&EPA
United States
Environmental Protection
Agency
Industrial Environmental Research
Laboratory
Research Triangle Park NC 27711
EPA-600 9-80-039b
September 1980
Research and Development
Second
Symposium on the
Transfer and
Utilization of
Particulate Control
Technology
Volume II.
Electrostatic
Precipitators
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/9-80-039b
September 1980
SECOND SYMPOSIUM ON THE
TRANSFER AND UTILIZATION OF
PARTICULATE CONTROL TECHNOLOGY
VOLUME II. ELECTROSTATIC PRECIPITATORS
by
F.P. Venditti, J.A. Armstrong, and Micbael Durham
Denver Research Institute
P.O. Box 10127
Denver, Colorado 80210
Grant Number: R805725
Project Officer
Dennis C. Drehrnel
Office of Energy, Minerals, and Industry
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
INDUSTRIAL ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NC 27711
-------
DISCLAIMER
This report has been reviewed by the Industrial Environmental Research
Laboratory-Research Triangle Park, North Carolina, Office of Research and
Development, U.S. Environmental Protection Agency, and approved for publi-
cation. Approval does not signify that the contents necessarily reflect
the views and policies of the U.S. Environmental Protection Agency, nor
does mention of trade names or commercial products constitute endorsement
or recommendation for use.
ii
-------
ABSTRACT
The papers in these four volumes of Proceedings were pre-
sented at the Second Symposium on the Transfer and Utilization of
Paniculate Control Technology held in Denver, Colorado during 23
July through 27 July 1979, sponsored by the Particulate Technology
Branch of the Industrial Environmental Research Laboratory of the
Environmental Protection Agency and hosted by the Denver Research
Institute of the University of Denver.
The purpose of the symposium was to bring together research-
ers, 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 combina-
tions 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, and
measurement techniques.
These proceedings are divided into four volumes, each volume
containing a set of related session topics to provide easy access to a
unified technology area.
ill
-------
CONTENTS
Paqe
VOLUME I CONTENTS viii
VOLUME III CONTENTS ix
VOLUME IV CONTENTS xvi
Section A - Fundamentals
COLLECTION EFFICIENCY OF ELECTROSTATIC PRECIPITATORS
BY NUMERICAL SIMULATION 1
E.A. Samuel
THE EFFECTS OF CORONA ELECTRODE GEOMETRY ON THE
OPERATIONAL CHARACTERISTICS OF AN ESP 31
G. Rinard, D. Rugg, W. Patten and L.E. Sparks
THEORETICAL METHODS FOR PREDICTING ELECTRICAL CONDITIONS
IN WIRE-PLATE ELECTROSTATIC PRECIPITATORS 45
R.B. Mosley, J.R. McDonald and L.E. Sparks
LATERAL PROPAGATION OF BACK DISCHARGE 65
S. Masuda and S. Obata
THEORETICAL MODELS OF BACK CORONA AND
LABORATORY OBSERVATIONS 74
D.W. VanOsdell, P.A. Lawless and L.E. Sparks
CHARGE MEASUREMENTS ON INDIVIDUAL PARTICLES
EXITING LABORATORY PRECIPITATORS 93
J.R. McDonald, M.H. Anderson, R.B. Mosley and L.E. Sparks
OPTIMIZATION OF COLLECTION EFFICIENCY BY VARYING PLATE
SPACING WITHIN AN ELECTROSTATIC PRECIPITATOR 114
E.J. Eschbach and D.E. Stock
INTERACTION BETWEEN ELECTROSTATICS AND FLUID DYNAMICS
IN ELECTROSTATIC PRECIPITATORS 125
S. Bernstein and C.T. Crowe
PARTICLE TRANSPORT IN ELECTROSTATIC PRECIPITATORS ... 146
G. Leonard, M. Mitchner and S.A. Self
Section B - Operation and Maintenance
THE "HUMAN ELEMENT" - A PROBLEM IN OPERATING
PRECIPITATORS 168
W.J. Buchanan
ELECTROSTATIC PRECIPITATORS - ELECTRICAL PROBLEMS
AND SOLUTIONS 173
R.K. Raymond v
-------
ELECTRODE CLEANING SYSTEMS: OPTIMIZING RAPPING
ENERGY AND RAPPING CONTROL 189
M. Neundorfer
COMPOSITION OF PARTICULATES--SOME EFFECTS ON
PRECIPITATOR OPERATION 208
J.D. Roehr
INCREASING PRECIPITATOR RELIABILITY BY PROPER LOGGING
AND INTERPRETATION OF OPERATIONAL PARAMETERS - AN
OPERATORS GUIDE 219
P.P. Bibbo and P. Aa
ELECTROSTATIC PRECIPITATORS - START-UP, LOW LOAD,
CYCLING, AND MAINTENANCE CONSIDERATIONS 242
F.A. Wybenga and RJ. Batyko
ELECTROSTATIC PRECIPITATOR EMISSION AND OPACITY
PERFORMANCE CONTROL THRU RAPPER STRATEGY 256
W.T. Langan, J.H. Oscarson and S. Hassett
RAPPING SYSTEMS FOR COLLECTING SURFACES IN AN
ELECTROSTATIC PRECIPITATOR 279
H.L. Engelbrecht
LOW POWER ELECTROSTATIC PRECIPITATION - A LOGICAL
SOLUTION TO COLLECTION PROBLEMS EXPERIENCED WITH
HIGH RESISTIVITY PARTICULATE 296
J.H. Umberger
Section B - Advanced Design
HIGH INTENSITY IONIZER TECHNOLOGY APPLIED TO
RETROFIT ELECTROSTATIC PRECIPITATORS '314
C.M. Chang and A.I. Rimensberger
BOXER-CHARGER - A NOVEL CHARGING DEVICE FOR HIGH
RESISTIVITY DUSTS 334
S. Masuda and H. Nakatani
PRECIPITATOR ENERGIZATION UTILIZING AN ENERGY
CONSERVING PULSE GENERATOR . ' 352
H.H. Petersen and P. Lausen
PRECHARGER COLLECTION SYSTEM - DESIGN FROM THE
LABORATORY THROUGH FIELD DEMONSTRATION 359
M. Nunn, D. Pontius, J.H. Abbott and L.E. Sparks
-------
Page
TOWARDS A MICROSCOPIC THEORY OF ELECTROSTATIC
PRECIPITATION 374
C.G. Noll and T. Yamamoto
ION CURRENT DENSITIES PRODUCED BY ENERGETIC ELECTRONS
IN ELECTROSTATIC PRECIPITATOR GEOMETRIES 391
W.C. Finney, L.C. Thanh and R.H. Davis
EXPERIMENTAL STUDIES IN THE ELECTROSTATIC
PRECIPITATION OF HIGH-RESISTIVITY PARTICULATE 399
J.C. Modla, R.H. Leiby, T.W. Lugar, and K.E. Wolpert
PILOT PLANT TESTS OF AN ESP PRECEDED BY THE
EPA-SoRI PRECHARGER 417
L.E. Sparks, G.H. Ramsey, B.E. Daniel and J.H. Abbott
Section C - Industrial Applications
PILOT PLANT/FULL SCALE EP SYSTEM DESIGN AND
PERFORMANCE ON BOF APPLICATION 427
D. Ruth and D. Shilton
THE SELECTION AND OPERATION OF A NEW PRECIPITATOR
SYSTEM ON AN EXISTING BASIC OXYGEN FURNACE 441
D. Ruth and D. Shilton
CONTROL OF FINE PARTICLE EMISSIONS WITH WET
ELECTROSTATIC PRECIPITATION 452
S.A. Jaasund
TUBULAR ELECTROSTATIC PRECIPITATORS OF TWO
STAGE DESIGN . . " 469
H. Surati, M.R. Beltran and I. Raigorodsky
PRESENT STATUS OF WIDE-SPACING TYPE^ PRECIPITATOR
IN JAPAN 483
S. Masuda
LOW FREQUENCY SONIC CLEANING APPLIED TO
ELECTROSTATIC PRECIPITATORS 502
S.B. Smith and J.A. Schwartz
AUTHOR INDEX 514
-------
VOLUME I
CONTROL OF EMISSIONS FROM COAL FIRED BOILERS
Section A - Electrostatic Precipitators
COST AND PERFORMANCE OF PARTICULATE CONTROL
DEVICES FOR LOW-SULFUR WESTERN COALS 1
R.A. Chapman, D.P. Clements, L.E. Sparks and J.H. Abbott
CRITERIA FOR DESIGNING ELECTROSTATIC PRECIPITATORS ... 15
K. Darby
EVALUATION OF THE GEORGE NEAL ELECTROSTATIC
PRECIPITATOR 3.5
R.C. Carr
EPA MOBILE ESP HOT-SIDE PERFORMANCE EVALUATION .... 56
S.P. Schliesser, S. Malani, C.L. Stanley and L. E. Sparks
PRECIPITATOR UPGRADING AND FUEL CONTROL PROGRAM
FOR PARTICULATE COMPLIANCE AT PENNSYLVANIA
POWER & LIGHT COMPANY 80
J.T. Guiffre
MODIFICATION OF EXISTING PRECIPITATORS TO RESPOND TO
FUEL CHANGES AND CURRENT EMISSION REGULATIONS 100
D.S. Kelly and R.D. Frame
PERFORMANCE OF ELECTROSTATIC PRECIPITATORS WITH
LOAD VARIATION 117
W.T. Langan, G. Gogola and E.A. Samuel
FLY ASH CONDITIONING BY CO-PRECIPITATION WITH
SODIUM CARBONATE 132
J.P. Gooch, R.E. Bickelhaupt and L.E. Sparks
PREDICTING FLY ASH RESISTIVITY - AN EVALUATION 154
R.E. Bickelhaupt and L.E. Sparks
S03 CONDITIONING FOR IMPROVED ELECTROSTATIC PRECIPITATOR
PERFORMANCE OPERATING ON LOW SULFUR COAL 170
J.J. Ferrigan, III and J. Roehr
DOES SULPHUR IN COAL DOMINATE FLYASH COLLECTION IN
ELECTROSTATIC PRECIPITATORS? 184
E.G. Potter and C.A.J. Paulson
ANALYSIS OF THERMAL DECOMPOSITION PRODUCTS OF FLUE
GAS CONDITIONING AGENTS 202
R.B. Spafford, H.K. Dillon, E.B. Dismukes and L.E. Sparks
viii
-------
VOLUME I CONTENTS (Cont.)
Page
BIOTOXICITY OF FLY ASH PARTICULATE 224
A.R. Kolber, TJ. Wolff, J. Abbott and L. E. Sparks
Section B - Fabric Filters
FABRIC FILTERS VERSUS ELECTROSTATIC PRECIPITATORS ... 243
E.W. Stenby, R.W. Scheck, S.D. Severson, F.A. Homey
and D.P. Teixeira
DESIGN AND CONSTRUCTION OF BAGHOUSES FOR
SHAWNEE STEAM PLANT 263
J.A. Hudson, L.A. Thaxton, H.D. Ferguson, Jr., and N. Clay
OPERATING CHARACTERISTICS OF A FABRIC FILTER ON A
PEAKING/CYCLING BOILER WITHOUT AUXILIARY PREHEAT
OR REHEAT 297
W. Smit and K. Spitzer
OBJECTIVES AND STATUS OF FABRIC FILTER
PERFORMANCE STUDY 317
K.L. Ladd, R. Chambers, S. Kunka and D. Harmon
START-UP AND INITIAL OPERATIONAL EXPERIENCE ON A 400,000
ACFM BAGHOUSE ON CITY OF COLORADO SPRINGS' MARTIN DRAKE
UNIT NO. 6 342
R.L. Ostop and J.M. Urich, Jr.
DESIGN, OPERATION, AND PERFORMANCE TESTING
OF THE CAMEO NO. 1 UNIT FABRIC FILTER 351
H.G. Brines
EXPERIENCE AT COORS WITH FABRIC FILTERS - FIRING
PULVERIZED WESTERN COAL 359
G.L. Pearson
FABRIC FILTER EXPERIENCE AT WAYNESBORO 372
W.R. Marcotte
A NEW TECHNIQUE FOR DRY REMOVAL OF SO2 390
C.C. Shale and G.W. Stewart
SPRAY DRYER/BAGHOUSE SYSTEM FOR PARTICULATE AND
SULFUR DIOXIDE CONTROL, EFFECTS OF DEW POINT, COAL
AND PLANT OPERATING CONDITIONS 410
W.R. Lane
SELECTION, PREPARATION AND DISPOSAL OF SODIUM COMPOUNDS
FOR DRY SO SCRUBBERS 425
D.A. Furlong, R.L. Ostop and D.C. Drehmel
ix
-------
VOLUME I CONTENTS (Cont.)
Page
HIGH VELOCITY FABRIC FILTRATION FOR CONTROL OF
COAL-FIRED BOILERS 432
J.C. My cock, R.A. Gibson and J.M. Foster
EPA MOBILE FABRIC FILTER - PILOT INVESTIGATION OF
HARRINGTON STATION PRESSURE DROP DIFFICULTIES 453
W.O. Lipscomb, S.P. Schliesser and V.S. Malani
PASSIVE ELECTROSTATIC EFFECTS IN FABRIC FILTRATION ... 476
R.P. Donovan, J.H. Turner and J.H. Abbott
A WORKING MODEL FOR COAL FLY ASH FILTRATION 494
R. Dennis and H.A. Klemm
Section C - Scrubbers
PARTICULATE REMOVAL AND OPACITY USING A WET VENTURI
SCRUBBER - THE MINNESOTA POWER AND LIGHT EXPERIENCE . . 513
D. Nixon and C. Johnson
PERFORMANCE OF ENVIRONMENTALLY APPROVED NLA
SCRUBBER FOR SO2 529
J.A. Bacchetti
DESIGN GUIDELINES FOR AN OPTIMUM SCRUBBER SYSTEM. ... 538
M.B. Ranade, E.R. Kashdan and D.L. Harmon
TESTS ON UW ELECTROSTATIC SCRUBBER FOR PARTICULATE AND
SULFUR DIOXIDE COLLECTION 561
M.J. Pilat
EPA MOBILE VENTURI SCRUBBER PERFORMANCE 570
S. Malani, S.P. Schliesser and W.O. Lipscomb
THE RESULTS OF A TWO-STAGE SCRUBBER/CHARGED
PARTICULATE SEPARATOR PILOT PROGRAM 591
J.R. Martin, K.W. Malki and N. Graves
AUTHOR INDEX 616
-------
VOLUME III
PARTICULATE CONTROL DEVICES
Section A - Scrubbers
Page
FLUX FORCE/CONDENSATION SCRUBBER DEMONSTRATION
PLANT IN THE IRON AND STEEL INDUSTRY 1
R. Chmielewski, S. Bhutra, S. Calvert, D.L. Harmon, J.H. Abbott
COLLECTION CHARACTERISTICS OF A DOUBLE STAGE SCRUBBER
TO ELIMINATE THE PAINT MIST FROM A SPRAY BOOTH .... 16
T. Isoda and T. Azuma
APPLICATION OF SLIPSTREAMED AIR POLLUTION CONTROL
DEVICES ON WASTE-AS-FUEL PROCESSES 25
F.D. Hall, J.M. Bruck, D.N. Albrinck and R.A. Olexsey
EVALUATION OF THE CEILCOTE IONIZING WET SCRUBBER ... 39
D.S. Ensor and D.L. Harmon
DEMONSTRATION OF A HIGH FIELD ELECTROSTATICALLY
ENHANCED VENTURI SCRUBBER ON A MAGNESIUM FURNACE
FUME EMISSION 61
M.T. Kearns and D.L. Harmon
DROPLET REMOVAL EFFICIENCY AND SPECIFIC CARRYOVER
FOR LIQUID ENTRAINMENT SEPARATORS 81
J.H. Gavin and F.W. Hoffman
AN EVALUATION OF GRID ROD FAILURE IN A MOBILE
BED SCRUBBER 95
J.S. Kinsey and S. Rohde
OPERATION AND MAINTENANCE OF A PARTICULATE SCRUBBER
SYSTEM'S ANCILLARY COMPONENTS 104
P.A. Czuchra
LOWERING OPERATING COSTS WHILE INCREASING THROUGHPUT
AND EFFICIENCY OF REACTORS AND SCRUBBERS 117
R.P. Tennyson, S.F. Roe, Jr. and R.H. Lace, Sr.
OPTIMIZING VENTURI SCRUBBER PERFORMANCE THROUGH
MODELING 127
D.W. Cooper
THE IMPACT OF HUMIDIFICATION CHAMBER PHYSICS ON
WET GAS CLEANUP SYSTEMS 145
D.P. Bloomfield, M.L. Finson, G.A. Simons and K.L. Wray
xi
-------
VOLUME III CONTENTS (Cont.)
IMPROVING THE EFFICIENCY OF FREE-JET SCRUBBERS 162
D.A. Mitchell
Section B - Fabric Filters
HIGH VELOCITY FIBROUS FILTRATION 171
M.J. Ellenbecker, J.M. Price, D. Leith and M.W. First
THE EFFECT OF DUST RETENTION ON PRESSURE DROP IN
A HIGH VELOCITY PULSE-JET FABRIC FILTER 190
M.J. Ellenbecker and D. Leith
ROLE OF FILTER STRUCTURE AND ELECTROSTATICS
IN DUST CAKE FORMATION 209
G.E.R. Lamb and P.A. Costanza
PRESSURE DROP IN ELECTROSTATIC FABRIC FILTRATION .... 222
T. Ariman and D.J. Helfritch
EXPERIMENTAL ADVANCES ON FABRIC FILTRATION TECHNOLOGY
IN JAPAN - EFFECTS OF CORONA PRECHARGER AND RELATIVE
HUMIDITY ON FILTER PERFORMANCE 237
K. linoya and Y. Mori
BAGHOUSE OPERATING EXPERIENCE ON A NO. 6
OIL-FIRED BOILER 251
D.W. Rolschau
NEW FABRIC FILTER CONCEPT PROVEN MORE FLEXIBLE
IN DESIGN, EASIER TO MAINTAIN, AND UNSURPASSED
FILTRATION 260
B. Carlsson and R.J. Labbe
EPRI'S FABRIC FILTER TEST MODULE PROGRAM: A REVIEW
AND PROGRESS REPORT 270
R.C. Carr and J. Ebrey
Section C - Granular Beds
ELECTROSTATIC ENHANCEMENT OF MOVING-BED
GRANULAR FILTRATION 289
D.S. Grace, J.L. Guillory and F.M. Placer
ELECTRICAL AUGMENTATION OF GRANULAR BED FILTERS ... 309
S.A. Self, R.H. Cross and R.H. Eustis
xii
-------
VOLUME III CONTENTS (Cont.)
Page
THEORETICAL AND EXPERIMENTAL FILTRATION EFFICIENCIES
IN ELECTROSTATICALLY AUGMENTED GRANULAR BEDS 344
G.A. Kallio, P.W. Dietz and C. Gutfinger
AEROSOL FILTRATION BY A CONCURRENT MOVING
GRANULAR BED: DESIGN AND PERFORMANCE 363
T.W. Kalinowski and D. Leith
DEEP BED PARTICULATE FILTRATION USING THE
PURITREAT (TM) PROCESS 382
L.C. Hardison
Section D - Novel Devices
PILOT-SCALE FIELD TESTS OF HIGH GRADIENT
MAGNETIC FILTRATION 404
C.H. Gooding and C.A. Pareja
EXPERIENCES WITH CONTROL SYSTEMS USING A UNIQUE
PATENTED STRUCTURE 416
G.C. Pedersen
ELECTROSTATIC EFFECTS IN VORTICAL FLOWS 429
P.W. Dietz
CONDENSATIONAL ENLARGEMENT AS A SUPPLEMENT TO
PARTICLE CONTROL TECHNOLOGIES 439
J.T. Brown, Jr.
