.J  |"--4
        ii    r i
 ii  I    ''•'    U-!    I
•Ml--IV" 'If Li.  -(!
                                           60.           120.
                                              TIME (rain.)
   Figure 6:  Voltage, Current and Control  Signal of Precipitator
              with Microcomputer in Manual  Operation (1 Zone).
                                                   180.
4\if
38.-
36-
34.-
32*-
30-
_ 28.-
" 26-
^ 24.-
* 22-
« 20-
<= 18.-
•* 16--
» 14.-
« 12r|
"a
43
ll
5TBRT TIME: 13:25:















	 1
. . . ,fhi


















k/i

















83 Z
















,
DNE

















3) PflCE 04
















1
    "0.
                                                        Key

                                                        	 Spark
                                                            Rate
                                                        	 Reference
                                                            Spark Rate
60.           120.
   TIME  (rain.)
                                               180.
      Figure 7:  Spark Rate (sparks/minute) During  the Same Period.
                                   496

-------
                                   POGE  11
 STRRT TJ.ME: 13:25:  2/9/1982
 0.
60.           120.
   TIME  (ntln.)
                                          180.
                                                   Key
                                                       Averaged Dust
                                                       Output
                                                       Instantaneous
                                                       Dust Output
Figure 8:  Dust Monitor Output (grams/m3) During the  Same  Period.
                              497

-------
                                 REFERENCES
1.  White,  H.J.   Industrial electrostatic precipitation,  Reading,  Mass.,
    Addison-Wesley,  1963.

2.  Hall, H.J.   Design and application of high voltage power supplies in
    electrostatic precipitation,  J.  Air Poll.  Cont.  Ass., 25:  132, 1975.

3.  Oglesby,  S.  and  Nichols,  G.B.  Electrostatic precipitation,  New York,
    M. Dekker,  1978.

4.  Hall, H.J.   High voltage power supplies  and microprocessor controls for
    electrostatic precipitation.   In:  Proceedings of International Conference
    on Electrostatic Precipitation,  Monterey,  CA, 1981.

5.  Schummer, H.W.  Process and energy optimization  of precipitator plants
    using microcomputers.   In;  Proceedings  of International Conference on
    Electrostatic Precipitation,  Monterey, CA, 1981.

6.  McLean, K.J., Ng, T.S., Herceg,  Z.  and Rana, Z  A new microcomputer
    system  and strategy for the control of electrostatic  precipitators.
    Paper presented  at 3rd EPA Symposium on  Transfer and  Utilisation of
    Particulate Control Technology,  March 9-13,  1981.
                                    498

-------
           ELECTROSTATIC PRECIPITATOR ENERGIZATION MD CONTROL SYSTEMS

               by:   K. M. Bradburn and K. Darby
                    Lodge-Cottrell Dresser
                    Houston,  Texas 77002
                                    ABSTRACT

      This paper reviews the development of systems for the energization and
.control of electrostatic precipitators, ranging from heavy duty thyristors
 used to control the input to the T/R set to the use of microprocessor tech-
 nology which gives the facility for a precipitator total energy management
 system (TEMS).   The impact of these changes is described first on the T/R
 set, then on the automatic voltage control system to maximize precipitator
 efficiency.

      Reference  is made to the electrical demands of a precipitator,  under
 varying operating conditions, to justify the advantages of the TEMS  concept.

      The system comprises a local control and power unit for each T/R set,  a
 supervisory control unit and a control room monitor which allow visual (CRT)
 display and print out of operating data and parameters remote from the
 precipitator.

      In addition to improved automatic voltage control, TEMS offers  the
 facility for integrating additional operations and parameters such as stack
 opacity,  rapping,  hopper heating, hopper ash level alarm,  precipitator start-
 up and shutdown.  This results in complete system control and improved
 operation with  lower power consumption and operating costs.

      Using a telephone modem, precipitator engineers remote from the plant
 can accurately  and instantaneously monitor and adjust precipitator operation
 and give technical support to plant engineers for prompt correction
 of any operating problems.
                                     499

-------
INTRODUCTION

     Nearly one hundred years ago, Sir Oliver Lodge in England made the first
attempt to apply electrostatic precipitation to an industrial application.
Inadequacies in the power supply equipment, however, severely limited the
success of this first installation.  To commemorate this significant
occasion, the Coutonwealth .Scientific and Industrial Research Organization of
Australia 2nd Electrostatic Precipitator Conference will be held in 1983, the
centenary year of electrostatic precipitation.

     It was not until 1907 that Frederick Cottrell used the then newly
developed synchronous mechanical rectifier as the power source for the first
commercial precipitator.

     Throughout the years both mechanical and high vacuum tube rectifiers
were commercially used in precipitators, but the associated system was
limited to one of manual operation.  It was not until the late 1940's that
the first automatic transformer/rectifier control became a reality.

     Another decade passed before high voltage silicon rectifiers and sat-
urable reactors came into common use.  These components remained the basic
power equipment until the emergence, in the mid 1960's, of modern SCR
(thyristor), linear reactor and solid state automatic voltage control.

     Today's precipitator control systems have now entered the computer age
with the use of microprocessors.  Such technology has provided the oppor-
tunity for the establishment of a fully integrated precipitator control
system.

     One particular area of interest which the microprocessor system in-
cludes is the ability to use an opacity feedback signal to adjust the power
consumed by the precipitator, thereby providing a basis for some power
economy when precipitator operating conditions are more favorable than the
design conditions.

     This Paper gives an illustration of the power/opacity relationship for a
full scale precipitator and indicates the level of power savings which may be
possible.  In addition, a description of a microprocessor system is given
which incorporates the facility to realize such saving as one of its many
features.

PRECIPITATOR PERFORMANCE - POWER CHARACTERISTICS

     Until microprocessors became available for control purposes,, electro-
static precipitators were designed with a T/R set for each electrical bus
section and usually incorporated an automatic control, either in the form
of a flash or spark counter or, in the case of Lodge-Cottrell, a device which
set the mean discharge electrode voltage at the highest possible level.  The
aim of such controls was to ensure that at all times and under varying condi-
tions, the precipitator was giving the maximum possible efficiency.
                                    500

-------
     The Lodge-Cottrell system operated on the principle illustrated in
Figure 1.  This shows that if the electrical input is set at a condition
where the mean discharge electrode voltage, including the effect of the loss
of field voltage during flashover, is at the maximum level possible, then the
effective migration velocity, hence, the particulate removal efficiency will
be at the highest level.  In the original patent for this type of control,
recognizing that opacity would also be at a minimum for maximum particulate
removal efficiency, an alternative to the use of the maxiinum value of dis-
charge electrode voltage was included, whereby the control was achieved by
using a signal from an opacity meter; in this case the system aimed at mini-
mum opacity.  The minimum opacity and the maximum mean value of discharge
electrode voltage were shown to correspond to the same condition of precipi-
tator efficiency.  Both systems had the advantage that the absolute accuracy
of the control signal input was not critical, since the control system would
seek maximum voltage, or minimum opacity, regardless of the absolute value.

     The overriding design requirement of an electrostatic precipitator is
that it shall meet the environmental particulate emission standards under the
most onerous of its design conditions.  These may be summed up as repre-
sented by the maximum gas flow rate which the precipitator is required to
treat, the maximum inlet particulate concentration and the properties of ash
and gas as determined by the properties of the fuel.  The automatic voltage
control system will for any of these conditions seek to give the best
possible particulate removal efficiency under the conditions existing in the
precipitator at any particular time.  For the different conditions of gas
volume, particulate properties, etc., the voltage and current corresponding
to the optimum efficiency vary considerably, and in many cases, the removal
efficiency and the power applied will be considerably greater than that
needed to satisfy particulate emission requirements.

     Figure 2 illustrates the characteristics of a typical precipitator
system attached to a 120 MW power station boiler.  Each precipitator had
two chambers in parallel and three electrical fields in series.  The graphs
show a number of important characteristics of the plant under a wide varia-
tion of load conditions.  Note first the central Graph B, which shows the
D.C. power input to one precipitator chamber.  It will be seen that for a
gas flow from below 300,000 to one in excess of 400,000 cfm actual, there is
actually a slight decrease in total power absorbed for optimum precipitator
efficiency obtainable at those flow rates.

     Refer next to Graph A.  This shows the power input per unit of gas
treated.  It will be seen that from the full load to the lowest illustrated,
there is approximately a 50% increase in specific power usage.  Graph C shows
the mass emission of particulate.  This reduces by 75% due to the longer
exposure of the gas to the electric field.  While this reduction in mass
emission may seem a very desirable situation, it should be noted that the
actual power absorbed has increased, if only by a small percentage, and the
specific power input, the power per unit volume treated, has increased by
more than 50%.  Bearing in mind that the particulate emission at the full
load condition was, in the case of this particular plant, already well within
the limit set by the environmental laws, then it could be argued that there

                                    501

-------
was little point in reducing the mass emission  at  the  expense of very sub-
stantial increases in specific power.  The mass emission could have been
maintained at a constant value as the load diminished  with  a substantial
saving in electric power by reducing the power  input to the precipitator.

