&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
------- |