Section E - Specific Applications
WELDING FUME AND HEAT RECOVERY - THE PROBLEM,
THE SOLUTION, THE BENEFITS 448
R.C. Larson
PARTICULATE REMOVAL CONSIDERATIONS IN SOLVENT
EMISSION CONTROL INSTALLATIONS 472
E.A. Brackbill and P.W. Kalika
ARSENIC EMISSIONS AND CONTROL TECHNOLOGY - GOLD
ROASTING OPERATIONS 484
J.O. Burckle, G.H. Marchant and R.L. Meek
CONTROL OF SALT LADEN PARTICULATE EMISSIONS FROM
HOGGED FUEL BOILERS 508
M.F. Szabo, R.W. Gerstle and L. Sims
AUTHOR INDEX 526
X11X
-------
VOLUME IV
SPECIAL APPLICATIONS FOR AIR POLLUTION
MEASUREMENT AND CONTROL
Section A - High Temperature High Pressure Applications
Page
FUNDAMENTAL PARTICLE COLLECTION AT HIGH
TEMPERATURE AND PRESSURE 1
R. Parker, S. Calvert, D.C. Drehmel and J.H. Abbott
PARTICULATE COLLECTION IN A HIGH TEMPERATURE CYCLONE . . 14
K.C. Tsao, C.O. Jen and K.T. Yung
EVALUATION OF A CYCLONIC TYPE DUST COLLECTOR FOR HIGH
TEMPERATURE HIGH PRESSURE PARTICULATE CONTROL .... 30
M. Ernst, R.C. Hoke, V.J. Siminski, J.D. McCain,
R. Parker and D.C. Drehmel
CERAMIC FILTER TESTS AT THE EPA/EXXON PFBC MINIPLANT . . 42
M. Ernst and M.A. Shackleton
HOT GAS CLEAN-UP BY GLASS ENTRAINMENT OF
COMBUSTION BY-PRODUCTS 64
W. Fedarko, A. Gatti and L.R. McCreight
THE A.P.T. PxP DRY SCRUBBER FOR HIGH TEMPERATURE AND
PRESSURE PARTICULATE CONTROL 84
R.G. Patterson, S. Calvert and M. Taheri
GAS CLEANING UNDER EXTREME CONDITIONS OF
TEMPERATURE AND PRESSURE 98
E. Weber, K. Hubner, H.G. Pape and R. Schulz
PROGRESS ON ELECTROSTATIC PRECIPITATORS FOR USE
AT HIGH TEMPERATURE AND HIGH PRESSURE 126
G. Rinard, D. Rugg, R. Gyepes and J. Armstrong
REDUCTION OF PARTICULATE CARRYOVER FROM A
PRESSURIZED FLUIDIZED BED 136
R.W. Patch
COMPARATIVE ECONOMIC ANALYSIS OF SELECTED PARTICULATE
CONTROL SYSTEMS FOR ADVANCED COMBINED CYCLE POWER
PLANTS 154
J.R. Bush, F.L. Blum and P.L. Feldman
CONCLUSIONS FROM EPA'S HIGH TEMPERATURE/HIGH
PRESSURE CONTROL PROGRAM 170
D.C. Drehmel and J.H. Abbott
xiv
-------
VOLUME IV CONTENTS (Cont.)
Section B - Fugitive Emissions
Page
WATER SPRAY CONTROL OF FUGITIVE PARTICULATES: ENERGY
AND UTILITY REQUIREMENTS 182
D.P. Daugherty, D.W. Coy and B.C. Drehmel
THE CONTRpL OF DUST USING CHARGED WATER FOGS 201
S.A. Hoenig
SPRAY CHARGING AND TRAPPING SCRUBBER FOR FUGITIVE
PARTICLE EMISSION CONTROL 217
S. Yung, S. Calvert, and D.C. Drehmel
CONTROL OF WINDBLOWN DUST FROM STORAGE PILES 240
C. Cowherd, Jr.
THE CONTRIBUTION OF OPEN SOURCES TO AMBIENT
TSP LEVELS 252
J.S. Evans and D.W. Cooper
FUTURE AREAS OF INVESTIGATION REGARDING THE
PROBLEM OF URBAN ROAD DUST 274
E.T. Brookman and D.C. Drehmel
STATUS OF CONNECTICUT'S CONTROL PROGRAM FOR
TRANSPORTATION-RELATED PARTICULATE EMISSIONS 291
J.H. Gastler and H.L. Chamberlain
NEW CONCEPTS FOR CONTROL OF FUGITIVE PARTICLE
EMISSIONS FROM UNPAVED ROADS 312
T.R. Blackwood and D.C. Drehmel
DEVELOPMENT OF A SAMPLING TRAIN FOR THE ASSESSMENT
OF PARTICULATE FUGITIVE EMISSIONS 321
R.L. Severance and H.J. Kolnsberg
SECONDARY NEGATIVE ELECTRON BOMBARDMENT FOR
PARTICULATE CONTROL 333
W.E. Stock
Section C - Measurement and Analysis
HIGH TEMPERATURE AND HIGH PRESSURE SAMPLING DEVICE
USED FOR PARTICULATE CHARACTERIZATION OF A FLUIDIZED
BED COAL GASIFICATION PROCESS 338
S.P. Tendulkar, J. Pavel and P. Cherish
ON-STREAM MEASUREMENT OF PARTICULATE SIZE
AND LOADING 351
E.S. VanValkenburg xv
-------
VOLUME IV CONTENTS (Cent.)
ANALYSIS OF SAMPLING REQUIREMENTS FOR CYCLONE
OUTLETS 368
M.D. Durham and D.A. Lundgren
ELECTROSTATIC EFFECTS ON SAMPLING THROUGH
UNGROUNDED PROBES 387
W.B. Giles and P.W. Dietz
OPTICAL PARTICULATE SIZE MEASUREMENTS USING A
SMALL-ANGLE NEAR-FORWARD SCATTERING TECHNIQUE .... 396
J.C.F. Wang
IN-STACK PLUME OPACITY FROM ELECTROSTATIC
PRECIPITATOR SCRUBBER SYSTEMS 411
L.E. Sparks, G.H. Ramsey and B.E. Daniel
TI-59 PROGRAMMABLE CALCULATOR PROGRAMS FOR
IN-STACK OPACITY 424
S.J. Cowen, D.S. Ensor and L.E. Sparks
UTILIZATION OF THE OMEGA-1 LIDAR IN EPA
ENFORCEMENT MONITORING 443
A.W. Dybdahl and F.S. Mills
EFFECTS OF PARTICLE-CONTROL DEVICES ON ATMOSPHERIC
EMISSIONS OF MINOR AND TRACE ELEMENTS FROM COAL
COMBUSTION 454
J.M. Ondov and A.H. Biermann
A SOURCE IDENTIFICATION TECHNIQUE FOR AMBIENT
AIR PARTICULATE 486
E.J. Fasiska, P.B. Janocko and D.A. Crawford
PARTICLE SIZE MEASUREMENTS OF AUTOMOTIVE
DIESEL EMISSIONS 496
J.D. McCain, and D. Drehmel
CONTROL STRATEGIES FOR PARTICULATE EMISSIONS FROM
VEHICULAR DIESEL EXHAUST 508
M.G. Faulkner, J.P. Gooch, J.R. McDonald,
J.H. Abbott and D.C. Drehmel
AN EVALUATION OF THE CYTOTOXICITY AND MUTAGENICITY OF
ENVIRONMENTAL PARTICULATES IN THE CHO/HGPRT SYSTEM . 524
N.E. Garrett, G.M. Chescheir, III, N.A. Custer, J.D. Shelburne,
Catherine R. De Vries, J.L. Huisingh and M.D. Waters
AUTHOR INDEX 536
xvl
-------
COLLECTION EFFICIENCY OF ELECTROSTATIC PRECIPITATORS
BY NUMERICAL SIMULATION
BY:
Eric A. Samuel
Buell Emission Control Division
Envirotech Corporation
Lebanon, Pennsylvania 17042
ABSTRACT
A numerical simulation code has been developed to analyze the coupled
electro-aerodynamic phenomena within an electrostatic precipitator operated
in both the positive and negative corona modes. The space charge and
current densities, as well as the static and space charge components of the
electric field, are first obtained using the known cross-sections for ioniza-
tion and negative ion formation. Data includes the effects of fluid flow.
The above quantities (except the static field) are then corrected in a subse-
quent iteration for the presence of dust particles of known size distribution,
charged predominantly by the field charging mechanism. The precipitator
collection efficiency is calculated from the trajectories of the charged dust
particles. For theory comparison, the collection efficiencies of model
precipitators were measured by light-scattering techniques using oil drops as
test particles. The variation of efficiency with wire-to-wire spacing will
be discussed as a case study.
-------
1. INTRODUCTION
Introduction
The sizing of electrostatic precipitators for a given application is
conventionally based on statistics drawn from performance data on existing
precipitators for that application. Such an approach will yield a reliable
sizing criterion only if the performance data base includes a sufficiently
large number of units. The method, however, is limited for sizing precipa-
tors for new applications and for design improvement. Computer-assisted
modeling of the precipitation process is an alternative sizing procedure which
overcomes the above limitation.
In this paper, we discuss a computer program which simulates the important
physical phenomena within an idealized electrostatic precipitator (to be defined
below) without the use of differential equations. Obtained by simulation are
the corona space charge density in the presence of air flow, the charging of
particulates and their subsequent drift in the total electric field, the effect
of the interaction among charged particulates (the concentration effect and,
eventually, the collection efficiency) by number density of the idealized
precipitator containing N wires in the gas flow direction. Also described in
the paper are the results of experiments performed using laser light scattering
techniques to test the predictions of the computer simulation. Because of the
good agreement between theory and experiment within the assumptions of an
idealized precipitator, our simulation code has the potential of becoming the
starting point for a computer model which includes the effects of non-ideal
behavior in industrial precipitators.
In order to demonstrate the use of our simulation code in design improve-
ment, we discuss as a case study the effect of varying one geometric parameter;
namely, the wire-to-wire spacing on the performance of the idealized precipi-
tator. We have found that for a given plate-to-plate spacing, there is an opti-
mum value of the wire-to-wire spacing. The efficiency maximum, however, is not
very sharp. The above finding, which we also confirmed experimentally, is
significant since it suggests that an increase in the number of wires per gas
passage over the current industry standard will result in increased performance
with the same plate area even though the wire-to-wire spacing is smaller than
the optimum value.
An Idealized Electrostatic Precipitator
Figure 1 shows a cross-sectional view of the studied model precipitator to
establish the scaling of the wire spacing (2L) to plate spacing (2W) ratio. The
precipitator extends to a height H perpendicular to the plane of the page such
that L/H«1. The electrodes are of smooth conducting sheets at the uniform
ground potential. The corona wires are under sufficiently large tension and
assumed to be free of oscillations in order to maintain uniform distances to
adjacent wires and plates. (See Reference 1 for the analysis of the wire
-------
oscillation.) The wire diameter D is chosen so that the ratio, D/W, is in the
same range as that for industrial precipitators; i.e., D/W ~3 x 10~2.
Ground Plate Electrode
I J
JZ,
aw i — - £
t ^
w +
'
+
single-
wire
section
•"•••
+t°+ + +
Qround Plate Electrode
i
r—-
[— — Fl
I
cr
i
Figure 1. A top view of the model precipitator.
The wires and plates extend to a height
H, normal to the plane of the figure.
The inlet of the precipitator consists of a long, leading section of
parallel plates without wires. The mean flow speed is comparable to those of
industrial precipitators. However, the flow Reynolds number is kept smaller
than the critical value for transition to turbulent flows by keeping the plate
spacings smaller than the industrial values. As a result of these restrictions,
the flow acquires the fully developed Poiseuille profile upstream of the first
corona wire. Depending on the wire spacing, the flow may recover the Poiseuille
profile downstream of a wire before encountering the next wire, and so on. The
exit also has a long trailing section of parallel plates without wires to insure
the desired flow field characteristics. The flow field has been measured in
great detail by photographing tracer particles under illumination by an argon
ion laser beam.
The host medium is synthetic air consisting of 80% nitrogen and 20% oxygen
at room temperature and one atmosphere. The corona operating current is chosen
in the range about a factor of 5 above the industrial range of around 20 yA/ft
of a corona wire. The corresponding potential difference depends sensitively
on the ratio L/W for small spacings, and this is included in the analysis.
Polydispersed oil droplets have been used for this study. Measurement
shows a Nukiyama-Tanasawa distribution of the respirable range with the maximum
at 0.3 ym in radius. The particulate number density ranges from 10^ through
10^ particles/cm^.
The precipitator is operated under clean electrode condition, both in
experiment and simulation in order to insure reproducibility. This is also
necessary because of difficulty in modeling the so-called aging of the precip-
itator.
The corona operates in both the positive and negative D.C. corona mode.
Arcing and current modulation are assumed to be absent. In experiment, this is
realized by using current limiting techniques with high voltage power supplies.
-------
The differences between the idealized precipitator and industrial precip-
itators obviously are numerous. The industrial precipitators have the follow-
ing additional features.
(a) Considerable level of turbulence in the flow due to higher Reynolds
number and the baffles and gaps in the ground plate.
(b) Substantial thickness of weakly conducting dielectric layers of
collected particulates on all electrodes.
(c) Frequent arcing between wires and plates, and the so-called back
corona resulting from breakdowns in the dielectric layers.
(d) Large amplitude oscillations of wires.
(e) Re-entrainment of collected particulates during removal by rapping
of the plates.
(f) Complex shapes and the interior structures of particulates.
(g) Contamination of the flue gas by a large number of gaseous combustion
products which affect space charge formation.
(h) Deliberately added conditioners to influence conductivity and adhesion
properties of particulates.
(i) Full wave rectified but unfiltered power supplies.
(j) Higher operating gas temperature, and pressure variations.
(k) Large fluctuations in the particulate number density at the inlet.
(1) Sedimentation of particulates.
(m) Particulate agglomeration.
Once agreement between theory and experiment is established for the ideal-
ized precipitator, the non-ideal effects listed above can be included in the
computer program; one at a time, starting with the most significant. Experi-
ments must be designed to test the predictions of the program after inclusion
of each non-ideal effect.
An Outline of the Paper
Particulate precipitation takes place in a precipitator in the following
manner. The particulate is carried into the precipitator by its surrounding
fluid element at the local flow speed. (Particulates in the respirable size
range follows the streamlines accurately, except around the corona wire where
the inertia effect is strong. See Reference 2 for detailed analyses.) Upon
entering the corona field of the first wire, the particulate acquires charges
in the form of ions or electrons. The corona creates a region of positive ions
by electron impact ionization of air molecules (positive corona), or negative
4
-------
ions by attachment of electrons to molecules (negative corona) around the
wire. The particulate, now polarized in the electric field of the high volt-
age wire, attracts the ions and electrons. Thermal fluctuations of
the ions and particulate, regardless of the wire electric field, also contrib-
utes to charging. In principle, the charging continues throughout the partic-
ulate motion until the Coulomb repulsion of the acquired charge can hold off
further diffusion of ions and attraction by polarization.
As the amount of electrical charge on the particulate increases, the
Coulomb force on the particulate begins to accelerate it against the drag of
the fluid. A terminal velocity results in the direction of the local electric
field line, while the velocity along the streamline persists. The eventual
particulate trajectory evolves out of the competition between these two velo-
city components, which in turn change according to how the particulate trajec-
tory is unfolded.
Since the maximum amount of charge depends on the particulate size, elec-
tric field strength, and space charge density, the particulate may not reach
the ground plate within the distance 2L. If it does not, the drifting can
continue in subsequent sections until the particle reaches the plate or leaves
the precipitator still entrained in the flow.
In finer details, the trajectory of the particulate depends on other
surrounding particulates; whereas, the space charge density is modified by the
spatial and size distribution of the particulates and the flow field. In
short, precipitation results from a set of strongly coupled physical processes.
The present investigation systematically separates individual processes, and
then combines them and evaluates the effects of their coupling on the precipi-
tation efficiency.
In Chapter 2, we describe calculations of the space charge density and
electric field everywhere in a two-dimensional section defined by 2L and 2W,
with a corona wire located at the center (see Figure 1). It includes detailed
descriptions of such calculations for both the positive and negative corona in
the presence of the flow field. A check of our numerical simulation technique
is described, which calculates a coaxial positive D.C. corona and compares the
result with the analytical solution from Poisson's equation. In this test, of
course, the flow has been turned off.
In Chapter 3, we discuss particulate trajectory calculations for a single-
wire precipitator, including the charging process and the effects of the partic-
ulate number density and size distribution. Extensions to multi-wire precipi-
tators are described next. The results of the multi-wire calculations are then
compared with the Deutsch equation. A framework for reducing a single-wire
efficiency from a multi-wire precipitation efficiency is formulated, which
becomes the basis for analyzing the scaling of wire-to-plate spacing ratio.
We then present the experimental results for the collection efficiencies of
several multi-wired research precipitators obtained by light scattering tech-
niques, and discuss the scaling of the L/W ratio on the basis of both experim-
ental and theoretical data. Concluding remarks are given in Chapter 4.
-------
2. CALCULATION OF SPACE CHARGE DENSITY AND ELECTRIC FIELD
FOR A PLATE-WIRE CORONA WITH TRANSVERSE FLOW
Introduction
Prior to treating the fundamental processes that sustain the positive and
negative corona in air, we first obtain expressions for the electric potential
and the electric field in the region between the wire and the plates using the
method of conformal transformations. This method is advantageous because the
expressions are in closed forms and are easily extended to the case of a multi-
wire precipitator.
Electrons play the central role in initiating and sustaining both the
positive and negative corona. In describing the positive and negative corona,
especially the negative corona, knowledge of the average electron energy in an
electric field is essential because it determines the rates that ions are
generated either by electron impact ionization of air molecules or by electron
attachment. It is generally known that the drift speed of an electron is
proportional to the electric field strength over a large range of electric
fields and gas densities. This is because an electron can be accelerated in
the field until the energy gain of the electron becomes equal to the energy
loss by collisions with molecules. When the molecules are monatomic, the
electron-molecule collisions are mostly elastic. When the gas molecules are
polyatomic, as in air, the electron-molecule collisions are more often inelastic
and the average electron energy evaluation requires attention to the specific
molecular structure. A detailed analysis is given in Reference 1. A simpli-
fied procedure for calculating the average electron energy in polyatomic gases
is discussed.
We will then discuss the main processes that operate in a positive corona
and the way these processes are simulated in our computer code. A similar
discussion follows for the negative corona. In both simulations, we assume that
air consists of 80% nitrogen and 20% oxygen only, and ignore other trace gases.
In the negative corona, it is assumed that only oxygen molecules can form stable,
long-lived negative ion species. The cross-sections for both ionization and
negative ion formation by electron impact are taken from the literature. The
mobilities of nitrogen and oxygen ions are also known.
In the numerical code for the corona, we also include the effect of fluid
flow. Analytically, one has to solve Poisson's equation, the continuity equation
and the Navier-Stokes equation simultaneously. Since this approach is rather
laborious, the flow field is measured instead, and the space charge calculation
carried out in that flow field. Due to the relatively low corona current range,
the influence of the corona wind on the flow field can be neglected. The ability
to include the effects of flow in this way is one of the chief advantages of our
numerical code.
In each simulation, the corona current and voltage are inputs. This means
prior knowledge of the corona characteristics. They are obtained from measure-
ment using the actual multi-wire research precipitator. We note that the current-
voltage characteristic depends on the wire spacing as well as the wire radius
plate spacing and electrode materials. Generally, the characteristic curve is
parabolic near the corona threshold.
6
-------
A detailed description of the algorithms used in the numerical code is
given next. This is followed by a discussion of the methods by which we
verify that the results of our numerical simulations are consistent with the
fundamental laws of physics. The final section contains a discussion of the
salient features of the electric field distribution within a precipitator.
The Solution of Laplace's Equation for the Plate-Wire-Plate Geometry
A solution to Laplace's equation in to independent variables x and y can
be formulated in terms of analytic functions of a comples variable since the
real and imaginary parts of such functions are harmonic. Details of the calcu-
lation using the theory of complex variables fot the potential cj)(x,y) within
a plate-wire precipitator, sketched in Figure 1, is given in Reference 3-
The result is:
N-l A - B
4>(x,y) = -A E £n ( " _ ),
p=0 P
where Ap = cosh j(y-2pL)/2W and B = CQs(rrx/2W). The x and y components of the
electric field, E, are obtained from E = - grad 4>(x,y):
ATT B /1-
B
*"" W A2-B2
P
T, , *K ^ 1~AP
Ey(x,y) = jr- -5 »~
A - R
P B (2)
The Average Electron Energy in the Electric Field
Evaluation of the average electron energy C(E) in the electric field of
strength E is, in general, very complicated and requires attention to the
details of energy losses during the Coulomb acceleration of the electrons. The
energy losses stem from (1) elastic collisions with molecules in which transfer
of a small fraction of the electron kinetic energy to a target molecule takes
place, (2) inelastic collisions with molecules in which an electron gives up a
fixed amount of energy to a molecule and the molecule becomes excited internally.
A proper accounting of the inelastic collisions can be done in principle but it
requires full knowledge of the molecular energy levels of a given target molecule
and the probability of an electron-molecule collision resulting in excitation to
each molecular energy level.
An approximate, but considerably more detailed, analysis of £(E) in air is
described in Reference 1. We have found that for consideration of both ioniza-
tion and electron attachment in the corona of air, the maximum distance of ten
collisional mean free paths is a good compromise for the distance of electron
-------
acceleration beyond which further acceleration is offset by inelastic energy
losses. Thus, throughout the space charge density calculation, we calculate
the average electron energy as:
= lOXoeE, Xo - (n a)'1 (3)
where n is the number of target molecules per unit volume and Xo is the cross-
sectional area of the target molecule. An experimental confirmation for the
rather gross simplification of equation (3) exists in that the radius of the
ionization zone calculated using equation (3) agrees with laboratory measure-
ments.