     The implication of the magnitude of possible  saving is illustrated by
the table below.  This shows the electrical  supply system for a  600 MW boiler
designed to  fire U.S. midwestern sub-bituminous coal to give particulate
removal of 99.85% efficiency.  Each of two precipitators has 6 fields in
series and 36 isolatable bus sections.  The  important  numbers here  are the
power rating of the T/R sets, the total installed,  capacity  is 4220  kVA. A.C.
or 2970 KW D.C.  While in practice the absorbed power  of the precipitator
system may only be 70% of the installed capacity,  this represents a signifi-
cant portion of the output generated by the  boiler and turbine.

                                    TABLF

     Boiler  Output                            600  MW
     Coal Fired                               Midwestern Sub-bituminous

     Precipitator Installation

         Design Efficiency                    99.85%
         Number of Precipitators              ,?
         Number of Fields                     6
         Number of Isolatable Bus Sections    72 Total

      T/R Sets

         Number of Sets                       72
         Design Rating                        36 at 55 kV,  600 mA.
                                              36 at 55 kV,  900 mA
         Total Rectifier Installed Capacity   2970 KW  D.C.  - 4220 kVA A.C.

     To demonstrate the possibility of power saving, a program of tests was
carried out  on a 120 MW boiler at Agecroft Power Station in the  U.  K.

     Each precipitator had two chambers in parallel and three electrical
fields in series, and the tests were carried out at full boiler  rating.  The
coal fired was a. hard bituminous coal of a relatively  high  sulphur  content
(1.8%) and was such that it gave very favorable precipitation conditions,  in
fact a typical condition which is normally associated  with  a high absorbed
corona power.  The precipitation characteristics of the ash were much better
than that on which the design was based.  With  rectifiers on all three fields
operating at optimum, power consumption was  approximately 23 kVA and opacity,
that is base or mean opacity, was about 5% corresponding to a particulate
emission of  15 mg/Nm  .  Apart from the mean  opacity there are periodic
short duration spikes on the opacity meter trace due to reentrainment when
the plates are rapped.  The maximum height of spike results from the rapping
of the outlet field.  The effect of reentrainment  is included in the emis-
sion concentration quoted.
                                       502

-------
      Two series of tests were performed.   In the  first  series of tests  illus-
 trated in Figure 3,  each of the three fields in turn was  reduced in power
 input until it approached  the minimum possible with the other two fields
 maintained at optimum.  This resulted in the opacity increasing only to 10%,
 equivalent to 35 mg/Nn , with a reduction in power input  of over 60%.   The
 significant difference was that the spikes due to rapping reentrainment were
 minimum when power reduction was carried out only on the  inlet, and greatest
 when  on the outlet field.

      In Figure 4 the same  series of tests were repeated,  but in this case one
 out of three fields was maintained at optimum power input, while the other
 two fields  were gradually  reduced by the sane amount.   Under these conditions
 the opacity increased to a maximum average value  of 20%,  for a  reduction in
 power input of 80%.   Thus  the guarantee for this  plant, which required  an
 opacity of 20%,  would still have been met with only 20% of the  power input
 which the  automatic voltage control systems would normally feed to the
 system.  Of particular interest here is the greater amplitude of the rapping
 spikes.  These approached  almost total opacity when either the  center and
 outlet or  the inlet and outlet  fields were reduced.  The  maximum height of
 spike again resulted from  rapping of the plates of.the  outlet field.  While
 of considerable amplitude,  in a large system,  since they  are only a few
 seconds  in duration,  spaced a considerable time apart and occur in only one
 section at a time, when mixed with the gases from other units,  the effect of
 even  the large rapping spikes did not result in any visible plums at the
 stack.   Even so the evidence that these large spikes occur could be a reason
 for limiting the power reduction.

      It can be seen from Figure 4 that it would be better, if such an operat-
 ing approach were adopted,  to limit the reduction of voltage to the first
 two fields  since this reduces the rapping spikes  to roughly half of the amp-
 litude of  the other two conditions.

      This particular operating  condition with an  extremely favorable fuel
 for the  precipitator, which in  practice gave migration  velocities in excess
 of 13  cm/sec., probably represents close to the upper limit of  power that
 can be saved by reducing the performance such that the precipitator main-
 tains  the guaranteed opacity or mass emission at  full boiler rating.

     The reason this extreme power change takes place is  illustrated in
 Figure 5(A).   This  shows the variation of corona  current  with increasing
 discharge electrode  voltage, and it will  be noted that  the upper half of the
 curve  shows  a  doubling of  corona current  for relatively small increases in
 discharge electrode  voltage.  In an electrostatic precipitator, very little
 of the current actually flowing is used to charge the particulate, but  is
 transported  across the interelectrode space in the form of ionized gas.
According to precipitator  theory,  once the particles have acquired their
 limiting charge,  then the  rate  of removal from the gas  stream should be
 roughly  proportional to the square of the voltage applied to the discharge
 electrodes.  This is illustrated in Figure 5(B) for the conditions recorded
 in Figure  5(A),   it will be seen that there is an approximate linear rela-
 tionship between the square of  the electrode voltage and  the opacity.   The
                                    503

-------
opacity increased from 5% to 15% for a 60% decrease  in the  square  of  the
applied voltage.  The rapidly increasing corona current,  as the applied vol-
tage is increased, contributes little to the efficiency of  the  precipitator.

     It is stressed that this particular example  is  for a type  of  fuel where
the maximum economy and power input is possible due  to the  very good  precipi-
tation conditions which would be encountered.  There would  be some power
economies for most fuels.  If particulate mass emission rather  than opacity
is regarded as the regulating condition, since for the same opacity at re-
duced boiler loads the mass emission reduces proportionally to  the gas flow,
then further power reductions are possible.  Illustrated  here,  however,  is a
way in which the use of microprocessor control can substantially reduce the
power consumed.  If the control system includes the  input of a  signal from
the opacity meter, then this can be used to override the  tendency  for the
automatic voltage control systems to strive for maximum power and  maximum
efficiency while still maintaining a constant stack  opacity sufficient to
meet environmental regulations with the advantage of very substantial econo-
mies in operating costs.

     The tests described above were, of necessity, of short duration,  and
conditions could change if the plant were operated for a  substantial  period
at reduced power input; that is, more than a few  hours, at  any  one of the
conditions shown.  For example, with the precipitator operating at the
optimum value, there is a considerable rate of spark-over within the  preci-
pitator; and it has been observed by the Authors  that if  the rate  of  flash-
over is maintained at a sufficiently high level,  then the efficiency  of the
precipitator can also be maintained for a considerable period even without
rapping.  On one installation this was demonstrated  over  a  period  of  seven
days with the rapping switched off.  This results from the  fact that  each
flashover, which is the discharge of the electrode formed condenser,  is the
equivalent of a mechanical blow on the collector  and discharge  system.   In
another paper presented at this Symposium, this will be discussed  in  some
detail, but in practice rapping forces of the order  of 20 to 30 'g' were
measured in the region of the plate corresponding to the  flashover.   The
implications of this are important since with a precipitator field being
subjected to random flashover as would be normal  with a well aligned  precipi-
tator, each of the plates is being subjected at some point  or other,  due to
the flashover, to blows sufficient to remove at least the freshly  deposited
ash.  This means that in contrast to the idea that the thickness of the ash
layer on the plates builds up progressively between  the mechanical rapping
blows, there is a continual cleaning process of the  collector system,  with
the result that though the mechanical rapping is  cyclic,  the variation in
cleanliness of the plate between rapping blows does  not vary as much  as might
be expected.

     Figure 6 shows photographs of a laboratory rig  built to demonstrate the
'effect of flashover on a layer of electrostatically  deposited ash  about 5
mm. thick.  The area over which ash is removed is, as expected,  quite large
relative to the spark point contact.
                                    504

-------
     It would appear, therefore, that the method of power economy whereby the
input to the fields is reduced so that flashover is virtually eliminated,
could have undesirable side effects from the point of view of plate build-up
and the effect of this build-up on the electrical field and hence, precipita-
tor efficiency.  This would of course be dependent on the rapping efficiency
of the mechanical system also, but this could not conpletely eliminate the
gradual build-up effect between rapping blows which would be reduced by the
flashover effect.  An alternative to gradual reduction of power input would
be to switch off sections or fields in sequence so that the resulting opacity
still achieved the required level.  This was the basis of the second series
of tests in which different numbers and combinations of fields were switched
off.  This is obviously a much coarser control than the first series and with
the three fields of the unit tested, there was very little flexibility
possible in the system.

     On this three field system, switching off the inlet field showed an
opacity increase of from 5 to 10%, while switching off of the inlet and
center fields increased the opacity to 35%.  Any combination of deenergizing
fields which included the outlet field, had a much greater effect on the in-
crease in opacity; and it would seem logical, therefore, that with such a
system, deenergization should begin with the first field.  On the six field
precipitator, for example, it could be possible to switch off two or more
fields and still meet the required opacity at low loads and favorable condi-
tions.  This could halve the power absorbed by the precipitator if environ-
mentally acceptable.