The Positive Corona
The positive corona is initiated by residual electrons in the gas, created
by cosmic rays. This electron moves towards the wire at positive high voltage
with respect to the plates and reaches the ionization zone around the wire
where £(E) exceeds the ionization potential of molecular oxygen (A£j = 12.5 eV) .
(A£i for molecular nitrogen is 15.7 eV.) An ionizing collision produces a
positive ion of the target molecule and an additional electron. This results
in electron multiplication or avalanche. In this way, after traversing several
ionization mean- free paths, the electrons multiply in number, until at the wire
the total corona current is carried entirely by electrons.
The electron multiplication factor, ot, for an electron passing through the
entire ionization zone is simply given by the total number of ionizing colli-
sions over Rj, the radius of the ionization zone:
RI
1 dr/XT
a = 2 D/2
(4)
_i
where X j = [n0 i(?)l • °l(C) is the ionization cross-section and depends on
the electron energy £(E) , thus becoming position dependent.
Of course, the ions are accelerated too, but due to their large mass, their
drift velocity remains much smaller than the electrons. For this reason, the
space charge density, inside and outside the ionization zone, is determined by
the ions. While the total corona current is carried entirely by electrons at
the wire, outside the ionization zone it is carried entirely by the positive
ions which are produced in the ionization zone.
In the simulation, we divide a single wire section of the multi-wire precip-
itator into a 20 bin x 20 bin mesh and calculate the x and y components of the
electrostatic field of Equation (2) at centers of the 400 cells. The charge per
unit length of the wire, X, is obtained from the boundary condition that the
potential at r=D/2, is the corona voltage, Vc, as given by the power supply.
D is the wire diameter.
The space charge density and the electric field due to the space charge
are then obtained as follows: The corona current is represented by a certain
8
-------
number 2 An, of line charge packets introduced into the field at the outer
surface of the ionization zone in a time interval At. Each charge packet is
assigned a number of elemental charges amounting to the magnitude of q. An,
At and q are related by:
where ic is the corona current. The actual values of An and At are chosen as
a compromise between the computing time and the desired accuracy of the result.
The charge packets are created within the ionization zone and their exact
departure point on the surface of the zone is determined randomly by a random
number generator. While the charge packets are long lines parallel to the wire
and interact among themselves as line charges, their motion is governed by the
rules of single ions because the ions in a packet are not bound to each other.
We assume that negative ion production within the ionization zone by electron
attachment is negligible.
The positive ion packets created in a time step as well as all previously
created positive ion packets are initially moved under the influence of the
electrostatic field due to the wires and the flow field. The local flow velo-
city is vectorially added to the electrostatic drift velocity of an ion to get
the total ion velocity. At every few time steps, the ion packets are frozen
at the closest grid points and the electric field, due to all of the ion
packets and their images at the plates, are calculated and added to the electro-
static field of the wires. The charge packets are then allowed to move during
the next time interval At in this new, total electric field coupled to the flow
field. At the end of the interval, the field is recalculated. The iteration
is continued until a steady state is reached; i.e., when the total number of
ion packets in the region between wire and plate is constant and 2An packets
per time step are collected at the plates on the average. The steady state
total electric field due to the ion packets, and the spatial distribution of the
ion packets (i.e., the space charge density), are averaged over several itera-
tions and retrieved in IBM cards for use in the second phase of the simulation.
The Negative Corona
Since the algorithms and numerical techniques for simulating the negative
plate-wire corona are essentially the same as for the positive corona, we will
only describe how the physical processes are taken into account in the simula-
tion. When the wire is maintained at a potential negative with respect to the
plates, the positive ions move toward the wire and electrons, and the negative
ions move toward the plates held at zero potential. The negative corona is
initiated by stray electrons in the ionization zone. As the electrons move
outward along the field lines, they will produce more electrons in the ioniza-
tion zone.
As these electrons leave the ionization zone, their average energies will
progressively decrease because of the decreased electric field strength. The
electrons reach a region where their energies are in the range favored for the
dissociative attachment reaction:
9
-------
+ e -K) + 0
There are many reactions by which these 0~ ions can form neutral molecules,
electrons, or other negative ions. We assume, however, that the attachment
rate is large enough to ignore depletion of the 0~ ions by secondary reactions.
Some electrons reach still another region in which their energies are suitable
for the direct attachment reaction:
Again, the electrons, because their mobilities are so much larger than
those of ions, do not contribute to the space charge density. But they induce
the formation of negative ions which do contribute to the space charge. Our
consideration of only dissociative and direct attachment to oxygen molecules to
the exclusion of all other negative ion producing reactions is by_no means an
accurate picture. Many other negative ion species such as 0^, NO , N02, have
been detected. It is expected that inclusion of the most abundant negative
ions is adequate to yield a representative field due to the negative ion space
charge density in the precipitator.
The simulation proceeds as follows. 2An packets of electrons per time
interval, At, are produced at the surface of the wire. The charge carried by
a packet is initially given by Jq, where q satisfies Equation (5). The factor,
y, is related to the rate of the electron multiplication in the ionization zone.
The electron packets are moved within the ionization zone radially along field
lines. The ionization zone is divided into 20 concentric rings and 16 equally-
spaced radial lines, superimposed in the 20 x 20 mesh field of the single wire
section, as shown in Figure 3. A grid point in the ionization zone has the
polar coordinates (r, 6):
r(i) = (Ar/2) (21-1), 1=1, ..., 20
6(j) = (A9/2) (2j - 1), j = 1, ..., 16, (6)
where Ar = (RT -D/2)/20 and A6= TT/8.
10
-------
-10
-5 -11
Ground Plate
10
Figure 2. The plan for division of the ionization
zone and the 20x20 rectangular mesh field
of the plate-wire-plate corona.
As an electron packet passes each ring, the number of ionizing mean-free
paths in the annulus between successive rings is calculated, assuming the
field to be uniform over the annulus. The number of additional electrons
produced in this annulus is 2^/^-1. To include this increase in the number of
electrons, the charge on an electron packet is increased by 2-L/^jq, instead of
increasing the number of electron packets. The electron packet, in passing
the annulus, also leaves behind (2l/^i -1) of positive ions.
The positive ions created in this way are randomly distributed around the
grid point closest to the position of the electron packet. Each electron
packet is moved until it reaches the edge of the ionization zone. The value of
the charge, when the charge packet arrives at the ionization zone edge, is com-
pared with the q of Equation (5). The factor yis initially started at a low
value, say, 10~^, and is successively doubled until the electron packet, which
left the wire surface with a charge , arrives at the edge of the ionization
zone with the charge q. That is, Y is increased during the iteration self-
consistently until one satisfies the relationship:
= 1
(7)
The electron packets are now moved outside of the ionization zone along
field lines. As they cross a line of centers of the cells or grid line of
constant x, their average energy is calculated from Equation (3) and the number
of negative ion packets they produce by attachment to molecules is calculated
11
-------
using the known cross sections. The negative ion packets formed along each
step are randomly distributed in the cell where the grid point is located.
Thus, each timestep consists of moving 2An electron packets, randomly
formed on the surface of the wire, through the ionization zone while^increas-
ing in the amount of charge per packet up to q or through the remaining space
to the plates with a fixed amount of charge q per packet. In the same timestep,
the positive and negative ions created during all preceding timesteps are also
moved. At regular time intervals, the electric field due to the positive ions
is calculated. Since the radius of the ionization zone is small compared to L
or W, the contribution from the positive ions inside the ionization zone is
calculated in the form of the increasing shielding of the charge on the corona
wire (-X). A symmetric distribution of the positive ions around the wire is
assumed for this purpose. The field due to the negative ion packets is found
by the same image charge method as for the positive corona. The calculations
are continued until steady state is reached. By keeping track of the average
number of times an electron packet crosses a given grid point (&,m) in At, we
can obtain the current density J(£,m), of electrons at the grid point. Again,
the total current density is given by the electrons while the space charge
density is by the ions. Note that the flow velocity is again added vectordaily
to the electrodynamic drift velocity in order to determine the total ion velo-
city.
For the purpose of calculating the rate at which particulates are charged
in the precipitator, we retrieve from the simulation code and sto^re the nega-
tive ion space charge density p (£,m), the total electric field E(&,m) and the
ratio J(£,m)/E(&,m). The ratio J(£,m)/E(5,,m) is equal to pe(£,m) ke where
Pe(&,m) is the electron space charge density and ke is the electron mobility.
The product p~(5,,m)k~ is also stored, where k~ denotes the negative ion mobility.
Computer Algorithms for the Simulation Code
Definition of Grid Points. A grid point refers to the center of a cell
and is identified by indices (£,m). Its x and y coordinates are:
x(£) = (2£-l)ax/2 and y(m) = (2m-l)ay/s (8)
where ax = W/10 and a = L/10
Electrostatic Field of the Wires. In order to commence on the calcula-
tion of the x and y components of the electrostatic field due to the
wires, we first determine A, the charge per unit length of the wire, from
the corona voltage Vc and Equation (5). From the A-value, the x and y
components of the electric field, Ex(£,-m) and Ey(£,m) are calculated at
each of the 400 grid points using Equation (2).
Algorithm for Moving Charge Packets. The x and y coordinates of all
charge packets at the current iteration (x1( y^), and the iteration immedi-
ately preceding (xo, y0) are stored in the program. A charge packet is
created randomly within the ionization zone at a point (XQ, y ). In the
12
-------
same time step it is created, the packet is moved by the algorithm:
K! = x0 + vx(x0) At
yi = y0 + Vy(y0) At
In subsequent time steps, the packet is moved by the slightly different
algoritym:
x2 = x0 + vx(Xl) (2At)
X3 = xl + Vx(x2> (2At)
and so on, where
xl = xo +fvx(xo) + vx(x1)]At/2
The same is done to calculate the y- components.
The velocities of equations (9) and (10) are determined as follows:
vx(x,y) = k± [Ex(Ji,m) + E^(H,m)]
vy(x,y) = k± [Ey(£,m) + Ey(£,m)] (11)
where ki is the mobility of the positive or negative ions, Ex (&,m) and
Ey(£,m) are the components of the electrostatic field of the wires alone,
and Ex(£,m) and E'(£,m) are the space charge field components. Since the
ions move from the high field region near the wire to the low field
region near the plates, Equation (11) is designed to correct for initial
overestimation of the displacement of the ion packet in the preceding time
step.
Electric Field Due to the Space Charge Density. This particular segment
of the calculation takes up most of the processing time in a typical run.
At regular time intervals, the ion packets throughout the region between
the wira and the plates are frozen and assigned to the nearest grid points
in order to find the distribution of the number of the charge packets,
Q(£,m). The space charge density is then obtained:
pa,m) = Q(£,m) q/axay (12)
Figure 4 illustrates the method of calculating the space charge electric
field in a single wire section of the precipitator. The section is divided
into four quadrants labeled I, II, III, and IV. Let us consider two charge
packets at grid points (&,m) and (n,k) . For the two real positive charge
packets in quadrant I, we show two real positive charge packets that can
always be found in quadrant II from the symmetry about the plane of wires.
Also shown are the image packets on the bottom side of the lower conducting
plate, which is labeled as quadrant V. Of the infinite number of image
charges that exist for the pair of the parallel conducting plates, we
consider only those images closest to the charge packets in quadrants I to
calculate the field in the quadrant of interest.
13
-------
symmetric!
partners '
T rcn ' - ,
i i ' '(*»)n^ i
real j
charges __i_
V image !
charge^
(l,m) |
(|,k>
Figure 3. The arrangement of image and symmetric
partner charge packets for two ion packets
in a plate-wire corona section.
The electric field at the grid point (£,m) of Quadrant I is calculated
from five charge packets: the real packet at (n,k) in Quadrand I, its
symmetric partner in Quadrant II and its image in Quadrant V; both the
symmetric partner in Quadrant II and the image packet in Quadrant V of
the real charge packet located at the grid point (£,m). Writing out the
x and y components of the electric field, we have
X
a
(£,m) = 2q -|
a <2N + 1-fc-n) a (jj+n-l)
a2(2N +l-£-n)2 + a2(m-k)2 a2(£+n-l)2 + ay(m-k)2
x o y x
1 + 1
ax(2A-l) ax(2No-l-2£) (13)
14
-------
and
ay (m-k)
- 2q ~2 ^
/ rt \ *• i "/ 1 \
a (Jo-n) + a (m-k)
a (m-k) a (m-k)
a2(2No+l-£-n)2 + a2(m-k)2 a2(£+n-l)2 + a2(m-k)2 '(14)
where N is the number of bins in L or W; i.e., N = 10 in this case.
Notice that the charge packets are infinitely long lines of charges. One
can similarly determine the field components at the grid point (n,k).
The total space charge field due to all charge packets in the entire
region is found by summing all individual charge packet contributions of
the type shown above.
When the corona wires are close together; i.e., L/W ^1, it is necessary to
include the field due to the space charge in the neighboring sections also.
To ease the use of computer processing time, we make the approximation
that most of the contribution to the field in Quadrant I comes from the
space charge in the left half of Quadrant VII of the section to the right,
and that in Quadrant IV from the right half of Quadrant VIII of another
section on the left. Expressions of Equations (13) and (14) are easily
adapted to include the effect of the space charge in the neighboring
sections.
The Flow Field. In this work, we use the velocity flow field obtained
experimentally using the tracer particle method. The experimental setup
and technique are given in Reference 1. Results for a single wire precip-
itator are shown in Figures 4(a) and 4(b). The x and y components of the
velocity, respectively, are shown in a dimensionless form. Since only the
y component of the velocity exists upstream and far downstream from the
wire and it obeys the Poiseuille law, the velocity scale on the graphs is
determined by requiring that the flow rate for unit plate spacing is unity;
i. e.,
u J (1 - x2) dx = 1,
Or, uo = 3/4. Thus, the maximum value of the y component of the normalized
velocity is 0.75. The actual velocities for a given flow rate are obtained
by multiplying the normalized velocities by Q/HW, where Q is the volume
flow rate.
It can be seen from Figures 4(a) and 4(b) that recovery of the flow down-
stream of the wire to the Poiseuille profile requires a distance compara-
ble to the plate spacing (2W). This holds for the full range of laminar
flows we have considered in the present study. As a result, the multi-wire
flow field
15
-------
Fig. 4(a). The measured laminar velocity
flow field for a single-wire
precipitator section. The
normalized x-component.
///// ///
// fi
fj /////// i/
--
u
iiiiiiliiiiiir,
Fig. 4(b). The measured laminar velocity
flow field for a single-wire
precipitator section, the
normalized y-component.
16
-------
is obtained by simply piecing together the single-wire flow field for all
values of L/W considered here. The practice should be reasonable down to
L/W =1, but below this the accuracy of the simulation suffers noticeably
as will be seen later.
Verification of the Simulation Code. The algorithms used in our numeri-
cal code are checked by two methods. The first method verifies that the
continuity equation is satisfied. The second method compares the electric
field and space charge density given by the analytical solution of
Poisson's equation for a positive coaxial corona configuration with those
obtained by the simulation. In the first case, the checking is done
computationally by verifying that all the charge packets crossing planes
parallel to the plate electrodes of the plate-wire-plate corona add up to
the total corona current. In fact, this check is made routinely while
the simulation is in progress. Details of the second check are given in
Reference 3. The agreement between the analytical results and those by
numerical simulation is extremely good outside the ionization zone, and
this is taken as evidence that our numerical code is consistent with the
fundamental laws of electro-magnetism and kinetic theory.
A Discussion of the Calculated Electric Field. Generally, the electro-
static field of the wire decreases monotonically as one goes away from the
wire, while the space charge electric field increases as one approaches
the plates. In addition, the space charge field shows two peaks; one on
the upstream side of the wire and another on the downstream side. The
downstream field is of flightly larger magnitude than the upstream. The
latter feature is due to the streaming of ions at the local flow velocities.
The most interesting aspect of the results shown has to do with the fact
that the electric field near the plate electrodes gets most of its contri-
bution from the space charge density. This means that the final collection
of particulates depends to a large extent on the space charge field, and
it underscores the importance of those molecular species susceptible to
ionization (positive corona) or electron attachment (negative corona) in
electrostatic precipitation.
3. CALCULATION OF THE EFFICIENCY OF A MULTI-WIRE ELECTROSTATIC PRECIPITATOR
Introduction
In this phase of calculation, the charging of initially uncharged dust
particles is considered and their subsequent trajectories in the precipi-
tator are calculated. From these trajectories, the collection efficiency
of the precipitator is determined. In this chapter, we first discuss the
rate at which a dielectric sphere becomes charged in the corona field.
The size distribution function of test particulates — namely, oil
droplets produced by an atomizer used in our experiments — will be dis-
cussed, together with the way the size distribution is incorporated in
the calculations. The latter discussion will include a description of
how the effect of charged particulates is accounted for in the space
17
-------
charge density and the electric field (the concentration effect). The
procedure for obtaining the collection efficiency of a single-wire
precipitator is given, followed finally by the procedure for obtaining
the collection efficiency of a multi-wire precipitator.
The Charging Rate of Particulates
Particulate charging is usually discussed for two mechanisms: charging
by polarization of particulates in the electric field, commonly known as
ion bombardment charging, and charging by thermal motion of ions, commonly
known as diffusion charging. A rigorous theory includes both mechanisms
and gives quite different results than shown here.l
An approximation considered in Reference 2 gives the following result
for the rate of field charging of a dielectric sphere of radius R:
qq(O = q_(») (1 - e~t7tc)
2
where t£ = (2neQ nQk )~ and qg (°°) = ^ £ +^2c ° (15>
no being the ion density far away from the dust particle, eo being the
electronic charge, k being the mobility of the ions and £o, £ being the
dielectric constant of the fluid and material of the sphere. The relaxa-
tion time, tc, is of the order 10~4 seconds under conditions considered
here, and this justifies the instantaneous charging of the particulates
to qs(°0 in the simulation.
Charging by diffusion involves collisions of ions with the target par-
ticulate by random notion of ions and subsequent adhesion due to attraction
by their image charges. The size distribution of the particulates consid-
ered in the present study can be shown to be in the range where field
charging predominates. We, therefore, ignore diffusion charging in our
study.
The Particulate Size Distribution Function
The particulates used for the present study have been the droplets of
heavy mineral oil produced by an atomizer. By measuring the sedimentation
velocities of several hundred oil droplets, the size distribution of the
oil droplets has been obtained previously in our laboratory.^ A least-
squares fit of the experimental distribution to the Nukiyama-Tanasawa
distribution has been made:
F(R) = A R2 exp (-BRS) (16)
where B = 2R~s/s, s =#1 and Rm = 2.58xlCT5 cm. Rm is the particulate
radius at which the distribution function has a maximum.
18
-------
0 0.2 0.1* 0.6 0.8 1.0
Fig. 5.
The measured particulate size
distribution function and a
discretized fit. The vertical axis
gives the normalized probability of
finding a particulate of given radius.
in i i i i i rn i i
Fig. 6. Calculated particulate trajectories over four
single-wire precipitator sections for L/W = 3,
i£ = 33 yA/wire, Q = 660 cm3/sec and np(0) =
10? cm~3. Four different size particulates are
introduced at each inlet position. Notice that
the largest of them reaches the plate first.
19
-------
In our calculations, we have approximated the measured distribution
with a histogram containing only 4 size divisions, as shown in Figure 5»
This approximation was dictated by the limited computer memory space.
In the histogram, the probabilities of finding particles with radii
2 x 10 , 4 x 10~5, 6 x 10~5 and 8 x l(T5cin are 0.418, 0.354, 0.167 and
0.060, respectively. We denote these probabilities by f]_, ±2t £3 and £4,
respectively. In order to simulate a random array of different size
particulates in space, a random number generator is employed. Since all
numbers between 0 and 1 are produced with equal likelihood by the random
number generator, it is first decided which of the four ranges, 0 to f^,
fl to f± + f.2, ±i + f£ to f! + f2 + f3, and f± + f2 + £3 to 1, a given
random number belongs to. When this number lies in the ith range (i =
1,2,3,4), a particulate of the ith radius is called out into the precipita-
tor under simulation. The distribution of particulate sizes produced in
this way indeed becomes identical with the histogram of Figure 5 when
averaged over a large number of particulates. The particulate number
density in the flow, of course, is determined by how frequently such
particulates are called out, each by a newly generated random number.
A Single Wire Precipitator
From a given flow rate, Q, the maximum flow velocity, uo, of the
Poiseuille profile is determined by uo = Q/HW. The normalized velocity
flow field discussed in Chapter 2 is multiplied by uo in order to estab-
lish an actual flow field for the full simulation.
Initially uncharged particulates are introduced at the entrance of the
first stage (y = -L) as packets in much the same manner as in the corona
field calculation. Each packet of particulates contains the same number
of uniform-size particulates arranged along the z-axis (i.e., the wire
axis). The number of particulates per unit length of a packet, 6nz, is
given as follows:
Qnp = 2^- H 6nz (17)
where n- is the number density of the particulates and AN is the number of
packets introduced in the inlet region between x = 0 and x = W at y = -L
in a time interval At.
W, half of the precipitator inlet width, is divided into ten sections.
During each time interval AN packets are introduced into the sections
sequentially; i.e., the first packet enters the first section closest to
the symmetry plane, the second packet enters the second section, and so
on, until the (AN)th packet has been introduced. In the next time step,
the first packet is introduced in the section next to that which the
(AN)th packet entered in the preceding time interval. When all ten
entrance sections have been used, subsequent packets are introduced sequent-
ially, starting with the first section.
Using the procedure outlined earlier, each particulate packet is
identified with one of the four diameters in the histogram of Figure 5,
such that the size distribution of the particulate packets is the same
as in the histogram. While 6 nz is the same for all particulate packets,
20
-------
the electric charge they carry will depend not only on the radius of the
particulates in a packet, but also on their starting positions as well
as their current positions in the precipitator.
Whenever a particulate packet enters a region, where the space charge
is greater than one tenth the charge density corresponding to one ion
packet per cell (i.e., Q(£,m)^ q/10 axay, where q is the charge carried per
ionic packet in the corona simulation program), we assume that each of the
particulates of the packet is instantaneously charged, according to Equa-
tion (15), provided that At»tc.
The particulate packets introduced in a time step and all the previ-
ously introduced packets are moved in the precipitator using the algori-
thms outlined in Chapter 2. Initially, a packet moves in a space charge
free region and follows the streamlines. Once the packet reaches a
region in which the space charge density is non-zero, as measured by
packets/cell, each particulate within it is given qs (°°) corresponding to
the local field strength.
Upon charging, the motion of a packet becomes coupled to both the flow
field and the corona field. The packet behaves according to the mobility
of a single particulate in the form of
kp = qs («)/6TmR (18)
because each particulate of the packet individually experiences the Coulomb
force and the Stokes drag, both of the same magnitude but opposite direc-
tions during steady state motion. During each time step, the charge on
the particulates within the packet is compared with qs(°°) of the new field
and the packet is allowed to have the higher of the two charge values.
This practice is justifiable for motions toward increasing as well as
decreasing field strength.
At regular time intervals of a chosen number of At's, the electric
field, due to the particulates, is recalculated and added to the total
corona field. During recalculation of the field, the particulate packets
are frozen on the grid points closest to them, as in the corona field
simulation. The field due these packets are calculated according to the
same procedure as outlined in Chapter 2 by the image charge method. Since
each particulate in a packet carries a charge qs(°°), the packet has an
effective charge per unit length of qs(°°) 6nz.
The calculation is continued until the total number of particulate
packets within the precipitator section is constant; i.e., a steady state
is reached. At this stage, the packets introduced during a chosen number
of At's are tagged, and the calculations are continued until all of the
tagged packets either end up on the plates, meaning collection, or exit
the precipitator. By keeping track of the total number of the tagged pack-
ets introduced and the numbers of those packets that are collected accord-
ing to particulate size, we can calculate the collection efficiency for
each of the four size as well as the overall efficiency representative of
the full size distribution.
21
-------
For later use in calculations for multi-stage precipitators, the
x-coordinates at the precipitator exit (y = + L) of those uncollected
particulates are stored in a 10 by A array, according to their starting
positions and sizes. Because the exit positions fluctuate due to fluctu-
ations in the electric field, stemming from variations in the spatial
arrangement of charged particulates, we only store the exit positions
averaged over those particulates that have been tagged at the inlet.
When calculating the efficiency for single section precipitators with
varying L/W ratio, it becomes necessary to change At. Whenever such a
change is made, we also change the number of particulate packets intro-
duced per time step so that the rate, AN/At, is the same for all L/W
ratios. In this way, 6nz, the number of particulates per unit length of
the packet along the z-axis is not changed, and this allows the direct
comparison of the efficiencies among single section precipitators of
different L/W ratios.
When the particulate number density n in the precipitator reaches
sufficiently high values, the electric field due to the charged particu-
lates become comparable to the electric field due to the ionic space
charge. When this condition prevails, the ionic space charge distribution
has to be recalculated in order to allow for the presence of the charge
particulate field. This is accomplished by freezing the electric field
due to the charged particulates at values obtained from the steady state
precipitation simulation, and re-running the simulation of the corona
field against this background charged particulate field. The corona
field obtained from this iteration is employed in the final simulation
for particulate trajectories and the collection efficiency.
It has been found that for particulates of the size distribution which
are predominantly in the sub-micron range, the charged particulate field
remains smaller than the ionic space charge field for nD <10^ cm~3. For
densities of 5 x 10 cm"-* or higher, some readjustment of the ionic space
charge distribution becomes necessary.
A Multi-wire Precipitator
We calculate the collection efficiency of a multi-wire precipitator by
performing exactly particulate trajectory calculations for the first two
sections and inferring from the results the exit positions of particulates
for subsequent sections. From the first section calculations, we obtain
for uncollected particulates not only their exit positions XIT^n^,^) ,
but also their charges q1(ni,n2> and mobilities kpl(n1,n2> witn which they
exit the first section. Note that each of these leads to a 10 x 4 array
of numbers because there are ten first section entrance positions (n^) and
four particulate radii (n2). The subscript 1 denotes the first section of
the precipitator. In the second section calculation, the particulates
enter with the same charges and mobilities at the same x-positions as they
exit the first section with. AN, the rate at which the particulate packets
are introduced, is the same as in the trajectory calculation for the first
section, except that only those uncollected packets from the first section
22
-------
are used. In other words, the particulate number density decreases in
subsequent sections as it should. When a steady state is reached with
the number of packets in the section, those packets introduced during a
certain number of At's are tagged and the calculation is continued until
all of the tagged packets either are collected or leave the second
section.
From this calculation, we obtain the exit positions for the second
section Xl^C!,^,^) , of those particulate packets which entered the
first section at the entry position designated by n^ and with a radius
designated by n2. The first index, 1, signifies the exit positions
computed for the packets having the same charges as at the exit of the
first section. At this point, two additional calculations are performed
in order to obtain the second section exit positions of the packets with
one-half and twice the original charges. The resulting exit positions
are designated by XIT2(2,n1,n2) and XIT2(3,n-L,n2>, the index 2 standing
for half charges and the index 3 standing for double charges. The exit
positions for subsequent sections are obtained by linear interpolation of
these three 10 x 4 arrays of exit positions, according to the charge and
exit position of each packet at the exit of the preceding section.
When the number density of particulates is high enough to influence
the ionic space charge distribution, the interpolation involves one more
variable; namely, the number density of particulates at the exit of the
preceding section. The first two section simulations are done for two
densities, one at 10' and another 10 cm , for this purpose. Even for
very high first section densities, such an interpolation is needed only
for the first few sections of the multi-wire precipitator.
The above procedure is repeated for all succeeding sections until each
particulate packet has either hit the plates or exits the entire multi-wire
precipitator. By carrying out this procedure for many particulates repre-
sentative of the size, distribution, we obtain the collection efficiency as
a function of particulate size and also as a function of the number of the
single-wire sections.
The Relationship Between the Single-Wire and Multi-Wire Precipitator
Efficiency
In the manner prescribed in the preceding discussion, it is simple to
determine the efficiencyn(l) of a single-wire precipitator. Here, we are
interested in expressing the efficiency of a multi-wire precipitator in
terms of the efficiencies of constituent single-wire sections. If No
particulates enter the multi-wire precipitator per unit time, the number
of particulates that leave the first single-wire section per unit time is
Nx = N0[l -
where the subscript i of r^Cl) denotes the single-wire section number.
23
-------
Similarly, the number after the Nth section is
N
NN = NQ TT [1 - HiCD] (19)
i=l
The collection efficiency of the N-wire precipitator is obtained from
Equation (19) as follows:
n(N) = (NQ - NN)/NQ
N
= 1 - TT [1 - 1(1)1 (20)
Once MN) is known as a function of N, it is possible to determine 1
as a function of the section number i by best fitting Equation (20) to
the data.
If n^(l) is assumed to be the same for all sections, Equation (20)
reduces to
n(N) = i - f i - m(i)l (2D
which is an analog of the Deutsch equation^. Equation (21) becomes identi-
cal to the Deutsch equation in the limit of large N:
D
- Hjd) N
nD(N) = 1 - e (22)
D A
Where N is integer values of y/2L and ru (1) = to—
and A is the plate area per single section of the precipitator and u the
average drift in a direction velocity perpendicular to the plates Q is the
flow rate.
As will be seen in the following section, the notion of uniform single-
wire efficiency is incorrect for multi-wire precipitators. The Deutsch
equation, which is based on such a notion as well as other simplifying
assumptions, naturally fails to represent the behavior of the multi-wire
precipitator.
A Discussion of the Results from the Numerical Simulation
Trajectories of forty particulates of four different sizes are displayed
in Figure 6 for L/W =3, i+ = 33 uA/wire, Q = 660 cm 3/sec and n (0) = 10'
cm ""3. Four different size particulates introduced at each inlet position
are seen to drift toward the plate at different speed; the largest being
the fastest as expected from the mobility consideration. Notice that the
figure has been compressed along the direction of the flow in order to
display the trajectories for four consecutive single-wire sections. We
emphasize that these particulates have been selected and followed out of
approximately one thousand particulates present in the field of each
section at any given time.
24
-------
The figure concisely sums up how a multi-wire precipitator behaves.
Those particulates near the plate are collected rather quickly, thus
contributing strongly to the collection efficiency of the first two
sections. The space charge density is large, as is the space charge
field strength in the region near the plate. There is a stretch of
distance where the particulates introduced in the middle between the plate
and the plane of wires drift toward the plate, but the drift is not
enough to effect collection. Thus, the efficiency is low. In the next
few sections, the efficiency increases due to the arrival of those partic-
ulates at the plate. Collection of the particulates introduced nearer to
the plane of wires takes place slowly over a long stretch of sections.
The electric field strength can be very high in this band of space very
near the wires, but the space charge density is very low. At the same
time, the flow velocity is the highest in this band and the streamlines
are such that the charged particulates are quickly taken out of the strong
field region at each wire. The net result is these particulates take a
long flow time to reach the region where the collection is favorable.
Figure 7 shows the collection efficiency as a function of the number
of single-wire sections for the case of L/W = 3, ic+ = 33 yA/wire, n (0)
= 10' cm"" 3 and Q = 660 cm 3/sec. xhe efficiency f)(N) is shown for each
particulate size as well as for the full distribution. As expected, ri(N)
cannot be represented by a uniform single-wire efficiency function of
Equation (21) or Equation (22). In Figure 8, we show ri-^(l) as a function
of the section number i, obtained by smooth curve fitting r)(N) of Equation
(20) to the data displayed in Figure 7. It is interesting to observe that
in each case of four different size categories as well as the full distri-
bution, the single-wire efficiency for the first section is larger than for
the immediately following ones. Of course, the fact that n^(l) eventually
becomes unity is the inevitable consequence of the property of Equation
(20). That is, n(N) cannot become unity for finite values of N without
one or all of rii(l)'s becoming unity.
The effect the particulate number density has on the precipitation effi-
ciency is seen quantitatively in Figure 9. ri(l) and ri(2) are shown for
three values of np for a precipitator with L/W = 3, i+ = 33 yA/wire and Q
= 660 cm -Vsec. This phenomenon, which we call the concentration effect,
has its origin in the Coulomb repulsion of charged particulates in the
corona field. As the number density is increased sharply, their repulsion
becomes significant and drives the particulates toward the plates. The
net result is a considerable increase in the efficiency.
Comparison of Theoretical Predictions with Experiment
In order to make an objective test of the results from the simulation,
a series of measurement has been carried out of the particulate collection
efficiency in 6-wire model precipitators. Altogether, five separate 6-wire
precipitators have been used. All have exactly the same wires and dimen-
sions except that the wire spacing (2L) varies from one unit to another to
give the wire-to-plate spacing ratio L/W of 3/4, 1,2,3 and 4, respectively.
The collection efficiency is measured by a light-scattering method at the
He-Ne laser, as well as argon ion laser wavelengths. Experimental details
25
-------
1.0
5 10
jjuniber of wires, N
Fig. 7. n(N) vs. N for L/W =3, i+ = 33 yA/wire, Q =
660 cm3/sec,and np(0) = 107 cm"3. Also shown
are %(N) for four different values of R: R =
2x10-5, 4x10-5, 6x10-5 and 8x10-5.cm. n(N) of
eq.(21) is plotted with n(l) = 0.145 for
comparison.
0.5
~Tiii
O best fit by eq. (21)
A best fit by eq. (22)
o-
O
o o
I I I
o
5 10
Section nunber, i
15
Fig. 8. rii(l) vs. the section number i, as obtained by
best fitting eqns. (20) and (21) to the data
of Fig. 7.
26
-------
0.7 r
0.6
0.5
O.li
IKS)
0.3
o.a
0.1
i-i
T T 1
- O'N-i M
__A:H = 2 J
r A J
h- (^1
L o 1
r Q -j
L-j-j-^--^-i--t_^-__L___LJ
5x10'
10
n (0), cm
-3
i.o
0.9
i.o
i
6
-
-
- 6
i
i i i
6 6 *
O
A He-ITe laser
O argon ion laser
1 1 1
-
-
-
-
n(£)
0.5
0.7
0.6
0.5
A
A
O
1
2 3
L/W
0.3
0.2
O calc-olation, R = 0.6 ^m
A measurement, He-He laser
Fig. 9.
n(l) and n(2) for three
values of the particu-
late number density at
the inlet, np(0), for
L/W = 3, i+ = 33yA/wire
and Q = 660 cm^/sec.
Fig. 10.
Measured Ti^(2) and
nR(6) vs. L/W of the
model precipitators
operating at i+ = 33
yA/wire and Q = 660
cm3/sec. Both He-Ne
and Arl laser data are
shown.
Fig. 11.
Calculated and measured
DR(1) vs. L/W for Q =
660 cm3/sec, i+ = 33
yA/wire and R = 0.6 ym.
T~|R(I) has been obtained
from n^(6) by eq. (21).
Note that n^O) is 107
and 10^ cm~3 for theory
and experiment respect-
ively, but this differ-
ence has no effect
here.
27
-------
may be found in Reference 3. In the following, we compare the experi-
mental results with the theoretical predictions. In the final section,
the scaling of the efficiency as a function of the L/W ratio is analyzed
on the basis of data from both experiment and numerical simulation.
The large majority of particulates used for this study are consider-
ably smaller in radius than the wavel ngth of the He-Ne laser (6328S) and
those of the argon ion laser (4880 and 5145 &). Studies of light
scattering (5) show that the total cross-section is the largest at the
He-Ne wavelength of 6328 8. for particulates of 0.59 ym (microns). At the
Ar ion wavelengths of 4880 and 5145S, the largest total scattering cross-
section is found for particulates of 0.49 ym. Figure 10 displays a plot
of measured total efficiencies at the end of 2 sections, 11^(2) and at the
end of 6 sections, T]R(6), versus the L/W ratio at Q=660 cm^/sec., obtained
with both the He-Ne and Arl laser; i.e., for R=0.59 and 0.45ym. A compar-
ison of the above data with those from numerical simulation in Figure 11
demonstrate a good agreement between theory and experiment in the case of
precipitator operation in the positive corona mode.
The Variation of the One-wire Efficiency with the Wire-to-plate Ratio
Figure 11 shows the experimental verification of the theoretically
predicted existence of an optimum value of the wire-to-plate ratio, L/W,
for a given flow rate. This optimum L/W ratio may be understood as arising
from the competition between two effects: larger L/W ratios give higher
efficiencies because of larger treatment times, while smaller L/W ratios
give higher efficiencies because of higher charging and collecting fields.
From this explanation, it is clear that the optimum L/W ratio will be
different for different flow rates. The significant fact is that the
one-wire efficiency maximum is somewhat shallow, which means that over a
large range of L/W, the one-wire efficiency can be regarded as approximate-
ly constant. The above finding leads to the important conclusion that it
is the number of wires per gas passage rather than the plate area which
determines the collection efficiency of the precipitator.
4. CONCLUDING REMARKS
In the preceding chapters, we have described a versatile and powerful
technique of numerical simulation ideally suited to solve a class of
coupled kinetic-electro aerodynamic problems found in large-scale electro-
static precipitators. A narrowly definable problem of the scaling of the
wire-to-plate spacing ratio has been identified and analyzed by this
technique. An independent experiment has been designed and carried out to
confirm the conclusions from the numerical simulation. A good agreement
between the simulation and experiment has been demonstrated.
The existence of an optimum wire-to-plate spacing ratio constitutes a
new and basic finding. Our understanding of the physics of the phenomenon
appears reasonable. While the analysis has been limited to the idealized
precipitators and to the positive corona configuration, it is most likely
that this optimal behavior is in effect in all sizes and variations of
the plate-wire precipitator. Detailed analysis has also been made for the
28
-------
negative corona configuration. It shows the same optimal behavior as for
the positive corona. The negative corona results are not presented here
because the efficiencies predicted from the numerical simulation fail to
agree with the experimental data on an absolute scale. This is due in
large part to the lack of reliable and detailed data on collision cross
sections for various electron sttachment processes in air.
This study has revealed in anatomical detail how the electrostatic
precipitation process takes place and, thereby, established the roles of
various operating parameters or features in terms of their influence on
the collection efficiency. An example is the concentration effect.
Another is the way the space charge density affects the precipitation
efficiency. Its primary role is not in the particulate charging rate as
commonly thought, but in contributing to the total electric field
strength near the plate where the electrostatic field of the wire is weak
and the particulates must gain significant drift speed for final collection
by virtue of high field strength. Charging of the particulates takes place
elsewhere, near the precipitator inlet where the space charge density is
generally low. It is noted that the assumption we adapted of instant
charging becomes questionable only in very small precipitators such as our
model units operating at the highest rate quoted. Its reasonableness
improves drastically for large scale precipitators.
Another basic finding of rather subtle nature is the role the particu-
late size distribution function plays in influencing the collection effi-
ciency. We have seen in Chapter 3 that the single wire efficiency can be
broken down to its dependence on the particulate size. Careful examination
of numerically generated trajectories shows that the makeup of the nearest
neighbor particulates of a test particulate strongly influences the event-
ual course of its trajectory. This leads to the observation that the
single wire efficiency for a given particulate size depends in a signifi-
cant way on the full-size distribution function of the particulates in the
precipitator. This effect contributes, in part, to the position depend-
ence of the single wire efficiency.
One can now ask how much one can improve the efficiency of an N-wire
precipitator by optimizing the wire-to-plate spacing ratio. Judging from
the results of Chapter 3, it is reasonable to anticipate 10 to 15% improve-
ment in the single-wire efficiency. Assume now for the sake of simple
estimation that r)(l) is constant throughout the precipitator. In Table 1
TABLE I. n(N) as calculated from Equation (21)
for several values of n(l) and N
n(D n(2Q) n(3Q) n(5Q)
0.06
0.07
0.08
0.09
0.10
0.710
0.766
0.811
0.848
0.878
0.844
0.887
0.918
0.941
0.958
0.955
0.973
0.985
0.991
0.995
29
-------
We list values of n(N) for five different values of n(l) and three of N.
A 14% increase in f|(l) from 7 to 8%, for example, results in a 6%
increase in T](20 from 76.6% to 81.1% and only a 1% increase in n(50)from
97.3% to 98.5%. The results of Table I also indicate the extent of effi-
ciency enhancement that may be expected by increasing the number of wires
per gas passage. This simply illustrates the magnitude of the task
in making the necessary improvement in the actual N-wire precipitator
efficiency. It also underscores the optimization of the wire-to-plate
spacing ratio as one area where a few precious percent can be gained in
the total efficiency.
This study benefited a great deal from discussions with Peter Sincerny
and Yong Kim of Lehigh University and William Langan, Gordon Gogola and
John Modla of Buell/Envirotech. Olga Garnet's help with the manuscript
is deeply appreciated.
References
1. Kim, Y. W. Electrostatic Precipitators I. Wire Oscillation,
Particulate Charging Rates, Flow Field and Imparity Effects", Physics
of Fluids Technical Report No. 26, Department of Physics, Lehigh
University (1978).
2. Kim, Y. W. An Analytical Consideration of the Particle Inertia
Effect with an Application to Aerosol Sampling Efficiency Calculation,
Physics of Fluids Technical Report No. 24, Department of Physics,
Lehigh University (1974).