     In the various tests described above, although all of short duration, a
period, of weeks was needed to collect the information given.  It is prefer-
able to repeat these tests on a variety of other types of coal ashes, includ-
ing the midwestern sub-bituminous widely used here in the United States, and
the very highly resistive ashes which result from the firing of hard bitumi-
nous coals in New South Wales, Australia.  The facility exists for extending
the role of the microprocessor to control precipitator corona power such that
the opacity or mass emission is maintained at a constant value with substan-
tial reduction in operating costs.  The correlation of mass emission and
opacity can be obtained from performance testing.  In the tests described
above, the rapping frequency was not varied as fields upstream were turned
off or reduced in efficiency.  The microprocessor is capable of making
changes to the rapping cycle to compensate for the variation of particulate
concentration entering the field.

ENERGY MANAGEMENT SYSTEM

     The advent of microprocessor technology has afforded the opportunity to
not only give improved transformer/rectifier control and monitoring but also
the facility to integrate many additional operations, parameters and fea-
tures, giving a total energy management system.

     Some of the features of such a system are:

     a)  Improved transformer/rectifier control and precipitator performance


                                    505

-------
     b)   High level system reliability

     c)   Convenient remote control,  monitoring and data logging

     d)   Closed loop control of stack opacity

     e)   Integrated rapping frequency control

     f)   Hopper ash level monitoring

     g)   Hopper heating monitoring and control

     h)   Maintenance scheduling and system fault diagnosis

     i)   Modem coinmnications

     The Lodge-Cottrell energy management system which has the facility to
incorporate such features comprises:

     1)   Local automatic voltage controller and power units for each
         transformer/rectifier set

     2)   Supervisory controller

     3)   Boiler control room monitor.

    The overall system is designed to be tolerant of individual component
failures.  There is no critical component in the system.  Failure of an
individual Local AVC Unit or Power Unit will result in the loss of one T/R
set.  Upon the failure of a Supervisor Unit, the associated Local Units
automatically revert to local operation.

LOCAL AVC UNIT

     The primary component of the system is the Local Automatic Voltage
Control Unit, which processes all precipitator signals from an electrical
section, provides local display and annunciation capability, provides switch
inputs for the user, conmunicates with the Supervisor Unit, receives ALL "ON"
and ALL "OFF" connands from a remote source and provides the control signal
required by the Power Unit.

     The Local Unit employs two independent microcomputers.  One is used to
perform all control functions, the other is used to derive the signals for
front panel display.  The. software for these two computers has been designed
to be self-starting.

     The automatic voltage control strategy employed by the microprocessor
AVC is more advanced than earlier designs, while maintaining the same funda-
mental control strategy.  This is to maintain the highest possible average
T/R unit output voltage, under all operating conditions, except when func-
tioning with an opacity signal override.


                                    506

-------
      In the automatic operating mode,  the AVC has a quick run up feature
which operates when the unit is brought into service.   The T/R unit output
voltage increases at approximately four times the normal rate until the first
drop in the average voltage is sensed.   Thereafter,  the normal control stra-
tegy takes  over.

      Automatic current limiting is provided to protect the T/R set and preci-
pitator components and predominates over both automatic and manual control.
Unlike  earlier designs that sense the  average value of the T/R unit primary
current, this  AVC Unit senses the rms  value.

      The rms value of the  T/R primary  current is digitally filtered to pro-
vide optimum closed-loop stability while the system is current limiting.

      The T/R voltage divider signal is digitally filtered to provide optimum
performance.   The parameter RESPONSE is used to control the filtering pro-
cess, matching control performance to  the characteristics of a specific
application.   In  addition,  the parameter STEP is used  to set the sensitivity
of control.

      The manual control mode is functionally identical to that provided by
earlier AVC's. ' When the system is in  the MANUAL mode,  the T/R voltage can be
manually raised or lowered by the operator.

      Each Local AVC Unit (Figure 7) continuously displays five basic operat-
ing  values:  secondary bushing voltage  (kv),  secondary  bushing current (mA),
line voltage (volts),  line current (Amps), power consumption (KW).   Parameter
settings can also be selected and displayed  as required and standard alarm
conditions  are indicated.

      In addition  to its basic function,  the  Local AVC  also performs rapper
control functions and implements an additional important new performance
feature: the ability to regulate the frequency of rapping associated with
the  electrical section the AVC regulates.

      The rapper control system is automatically adjusted to accommodate Local
Units that  are out of service,  i.e., the Local Control Unit will continue  to
operate even though its associated Power Unit or T/R set may not be capable
of operation.   Should a Local Unit fail completely,  the adjacent AVC in that
field would automatically  control the  rapping.   In addition,  the frequency of
rapping of  the downstream  field would  be adjusted accordingly for  the
increase in particulate loading.

LOCAL POWER UNIT

      The respective Power  Unit for each automatic voltage control  unit modu-
lates the line voltage supplied to the  T/R set in accordance with  commands
supplied from  that Local AVC Unit.   The Power Unit can be located  up to 1500
meters  away from  its associated Local AVC Unit.   The SCR power control ele-
ments in the Power Unit are conservatively rated and adequately cooled for
plant operating conditions.
                                       507

-------
     Phase modulation, as opposed to integral cycle switching, is employed to
provide the fine control required for optimizing secondary voltage.  A single
phase configuration is employed.

     The SCR's are of disc package construction and in order to provide
reliable operation in the rigorous precipitator application, exceptionally
high I2T ratings are used.  To protect the SCR, an R-C snutber network is
provided along with a high-power metal oxide varistor; in addition, a high-
speed fuse is provided to give ultimate overcurrent protection.

SUPERVISOR

     Broadly conceived, the Supervisor is a loop in the precipitator control
system  (Figure 8) that is closed by operator response to concisely presented
information.  As a complete system, it is a computer in communication with
up to 128 local Automatic Voltage Controllers which control the individual
sections of the precipitator.

     Local AVC units are "intelligent" in the sense that once they are pro-
grammed with a set of initial parameters, they will independently control
their bus sections.  The addition of the supervisor computer brings to a
central location the configuration of local unit parameters and operational
data.  Hence, information on the whole precipitator plant can be concisely
displayed.  Full control of all Local Units is available at this central
location, abnormal operation which the Supervisor Unit is programmed to
recognize is instantly displayed allowing correction or remedial action to
be taken immediately.

     As with any complex system, various parts of the system have distinct
boundaries and specific communication requirements.  Choosing the boundaries
and the communication wisely means:

     a.  The functional requirements can be met.

     b.  Failure of any components will be known and have minimal effect.
         Repair can be made without a total system shutdown.

     c.  The system is structured to permit progressive expansion.

     The following basic specifications satisfy these requirements (Figure 9):

     a.  The Supervisor is a microcomputer board system, which has a well
         defined structure of communication among core parts of a computer.

     b.  The Supervisor communicates with Local Unit controllers through RS-
         422 communications protocol.  This allows high speed communication
         on a simple coax cable connecting all Local Units and the Super-
         visor.  Each unit has its own address on this bus.  Additional
         Local Units and other peripherals can be added.
                                    508

-------
      c.   The Local Units can control the bus sections independently of the
          Supervisor.  Failure of the Supervisor will not affect the Local
          Control Units, and failure of a Local Control Unit only affects
          that  respective energized section.

      The  Supervisor Console consists of a color CRT, keyboard,  printer,
telephone modem, two disc drives, and microcomputer.  The Supervisor com-
puter conmunicates with the local AVC's and Control Room tfcnitor via PS-42?
buses.  Also included in the hardware structure is the ability  to accommodate
16 single ended A/D channels (4-?0 mA in, minimum 1O bit resolution),  and
facility  to address 16 parallel lines of I/O - 8  in, 8 out.  This parallel
interface is through 110 VAC lines.

      The  Supervisor CRT will display any of the following parameters at one
time  for  the precipitator system:  secondary bushing voltages,  secondary
bushing currents, line voltages, line currents, unit power consumption, and
the status of  each transformer/rectifier:

      a.   Unit  "TRIPPED" - designated by flashing background.

     b.   Unit  "OFF" - designated by red background.

      c.   Unit  "CUT OF SERVICE" - designated by blue background.

      The  Supervisor CRT will display all of the following for each Local AVC
Unit:  secondary voltage, secondary current, line voltage, line current, unit
power consumption, voltage drop counter, operation status.

      In the system display mode, the total system can be placed in service,
taken out of service, placed in system start-up or shutdown mode.

      In the individual local AVC display mode, all functions of the Local
Unit  are  available at the supervisor control.

     The  Supervisor has complete facility for data acquisition; system or
local unit history, voltage/current relationships for local units, histori-
cal trends for all basic parameters can be tabulated, plotted, printed on the
data  logger or transferred to the portable storage media.

      In addition to the data acquisition facility, reference information can
be stored.  Operating and maintenance instructions can be stored and dis-
played on the  CRT as required.  Trouble diagnostic guides can be stored in
the form  of logic diagrams to assist in maintenance and trouble shooting of
the equipment.  Maintenance schedules, records and replacement parts lists
can be stored  to provide a permanent record.

SUPERVISOR TELEPHONE MOEiM

     The  Supervisor Telephone Modem gives the system the ability to comnuni-
cate with a remote supervisor console located in a utility' s main office or
the precipitator supplier's home office such as here in Houston (Figure 8).
                                     509

-------
At this location, which may be several thousands of miles  from the. plant,  the
precipitator can be monitored and adjusted  if  required.  The availability of
accurate and instantaneous operating data to precipitation engineers located
in the home office allows technical support to be given  to plant engineers
for speedy correction of any operational problems which  may develop.