3. Kim, Y. W. and E. A. Samuel. Electrostatic Precipitators II. The
Efficiency and Wire-to-Plate Spacing Ratio, Physics of Fluids
Technical Report No. 27, Department of Physics, Lehigh University (1978),
4. Rose, H. E. and A. J. Wood. An Introduction to Electrostatic Precip-
itation in Theory and Practice, Constable and Company, Ltd, London
(1956).
5. Kerker, M. The Scattering of Light and Other Electromagnetic Radia-
tion, Academic Press, New York (1969).
30
-------
THE EFFECTS OF CORONA ELECTRODE GEOMETRY
ON THE OPERATIONAL CHARACTERISTICS OF AN ESP
by
George Rinard, Donald Rugg and Whitney Patten
Denver Research Institute
Denver, Colorado 80208
and
Leslie Sparks
Industrial Environmental Research Laboratory
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
Experiments were run in a pilot scale electrostatic precipitator
(ESP) to determine the effects of corona wire-to-wire spacing on the
operating conditions. Tests were run, using a reentrained low sulfur fly
ash at both hot- and cold-side conditions. The effects of varying wire-
to-wire spacing were determined. Results are given which show that vary-
ing wire spacing at cold-side conditions has little operational effect on
the ESP while improved efficiency can be obtained at hot-side (low resis-
tivity) conditions by reducing wire spacing. The increased efficiency
results from a higher average operating voltage. The effects of back
ionization are clearly demonstrated by a set of experiments in which dust
was selectively removed from the wires or plate. These tests show that
the lower operating voltage caused by back ionization is a combined effect
of high resistivity dust on both the wires and plate.
31
-------
THE EFFECTS OF CORONA ELECTRODE GEOMETRY
ON THE OPERATIONAL CHARACTERISTICS OF AN ESP
INTRODUCTION
An electrostatic precipitator (ESP) collecting ash with high electri-
cal resistivity must be operated at low corona current density to avoid
back ionization. Usually this means that the operating voltage will also
be very low. If the operating voltage for a given current density could
be increased, ESP performance should improve. It is well known that
changes in the geometry of the corona electrode will alter the clean
plate VI characteristics. Either an increase in the size of the corona
wires or a decrease in the spacing between the wires will result in an
increased corona onset voltage and higher voltages for a given current
density (White, 1963). If the dirty plate VI characteristics of an ESP
could be modified to give the same result, improved ESP performance could
result.
Experiments were conducted on a pilot scale ESP at EPA's Industrial
Environmental Research Laboratory, Research Triangle Park, to determine
the effects of corona electrode geometry on the clean and dirty plate
operating characteristics of an ESP collecting high resistivity dust. The
results of these tests, regarding how the electrode geometry affected the
electrical characteristics, efficiency and opacity, are given below.
Experimental Procedure
The pilot scale ESP is a one lane wire duct ESP and is fully de-
scribed by Lawless et al. (1979). There are four electrical sections
consisting of two 1.2 m square collector plates. The plate-to-plate
spacing for these experiments was 22.9 cm. Temperature is controllable
from ambient to 400°C by means of propane burners. A draft fan provides
air flow. The aerosol used in these experiments was fly ash which was
injected through sandblast guns. This method of injection produced a
stable test aerosol.
The test procedure included variation of electrical conditions and
corona wire-to-wire spacing, and measurement of mass efficiencies at both
cold-side and hot-side temperatures utilizing a low sulfur ash. Tests
were also performed to determine the relative effect, on VI characteris-
tics, of ash on the plates and the wires.
The resistivity of the fly ash, as measured in the laboratory, is
given in Figure 1. The laboratory measurement for 150°C agrees well with
i_n situ measurements.
Sections 1 and 2 of the ESP had 0.32 cm corona wires spaced at
22.9 cm. Section 3 and 4 had 0.32 cm wires spaced at 7.62 cm.
This work was supported under Grant R805733 through the EPA's Industrial
Environmental Research Laboratory, Research Triangle Park, North Carolina.
32
-------
to
CO
I
tO
i 2
o
t io'
> 8
I- 6
in 4
u
cc
K>«
6
200
T—HT—r
TEMPERATURE, °F
400 500 600 700 800
300
I 1 1 Ll 1 \ 1 1 Hn 1 1 1 1—I 1—I I I |'l I I I I t M I I I'l I I I I
w-
9
%H20(VOL)O5% I
FIELD STRENGTH 4 Kv/cm
a
80 90 KX)
200
TEMPERATURE,°C
300
400
500
Figure 1. FLY ASH RESISTIVITY
-------
RESULTS
Previous tests had shown that the mechanical condition of the ESP is
good and thus there is no section to section variation in electrical
conditions for a given wire-to-wire spacing. This was verified in these
experiments, i.e., the electrical conditions in any sections with the same
wire-to-wire spacing were the same under all operating contions.
Clean-plate/clean-wire (CP/CW) and dirty-plate/dirty-wire (DP/DW) VI
curves for Section 1 (22.9 cm wire spacing) at 148°C are shown in Fig-
ure 2. The corona onset voltage is much lower for the DP/DW case. Also
note that the current for a given voltage is much higher in the DP/DW
case.
CP/CW corona onset voltage and DP/DW VI curve for Section 3 (7.62 cm
wire spacing) are shown in Figure 3. In this case the corona onset volt-
age for DP/DW case is much less than corona onset for clean conditions.
Also note that the DP/DW VI curve is similar to the DP/DW VI curve for
Section 1 (Figure 2) although the CP/CW curves are much different.
Also shown in Figures 2 and 3 are the operating points for Sections 1
and 3 for two levels of current density. The current was manually con-
trolled around the two set points. The data were obtained from the data
printout for the ESP which printed operating conditions once every minute.
The operating points for both sections drift but are essentially the same
and are independent of wire spacing. Curves similar to those in Figures 2
and 3 but for hot-sides (350°C) are shown in Figures 4 and 5.
For hot-side conditions the operating point was much more stable and
remained very close to the CP/CW characteristics for each of the four
sections. For the 22.9 cm wire-to-wire spacing, while operating at a
current of about 1 mA, the voltage remained in the vicinity of 26 kV.
However, for the 7.62 cm wire-to-wire spacing while operating at the same
current the voltage was maintained in the vicinity of approximately 33 kV.
Comparisons of these four figures indicate that at cold-side tempera-
tures the ultimate operating point of the precipitator is nearly indepen-
dent of the corona electrode geometry. However, while operating under
hot-side conditions, the operating point was quite dependent upon the
corona electrode geometry.
Figure 6 shows the change in precipitator operating voltage as a
function of temperature for a constant operating current of 1.5 mA. As
can be seen, the operating voltage for both 7.62 and 22.9 cm wire-to-wire
spacing is very nearly the same and is independent of electrode geometry
for temperatures in the range of 148°C to 204°C. However, the operating
voltage for 7.62 cm wire-to-wire spacing is about 8 kV higher than that
for the 22.9 cm wire-to-wire spacing in the temperature range above 260°C.
In this range the operating characteristics are very nearly those of
CP/CW. In the range of 204°C to 260°C the operating voltage is somewhat
lower than would be expected from the CP/CW characteristics for both
7.62 cm and 22.9 cm wire-to-wire spacing. At around 233°C the operating
voltage for the 7.62 wire-to-wire spacing makes a dramatic jump of approx-
imately 5 kV. 34
-------
oo
en
1.8
1.6
---5O
1.4
i.2
I 0
0.6
0.4
26 2!
Figure 2. OPERATING CHARACTERISTICS OF SECTION 1
WITH 22.9 cm SPACING AT COLD-SIDE TEMPERATURES
-------
oo
01
20 21
Figure 3. OPERATING CHARACTERISTICS OF SECTION 3
WITH 7.62 cm WIRE SPACING AT COLD-SIDE TEMPERATURES
-------
GJ
0.6
0.4 -
0.2 -
20 21
Figure 4. OPERATING CHARACTERISTICS OF SECTION 1
WITH 22.9 cm WIRE SPACING AT HOT-SIDE TEMPERATURES
-------
It is evident from Figure 6 that the operating characteristics of the
precipitator in the temperature range of 148°C to 204°C are determined
primarily by the high resistivity of the dust (approximately 1012 ohm-cm).
At 148°C the operating points for 7.62 wire-to-wire spacing and 22.9 cm
wire-to-wire spacing very nearly coincide. As the temperature is increas-
ed above 148°C the operating voltage for the precipitator increases as the
dust resistivity decreases. In the temperature range 204°C to 260°C the
dust resistivity becomes low enough (approximately 1011 ohm-cm) so that it
is no longer the predominant effect. As the temperature is increased
beyond this range the operating voltage decreases due to the decreased
density of the air at the higher operating temperature. Similar results
were obtained by Tachibana et al. (1978) in tests where the voltage rather
than current was maintained constant as temperature was increased.
Figure 7 gives the voltage current characteristics of a section of
the precipitator with 7.62 cm wire-to-wire spacing and without dust on the
wires and/or the plates. These characteristics were all obtained at
cold-side temperatures at around 148°C. To obtain these characteristics,
the ESP was opened after several hours of operation with dust and either
plates or wires were cleaned. Then the ESP was closed and operated with
no reentrained dust.
The corona onset voltage for DP/DW is about 12 kV lower than that of
CP/CW. This voltage difference was maintained between these two curves as
the current increased. As the voltage was decreased for DP/DW the current
decreased more slowly, resulting in hysteresis in the operating character-
istics. The corona onset voltage for dirty-plate/clean-wire (DP/CW) is
very nearly the same as that for CP/CW. As long as the operating voltage
for DP/CW was maintained slightly above corona onset, stable operating
conditions could be maintained. However, as the voltage was increased
beyond this point the current increased abruptly and the voltage started
to decrease. As the voltage was decreased the current remained rather
high resulting in considerable hysteresis for this situation. For the
clean-plate/dirty-wire (CP/DW) case the corona current started at a volt-
age below 25 kV. The current for this situation then increased up into
the range of about 38 kV. At this voltage, the corona current increased
abruptly with voltage. As the voltage was decreased from the point of
maximum current, very little hysteresis was observed. Other investigators
suggest that high resistivity dust on the discharge electrode wire will
decrease the corona current (Robinson, 1971) which is not consistent with
the data in Figure 7. The effect of a high resistivity coating on the
wires is not fully understood. It is clear that the dust affects the
roughness factor but other mechanisms also appear to be involved.
Figure 8 gives the results of mass efficiency measurements as a
function of the average voltage of -all four of the precipitator electrical
sections for the period of time that the mass samples were taken. The
four efficiency measurements in Group I are for 22.0 cm wire-to-wire
spacing in all four electrical sections while the precipitator is operat-
ing at cold-side temperatures. The three efficiency measurements in
Group II are for the same precipitator configuration except when operating
at hot-side temperatures. The efficiencies for hot-side conditions were
considerably higher than for cold-side conditions. In both cases the
efficiency was greatly increased by the increase in operating current and
38
-------
CO
1.8
1.6 -
--•50
1.4
-45
1.2
1.0
0.8
-40
-35
~3O
-25
0.6
0.4
0.2
-20
-15
-10
20 21
24
Figure 5. OPERATING CHARACTERISTICS OF SECTION 3
WITH 7.62 cm WIRE SPACING AT HOT-SIDE TEMPERATURES
-------
50
40
45.
O
30
ao'—
200
O CW-CP 3" WIRE TOWIRE SPACING
0 CW-CP 9" WIRE TO WIRE SPACING
x DW-DP 3" WIRE TO WIRE SPACING
® DW- OP 9" WIRE TO WIRE SPACING
3OO
400
500
600
T «F
Figure 6. OPERATING VOLTAGE vs TEMPERATURE
FOR CONSTANT CURRENT OF 1.5 mA
-------
8.0 -
4.6
1.0
35
ISO
< 3.0
£
r
-100
Z.5
Ul
K
3
1.5
I.C
0.5
50
-20
» ° CP-CW £('70-3l00Ufi4r-l849) 8/15
A—-A CP- DW 263C>-2900{!506-I5II) 8/^3
X X DP-CW 2o3"-283°(l 150-1157! B/22
• . DP DW 2790-E8J>°tl50e-ISI5 ( d/ZI
-25
-30
735 --50
VOLTAGE - kV
-45
-50
Figure 7. VOLTAGE/CURRENT CHARACTERISTICS
WITH/WITHOUT DUST ON WIRES AND/OR PLATE
-------
-p.
ro
1 X Z2.9cm, I35°C
H • 22.9cm, 3I6°C
m Q 7.62cm, I35°C
EC ® 7.62cm, 3I6°C
GROUP
I
Figure 8. EFFICIENCY & PENETRATION vS
AVERAGE VOLTAGE DURING MASS SAMPLE RUNS
-------
the resultant increase in voltage. The two mass efficiency measurements
in Group III are for the precipitator with 22.9 cm wire-to-wire spacing in
the first two electrical sections and 7.62 cm wire-to-wire spacing in the
last two electrical sections. In this case, the efficiency was not dra-
matically increased by the tenfold increase in current. However, the ESP
was somewhat cleaner during the test at 1.5 mA and may have influenced the
efficiency measurement at this current level. The mass efficiency mea-
surement in Group IV was for the same precipitator configuration except
while operating under hot-side conditions. This resulted in the highest
efficiency of all tests made.
Opacity measurements were made during all runs using a plant process
visiometer (PPV). The opacity varied directly in all tests with efficien-
cy.
DISCUSSION OF RESULTS
From the results given in the previous section, the following is
evident:
Cold-Side Temperatures
1. At these temperatures the dust electrical resistivity is high
(1012 ohm-cm) and the predominant effect is back ionization. Thus,
the steady state operating voltage is very nearly independent of the
electrode geometry.
2. Back ionization onset is caused by a combination of the dust on the
wires and the dust on the plates. When the wires are clean and the
plates are dirty, VI characteristics are similar to clean conditions
for low operating currents. However, when there is dust on the
wires, corona onset occurs at a much lower voltage causing back
ionization to occur and the steady state operating point to be main-
tained at a much lower voltage.
3. The beneficial effects of corona electrode geometry can be realized
only if the corona electrodes can be maintained essentially dust
free.
4. Wire spacing of 7.62 cm appears to offer some advantage with regard
to efficiency of dust collection for high resistivity dust when
operating at low current levels. However, further tests are indi-
cated.
5. Theoretical understanding of why dust on the wire is so important is
lacking. Roughness factor explanations are insufficient especially
in light of hot-side temperature results.
Hot-Side Temperatures
1. The operating point for the precipitator is determined primarily by
the wire-to-wire spacing of the corona electrode.
43
-------
2. Higher operating voltage and resultant higher efficiency can be
obtained by decreasing wire spacing.
3. Dust on the wires has little effect.
CONCLUDING REMARKS
Some improvements in efficiency may be obtained at cold-side tempera-
tures by using closer wire spacing. However, the improvement is not
dramatic and the electrical operating conditions are practically unchanged
by decreasing wire spacing.
At hot-side temperatures the operating voltage and efficiency can
both be increased by closer wire spacing.
It is apparent from these tests that back ionization once started,
becomes more important than corona wire spacing in determining the elec-
trical operating conditions. It is also apparent that the dust on the
corona wires plays an important role in the formation of back ionization.
REFERENCES
1. Lawless, P.A., B.E. Daniel, and G.H. Ramsey. "Characterization of
the EPA/IERL-RTP Pilot-Scale Precipitator." EPA 600/7-79-052,
February 1979.
2. Robinson, M., Air Pollution Control. Part I, Ed. W. Strauss, Wiley-
Interscience, NY, 1971.
3. Tachibana, N., Y. Matsumoto, and N. Sakamoto. "Vanishing Temperature
of Back Corona with High Resistivity Fly Ash in Hot-Side Precipita-
tion." CSIRO Conference on Electrostatic Precipitation, August 1978.
4. White, H.J. Industrial Electrostatic Precipitation. Addison-Wesley,
Reading, MA, 1963.
44
-------
THEORETICAL METHODS FOR PREDICTING ELECTRICAL
CONDITIONS IN WIRE-PLATE ELECTROSTATIC PRECIPITATORS
by
Ronald B. Mosley
Jack R. McDonald
Southern Research Institute
2000 Ninth Avenue South
Birmingham, Alabama 35205
and
Leslie E. Sparks
Industrial Environmental Research Laboratory
Environmental Protection Agency
Research Triangle Park, North Carolina 27711
ABSTRACT
A new semi-empirical, approximate theory for predicting electrical con-
ditions is described. In the approximate theory, analytical expressions are
derived for calculating voltage-current characteristics and electric potential,
electric field, and space charge density distributions. Comparisons of numer-
ical and approximate solutions over a wide range of possible precipitator
geometries and electrical operating points indicates that for practical pur-
poses the approximate theory can be used in lieu of the more rigorous numer-
ical theory. This saves large amounts of computer time and makes possible
hand calculator usage. Recent in situ gaseous ion mobility data which are
needed in the models are presented. For coal fired power plants, the reduced
effective ion mobility in positive corona is found to be 1.6 times that for
negative corona. Approaches for describing particulate space charge effects
in the gas and electrical conditions in the collected particulate layer are
briefly discussed.
45
-------
THEORETICAL METHODS FOR PREDICTING ELECTRICAL
CONDITIONS IN WIRE-PLATE ELECTROSTATIC PRECIPITATORS
INTRODUCTION
An accurate representation of the electrical conditions in an electro-
static precipitator is one of the most crucial steps in developing a mathe-
matical model of the electrical precipitation process. A realistic model
must describe the electrical properties of both the gaseous region of the
precipitator and the dust layer collected on the plates. The model should
be capable of describing the electrical properties, not only of the clean
gas with ionic charge carriers, but also of the dust-laden gas in which the
charged particles alter the electrical processes. In order to model the
current-voltage characteristics of the clean gas it is necessary to know the
mobility of the ionic charge carriers. For this reason, some in situ mea-
surements of effective ion mobilities are presented. An approach for in-
corporating the space charge effects of particles into the model is discussed.
An approach for describing the electrical properties of the dust layer is also
presented.
APPROXIMATE SOLUTION FOR SPACE CHARGE LIMITED CURRENTS IN WIRE-DUCT PRECIPITATORS
Development of Equations
In order to model the behavior of an electrostatic precipitator, it is
necessary to compute the electric field and potential in the interelectrode
space. In one mathematical model1 of the electrostatic precipitation process,
a relaxation method is used to solve Poisson's equation coupled with the charge
transport equation. This numerical procedure2 requires a great deal of computer
time. An analytic solution would be very advantageous in circumventing this
difficulty. Since an exact solution does not seem probable, an approximate
solution is appropriate.
Using a conformal mapping technique, Cooperman3 demonstrated that the elec-
trostatic potential in the interelectrode space of a wire-duct precipitator with
2N+1 corona wires can be written as
N
V ^ V In
m=-N
with
coshir(y-2ms )/2sx - cos (irx/2sx)
£,-*_
*—' coshir (y-2ms )/2sv + cos (irx/2s^J
y X X
(i)
46
-------
Eo
st
ir Sin(itao/2s
N
m=-N
cosh (mns
cosh* (nnrs /sx) -'cos2 (Tra0/2sx)
-1
(2)
E0 = the electric field at the surface of the wires (V/m),
a0 = radius of the wires (m),
sx = wire-to-plate spacing (m),
s = half wire-to-wire spacing (m).
Figure 1 shows the region of the precipitator that is of interest. The origin
is taken at the wire, while x is measured toward the plate and y is measured
toward an adjacent wire. When a current flows in the presence of space charge,
equation 1 does not apply. This solution can be used to advantage, however, in
obtaining an approximate solution for the dynamic case in which the current is
space charge limited. Suppose a steady current flows under the action of an
applied potential for the case illustrated in Figure 1. The electrical con-
figuration will be simplified by using the principle of superposition. Con-
tributions to the potential will be separated and described below.
m»IM
I
I
i 1
,V = O ON
*• THE PLATES
V " V0 - APPLIED POTENTIAL ON THE WIRES
Figure 1. Wire-duct precipitator.
First consider just the electrodes. For a negative corona, the wire will
be charged negatively while the plates will be charged positively. The charge
on the plates will exceed the magnitude of the charge on the wire by an amount
equal to the quantity of negative space charge in the interelectrode space.
The negative charge on the wires coupled with an equal amount of positive
charge on the two plates constitutes a situation similar to the static case of
Cooperman. Mathematically, the solution for this case is identical in form to
47
-------
the electrostatic case, but the amount of charge on the wires is not necessarily
the same as when the current is zero. Suppose that a small increase in the
electric field at the surface of the wire gives rise to a large increase in
the number of electrons emitted from the wire during the corona discharge.
The field at the surface of the wire when a current flows, then would be nearly
the same as Peek's value of the field required for corona start. In this case,
the charge on the wire will be computed in terms of Peek's condition on the
field at the wire. The static-like solution would then be completely deter-
mined. It remains to specify the contributions to the potential from the
negative space charge coupled with an equal amount of positive charge on the
plates and from the corona discharge process itself. These last contributions
must be approximated since closed-form solutions are not available.
The primary considerations to be used in obtaining approximate solutions
will be the apparent symmetry of the arrangement. For example, the equi-
potential lines near the discharge wires look very much like those for a wire-
cylinder geometry,1 while the equipotential lines near the plates look very
much like those for a parallel plate geometry.5'6 Using these observations,
it will be assumed that the space charge near the wire is distributed in the
same manner as that in a wire cylinder geometry, and that the space charge
near the plate is distributed in the same manner as for a parallel plate
geometry. These two space charge distributions along with their associated
surface charges are illustrated in Figures 2 and 3, respectively. Potentials
computed from these two geometries will be matched at some point (x=R, where
R is the radius of the cylinder of charge) in the interelectrode space to
yield continuity of potential. The electric field is also required to be con-
tinuous at the matching point.
SURFACE CHARGE
x-O » - R
SPACE-CHARGE
SURFACE-CHARGE'
- \
SPACE CHARGE .
Figure 2. Cylindrical distribution
of space-charge with
associated surface-charge.
Figure 3. Parallel plane distri-
bution of space-charge,
The contributions to the potential by the corona region will be assumed to
be small. According to Loeb,7 the potential drop across the ionization region
48
-------
of a corona is essentially independent of the current. The effect of the corona
on the field at the surface of the wire at corona start is included in the mea-
surement of Peek's field since the measurement is made with a minimum steady
current. Based on the above discussion, it is assumed that the only role which
the corona plays is to supply ions to carry the current in the interelectrode
space.
Negative charge distributed uniformly in planes parallel to the plate
which has an equal amount of positive charge, as shown in Figure 3, will yield
a potential given by
V = - | (^)h [ (Sx - R) »/2 - (X - R) 3/2 ] (3)
where
J = current density (A/m2),
b = ion mobility (m2/V-sec),
eo = permittivity of gas (A sec/Vm),
R = point at which the solutions are to be matched.
In this same region, near the plate, the potential produced by the cylinder of
negative charge coupled to an equal amount of positive charge on the plates
must be taken into account. For points which lie outside the cylinder of
charge shown in Figure 2, the solution will be the same as for a line of charge
at the center with the same quantity of charge as contained in the cylinder.
This problem has a solution similar to the electrostatic case mentioned before.
When all 2N+1 wires are considered, the contribution to the potential for x
greater than R will be
N r coshir (y-2ms )/2s - cos (irx/2s )T
Zi y x ^ I
In (A)
Lcoshir (y-2ms )/2s + cos(T;x/2s )J
where Vi depends on the amount of charge in the cylinder.
At points inside the cylinder of charge, the potential due to the other
wires is given by an expression which is identical to equation (4) except that
the term for m=o is absent. To this we must add the potential due to the charge
inside the cylinder as well as that due to the two images shown in Figure 2.
To obtain the potential due to the negative space charge in the cylindrical
distribution, we subtract the electrostatic solution for a wire and cylinder
with no space charge from the solution of the same wire and cylinder with space
charge.8 The two images are used to account for the coupling of the space
charge in the cylinder with the associated charge on the plates. The potential
at a point inside the cylinder due to space charge and the two images is given
by a complicated expression of x and y. This expression can be identified in
the equation for the potential shown in the appendix. Mathematical expressions
for the total potential as well as for the x and y components of the electric
field are presented in the appendix.
49
-------
Comparisons of Approximate Model with Numberically Integrated Solutions and
Experimental Observations
In general, the agreement of the approximate model with the numerically
integrated solution is found to be quite good. Table 1 summarizes the com-
parison between the approximate model and the numerical solution for a wide
range of geometrical parameters and current densities. Columns 1, 2, and 3
list the corona wire radius, the wire-to-plate spacing, and the half wire-to-
wire spacing, respectively. While most combinations of geometrical parameters
in this table are typical of industrial electrostatic precipitators, examples
of both close and wide wire spacing are shown. Both large and small corona
wires are represented and a considerable range of plate spacings is given.
Column 4 lists the current densities that were used. The values of current
density were chosen to represent both moderate and relatively high space
charge situations. Column 5 gives the percent difference in applied voltage
for the approximate and numerically integrated solutions. Column 6 gives the
percent difference in the average electric field at the plate. Column 7 gives
the maximum percent difference in the potential profiles. Columns 8 and 9
give the maximum percent differences in the x-component and the y-component
respectively of the electric field profiles. Column 10 gives the maximum per-
cent difference in the charge density profiles.
In the approximate model, one is free to choose the position (x=R) at
which to match the functions derived from the two different symmetries. For
the calculations shown in Table 1. R was choosen to be 0.975S . This cor-
x
responds to using only a few millimeters near the plate in which the space
charge is distributed according to the rectangular geometry of a parallel
plate arrangement. The reason for using such a small volume with rectangular
symmetry is that it tends to over estimate the electric field near the plate.
That is, the parallel plate solution predicts too large an effect due to
space charge. The fact that rectangular symmetry plays such a small role in
the space charge contribution to the profiles does not necessarily contradict
the observations made earlier that the equipotential lines near the plate re-
flect a strong influence of planar geometry. The planar character of the
equipotential lines is largely provided by the static-like solution which
accounts rigorously for much of the charge on the plates. In all cases the
approximate model agrees well with the numerically integrated solution for
zero current. This case, of course, corresponds to the electrostatic solution
at conditions for corona start, the limit in which the approximate model be-
comes exact.
Predictions of the approximate model as well as those of the numerical
solution are compared with some experimental results of Penney and Matick6 in
Figures 4, 5, and 6. An effective ion mobility of 1.8 x W~k m2/V-sec in the
approximate model yields voltage-current curves which predict the experimental
operating conditions to within one percent. These voltage-current curves along
with the numerically integrated ones are shown in Figure 4. Figure 5 shows a
comparison of the potential profiles along a line connecting the wire with the
plate. The approximate model agrees with this measured potential profile to
within 3 percent. A comparison of the potential profiles along a line con-
necting a point midway between two wires with the plate is shown in Figure 6.
The approximate potential is as much as 15 percent larger than the measured
50
-------
Table 1. COMPARISON OF APPROXIMATION AND NUMERICALLY INTEGRATED SOLUTION
Wire
Radius
(1(T3 m)
1.1906
1.1906
1.1906
1.1906
1.1906
1.1906
1.1906
1.1906
1.1906
1.1906
3.00
3.00
0.500
1.016
1.016
0.1524
0.1524
Wire
Plate
Distance
(»)
0.127
0.127
0.127
0.127
0.127
0.127
0.127
0.127
0.200
0.200
0.127
0.127
0.127
0.1143
0.1143
0.1143
0 . 1143
Half Wire-
to-wire
Distance
(m)
0.03175
0.03175
0.0635
0.0635
0.127
0.127
0.15875
0.15875
0.0635
0.0635
0.0635
0.0635
0.0635
0.07348
0.07348
0.07348
0.07348
Current
Density
(nA/cm2)
10
100
10
100
10
100
20
100
20
80
20
100
10
20
200
5
180
Percent
Difference
In Applied
Voltage
1
3
0
2
3
1
2
2
3
4
2
2
3
1
2
22
9
Percent
Difference
In Field At
The Plate
6
12
8
9
1
0.1
5
6
1
6
4
2
9
3
1
6
10
Max. %
Difference
In Potential
Profiles
2
5
3
8
8
11
12
13
7
2
2
4
4
3
6
22
9
Max. %
Difference
In
x-component
of field1
12
12
12
11
12
5
10
16
15
7
7
9
15
9
10
18
16
Max. %
Difference
In
y-component
of field2
400
400
100
320
53
118
30
40
60
50
50
50
300
40
80
300
213
Max. %
Difference
In Charge
Density
Profiles3
23
32
18
12
8
10
19
22
24
10
10
12
14
9
15
6
13
Joints closer than two diameters to the wire and points near the plane connecting the centers of the wires with the
plates have been neglected.
2 Even though the percent difference in the y-components of the electric field seems quite large in some cases, the
absolute error is relatively unimportant because the y-component is much smaller than the total field for nearly all
such cases.
3Some points very near the plates and some near the centers of the wires have been neglected.
-------
potential along this line. The approximate potential is 10 percent larger than
the numerically integrated potential along this same line. The apparent reason
for this disagreement lies in the large current density used. A large current
density corresponds to the large space charge limit where the approximation
is expected to be less accurate. Also, this region, between the wires, is
where the approximate model is expected to be least accurate. It should be
noted that moderate errors in the potential profile in this region (between
the wires) have little influence on the predicted electrical operating con-
ditions of the precipitator device. The value of 1.8 x 10~ m /V-sec used for
effective mobility is consistent with measured values of ion mobility in nega-
tive corona discharges in air. It may be concluded that the predictions of
the approximate model agree quite well both with the numerically integrated
solutions and with experimentally measured electrical properties of precipita-
tor devices.
S * 0.1143m
x
V
b •
f •
O •
A-
0.073480
1.8x10"* mVv-s
1.0
Integrated
Approximation
Penney and
Mat let'a
Operating
Conditions
3 20
S, = 0.0734Sm
b - 1.8x10"* mVv-8
f - 1.0
- • Integrated
A - Approximation [ a - 1.524x10"
• Experimental
O - Integrated
A - Approximation] a - 1.016xlO~!n
D - Experimental
i
O
A
?
go
,AQ
*°
O
8*0
0.
30 40 50
Applied Voltage, kV
°\
1 1 1 1
4 6
Displacement (10~
8
Meters)
7/i
Figure 4, Theoretical voltage-current
curves for the geometries
used by Penney and Matick.
Figure 5. Theoretical and experimental
potential profiles along
a line from a wire to the
plate.
52
-------
Sy • O.QT^Bw
b • 1.8x10" n'/V-s
f • 1.0
• Inteerattd j
A - Approximation a .
• - Experimental |
0 - Integrated |
_ 3 - Approximation ' a •
O - Experimental |
A 0
o
D
1 1 _L 1
•\
Displacement (LO"' Ketere)
Figure 6. Theoretical and experimental potential profiles along a line from
a point midway between two wires to the plate.
MEASURED EFFECTIVE ION MOBILITIES
The effective mobility of ions created in a corona discharge can strongly
influence the current-voltage characteristics of the discharge device, the
electric field distribution in the gaseous region, and the particle charging
process. Thus the effective ion mobility can have a significant effect on
particle collection efficiencies in an electrostatic precipitator. In general,
the effective ion mobility depends on the temperature, pressure, and com-
position of the gas, as well as on the electric field strength. Consequently,
for purposes of modeling electrostatic precipitators, it is important to ob-
tain in situ measurements of effective ion mobility in industrial flue gases.
The technique is based on a measured current-voltage curve obtained for corona
discharge in a wire-cylinder geometry. Both the instrument and the method of
measurement have been described elsewhere. Some representative measurements
from three different plants are summarized in Table 2. Plant A is a gold
smelter, while Plants B and C are coal fired power plants. It will be noted
that the chemical analysis of the gas in Plant A is quite different from the
other two plants. It is also apparent that the positive reduced mobility
experiences a greater variation from one plant to another, as well as for
different conditions at the same plant than does the negative reduced mobility.
It is generally the case that the effective mobility of positive ions is more
sensitive to small variations in the gas composition than the mobility of the
negative ions.
The reduced effective mobilities for both negative and positive coronas
are somewhat smaller than previously reported.8 In filtered flue gas, the
reduced effective mobility for positive corona is found to be about 1.6 times
greater than for negative corona. Larger ratios were observed in previous
measurements. The measurements in this paper were made at pressures near that
53
-------
Table 2. ION MOBILITY FIELD TEST DATA
Maxi-
Multiple Average
Run
ID1
PLANT
13NG
14NG
15NG
16NG
17NG
18NG
19PG
20PG
21PG
22PG
23PG
24PG
PLANT
01NG
02NG
03NG
04NG
05NG
06NG
02PG
03PG
04PG
05PG
06PG
PLANT
140PG
141PG
142PG
143PG
144PG
145PG
146NG
147NG
148NG
149NG
150NG
151NG
PLANT
242NG
243NG
244NG
245NG
246NG
247NG
248PG
249PG
250PG
251PG
252PG
253PG
PLANT
2B4NG
285NG
286NG
287NG
288NG
289NG
290PG
291PG
292PG
293PG
294PG
2 9 SPG
Temp.
°c
A
184
185
185
185
185
185
186
186
186
186
186
186
B
125
133
141
142
143
143
142
144
144
144
144
C (NO
146
146
146
146
146
146
146
146
146
146
146
146
Absolute
Pressure
mm Hg
701
701
701
701
701
701
701
701
701
701
701
701
735
735
735
735
735
735
732
732
732
732
732
ADDITION OF
597
597
597
597
597
597
597
597
597
597
597
597
Effective
Mobility
K= cmVv-s
2.76
2.24
1.96
2.15
2.10
1.89
2.55
2.61
2.48
2.39
2.55
2.49
2.10
2.07
2.10
1.97
1.91
1.93
3.30
3.16
3.23
3.39
4.27
CONDITIONING
4.17
4.07
3.99
4.11
3.98
4.07
2.89
2.87
2.87
2.84
2.86
2.85
Reduced Corre-
Effective lation
Mobility Coeffi-
Ko cm2/v-s cient
1.52 0.9641
1.23 0.984 I
1.08 0.975!
1.18 0.984 /
1.15 0.990
1.04 0.970^
1.40 0.992\
1.43 0.989 j
1.36 0.995!
1.31 0.997 j
1.40 0.990
1.37 0.987J
1.39 0.996'
1.35 0.997 .
1.34 0.974 '
1.26 0.997 |
1.21 0.997
1.23 0.991)
2.09 0.997'
1.99 0.998 |
2.04 0.998
2.14 0.997 1
2.69 0.997 )
AGENTS)
2.14 0.992^
2.08 0.996 ,
2.04 0.998 '
2.10 0.998
2.04 0.997
2.08 0.998 >
1.48 0.9981
1.47 0.998
1.47 0.997
1.45 0.997
1.46 0.997
1.46 0.997 ,
Reduced
Effective
Mobility Volume Percent PPM
Ko cm2/v-s HzO Oz C02 SOa SOa
6.0 18.7 0.0 10* 6.0
6.0 18.7 0.0 10* 6.0
, ,n 6.0 18.7 0.0 10* 6.0
1'20 6.0 18.7 0.0 10* 6.0
6.0 18.7 0.0 10* 6.0
6.0 18.7 0.0 10* 6.0
6.0 18.7 0.0 10* 6.0
6.0 18.7 0.0 10* 6.0
, ,0 6.0 18.7 0.0 10* 6.0
1-JU 6.0 18.7 0.0 10* 6.0
6.0 18.7 0.0 10* 6.0
6.0 18.7 0.0 10* 6.0
(No Gas Analysis)
(No Gas Analysis)
. . ,n (No Gas Analysis)
• (No Gas Analysis)
(No Gas Analysis)
(No Gas Analysis)
i(No Gas Analysis)
(No Gas Analysis)
2.08 (No Gas Analysis)
(No Gas Analysis)
(No Gas Analysis)
ilO.5 4.5 14.7 593 0.5
10.5 4.5 14.7 593 0.5
, no 10.5 4.5 14.7 593 0.5
' 10.5 4.5 14.7 593 0.5
10.5 .5 14.7 593 0.5
10.5 .5 14.7 593 0.5
10.5 .5 14.7 593 0.5
10.5 .5 14.7 593 0.5
, .7 10.5 .5 14.7 593 0.5
• ' 10.5 .5 14.7 593 0.5
10.5 .5 14.7 593 0.5
10.5 .5 14.7 593 0.5
Maxi-
mum
Volt-
age
kV
22
21
21
21
22
22
20
20
20
21
20
20
40
38
37
33
33
31
24
24
24
24
23
20
20
20
20
20
19
30
30
30
30
29
30
mum
Cur-
rent
Den-
sity
MA/cma
0.88
0.69
0.69
0.70
0.79
0.69
0.54
0.56
0.55
0.66
0.54
0.53
8.1
7.0
6.4
4.6
4.5
3.8
2.5
2.6
2.5
2.6
2.2
2.1
2.1
2.1
2.0
2.1
1.7
5.6
5.4
5.4
5.4
5.0
5.5
Maxi-
mum
Average
E/p
V/cm-Torr
7.3
6.9
6.9
6.9
7.3
7.3
6.6
6.6
6.6
6.9
6.6
6.6
12.9
12.3
11.9
10.6
10.6
10.0
7.8
7.8
7.8
7.8
7.5
7.9
7.9
7.9
7.9
7.9
7.5
11.9
11.9
11.9
11.9
11.5
11.9
C (7 PPM SOj INJECTED)
146
146
146
146
146
146
146
146
146
146
146
146
C (100
157
158
160
161
161
163
157
157
157
157
157
157
597
597
597
597
597
597
597
597
597
597
597
597
Ibs/hr Na2
597
597
597
597
597
597
597
597
597
597
597
597
2.92
2.81
2.91
2.94
2.96
2.70
4.79
.77
.81
.81
.91
.80
CO 3 INJECTED)
2.99
3.02
2.99
2.98
3.04
3.04
4.96
4.76
5.00
5.03
5.12
5.18
1.49 0.998
1.44 0.997
1.49 0.997
1.50 0.996
1.52 0.996
1.38 0.997
2.45 0.996
2.44 0.998
2.46 0.998
2.46 0.997
2.51 0.996
2.46 0.998
1.49 0.999'
1.50 0.999
1.48 0.999
1.47 0.999
1.50 0.999
1.49 0.999
2.48 0.997
2.38 0.998
2.49 0.998
2.51 0.998
2155 0.998
2.58 0.997
10.5 4.5 14.7 593 7.0
10.5 .5 14.7 593 7.0
, ,, 10.5 .5 14.7 593 7.0
10.5 .5 14.7 593 7.0
10.5 .5 14.7 593 7.0
10.5 .5 14.7 593 7.0
10.5 .5 14.7 593 7.0
10.5 .5 14.7 593 7.0
, ., 10.5 .5 14.7 593 7.0
10.5 .5 14.7 593 7.0
10.5 .5 14.7 593 7.0
10.5 .5 14.7 593 7.0
10.5 .5 14.7 593 0.5
10.5 .5 14.7 593 0.5
, ., 10.5 .5 14.7 593 0.5
' 10.5 .5 14.7 593 0.5
10.5 .5 14.7 593 0.5
10.5 .5 14.7 593 0.5
10.5 .5 14.7 593 0.5
10.5 .5 14.7 593 0.5
, 50 10.5 .5 14.7 593 0.5
10,5 4.5 14.7 593 0.5
10.5 4.5 14.7 593 0.5
10.5 4.5 14.7 593 0.5
29
28
28
28
28
28
21
21
21
21
20
20
34
33
31
31
32
31
20
20
20
20
20
20
4.9
4.4
4.4
4.3
4.3
4.2
2.8
2.8
2.8
2.8
2.3
2.3
7.6
7.0
6.0
6.0
6.0
5.9
2.4
2.3
2.4
2.4
2.4
2.4
11.5
11.1
11.1
11.1
11.1
11.1
8.3
8.3
8.3
8.3
7.9
7.9
13.5
13.1
12.3
12.3
12.7
12.3
7.9
7.9
7.9
7.9
7.9
7.9
'The Run ID' s are coded in the following manner? The first 2 or 3 digit number is the run number at that test location. The
first letter indicates the polarity of the corona (P or N) . The second letter, G, indicates that the measurement was made in
filtered flue gas.
54
-------
of the flue gas, while the previous measurements were made at reduced pressures,
The differences in these sets of measurements are attributed primarily to the
differences in pressure. The values of effective mobility reported in this
paper are believed to be more representative of ions in industrial flue gases.
SPACE CHARGE EFFECTS OF PARTICLES
The effects of charged particles on the electrical properties of an elec-
trostatic precipitator can be taken into account by the use of the effective
mobility concept. For the case in which both ions and charged particles con-
tribute to the electrical conditions in the interelectrode space, the two
transport equations and Poisson's equation can be reduced to a system of
equations which is identical in form to the system of equations applicable to
ions alone. This system of equations differ from the case of ions alone in
that effective values of current density, mobility and charge density are used.
The effective current density is given by
Jeff * beff (Vbi + Jp/bP >
where
b ff = effective mobility (m2/V-s),
bi = ion mobility (m2/V-s),
b = particle mobility (m2/V-s),
J. = ion current density (AMP/m2) ,
J = particle current density (AMP/m2) .