CONTROL ROOM MONITOR

     The Control Room Monitor is a color CRT and keyboard,  and is the boiler
shift operator interface to the system.  The information displayed allows the
operator to see the status of all the precipitator controllers at a glance.
The control functions at this station are limited.  The  system can be turned
on or off, tripped sections can be reset, and.  placed in  or out of the start-
up/shutdown mode.

     Switches are provided for system and Local Unit control.   Four of  these
are control, namely:  "system on", "system  off", "start-up/shutdown"  and
"opacity control", as provided, on the supervisor.

     The remaining five control the information displayed  on the CRT.   The
standard mode of display is the kV of each  section, but  any one of the  five
parameters:  kilovolts, millamperes, line voltage, line  amperes and kilo-
watts, can be displayed in the same format.

     In addition, when the operator moves the  cursor arrow to a section on
which he wants more detailed information and activates the  cursor switch,  the
CRT will display all parameter readings specific to that section,  together
with a listing of additional control functions, such as  automatic,  manual,
raise, lower, manual rapping, which are also cursor activated.

     If a Local Unit is off line for any reason, this will  be noted on  the
CRT and the duration of failure will also be displayed.  The last set of
meter readings, prior to the trip, will be  retained until  the unit is
returned to operation.

SUMMARY

     A microprocessor based precipitator control system, such as outlined
above, has many features not formerly available.  One of these is the abi-
lity to control transformer/rectifier power through the  use of an opacity
feedback signal.

     While maintaining constant opacity economies in power consumption  can
result when:

     a)  Particulate properties are more favorable to precipitation than
         those on which the precipitator design was based.

     b)  Reduction in the gas flow rate to  the precipitator occurs.
                                     510

-------
     While power economy is functionally available, the ramifications of such
a control philosophy must be thoroughy investigated by the owner to ensure
that it is environmentally acceptable.

     The work described in this paper was not funded by the U. S. Environ-
mental Protection Agency and, therefore, the contents do not necessarily
reflect the views of the Agency and no official endorsement should be in-
ferred.
                                     511

-------
on
                -£°
                E
                ui
                            MEAN DE VOLTAGE 48 K V
                                 OPTIMUM
                                OPERATING LEVEL
NOTE HIGHEST MIGRATION VELOCITY
  CORRESPONDS TO HIGHEST MEAN
  VALUE OF ELECTRODE VOLTAGE
                     MEANDE
                     VOLTAGE 4O KV
                                                    MEANDE
                                                    VOLTAGE 45 KV
                    TOTAL CURRENT INTO ELECTRODE SYSTEM (SPARKOVER PLUS CORONA) (MILLIAMPS/5Q.METRE)
O.O5
OIO
              OI5
                                                                                  02O
                        Figure 1.   Effect of Power Input to Electrode System
                                       on Effective Migration Velocity

-------
  1<*0
- 130
        ( A )
- 120
                  BLYTH POWER STATION (U.K.)

                  120 M.W. BOILER

                  PREC1PITATORS - 3 FIELDS  IN  SERIES
• 110
SPECIFIC PDUEH
   INPUT        O


(U.A./1000  CFM)
           X   BOILER NO.  1
• 10D      +   BOILER NO.  2
           0   BOILER NO.  3
-  90
                          GAS LOAD  (CFM ACTUAL)

                  300.ODD	        I	
        ( B )
- 1.0
   D.C. POUER INPUT
    (K.U.A.)       X
                         GAS LOAD (CFM ACTUAL)
                 300,000	  i     	
                                                 (.00,000
  (.0    (  C )


      MASS EMISSION

  30  POUND/MINUTE




  20
  10
                           GAS LOAD (CFM ACTUAL)
               300,000
                                                 (.00,000
                                                —I	
           Figure 2.   Effect of Varying Gas Load on
               Ash Emission  and Precipitator Power
                                  513

-------
_ 100
 -  90
 .  60
 -  70
    OPACITY

   BO
   50
   30
   20

                    AGECROFT POUER STATION  (U.K.)

                    120 M.U. BOILER

                    PRECIPITATORS - 3 FIELDS  SERIES

                    TESTS AT FULL BOILER  RATING
X

0
INLET FIELD  ONLY REDUCED

CENTRE FIELD ONLY REDUCED

OUTLET FIELD ONLY REDUCED

SOLID LINE - AVERAGE  OPACITY

10% OPACITY  = 35 mg/Nro3

5%  OPACITY  » 15 mg/Nro3
              SHORT DURATION SPIKES DUE TO
                      RE-ENTRAINMENT
              Figure 3.   Effect  on Stack Opacity
               of Reducing Power Input  to Fields
                                514

-------
       AGECROFT POUER STATION (U.K.)  120 M.U.

       BOILER   PRECIPITATOR5 3 FIELDS SERIES
            OUTLET  RUN MAXIMUM
            INLET AND CENTRE REDUCED
" 100
i
' 90

• BO
• 70
OPACITY %
. so

. 50
• i.0
• 30
. 20 V
. 10


0
H





-
'


A
•?


a
4-
r
i

<


^
r


i
NV
1




«
?







r**
V
CENTRE RUN MAXIMUM
INLET AND OUTLET REDUCED
INLET RUN MAXIMUM
CENTRE AND OUTLET REDUCED
SOLID LINE - AVERAGE OPACITY
10X OPACITY - 35 mg/Nm3
5X OPACITY - 15 mg/Nm3
+•



SHORT DURATION SPIKES
^r DUE TO RAPPING RE-ENTRAINMENT
t
\

f
1
1
" ' T •* T
^••*-*».i ^ 	 1 j t
D.C. POUER INPUT 0-*— X
( K . V . A . ) 2.0 10
Figure 4.  Effect on Opacity of Reducing
                Power Input  to Fields
                     515

-------
200
               AGECRDFT POUIER STATION


               120  M.UI. BOILER


               PRECIPITATDRS  -  3  FIELDS  IN SERIES
               X   INLET FIELD
               0   CENTRE FIELD
               +   OUTLET FIELD
 CORONA CURRENT

   (MILLIAMPS)




100
                30
                                      ELECTRODE VOLTAGE (K.V.)
                                    <40
     (  B  )
 20
                       RELATIONSHIP BETWEEN STACK OPACITY


                       AND SQUARE OF VOLTAGE APPLIED TO  ELECTRODES
  10
tono
                           (ELECTRODE VOLTAGE)2
                 1500
2000
  I
? , 500
 I
           Figure 5.  Voltage -  Current Relationship
                                  516

-------
Ul
                    Figure 6.  Effect of Spark Over on Electrostatically Deposited Ash Layer

-------
en
00
                                DIGITAL L-M AVCjn
                                                                                                              I O MAIN BREAKER
                                                                                                               OCURRENT LIMIT
OL.OW KV.
                                                                                                               O HOT OIL
                                                                                                               ORAP ALARM
                                                                                                               O RAPPING
                                                                                                               OCOMM. ERROR
                                                                                                               OFREQ. UNSTABLE
                                                     Figure 7.   Local Panel  Display

-------
on
—i
10
                                   AUTOMATIC
                                   VM.TAM
                                   CONTHOl
                              M.ICM
                              CONTHOUID
                                                                                                                        -C HOU1TOM
                                                                                                                        PEMATION*
                                       tLgCTHOSTATIC
                                       PRECIPITATOR
                                                        Figure  8.   AVC System Loop Diagram

-------
                       A
                           MAX 32  UNITS/LINK
                  4 LINKS = 128 MAX UNITS
                  RS-232
(PHONE
 COMMUNICATION)
1200 BAUD DATA
LINK        	
    MASS DATA  STORAGE
    10MBYTE  CAPACITY

    (WINCHESTER  DRV.)
      •   PLANT
      j_COMPUTER
                            rJLOCAL  AVC \ "JLOCAL AVC
                              LOCAL AVCl ILOCAL AVC
                       ILOCAL AVC HOCAL AVC j
                       [LOCAL AVC IHLOCAL AVC |
                                     LINE  DRIVER MANIFOLD
                                  J
                              MICRO-COMPUTER SYST.

                              FUNCTIONALLY  & MEMORY

                                  EXPANDABLE
PORTABLE MEDIUM
MASS STORAGE
DOUBLE DENSITY
FLOPPYDISK(450K)
                             REAL TIME
                               CLOCK
                           W/BATT.BACK-UP
                                                                     CONTROL ROOM
                                                                       MONITOR
                                                /
                                                                         HOPPER
                                                                         HEATERS
                                                                       HOPPER ASH
                                                                    LEVEL INDICATORS
                                   A/D CONVERTER
                                FOR OPACITY MEASURE
                                       CONTROL
                                       (4-20ma)
                                           8-110VAC
                                        PLANT CONDITIONS
                                          LOGIC LINES
                                           (INPUTS)
                                        8-110VAC
                                     PLANT CONDITIONS
                                       LOGIC LINES
                                        (OUTPUTS)
                           Figure  9.   Supervisor  System Hardware
                                               520

-------
         APPLYING MODULAR MICROCOMPUTER CONTROL ELEMENTS
                     IN A PRECIPITATOR CONTROL SYSTEM
       by:    Ira M. Wexler
             Environmental Elements Corporation
             Baltimore, Maryland 21203
                                   ABSTRACT

      The declining cost of microcomputers allows the control designer to implement
microcomputers into individual components  of  a precipitator control system.  This
distributed processing technique allows the integration of the individual controls with
a  master computer, and also provides the  controls with  intelligence  required to
operate independently, should the need arise.  The individual microcomputer elements
used  to  control  transformer/rectifiers, rapping  systems  and  alarm systems  are
described.  Techniques used  to integrate these  elements with a master precipitator
controller providing data acquisition,  energy management, and control optimization
are also defined.  In addition, a fail-safe operating philosophy which takes advantage
of the intelligence designed  into  the  individual controls is  presented.  Finally, the
impact of this technology on the end user is discussed, using an actual installation at a
utility boiler.