The effective mobility is defined in the usual way
beff =
-------
necessary to iterate on the effective mobility. If the particle current
density is neglected in comparison with the ion current density, the following
algorithm could be used for a given particle mass loading and known size dis-
tribution.
1 . Specify the current density (hold fixed) .
2. Compute the potential and charge density for the clean gas. These
values will be used to compute an initial estimate of the charge on
the particles,
3. Compute the average particle mobility and the effective mobility.
4. Compute the effective current density.
5. Compute new potentials, fields and effective charge densities.
6. Compute the average ion density from the effective charge density.
7. Compute the new charge on the particles.
8. Compute new values of average particle mobility and effective mobility.
9- Repeat steps A through 8 until the effective mobility does not change
significantly .
AN APPROACH FOR DESCRIBING ELECTRICAL PROPERTIES OF PRECIPITATED DUST LAYERS
Background
Once a dust layer is deposited on the collection electrodes of an elec-
trostatic precipitator, the negative corona current must pass through the
particulate layer to the grounded collection electrode. The average electric
field in the particulate layer (E ) can be increased to the point that the
"
layer breaks down electrically. When this breakdown occurs, one of two possible
situations will ensue. If the average electrical resistivity of the particulate
layer is moderate (M). 1-1. 0x10 1 l ohm-cm), the applied voltage may be sufficiently
high so that a spark will propagate across the interelectrode space. The rate
of sparking will determine the operating electrical conditions in such a cir-
cumstance. If the average electrical resistivity of the particulate layer is
high (>101 1 ohm-cm) , the applied voltage may not be high enough to cause a spark
to propagate across the interelectrode space. In this case, the particulate
layer will be continuously broken down electrically and will discharge positive
ions into the interelectrode space. This condition is called back corona or
back discharge. The effect of these positive ions is to reduce the amount of
negative charge on the particles in the gas stream due to bipolar charging
which in turn reduces the space charge component of the electric field. Both
the magnitude of particle charge and the rate of particle charging are affected
adversely by back corona. Useful precipitator current is therefore limited to
values which occur prior to electrical breakdown whether the breakdown results
in sparkover or back corona. At the present, no comprehensive theoretical
model is available to describe the electrical conditions in the collected par-
ticulate layer. This type of model is needed in order to develop a complete
model for electrostatic precipitation and in order to provide a fuller under-
standing of the physical mechanisms which produce the observed electrical
phenomena .
In this paper, a theoretical approach for describing the electrical char-
acteristics of precipitated dust layers is proposed and discussed. The macro-
scopic electrical properties of the dust layer are described by coupling the
56
-------
current transport equation and Poisson's Equation. The electrical breakdown
of the dust layer is attributed to enhanced values of the local electric field
due to the combined effects of polarization in the dielectric layer and the
contribution to the electric field from the space charge distribution in the
layer.
Models formulated from the proposed approach offer the potential of pro-
viding the following information:
(1) voltage-current characteristics of the particulate layer,
(2) electric field, electric potential, and charge density distributions
in the particulate layer,
(3) average electrical resistivity and positional dependence of resistivity,
(4) electrical force holding the layer to the collection electrode,
(5) prediction of the average electric field strength at the onset of
electrical breakdown of the layer, based on a designated value of
breakdown strength, and
(6) prediction of what types of conditioning agents will be beneficial
in reducing the resistivity of the particulate layer.
In one commonly accepted picture of electrical breakdown of the dust layer,
the gas trapped in the interstitial space breaks down electrically.9'10 This
breakdown results from the acceleration of free electrons to the gas ionization
velocity which produces an avalanche condition similar to that at the corona
electrode. Experimental measurements show that particulate layers may experi-
ence electrical breakdown at average electric field strengths across the layers
of approximately 5-15 kV/cm.11'12 For temperatures and pressures encountered
in precipitators, it takes an electric field strength of approximately 15-30
kV/cm to cause electrical breakdown of the gas.13 This suggests that locally
high electric fields must exist in the particulate layer to produce breakdown
of the gas in the layer.
Polarization Effect
The effect due to polarization can be estimated by treating the layer as
a collection of spherical particles all of the same size and arranged in a cubic
array. The problem is then to evaluate the effect of placing dielectric spheres
in an applied electric field. For a parallel plate capacitor arrangement of
metal electrodes containing a dielectric material, the polarization charge in-
duced on the surface of the dielectric is neutralized by charge flowing in an
external circuit. The local electric field inside the volume of such a dielec-
tric is computed by combining the contributions from all the dipoles of the
particles with the applied electric field. Figure 7 shows the physical arrange-
ment and nomenclature that will be utilized in estimating the effect due to
polarization. A calculation of these effects has been presented by the present
authors.
Space Charge Effect
The effect of space charge on the macroscopic electric field can be in-
vestigated by solving the problem involving the transport of charge carriers
through the particulate layer. A simplified model is proposed in order to
demonstrate that space charge effects can result in a significant enhancement
57
-------
Figure 7. Polarization of particles in the layer.
of the electric field in the particulate layer. In this model, we assume that
only one species carries the charge and that the gradient term in the transport
equation can be neglected for all values of x. For a detailed calculation see
reference 14. When polarization and space charge effects are combined, the
maximum local electric field strength (E ) in the layer is given by
.
loc
s;1.5 [1 + 1.25 (K-l)] E
'Ave
(8)
This relationship depends on the dielectric constant (K) of the layer or,
more specifically, of the individual particles. Since, in the case of fly ash
particles, the dielectric constant may vary over a certain range from one source
to another, it is desirable to compute the maximum electric field strength as a
function of dielectric constant. The average value of the electric field for
which breakdown will occur can also be computed as a function of dielectric
constant by specifying a breakdown strength (ER) for the trapped gas. For
illustrative purposes, the breakdown strength of the gas will be assumed to be
that of air (approximately 30 kV/cm at T0 = 293°K and PQ = 760 mm Hg). It will
also be assumed that the breakdown strength of the gas at any arbitrary temper-
ature (T) and pressure (P) can be obtained by multiplying 30 kV/cm by the
quantity (To*P)/(T*Po)- Table 3 shows the results of calculations for several
cases of interest. Notice that the description of dielectric breakdown ob-
tained from the simplified model presented here is consistent with observed
phenomena.
If the dielectric constant and the average resistivity of a particular
fly ash are measured in the appropriate environment, then a practical upper
bound on the operating current density can be established to avoid dielectric
breakdown. If the dielectric constant is known, we can compute the average
58
-------
Table 3. AVERAGE ELECTRIC FIELD AT BREAKDOWN FOR VARIOUS
DIELECTRIC CONSTANTS, TEMPERATURES, AND PRESSURES
Ul
ED = 30 kV/cm at
15
293°K, 760 mm Hg
E = 20.8 kV/cm at
422°K, 760 mm Hg
EB = 11.4 kV/cm at
616°K, 608 mm Hg
K
1
2
3
4
5
6
7
8
9
10
E /E EAve f°r
loc Ave Breakdown (kV/cm)
1.50
3.38
5.25
7.13
9.00
10.88
12.75
14.63
16.50
18.38
20
8
5
4
3
2
2
2
1
1
.00
.88
.71
.20
.33
.76
.35
.05
.82
.63
Eloc
1
3
5
7
9
10
12
14
16
18
EAve for
Ave Breakdown (kV/cm)
.50
.38
.25
.13
.00
.88
.75
.63
.50
.38
13
6
3
2
2
1
1
1
1
1
.90
.15
.96
.92
.31
.91
.63
.42
.26
.13
E.. for
F /F
loc7 Ave Breakdown (kV/cm
1.
3.
5.
7.
9.
10.
12.
14.
16.
18.
50
38
25
13
00
88
75
63
50
38
7
3
2
1
1
1
0
0
0
0
.60
.37
.17
.60
.27
.05
.89
.78
.69
.62
-------
electric field required for breakdown from equation (8) . The current density
corresponding to this average electric field would constitute the upper bound
on the operating current. The condition which must be satisfied in order to
avoid breakdown is
qj < E/p (9)
where E, is computed for breakdown.
SUMMARY
An approximate model for computing current-voltage characteristics in a
wire-plate precipitator has been presented. This model uses a rigorous solu-
tion for the paired charges of opposite sign on the electrodes. The contribu-
tions to the potential and electric field from space charge are approximated.
Near the wires these approximations are based on cylindrical symmetry while
near the plates the approximations are based on rectangular symmetry. The
predictions of this approximate model agree quite well both with the numerically
integrated solutions and with experimental results.
Values of effective ion mobilities were measured at three different in-
dustrial plants. The values of reduced effective mobilities for negative
corona ranged from 1.2 to 1.5 m2/V-s. Values of reduced effective mobilities
for positive corona ranged from 1.4 to 2.5 m2/V-s. The positive mobility was
found to exceed the negative mobility by a factor of about 1.6. A method for
obtaining a self -consistent solution which incorporates the effects of charged
particles is proposed. This method is based on an effective current density,
an effective mobility, and an effective charge density. Iterations on the
effective mobility are required to obtain self -consistency.
A method for computing the electrical properties of the dust layer is
proposed. The approach is to solve the electrical transport equation coupled
with Poisson's equation. Electrical breakdown fields in the layer are com-
puted in terms of an enhancement by space charge and by dielectric polarization
of the particles in the layer. The combination of space charge and polariza-
tion are shown to give a sufficient enhancement of the average electric field
to explain electrical breakdown of trapped gas in the layer.
ACKNOWLEDGMENT
This work was supported by the Industrial Environmental Research Laboratory,
U.S. Environmental Protection Agency, Research Triangle Park, North Carolina,
under Contract Nos. 68-02-2193 and 68-02-2683, Dr. Leslie E. Sparks, Project
Officer.
REFERENCES
1. McDonald, J. R. A Mathematical Model of Electrostatic Precipitation:
Revision 1. Volume 1, Modeling and Programming. EPA-600/7-78-lllb, U.S.
Environmental Protection Agency, Research Triangle Park, North Carolina,
June 1978.
60
-------
2. McDonald, J. R., W. B. Smith, H. W. Spencer III, and L. E. Sparks. A Mathe-
matical Model for Calculating Electrical Conditions in Wire-Duct Electro-
static Precipitator Devices. J. Appl. Phys., Vol. 48, No. 6, June 1977.
3. Cooperman, P. Progress Report No. 46, Research Corporation Laboratory,
(1952).
4. Peek, F., Jr. Dielectric Phenomena in High Voltage Engineering (University
Microfilm International, Ann Arbor, Michigan, 1959), p. 73.
5. White, H. J. Industrial Electrostatic Precipitation (Addison-Wesley,
Reading, Mass., 19), p. 98.
6. Penney, G. W., and R. E. Matick, Trans. AIEE 79, Part I, 91 (1960).
7. Loeb, L. B., Electrical Coronas Their Basic Physical Mechanisms (University
of California Press, Berkeley, Ca. 1965), p. 46.
8. McDonald, J. R., S. M. Banks, and L. E. Sparks. Measurement of Effective
Ion Mobilities in a Corona Discharge in Industrial Flue Gases. Symposium on
the Transfer and Utilization of Particulate Control Technology: Volume 1,
Electrostatic Precipitators. EPA-600/7-79-044a, U.S. Environmental Pro-
tection Agency, Research Triangle Park, North Carolina, February 1979.
9. S. Masuda, A. Mizuno, and K. Akutsu, Initiation Condition and Mode of Back
Discharge for Extremely High Resistivity Powders. IEEE 1977 IAS Conference
Record, Printcraft, Inc., Conneaut, Ohio.
10. S. Masuda and A. Mizuno, Flashover Measurements of Back Discharge. Journal
of Electrostatics, 4, 215-231 (1978).
11. J. W. Baker and K. M. Sullivan, Reproducibility of Ash Resistivity Deter-
minations. Joint Power Generation Conference, Long Beach, California,
September 1977.
12. Work performed at Southern Research Institute under sponsorship of the U.S.
Environmental Protection Agency on Contract No. 68-02-2114 (to be published).
13. F. M. Peek, Jr., Dielectric Phenonema In High Voltage Engineering, Third ed.,
McGraw-Hill, New York, 1929, pp. 43-53.
14. McDonald, J. R., R. B. Mosley, and L. E. Sparks. An Approach to Describing
Electrical Characteristics of Precipitated Dust Layers. Submitted to the
Journal of the Air Pollution Control Association in May, 1979 for review
and consideration for publication.
APPENDIX
Electrical Solutions for the Approximate Model
An expression for the total potential which combines the separate contribu-
tions described in the text is shown below. The x and y components of the elec-
tric field are obtained by differentiating the potential expressions. After
the functions have been matched at x = R we obtain
61
-------
V(x,y) = - f
2 ,2J
DC-
3 /2
(Sx-R) + Vi In
- cos(irR/2s)
X
cos(TTR/2s
-a0Eo
In
(xz+y2-a02))
(In
(s -x) xz+y2+4s (s
x x ) + In ( x x
_____
(s -R)
XX
___________ ,_.-r-L.. __. __
R2+y2+4s (s +R)
X X
N
m=l
[coshir (y-2ms )/2s - cos(TX/2s )
I y x *^__
coshir (y-2ms )/2s + cos(irx/2s )
L y x x
+ In
"coshir (y+2ms )/2s - cos(irx/2s
coshir(y+2ms )/2s + cos(irx/2s
y x >
N
V
In
rcoshTr(y-2ms )/2s - cos (frx/2s
st / j "* [^coshTT (y-2ms ) /2s + cos(irx/2s )
m=-N
(Al)
Ex(x,y) =
- l
x-2s
x
x x
x+2s
X X
N
- ^ Sin(irx/2s )
sx
m=l
cosh-rr (y-2ms ) /2sx
cosh2ir(y-2ms )/2s - cos2 (irx/2s )
y x x
coshir (y+2ms ) /2sx
cosh2ir (y+2ras ) /2s - cos2 (irx/2s )
y x x
V TT
m=-N
coshTt (y-2ms )/2s
cosh2 IT (y-2ms )/2s - cos2 (irx/2s )
r(x,y) =
(A2)
- 1
62
-------
R2+y2+4s (s -R)
X X
R2+y2+4g
X X
N
COS(TTX/2S)
Sinhir(y-2ms )/2s
x' £—1 )cosh2ir(y-2ms )/2s - cos2 (rfx/2s )
m=l \ y x x
Sinhir(y+2ms )/2s
y x
cosh2ir(y+2ms )/2s - cos2 (irx/2s )
y x x
V IT
-p- cos(TTx/2sx)
N
SinhTr(y-2ms )/2s
y x
j cosh2TT (y-2ms )/2s - cos2 (irx/2s )
m=-N l y x x
(A3)
for x < R, and
1/2 3/2
(SX-R)
3/2
x
N
(Vst+Vl)
In
m=-N
coshir(y-2ms )/2s - cos(irx/2s )
y x x
coshn (y-2ms, ) /2s + cos(irx/2s )
J;
(x-R)
(A4)
N
- (Vgt+Vi)7r/sx Sin(irx/2sx)
m=-N
coshir(y-2ms )2sx
cosh2TT(y-2ms,J/2sv - cos2(Trx/2s
y x
Ey(x,y)= -
N
cos(Trx/2sx)
m=-N
SinhTr(y-2ms
____^ > «
cosh2ir(y-2ms )/2sx - cos2
(A6)
for x > R, where
Vi =
x
TTR
X
R
R~2s
R+2s
2 |R2+4sx(sx-R) R2+4sx(sx+R)
(A7)
63
-------
6 = 2Js
5 = (1 - Sa02) j,
Y = (i + S(R2-ao2))^,
E0 = electric field at the wire,
ao = radius of the wire.
o
Because the potential on the plane x = R must be constant, y terms are empiri-
cally added to the integration constants of the cylindrically symmetric solu-
tions. This would have little effect near the wire, but allows the boundary
conditions to be met by a slight modification to the cylindrical solution.
This solution not only yields a constant potential on the plane x = R, but
also leads to x and y components of electric field which satisfy the necessary
boundary conditions imposed on the solution by symmetry. Specifically, the
x-component of electric field must vanish in the plane of the wires, while
the y-component must vanish at the plates and at the two planes represented
by dashed lines connecting the plane of the wires with the plates in Figure
1. For cases in which the radius (R) of the cylinder of space charge is
larger than the half wire-to-wire spacing, adjacent cylinders of space charge
will overlap. In this situation the effects of space charge would be over-
estimated. To correct for that difficulty in this simple model we reduce the
contribution to the potential and electric field components from space charge
associated with wires other than the central one by the fraction of the volume
of the cylinder that does not extend beyond S . This fraction is given by
2 I
1 - - \ ArcCos(sy/R) -
' sy > R • (A8)
This factor would multiply Vi in the space charge terms arising from wires
other than the central one (m = 0). Surprisingly, this simplified approach
to computing a correction factor yields rather good agreement between the
approximate model and the numerically integrated solution.
64
-------
LATERAL PROPAGATION OF BACK DISCHARGE
By:
Senichi Masuda
Department of Electrical Engineering
Faculty of Engineering, University of Tokyo
7-3-1, Kongo, Bunkyo-ku, Tokyo, Japan 113
Sadaji Obata
Ishikawajima Harima Heavy Industries Ltd.
Technical Research Center
3-2-16, Toyosu, Koto-ku, Tokyo, Japan 135
ABSTRACT
In a tri-electrode type precipitator having the discharge, third, and
collecting electrodes, where the main voltage is applied between the third
and collecting electrodes and a dc or pulse voltage is applied between the
third and discharge electrodes for producing corona discharge, special pre-
cautions must be taken to avoid the lateral propagation of back discharge.
The primary back discharge occuring in a resticted region on the collect-
ing electrode facing to the discharge electrode tends to propagate in the
lateral direction, once it takes the form of streamer-mode, finally to co-
ver the whole surfaces of the collecting and third electrodes. The secon-
dary back discharge resulted by this propagation becomes self-sustained if
well developed, and does not disappear unless the main field strength is
lowered below a certain critical level. The propagation initiates when the
magnitude of the main field strength exceeds ca. 5 kV/cm in air at NTP, and
it is extinguished at a slightly lower field strength. The use of pulse
voltage for corona production decelerates the occurence of the propagation
to a great extent. When high resistivity dusts causing back discharge are
encountered, it is imperative to keep the main field strength below the ext-
inguishing threshold of the propagation, in order to obtain the inherent ad-
vantage of the tri-electrode precipitator. The use of pulse voltage for
corona production is also highly desirable. This lateral propagation occurs
in any precipitators having a parallel plane electrode construction, too.
65
-------
LATERAL PROPAGATION OF BACK DISCHARGE
1. INTRODUCTION
During the course of developmental efforts of the tri-electrode type pre-
cipitator as illustrated in Figure 1, vhere the main dc voltage is applied be-
tween the third and collecting electrodes for producing the collection field
and the dc or pulse voltage is applied for producing a controlled corona dis-
charge, a curious phenomenon was observed to occur when high resistivity dust
causing back discharge was being collected. The collection performance as
expected could be obtained only
when the main field strength was
kept below a certain critical
level, beyond which a severe imp-
airment of performance started to
occur. In this case the surface
glow of back discharge covered
the whole surface of both collec-
ting and third electrodes, not
only the restricted region on the
collecting electrode facing to
Collecting
electrode
Gas outlet
Switch
a,b,c,d: dc high voltage source
e: pulse high voltage source
Figure 1 Tri-electrode type precipitator.
the discharge electrode, but also
the other regions where no ions
from the discharge electrode were
arriving. The back discharge,
once occured in the region A fac-
ing to the discharge electrode because of its ionic current, propagated towards
the remote regions where no primary ions were existing, because of the secondary
ions emitted from the back discharge points. This phenomenon of lateral prop-
agation of back discharge was already observed to occur in the author's experi-
ments of electrostatic powder deposition using a parallel plate electrode system
and tribo-charged resin powders (Masuda et. al. (19T7)1)- In this paper are
reported the results of investigations made by the authors in order to
clarify this phenomenon in more details and to find out the condition to avoid
this trouble.
2. METHOD OF EXPERIMENTS
Figure 2 illustrates the model tri-electrode system used in the present
investigation, consisting of two parallel circular discs as the third and coll-
ecting electrodes and a needle located at the center of a small hole drilled at
the center of the third electrode. The probe electrodes are attached to both
the third and collecting electrodes for measuring current density distribution.