                                INTRODUCTION

      In  a  precipitator  control  environment,   where  reliability is a   paramount
consideration, the application of digital microcomputer  hardware and software must
be carefully thought out. In addition, a well designed control system should supply the
plant operator with confidence in its operation and maintainability.

      This paper presents the implementation  of such  a system  by  describing the
individual control elements  involved  and  the  techniques and  philosophies  used to
integrate the elements into an intelligent, user-friendly, reliable system.

      These  techniques and philosophies have been proven to be successful by five
years of  field experience with  several hundred microcomputer  based  controls in
operation.

                          HISTORICAL BACKGROUND

      Prior to 1976,  only analog  type precipitator  controls were in service.   The
emergence of the microcomputer spurred a  development effort to incorporate the
microcomputer intelligence into a precipitator control.  The result of this first effort


                                      521

-------
was  the DIGICON  Automatic Power Controller, used to  control the  power to an
individual Transformer-Rectifier set.  The next logical step in the development was
the OPTI-CON  H Data  Management System, to be used  as  a Master precipitator
control.  Concurrently, the Intelligent Rapper Control and the Intelligent Peripheral
Controller were  developed.

            INTELLIGENT CONTROL ELEMENT DESIGN  PHILOSOPHIES

      The design of microcomputer based equipment for precipitator controls requires
special consideration.  The equipment which will be subsequently discussed utilizes the
following criteria:

1.     The  equipment  is   designed  specifically  for  use  in  precipitator  control
       applications,  rather than  an  adaptation  or modification  of  off-the-shelf
       programmable controllers or minicomputers.

2.     All  operating programs  in the intelligent  control  elements are  stored  in
       non-volatile  EPROM (erasable programmable read only  memory).  This means
       that  the  operating  program  is resident on  a semiconductor  "chip" in each
       element.   The  data is always available  and  is retained on power loss.  The
       advantages of this  method are legion.  First, no programming knowledge  is
       required by plant personnel, since the programs have already been qualified and
       developed.  Secondly, program loading from cassettes, floppy diskettes or other
       media and the  unreliability associated with them  are completely eliminated.
       Lastly, the operating programs can still be process adaptive by using temporary
       volatile (read-write) memory  for the  storage of  certain data.  A sustained
       power loss will erase this  information, but the program in the  EPROM will
       replace  it with default values on power-up,  permitting  continued  operation
       without plant operator intervention.

3.     All hardware is designed specifically to operate in an industrial environment,
       freeing  the  control designer  from  the  restrictions  of  air  conditioned
       atmospheres and conditioned power lines.

4.     Incoming power and data lines are  hardware protected from voltage  transients
       typically  encountered in an industrial location.

5.     Linear  (as opposed  to switching) power supplies are used, which have been
       designed to permit a power loss of up to 100 miUiseconds with no discontinuity
       in operation.   Operation  after longer  power losses  is self-recovering, as
       described above.

6.     Error  detecting and correcting  software is  used to protect  against system
       control loss as a result of garbled or erroneous data.

7.     During power-up, each microcomputer element  will run thru a self check
       program. If a problem is found, an indication is given to the operator.
                                       522

-------
                           EQUIPMENT DESCRIPTION

INTELLIGENT T/R CONTROL ELEMENT

     The intelligent T/R (Transformer Rectifier)  control element  consists  of the
following equipment housed in a NEMA 12 control cabinet (Figure 1):
                                  FIGURE 1
                    CONTROL CONSOLE BLOCK DIAGRAM
       An SCR  (Silicon  Controller Rectifier) package  used to regulate  the  power
       supplied to the T/R.

       T/R analog metering circuitry and low voltage control, switching, and alarm
       circuitry.

       The DIGICONR H Automatic Power Controller (Figure 2). This unit, housed in
       a NEMA  12  type enclosure, monitors the primary and secondary voltage and
       current associated with the T/R and controls the power applied to the T/R as a
       function  of  its  operating  software (program).    In addition,  a  serial
       communications  link is  provided  to  the  OPTI-CON™-n Data  Management
       System Controller,  which  is used to provide  the  exchange of  information
       between  both  units.    Indicator  lights  and  thumbwheel  switches  on the
       Automatic Power Controller provide  the necessary interaction with the plant
       operator.

       The Control Console Annunciator.  This microcomputer based device, which
       operates independently of the Automatic Power Controller, actually performs
       several functions.  First,  the spark rate is counted and displayed  on a three
       digit  display.  Secondly, the annunciator monitors the  control operation for
       overspark and undervoltage conditions, and finally  the unit  annunciates the
       various control console trip mechanisms (overcurrent, T/R high temperature) on
       the  three digit display.   In addition, this  device  also  contains  a serial
       communications link to the Data Management System Controller.
                                      523

-------
                                                     •  •
                                  FIGURE 2
                      AUTOMATIC POWER CONTROLLER
INTELLIGENT RAPPER CONTROL ELEMENT

     The  Rapper Control Element  consists of the  following equipment housed in a
NEMA 12 control cabinet, mounted on or near the precipitator roof:

1.     Power conversion circuitry to supply the proper voltages and currents to fire
       the rappers.   The circuitry is configured such that various  types  of rapping
       devices (such as pneumatic, electric, and motor drives for mechanical rapping)
       can be intermixed in the same control cabinet.

2.     Digital switches and  indicators used  to  select  the rapping  intervals  and
       intensities and display the status of the system.

3.     A microcomputer element to control  the timing sequences and  single out
       faulted rappers on an individual basis.

4.     A serial communications   link to the Data Management System Controller used
       to report  the status  of the rapper controller and accept changes to  timing and
       intensity parameters.
                                      524

-------
INTELLIGENT PERIPHERAL CONTROL ELEMENT

     In  a precipitator control environment,  it  may  be desirable  to  monitor or
integrate related  control  parameters  (such  as  temperature,  contact  closures,
pressures, or other analog data) into the system.

     For example, hopper  level indicators,  flue gas temperatures or any number of
parameters may be monitored  and  displayed on the CRT and line printer.  Another
example would be the use of a  hopper temperature  differential measurement to
control a screw conveyor motor.

     If the number of peripheral points is small (less than 16), this function is carried
out in the Data Management System Controller.  However, if a large number of points
are to  be monitored,  or if  one  or more complex control loops are involved,  control is
delegated to an intelligent peripheral control element.

     This element consists of its  own  microcomputer, an  operating  program in
EPROM, signal conditioning circuitry, and a serial communications link  to the Data
Management System Controller.
                  DATA MANAGEMENT SYSTEM CONTROLLER

     All of the previously discussed  control elements are connected by their serial
communication links to the Data Management System Controller.  Figure 7 presents a
block diagram  of the entire system.   This  device, also housed in a NEMA 12 type
enclosure, consists of the following: (Figures 3 thru 6)

1.    A serial communications link to a CRT (Cathode Ray Terminal) and a printer.
      The latter devices are used to display, record,  and change data relating to
      precipitator control thru simple commands.

2.    A serial  communications link  to a  modem,  which, at the plant operator's
      discretion can transfer control to a remote location thru the telephone lines. A
      monitoring facility in Baltimore is presently  in  use to assist customers with
      precipitator troubleshooting by  using telephone access to the Data Management
      System at the customer's location.

3.    An optional communications link to the customer's plant computer.  When this
      link is used, the Data Management System becomes an intelligent "front end"
      for the precipitator, allowing the plant computer access to the commands and
      information used in the Data Management System.
                                      525

-------
4.     Operating programs in EPROM which perform the following functions:

      a.     Processes  and  formats the data gathered from the intelligent control
             elements and displays it on the CRT.

      b.     Allows the plant operator to alter various parameters (such as rapping
             cycle times, T/R control,  etc.) using the CRT.

      c.     Periodically instructs the  printer to record selected data. (Figure 8)

      d.     Integrates  stack transmissometer  data  with  the  precipitator control
             elements to define an energy management control algorithm.

      e.     Supervises  the  operation of peripherals connected to the  intelligent
             peripheral control element.

     Although  it may appear that this Data Management System  Controller is  the
heart of the  control system, it is important to remember that the rest of the control
elements  can operate independently due to  their  inherent intelligence.   The next
section  describes the philosophies used to integrate the individual  control  elements
into an intelligent system.
                                   FIGURE 3
                 DATA MANAGEMENT SYSTEM CONTROLLER
                                     526

-------
      OPTI-CON ' II
         riAT*
             FIGURE 4
     OPERATOR CRT TERMINAL

             FIGURE 5
DATA MANAGEMENT SYSTEM PRINTER
                527

-------
          FIGURE 6
      TELEPHONE MODEM
I
1

HOST
(PLANT)
COMPUTER

CRT
TERMINAL

PRINTER

*~l c°
I 1 MAN
ICONT
•I 1
1— «- DATA M/
SYSTEM (


T/R
[

NTROL CONSOLE
AUTOUATIC'
— -i r—f POWER
DAL 1 CONTROLLER
— — ' 1 INTELLIGENT
"ANNUNCIATOR

1 1
iNAGEMENT
:ONTROLLER


T/R
1
CONTROL CONSOLE
( up to 32 per controller )
1 — r-
r

*i
L

RAPPER
CONTROLS
MODEM
OTHER
PERIPHERALS


	 	 PHONE
LINES
          FIGURE 7
PRECIPITATOR BLOCK DIAGRAM
            528

-------
                  INTELLIGENT SYSTEM DESIGN PHILOSOPHIES

      Integrating the intelligent control elements into a reliable system again requires
 careful  consideration.   When  reliability is important, it  is  imperative that the
 malfunction of a control element will not cause either a snowball  or domino effect or
 a  gross  system  failure.   The  following techniques  have  been employed  in the
 OPTI-CON II Data Management System to carry out these requirements:

 1.     Each control element in the system can operate independently, as a result  of its
       inherent individual microcomputer intelligence.

 2.     Fail-safe  mechanisms  are incorporated in system software.   No hardware
       handshaking lines are  employed, since such lines could fail in  a  manner to
       indicate normal operation.  The use of software watch-dog timers, and periodic
       enquiry-response exchanges provide a true fail-safe mode of operation.

       An  analogy to  this philosophy can  be  illustrated  by two  people using  a
       telephone.   The  absence of a  dial  tone  (handshaking line) would normally
       indicate that someone is  on  the line.  If  this  were the only indication of a
       connection and if the  telephone hardware  should fail, both people would wait
       forever  for communication.  However, each party would realize  (due to  their
       inherent  intelligence)  that  the  absence  of  a  voice  response  indicates  a
       malfunction. The people would hang up and go about their normal business.

 3.     System  data is  checked  for parity,  framing  and overrun errors  to insure
       integrity.

4.      The system is  configured in a star pattern; data highways  are  not used.  The
       failure or  degradation  of no single cable can cause system failure.  Refer to
       Figure 7.

5.      System data is transferred serially in  an  RS422 format.  This format provides
       high speed, long distance transmission and possesses high noise immunity.  Data
       cables require  no special conduit runs and  may be placed  in cable  trays with
       480V power wiring.

                  USER FRIENDLINESS AND MAINTAINABILITY

     To  gain acceptance,  an  intelligent system must provide ease of  understanding,
operation, and maintainability. The  following techniques are used  to implement these
objectives:

1.      Since all programming  is intact in EPROM,  the plant operator is relieved  from
       this task.   Flexibility  is still maintained, however, by allowing the operator
       simple commands to alter system operation.  For example, the  energy-savings
       mode may be enabled or disabled by a single keystroke command. As another
       example, rapping cycle  times  may  be changed by a simple  sequence  of
       keystrokes.

2.      The CRT-operator interaction is menu driven.   The operator selects desired
       information  from a menu of  available  displays  and  commands.   Refer  to
       Figure 9 and 10.


                                       529

-------
3.     Operator generated commands into the system are checked and error trapped,
      thus preventing the entry of unrealistic or false parameters into the system.

4.     Printouts of data on a periodic or demand basis are provided to aid in record
      keeping.   Figure 8 is a typical display from an operating salt cake recovery
      precipitator.

5.     Equipment  is designed to be repaired by  the substitution of modular  circuit
      boards.  No special diagnostic equipment or skills are needed.

6.     To  insure added  reliability, all  major system elements  are  burned in for  a
      minimum of 100 hours.
*** CONSOLE STATUS ***
CONSOLE INDEX STATUS EP
i * * VOLTS
A INLET 1 ON 223
B INLET 2 ON 190
| A CENTER 3 ON 286
B CENTER 4 ON 241
A OUTLET 5 ON 369
j B OUTLET 6 ON 371
1
ENTER M FOR MENU
IP ES
AMPS KV
28 65
23 65
64 58
52 60
147 55
140 58

JUL. 01,
IS
MA
146
124
205
258
656
632

1982
SR
SPM
0
0
24
4
0
0

14:08
                                  FIGURE 8
                   PRINTOUT OF CONTROL CONSOLE DATA
                     FROM A SALT CAKE PRECIPITATOR
                                     530

-------
            *** COMMAND MENU ***
0. SYSTEM STATUS

1. CONSOLE STATUS

2, DIGICON SWITCH STATUS

3. RAPPER STATUS

4. RAPPER FAULT SUMMARY

5. ENERGY SAVINGS STATUS

6, I/O STATUS



ENTER COMMAND ?
                   FIGURE 9
    DATA MANAGEMENT SYSTEM COMMAND MENU
                    531

-------
 SET CLOCK
ENABLE LINE PRINTER
DISABLE LINE PRINTER
ABORT PRINTOUT
LOCK OUT REMOTE ACCESS
ENABLE REMOTE ACCESS
ABORT REMOTE ACCESS
PRINT CURRENT SCREEN
FLIP SCREEN PAGES
TURN A CONSOLE OFF
TURN A CONSOLE ON
TURN ALL CONSOLES OFF
TURN ALL CONSOLES ON
CHANGE THUMBWHEEL SETTING (Voltage)
CHANGE THUMBWHEEL SETTING DOWN ARROW NNN
                           FIGURE 10
                  COMMAND REFERENCE DISPLAY
                              532

-------
                                    RESULTS

     The first OPTI-CON system was installed on a flyash precipitator in a power
plant application 2 1/2 years ago. (Microcomputer based Automatic Power Controllers
have been in  the field  for 5 1/2 years).  In this application, the boiler experienced
varying loads,  due to seasonal and  usage demands.   The  customer conducted  an
evaluation program of  energy savings during several months  of OPTI-CON system
operation.  The following results were obtained:

1.     The average daily power savings were 125HP/hour or 93 KWH.

2.     The yearly savings amounted to $30,720, based on a power cost of $.04/KWH

3.     The present worth of  a 125  HP  savings  amounted to $125,000 based  on
       escalating fuel costs of 7% per year.

     Other systems  are  presently  in  operation  at  several  other locations, and
feedback has been positive both in operation and reliability aspects.


                         SUMMARY AND CONCLUSIONS

     From the previous discussions, it can be seen that the use of individual modular
intelligent control elements constitutes a reliable and versatile method of precipitator
control.   Proper hardware and software design philosophies also reinforce system
integrity and produce reliable,  intelligent and cost  effective solutions in precipitator
control.

     The work described  in this paper was not funded by  the U.S. Environmental
Protection Agency and  therefore the contents do not necessarily reflect  the views of
the Agency and no official endorsement should be inferred.

     This  work was  funded  by Environmental Elements Corporation which  accepts
total responsibility for its contents.  No official  endorsement by any  govermental
organization is either  sought or desired.
                                       533

-------
        THE CURRENT  STATUS, FUTURE  DIRECTIONS,  AND  ECONOMIC
               CONDITIONS  IN THE APPLICATIONS OF ESP's

                              Sabert Oglesby
                      Southern Research Institute
                       Birmingham, Alabama   35255
     The subject I  have  been asked to address today is the current status,
future directions,  and economic considerations in the application of electro-
static precipitators.  By  current status, I assume is meant the type of
equipment that would be  offered in response to a bid invitation today for
various dust control applications.  In the U. S., the majority of electro-
static precipitators offered today are of the conventional design with stan-
dard weighted-wire  or rigid-frame electrodes, conventional 10-12 inch plate
spacing and power supplies of conventional design.  There is a decided trend
in this country toward greater use of rigid-frame discharge electrodes and
flail-hammer rappers of  the type commonly used in European installations.
Power supplies are  conventional, but there are a variety of types of control
systems currently offered.  The trend in recent years has been toward faster
response and greater reliability in the control circuitry.

     There are a number  of exceptions to the more or less conventional
installations.  In  Europe  and Japan there have been a number of precipitators
installed with wider than  normal plate spacing.  The practice in Europe began
in the cement industry,  where aluminum plates were installed to minimize cor-
rosion from the alkali-laden gases from a cement kiln.  Inlet sections were
made with normal spacing and downstream sections were wider than normal to
reduce the cost of  collection electrodes.  There are also some wide-plate
precipitators installed  in Japan.

     The greatest advancement that has been made in electrostatic precipita-
tor technology in the past 10-12 years has been in better applications.  For
the most part, the  practice of using catalogs of migration velocity values to
arrive at the size  precipitator required to meet a given level of emission
has largely been overcome.  Both manufacturers and users have become much
more sophisticated  in the  sizing process, and considerable effort goes into
characterizing the  effluents to insure that emission limits will be met.  I
don't have any figures to  back up the contention, but based on our observa-
tions, one can be reasonably well assured that present-day ESP installations
will perform within their  guarantee and operate reliably.


                                    534

-------
      Precipitators  can  also more  than meet  the  new  source  performance  stan-
dards.  Emissions of  as low as  .001  Ibs/million BTU have been measured on  an
installation with an  SCA of about 500 ft2/1000  cfm.

      Many new concepts  for overcoming some  limitations of  electrostatic pre-
cipitator problems  or enhancing precipitator performance are in either the
research or advanced  development  stage,  and many of these  will be the  subject
of papers at this symposium.

      The subjects being most widely  discussed for advancements in electro-
static precipitators  are conditioning to modify dust resistivity, pulsed
power supplies, precharging, wide-plate  spacing, microcomputer controls,
alternative rapping schedules,  and different electrode geometries.  To these
must  be added advancements in fundamental knowledge of the precipitation pro-
cess.  Much progress  has been made in recent years  in developing equations to
predict particle charging, electric  field,  and  collection, and these have led
to mathematical models  which have been extremely useful in sizing and  trouble-
shooting,  Progress is  continuing in this area  as methods  are developed for
including electron  charging, gas  turbulence, and better information on reen-
trainment in the various models.

      Conditioning to  modify electrical resistivity  is certainly not new.  The
fitst such installation reported  in  the  literature  was made by Cottrell to
reduce the resistivity  of smelter dust.  However, there are some relatively
new concepts in conditioning, such as addition  of sodium compounds to  coal to
reduce resistivity  of fly ash for hot-side  precipitators,  and the use  of
ammonia compounds to  achieve better  performance in  some precipitators  by
reducing both resistivity,and reentrainment or  increasing  the collection elec-
tric  field by space charge enhancement.

      For the most part,  conditioning is  resorted to  for increasing the per-
formance of an existing precipitator that was sized  too small to collect a
high  resistivity dust or to accommodate  a change in  fuel supply or other fac-
tors  that change the  dust resistivity.   Conditioning, however, is becoming
much  better accepted  as systems become simpler  and more reliable, and  has
reached  a stage of development where it represents  one option that can be
more  cost effective for  new installations.

      Pulsed power supplies have also been used  primarily for upgrading an
existing system, as opposed to being supplied as a  part of a new installatibn.
One of the major advantages cited for a  pulsed  power  supply is that the num-
ber and uniformity  of emitting points along the discharge  electrode is
increased,  thereby resulting in  a more  uniform current density at the collec-
tion electrode.  Also,  it is suggested that back corona can be minimized
because the pulses  are  short compared to the relaxation time of the dust
layer.  Both of these factors permit operation  at higher electrical condi-
tions, especially if  the dust resistivity is high.   It has been reported that
improved electrical conditions achieved  with pulsing  can result in an  increase
in precipitation rate parameter by a factor of  2 for  a high resistivity dust,
but only a 10% or so  improvement  for lower  resistivity dusts.  Pulsed  power
supplies appear to  have  a definite possibility  for  some retrofit applications
where the improved  precipitation  rate parameter can  bring  the system in


                                     "535

-------
compliance.

     There is still some uncertainty as to how pulsed power supplies function
to give increased performance.  Measurements of particle charge show some
increase with pulsed power, but not enough to account for the decreases  in
emissions.  Further studies are planned to more precisely define the system
parameters.  Hopefully, pulsed systems can be designed to be cost effective
and find their way into new precipitator installations.

     Precharging represents another area of potential precipitator improve-
ment.  The concept of precharging is not new, but there are some new
approaches.  The potential advantage seems to be in the use of a combination
precharger and modified downstream collector which has a low current, high
field electrode system.  The advantage of precharging is that a short pre-
charging section can be used which permits some techniques to be utilized to
overcome high dust resistivities that would be impractical for a full-sized
precipitator.  Precharging has also been suggested as a means of achieving a
higher than normal charge on dust whether or not there is a resistivity  prob-
lem.  Examples of precharging systems include the tri-electrode charger  being
studied by Southern Research Institute under contract with EPA, a cooled
electrode system being studied by Denver Research Institute, and the high-
intensity ionizer system that was investigated by EPRI .

     I suppose the jury is still out on the practicality of prechargers.
There is no doubt of the correctness of the theory, but there may be some
practical problems in their application.  The tri-electrode precharger and
the high-intensity ionizer have both proven that particles can be charged
rapidly in a precharger.  In the case of the HII, a discharge of the dust
due to electrical breakdown seems to be a limiting factor in obtaining a
higher than normal particle charge.  The tri-electrode precharger does not
attempt to put a higher charge than could be achieved when collecting a  low-
resistivity dust, and the system is not plagued with discharge problems.  The
concept of enhanced downstream collection by providing a low-current, high-
field electrode system does give good results on a reasonably large  (30,000
cfm) pilot-scale system.

     There is considerable interest in wide-plate spacing for a number of
applications, especially in other than inlet sections.  From a theoretical
standpoint, wide-plate spacing would enhance the space charge contribution to
the field at the plate.  This should increase collection efficiency  about
proportional to the increase in field.  Another consideration is the more
uniform current density that should be achievable with wider-spaced  plates
and the higher voltage that can be attained due to the diminished influence
of electrode spacing variations.  Additional studies are currently being
planned to attempt to quantify the beneficial effects of wide-plate  spacing
and the conditions under which these benefits can be realized.

     Use of microcomputers for many precipitator control functions is another
new development that offers many possibilities for increased reliability and
performance.  Present day microcomputers are programmable to control voltage
input to maintain optimum electrification.  Because of the memory and adap-
tive capabilities, microcomputers can be programmed to maintain voltage  and


                                     536

-------
current at optimal values, thus improving performance over conventional ana-
log control circuits.  They can also be programmed to prevent development of
power arcs and other disruptions that reduce reliability.

     In addition to control of the TR sets, microcomputers can also be
arranged to control rapping sequence, ash discharge schedules, and addition
of flue gas additives for conditioning.  Microcomputers can also be used to
reduce energy consumption by optimizing the power input to meet opacity
requirements.

     Many of the advancements that have been made in precipitator technology
in recent years have been in improved understanding of the fundamentals of
the precipitation process.  Particle-charging equations have been developed
to a point that charge can be predicted with reasonable accuracy, except for
higher temperature operation where electron charging is significant.  Results
of tests to determine charge on dusts exiting from full-scale precipitators
agree quite well with calculated values using more recently developed charg-
ing equations.  Similarly, electric field measurements show good agreement
with calculated values for wire-plate electrode structures, and data are
being accumulated to account for various other types of discharge electrodes.
Considerable attention is also being given to collection mechanisms to
account for turbulent diffusion and other factors that influence collection.
All of these have been combined into precipitator mathematical models that
have helped to define the operating conditions and assisted greatly in both
sizing and analysis pf precipitator performance.

     One of the most significant advancements that has been made in recent
years is in development of better methods for measuring and predicting dust
resistivity, and in defining the conduction mechanisms.  Of the problems
associated with precipitators in the past, a very large percentage of them
are associated with failure to recognize or account for high-resistivity
dust.  This problem has been greatly reduced with the development of better
methods of predicting dust resistivity based on dust and flue gas composi-
tion.

     The future of electrostatic precipitators for many applications is
related to cost and reliability.  Precipitator cost is a direct function of
the plate area involved, and attempts to reduce costs must address the prob-
lems of reducing the collecting surface area.  For low-resistivity dusts, the
plate area requirements are minimal and precipitators constitute the most
cost-effective means of particulate emissions control.  However, for very
high resistivities, large plate areas are required and alternative particu-
late control systems are likely to be more cost effective.

     Economic factors are important considerations in the selection of the
type of dust control equipment for a given application.  Unfortunately, costs
are highly variable, depending upon resistivity of the dust, extent of duct-
work, size of hoppers, and other factors.  Perhaps because of these varia-
tions, there have been relatively few definitive cost studies.  A 1978 anal-
ysis by- Stearns-Roger for EPRI showed costs ranging from about $19 to $32 per
square foot of collecting area, the higher cost being associated with a small
precipitator.  These costs are for a complete system including ductwork, ash-


                                     537

-------
handling equipment erection, and engineering.  Costs, according to the
Stearns-Roger study, are broken down into, categories.,, with the precipitator
and supports representing about 35-40% of the total, ductwork about 10-20%,
and erection costs 35-40%.  All other costs represent about 10% of the total.
Since precipitator and erection costs represent 75-80% of the total, it is
apparent that major cost reductions must come from reduced precipitator size.

     As a general rule, electrostatic precipitators are competitive with
fabric filters at specific collecting surface areas of up to around 600
1000 cfm.

     The competitive position of precipitators is influenced by a number of
factors, including interest rates, energy costs, and labor costs.  Precipi-
tators have the advantage of lower energy consumption and lower maintenance
costs.  From a cost standpoint, the major disadvantage is the higher first
cost.  This has to be offset by reduced operating costs in order to remain
competitive.

     I have alluded to several developments that are currently being studied
and to some extent applied to full-scale precipitators.  These represent rea-
sonably near-term developments that should become state-of-the-art for many
applications, assuming they are successful.

     In the longer term, there are a number of areas for precipitator
improvement.  These constitute those areas where precipitators are constrained
to operate at less than theoretical limits.  One example is the electrical
operating conditions.  Theoretically, breakdown of the dust layer would occur
at much higher currents than present limits.  One of the reasons for this
limit is the non-uniformity of current density.  In negative corona, maximum
current occurs opposite a corona tuft.  If a more uniform corona can be
developed, much higher electrical operating conditions can be maintained.
Pulsed power supplies reportedly achieve a greater uniformity, but there may
be possibilities for achieving this uniformity at less cost through shaping
of conventional waveforms or through electrode design.

     Another area of potential development is in tailoring charging and col-
lection sections to achieve optimum collection.  Most present-day precipita-
tors have the same electrode structure throughout.  By tailoring a section' to
maintain a high field at low current just beyond a charging region, much
better collection rates might be achieved.

     Finally, more attention should be given to reducing rapping losses.   In
a typical precipitator, reentrained dust, primarily from the last field,
amounts to something around 30 to 70% of the emission.  A substantial  reduc-
tion in reentrainment would markedly improve precipitator performance.

     It is improvements such as these that may substantially reduce precipi-
tator sizes required, and in turn affect precipitator costs and their  compet-
itive position for particulate control.
                                      538

-------
                             AUTHOR INDEX
AUTHOR NAME                                                     PAGE
ADAIR, L	     1-460
ADAMS, R.L	     H-35
ANDO, T	    11-474
ARMSTRONG, J	   III-241
ARSTIKAITIS, A.A	    11-194
BALL, C.E	   III-370
BANKS, R.R	    1-37, 1-62
BARRANGER, C.B	     1-132
BAYLIS, A.P	    11-384
BELTRAN, M.R	     11-51
BENSON, S.A	    111-97
BERGMAN, F	   III-154
BIESE, R.J	     1-446
BOSCAK, V	    111-66
BRADBURN, K.M	    11-499
BRADLEY, L.H	    11-369
BRINKMANN, A	   III-211
BUCK, V	   III-335
BUMP, R.L	     11-17
CAPPS, D.D	     1-121
CARR, R.C	     1-148
CHAMBERS, R	 .  .  .     1-226, 1-239
CHANG, R*  	   III-271
CHEN, F.L	   III-347
CHEN, Y.J	     1-506
CHIANG, T	    11-184
CHRISTENSEN, E.M	    11-243
CHRISTIANSEN, J.V	    11-243
CILIBERTI, D.F	     III-282, III-318
CLEMENTS, J.S	     11-96
COE/JR, E.L	    11-416
COLE, W.H	     III-l
                                 539

-------
COOK, D.R	    11-349
COWHERD,JR, C	   III-183
COY, D.W	   III-370
CRYNACK, R.R	      II-l
CUSCINO, T	   III-154
GUSHING, K.M	     1-148
DAHLIN, R.S	     1-192
DARBY, K	    11-499
DAVIS, R.H	     H-96
DAVIS, W.T	     1-521
DAVISON, J.W	   III-166
DELANEY, S	     1-357
DEMIAN, A	    111-66
DENNIS, R	    1-22, 111-81
DIRGO, J.A	  111-26, 111-81
DISMUKES,  E.B	    11-444
DONOVAN, R.P	1-77, 1-107, 1-316, 1-327, 1-342
DORCHAK, T.P	   III-114
DRENKER, S	    III-271, III-282
DRIGGERS,  G.W	    11-194
DUBARD, J.L	    11-337
DUFFY, M.J	    11-489
DURHAM, M	    11-84, III-241
EBREY, J.M	    11-349
ENGLEHART, P.J	   III-183
ENSOR, D.S	   III-347
FAULKNER,  M.B	    11-204, 11-337
FINNEY, W.C	     11-96
FORTUNE, O.F	1-482, 1-494
FOSTER, J.T	     1-37,  1-91
FREDERICK, E.R	     1-536
FRISCH, N.W	   III-114
FURLONG, D.A	1-287, 1-342
GARDNER, R.P	    1-77, 1-107
GAWRELUK,  G.R	     11-17
GELFAND, P	     11-35
                                  540

-------
GIBBS, J.L	     11-430
GILES, W.B	111-41,  111-53
GOLAN, L.P	    III-226
GOLDBRUNNER, P.R	     11-401
GOLIGHTLEY, R.M	      1-164
GOOCH, J.P	     11-444
GOODWIN, J.L	    III-226
GRANT, M.A	     111-81
GREEN, G.P	      1-192
GREINER, G.P	1-287,  1-357
GRONBERG, S	    III-141
GRUBB, W.T	    1-62,  1-91,  1-179
HALL, H.J	     11-459
HALOW, J.S	      11-96
HANSON , P	      1-460
HARMDN, D	    1-226,  1-239,  III-131
HAWKINS, L.A	     11-194
HERCEG, Z	     11-489
HOVIS, L.S	   1-22, 1-77, 1-107, 1-287,  1-316,  1-327,
                                         1-342,  1-357,  111-81,  III-347

HOWARD, J.R	      1-164
INGRAM, T.J	      1-446
ISAHAYA, F.	     11-154
ITAGAKI, T	     11-322
JACOB, R.O	      1-446
JENSEN, R.M	      1-431
JONES, R	      1-303
KASIK, L.A	     11-430
KETCHUCK, M.   	      1-482
KINSEY, J	    III-154
KOHL, A.L	    III-300
KUBY, W	    III-271
KUNKA, S	      1-239
KUTEMEYER, P.M	    III-211
LAMB, G.E.R	      1-303
LARSEN, P.S	     11-243
                                 541

-------
LAWLESS, P.A	    11-271
LEE, W	     1-303
LEITH, D	    111-26
LEONARD, G.L	    11-230
LEWIS, M	     1-179
LIPPERT, T.E	III-280, III-318
LUGAR, T.W	    11-184
MARCHANT,JR, G.H	    11-444
MASON, D.M	   III-256
MASUDA, S	11-139, 11-169, 11-322, III-386
MATSUMOTO, Y	    11-474
MATULEVICIUS, E.S	   III-226
MCCAIN, J.D	   III-198
MCCOLLOR, D.P	    111-97
MCDONALD, J.R	    11-204
MCKENNA, J.D	     1-210
MCLEAN, K.L	    11-489
MENARD, A	     1-255
MILLER, M.L	     1-482
MILLER, R.L	     1-494
MILLER, S.J	    111-97
MITCHNER, M	    11-230
MOSLEHI, G.B	11-288, 11-306
MOSLEY, R.B	    11-204
MOVER, R.B	     1-460
MUSGROVE, J	     1-382
MYCOCK, J.C	     1-210
NAKATANI, H	    11-169
NG,  T.S	    11-489
NOVOGORATZ, D	    11-349
OGLESBY, S	    11-534
O'ROURKE, R	   III-318
PEARSON, G	     1-121
PETERS, H.J	,	     1-179
PIULLE, W	    11-401
PONTIUS, D.H	     11-65
                                 542

-------
PUDELEK, R.E	      1-521
PUTTICK, D.G	     11-126
QUACH, M.T	      1-506
RAMSEY, G.H	1-316, 1-327
RANADE, M.A	    III-347
REED, G.D	      1-521
REHMAT, A	    III-256
REIDER, J.P	    III-183
REISINGER, A.A	      1-179
RICHARDS, R.M	      1-255
RICHARDSON, J.W	      1-210
RINARD, G	11-84,  III-241
ROOP, R.N	      1-460
ROSS, D.R	      1-164
RUGENSTEIN, W.A	     11-430
RUGG, D	11-84,  III-241
RUSSELL-JONES, A	     11-384
SAIBINI, J	      1-132
SAMUEL, E.A	    1-1, 11-218
SANDELL, M.A	       II-l
SAWYER, J	    III-271
SEARS, D.R	    1-192, 111-97
SELF, S.A	11-230,  11-228, 11-306
SHACKLETON, M	    III-271
SHISHIKUI, Y	     11-139
SMITH, W.B	      1-148
SORENSON, P.H	    III-362
SPARKS, L.E	  11-204,  11-271, 11-337
SPENCE, N	      1-132
SPENCER,III, H.W	      1-506
STELMAN, D	    III-300
STOCK, D.E	     11-261
SUHRE, D	    III-335
SUNTER, T.C	       1-48
SURATI, H	      11-51
TACHIBANA, N	     11-474
                                 543

-------
TASSICKER, O.J	   III-271, III-282
THOMPSON, C.S	    111-12
THOMSEN, H.P	    11-243
TOKUNAGA, 0	     11-96
TKEXLER, E.C	     11-96
TRILLING, C.A	   III-300
TSAO, K.C	   III-256
VANN BUSH, P	     II-65
VANOSDELL, D.W	1-287, 1-342
WALSH, M.A	     1-482
WEBER, E	    11-111
WELLAN, W.G	     1-420
WEXLER, I.M	    11-521
WHITTLESEY, M	     1-482
WILOOX, K	   III-154
WILLIAMSON, A.D	   III-198
YAMAMOTO, T	   III-241
YEAGER, K.E	    III-XV
                                  544                 * U.S. GOVERNMENT PRINTING OFFICE: 19B5 - 559-111/10738

-------