Instead of dust samples, two kinds of towel paper samples are used: one with
1.2 mm thickness and 0.03^ g/cm3 volume density (low-density sample), and the
other out of 5 original low density papers pressed to 1.2 mm thickness (high
-density sample). The sample resistivity is changed in a wide range in air at
NTP by changing the ambient air humidity within "the humidity chamber in which
66
-------
Pulsa high voltage source
Discharge electrode
Third electrode
_ I'll
D.C. high voltagt
source for corona
L
,/
Collecting
— Z15 -^ »
1 mjS
J^aissiESia tssa
electrode
j
Probe electrode
O.C. high voltage
source for main field
the entire system is located. The inter-electrode field intensity between the
third and collecting electrodes (main field intensity ),E]_ = VJJ/D, and the pri-
mary corona current from the needle are changed by changing the values of VB
and Vp. However, the primary corona current is not used as a parameter as it
does not provide reproducible results, and the average field intensity between
the third and needle electrodes, E2 = Vp/R, is used instead. Between the th-
ird and needle electrodes is applied
either a negative dc voltage or a
negative pulse voltage. E2 in the
latter case represents the average
field intensity in terms of pulse
crest voltage.
At first, the mode of the lat-
eral propagation of back discharge
is observed under different condit-
ions using an image intensifier tube
(EMI type 9912). Then, the initi-
ation and extinguishing conditions
of the secondary back discharge are
measured. Finally, the current
wave form of the secondary back dis-
Figure 2 Test tri-electrode system charge is observed.
3. RESULTS OF EXPERIMENTS
3.1 Mode of lateral propagation of back discharge
The primary back discharge, occured in the region A in Figure 2, can gener-
ally propagate in the lateral direction only when it takes the form of space-
streamer-mode and the magnitudes of EI and £2 exceed respectively their own th-
reshold values, except a few cases to be described later. The development of
the space-streamer-mode back discharge requires not only a sufficiently high
main field intensity, Ej_, but also enough quantity of corona current obtainable
by raising E2- Raising of Ej_ at low E2 only results in sparking without caus-
ing the primary back discharge itself, whereas raising of E2 at low EI only cau-
ses the primary back discharge in a form of either surface-glow or very weak
space-streamer so that no lateral propagation can take,place. Except for the
lower dust resistivity case at p^ < 10-'--'- ohm-cm, once the secondary back disch-
arge occupied the whole inter-electrode region, it becomes self-sustaining and
does not disappear unless the main field intensity, EI, is lowered below a cer-
tain threshold, even if the primary back discharge is extinguished by lowering
E2- The propagation mode differs completely between the dc corona operation
and pulse corona operation, so that it is described separately in the following.
a) DC corona operation. Figure 3 (a) - (d) are the photographs of the second-
ary back discharge occuring in the dc corona operation when only the collecting
electrode is covered with the low-density sample. As shown in these pictures,
the lateral propagation can occur even when the third electrode is kept clean.
However, the secondary back discharge in this case always takes the form of sp-
ace-streamer bridging across the third and collecting electrodes. It is evid-
ent that the negative ions to maintain the secondary back discharge in the outer
region around A are being regenerated in the streamer channels. These ions are
conveyed through the field-drift and diffusion processes towards the regions on
67
-------
the collecting electrode remote from the original back discharge spots.
= 5.5 kV/cm E! - 5.1 kV/cm Ej_ = 5-1 kV/cm E]_ = 5-1* kV/cm
Pd = l.lxioll pd = 2,5x1012
ohm-cm ohm-cm ohm-cm ohm-cm
(a) (b) (c) (d)
Figure 3 Secondary "back discharge occured when only the collecting electrode
is covered with a low-density sample layer.
E-J_ = 6.9 kV/cm E! = 5-2 kV/cm E! = 5.2 kV/cm EI = 5.2 kV/cm
Pd = l.lxlOU pd = 2.5x1012 p = l.lixlQiS p = l.UxlO1^
ohm-cm ohm-cm ohm-cm ohm-cm
(a) (b) (c) (d)
Figure h Secondary back discharge occured when both the collecting and third
electrodes are covered with high-density sample layers.
In comparatively low resistivity region p - 10 L ohm-cm, below which no
lateral propagation occurs, the secondary streamers are very weak, unstable, and
move around rapidly in the inter-electrode spacing without covering its entire
region (Figure 3 (a)). The secondary streamers in this stage are not self-
sustaining. In the intermediate resistivity region at pd = ICr- ohm-cm, the
secondary streamers become well developed, self-sustaining, and can occupy the
entire inter-electrode spacing (Figure 3 (b)). In the high resistivity region
at p<3_ - 1Q13 ohm-cm, the leader channels begin to appear, stemming from the co-
llecting electrode and splitting into many streamers at their upper ends (Figure
3 (c)). However, the appearance of the leaders seems to be colsely related to
the increased breakdown strength of the sample which is raised at the same time
by lowering the ambient air humidity to get high sample resistivity. In the
very high resistivity range at p^ - 10 1** ohm-cm, all the secondary streamers
are originating from the leaders, as shown in Figure 3 (d), while the primary
back discharge turns into surface-glow-mode. In the high resistivity range
beyond lO-^ ohm-cm, several curious phenomena are observed to occur. When the
random sparking occurs repeatedly at very high value of E]_, the secondary strea-
mers can start to appear without the aid of the primary streamers. Secondly,
58
-------
even after the secondary back discharge has "been extinguished once by lowering
E]_, it can be reignited when E]_ is raised again in due time.
When the low-density sample is replaced by the high-density sample in Fig-
ure 3, the mode of secondary back discharge as indicated by (a) - (d) occurs at
the resistivity value one order of magnitude higher.
In case both the collecting and third electrodes are covered with the low-
density samples, no appreciable change occurs in the secondary back discharge of
each resistivity level. On the other hand, when the high-density samples are
used instead to cover both electrodes, a drastic change appears as indicated by
the photographs of Figure U (a) - (d). In the lower resistivity range at p -
1011 ohm-cm, the secondary back discharge appears only in a form of the scatter-
ed glow points primarily on the third electrode (Figure b (a)). Beyond 1012
ohm-cm the secondary streamers begin to appear and cover the entire electrode
region. However, these streamers are always stemming from the collecting ele-
ctrode, and do not bridge across the electrode gap (Figure k (b) and (c)). The
back discharge on the third electrode remains in a form of surface-glow. In
the very high resistivity range p - lO1^ ohm-cm, the secondary streamers do
completely disappear, and the back discharge in the surface-glow-mode covers
both electrodes. In addition, in the high resistivity range beyond l(P-3 ohm-cm
a new type of secondary back discharge appears on both electrodes at a fairly
low value of E^ at which the secondary streamers cannot occur. This is of the
surface-glow-mode and extremely feeble so that it can be detected only with the
aid of an image intensifier. Its current is also extremely small, so that it
may be doubtful whether this back discharge will cause a detrimental effect on
the collection performance.
b) Pulse corona operation.
: pulse voltage
' Of 1
r
i
~Ta *i
i i
nrnr
3 interval for occurence
bhe primary streamer
i
! f~20Hz- +• — Rmc;
f=100Hz; t =lms
Uf=500Hz;
t =0.2ms
i i _
D 2 4 6 8
time (ms)
Figure 5 Delay of time interval for
occurance of primary strea-
mer.
69
The pulse corona operation enables a large reduct-
ion in primary corona current without
lowering the main field intensity, ET_
(Masuda et. al. (l976)2(l9T7)3(l9T8)i:i:).
This current-limiting characteristics
hampers greatly the initiation of the
primary back discharge itself and also
its transition into the space-streamer
which is the origin of the lateral pro-
pagation. This tendency is pronou-
ce_d when both electrodes are covered
with the high resistivity samples.
The primary back discharge, once occur-
ed, usually remains in the surface-
glow-mode and hardly turns into the
space-streamer-mode to cause the later-
al propagation unless the main field
intensity, E^, and the primary corona
current are greatly increased. Here,
the raising of the corona current can
be made by either raising the pulse
crest voltage, its repetition frequency
f, or its duration time tp.
In addition to the current-limit-
-------
ing effect as described above, the pulse corona operation has another large
advantageous effect of hampering the lateral propagation, such that the pri-
mary streamers tend to occur in retard to the pulse voltage, as illustrated in
Figure 5- The time interval in which the primary streamers occur becomes more
and more retarded to the pulse duration interval with decreasing pulse duration
time, tp. As a result, ths situation- disappears that the primary .streamers can
be intensified enough to cause the lateral propagation by the strong space-
charge field of oncoming ions which are supplied only in the pulse duration
period. However, this effect becomes lost when the pulse repetition frequency
f is excessively raised, because the time interval of the primary streamer occ-
urance covers the entire time domain in this case, as indicated by the bottom
curve in Figure 5•
Owing to these circumstances, the lateral propagation of back discharge is
greatly decelerated when the pulse corona operation is used. However, this
does not always exclude the possibility that the lateral propagation may be
caused at very high values of E]_ and E2 after enough time has elapsed during
which ions accumulate on the sample surface in a sufficient amount. Actually
the initiation threshold of the secondary back discharge in terms of E]_ does
not change much from that in the dc corona operation. Moreover, once the lat-
eral propagation has developed in the inter-electrode space, any difference is
lost between the pulse and dc corona operations, because the secondary stream-
ers, then, become self-sustained.
3.2 Initiation and extinguishing conditions of lateral propagation
a) Initiation condition. The initiation condition of the lateral propagation
of back discharge does not differ much in all of, the cases as described, and it
is governed by the parameters E]_ and Eg. The curves in Figure 6 represent the
typical instances of the initi-
ation conditions for the dc and
corona pulse operations. It
can be clearly seen that the
I 5 j. " ^*—-K,. ^ _x_ ' threshold value of the main
field intensity, Ej_, for the
initiation of the lateral pro-
4 I. pagation, approximately 5 kV/cra,
coincides well with the condit-
ion of streamer propagation in
air at NTP.
o
i 1 * 1 .j _
8 10 b) Extinguishing condition.
2 cm There is no difference in
the extinguishing condition of
i: d.c. corona operation; only the counter electrode ^, ..
13 the secondary back discharge
covered with a low-density samples; r,= 1.4x10 ohm-cm
d between the dc and pulse corona
A: pulse corona operation; both electrodes covered
13 operations, and moreover, it is
with high-density samples; i,- 1.4x10 ohm-cm •*• "
n , ..d . . , , also independent of the magni-
O: pulse corona operation: both electrodes covered ^
... , . , , ... , , . Ini4 , tude of Ep. These are due to
with high-density samples; r,= 1.4x10 ohm-cm £-~
D: pulse corona operation: both electrodes covered the Self-SUStaining effect Of
with high-density samples; rd= i.oxio15ohm-cm the secondary back discharge as
described previously. The
'igure 6 Initiation condition of secondary extinguishing condition is pri-
back discharge , marily governed by the main
70
-------
4 •-
2 -•
1
io12
—I—
io14
H
io15
—\
1016
—o-
A :
1011
io13
resistivity (ohm-cm)
extinguishing threshold; low-density sample
extinguishing threshold; high-density sample
sparkinh threshold; low-density sample
sparking threshold; high-density sample
Figure 7 Extinguishing condition of secondary back discharge.
(only the collecting electrode covered with a sample)
10
17
8 I \
\^
%
6 •• ~fc====z-^
4 ., ° ^C
2 -•
0 | |
io11 io12
— • — : extinguishing
— D — : extinguishing
_— A — : sparking thres
A : sparking thres
— . — o — : extinguishing
fs"'
^'
» — — i-^zLiz: j-j^^_"J — -^^^
>"*~^> _-C3-
N8 S^-D — "^""
III!!
io13 io14 io15 io16 io17
resistivity ( ohm-cm)
threshold; low-density sample
threshold; high- density sample
hold; low-density sample
hold; high-density sample
threshold of secondary back discharge in
surface-glow-mode; high-density sample
Figure 8 Extinguishing condition of secondary back discharge.
("both the collecting and third electrodes covered with samples)
71
-------
field intensity, EI, and slightly by the sample resistivity, pd> too. The
solid curves in Figure 7 represent the extinguishing threshold of Ej when only
the collecting electrode is covered with the low- and high density samples, pl-
otted against pd. The solid curves in Figure 8 indicate the same threshold
when both the collecting and third electrodes are covered with the low- and
high-density samples. The dotted curves in Figures 7 and 8 represent the
sparking threshold. The broken line in Figure 8 indicates the extinguishing
condition of the very feeble secondary back discharge in the surface-glow-mode
with negligibly small corona current, as described in the preceding section.
It can be seen from Figures 7 and 8 that the extinguishing threshold of E^ does
not differ much in all of the cases, except a special case as described above.
It lies in the range of 3.5 - 5 kV/cm and is always slightly lower than its
initiation threshold under the same condition.
3.3 Current wave form
The current wave form of the secondary back discharge in the streamer-mode
is essentially the same as that of the primary streamer back discharge which
have been already reported (Masuda and Mizuno (1976/1977)5).
U. CONCLUSION
The experimental results thus far obtained provide the following conclu-
sions:
l) The lateral propagation of back discharge starts to occur when the sample
resistivity, p<}, exceeds the level of ca. lO1-*- ohm-cm, but it becomes domin-
ant only when p^ exceeds ca. 10^2 ohm-cm. It is resulted by the primary
back discharge in the space-streamer-mode, and propagates mostly in a form
of the secondary streamer. The secondary back discharge becomes self-
sustaining once well developed, and it does not disappear unless the main
field intensity, EI, is lowered below a certain threshold level.
2) The magnitudes of the main field intensity, EI, and primary corona current
expressed in terms of E2 play the most essential role in the initiation and
extinguishing of the secondary back discharge. Their threshold values for
the initiation and extinguishing do not differ much under different operat-
ing and dust resistivity conditions.
3) The threshold value of the main field intensity, E^, for the initiation of
the lateral propagation is about 5 kV/cm in air at BfTP, which agrees well
with the streamer propagation condition in atmospheric air. The threshold
value of EI for the extinguishing of the secondary back discharge is always
slightly lower than that for the initiation.
^) The lateral propagation of back discharge can be greatly decelerated by the
use of the pulse corona operation with the suitable crest voltage, £2, repet-
ition frequency, f, and duration time, tp. However, the possibility of the
propagation to occur after a sufficiently long time for ion accumulation may
not be excluded even in the pulse corona operation. This leads to the con-
clusion that the tri-electrode precipitator should be operated at the main
field strength, E-|_, below the extinguishing threshold as described above.
5) This extinguishing threshold of E]_, however, lies on a level (3.5 - 5 kV/cm
in air at 1TTP) which is still sufficiently high for a satisfactory collect-
ion performance to be obtained.
72
-------
The tri-electrode precipitator has its inherent large advantage to decrea-
se freely the primary corona current, without lowering the magnitude of the
main field intensity, down to the level at which no primary back discharge does
occur. The use of the pulse corona operation makes this current-limiting act-
ion more easily, and it also makes the distribution of the primary corona curr-
ent on the collecting electrode much more uniform, especially when the corona
current is to be kept at a very low level. However, in order to exploit this
advantage of avoiding back discharge, it is not sufficient to take precaution to
avoid the primary back discharge only by lowering the primary corona current
below a certain threshold level. Another additional precaution must be taken
against the occurence of the secondary back discharge. In order to avoid this
trouble under any circumstances, the tri-electrode precipitator should be ope-
rated at a main field intensity level below the extinguishing threshol as des-
cribed.
Finally, it should be pointed out that the secondary back discharge of the
similar character was obserbed, by the use of an image intensifier, to occur in
the conventional twin-electrode precipitators collecting high resistivity dust
at the frame-works supporting discharge electrodes. The secondary back dis-
charge is also expected to occur in any type of collecting field comprizing the
parallel plane electrode construction, such as the collection field of the two-
stage type precipitator having a separate precharging section. The same pre-
caution will be needed in such applications.
ACKNOWLEDGEMENTS
The author appreciates the assistance of his co-workers, Mr. Kensuke Akutsu
and Mr. Yoshio Ogura, given to this research. The financial supports given by
the Ministry of Education Japan and IHI Heavy Industries Ltd. are also appreci-
ated.
REFERENCES
(l) Masuda, S., A. Mizuno and K. Akutsu. Initiation Condition and Mode of
Back Discharge for Extremely High Resistivity Powders. Record of IEEE/IAS
1977 Annual Meeting, p. 867 (Los Angels, October 1977).
(2) Masuda, S., I. Doi, M. Aoyama and A. Shibuya. Bias-Controlled Pulse Char-
ging System for Electrostatic Precipitators. Staub-Reinhaltung der Luft
(Bonn). Vol. 36, p. 19 (1976).
(3) Masuda, S., I. Doi, I. Hattori and A. Shibuya. Utility Limit and Mode of
Back Discharge in Bias-Controlled Pulse Charging System. Record of IEEE/
las 1977 Annual Meeting. p. 875 (Los Angels, October 1977).
(h) Masuda, S. Novel Electrode Construction for Pulse Charging. Proc. of
1st EPA-Symposium on Transfer and Utilization of Particulate Control Tech-
nology. Vol. 1, p. 241 (Denver, July 1978).
(5) Masuda, S. and A. Mizuno. Light Measurement of Back Discharge. Journal
of Electrostatics (Amsterdam). Vol, 2 (1976/1977), p. 375-
73
-------
THEORETICAL MODELS OF BACK CORONA AND LABORATORY OBSERVATIONS
by
D.W. VanOsdell
P.A. Lawless
Research Triangle Institute
P.O. Box 12194
Research Triangle Park, N.C. 27709
and
L.E. Sparks
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, N.C. 27711
ABSTRACT
Several mechanisms for the generation of positive ions due to back corona
were postulated. These mechanisms were then used in conjunction with an elec-
trostatic precipitator model to produce calculated voltage-current (V-I)
relationships for each back corona mechanism. Multiple values of the
parameters of the back corona mechanism were used and the effect of space charge
was investigated. It was found that near-vertical V-I relationships could be
produced for most of the models and that the hysteresis present in actual data
was possible with certain models. In particular, the model postulating that
positive ions are generated in proportion to a threshold current density was
found to be capable of giving a theoretical V-I relationship which exhibited
the hysteresis and step structure of some experimental V-I relationships.
74
-------
THEORETICAL MODELS OF BACK CORONA AND LABORATORY OBSERVATIONS
1.0 INTRODUCTION
The undesirable effects of back corona in electrostatic precipitators (ESPs)
are well known. The emission of positive ions from the collection plate of an
ESP achieving a bipolar ionic equilibrium reduces the net particle charge,1'2
at a given current, and particle collection is degraded. The formation of the
positive ion current leads to a rapid increase in total current as voltage is
increased.3'4 The increase can be so large as to give near-vertical or even
negative slopes to the voltage-current (V-I) curves. This V-I characteristic
is the most common indication of the presence of back corona, because direct
detection of free ions of both polarities is very difficult. This paper
postulates mathematical formulations for back corona, and, through an ESP model,
relates these back corona mechanisms to calculated V-I curves. The implications
for particle charging and collection are examined.
2.0 PRECIPITATOR MODEL
As back corona is difficult to observe directly, the back corona
mechanisms were used as parameters in an ESP model to give calculated quantities
which could be compared to actual measured conditions. The model used is based
upon equations involving the local electric fields rather than the electric
potentials. The reasons for this are the charges in electric fields are
influenced only by the local conditions and that they are the natural
coordinates for calculating particle charges and the forces on charged
particles. It is fortuitous that the resulting equations are particularly
simple and easy to solve. This ESP model will not be described further in this
paper except to say that the numerical solution is relatively fast and that the
solutions appear to fit actual ESP data quite well. Figure 1 presents V-I data
taken on three identical sections of a pilot-scale precipitator. Comparisons
with other experimental measurements5 produce similar satisfactory agreements.
The point is that this ESP model gives reasonable results and is suitable for
comparison of back corona mechanisms.
3.0 BACK CORONA MECHANISMS
Although recent experimental work has begun to delineate the nature of
back corona mechanisms6'7 revealing streamer and glow models of propagation,
the mechanisms of generating positive ions in the particulate layer are still
uncertain. Electrical breakdown of the particles or the interstitial gases
in the particle layer are postulated,1' but detailed descriptions are lacking.
The approach taken in this paper is to assume several simple model mechanisms
for the generation of positive ions and to explore the consequences of each.
All of these mechanisms are described in terms of a critical current density,
since experimental evidence7 indicates the sudden onset of back corona as the
current density is increased. None of the mechanisms includes an effect due
to local electric field because in standard electrostatic precipitators, the
critical current density and the local electric field are jointly determined
by the resistivity of the particle layer. All the proposed mechanisms are
75
-------
assumed to be local, with back corona onset in a given location having no direct
effect on adjacent locations. Finally, all the mechanisms assume a continuous
generation of ions, as opposed to the pulsing effects observed experimentally.
The five mechanisms chosen are described below.
I. Positive ions are generated in proportion to the amount that the
total current density exceeds the critical current density. This
can be written:
P+ =
where p is the positive free ion space charge density (C/m3), 6 is
an adjustable dimensionless parameter, E is the electric field
strength (V/m) , b is the carrier mobility (m2/V-sec), and j , is the
threshold current density (A/m2). The "x" direction is between the
plate and wire. Since the generation of positive ions occurs in the
dust layer near the plate, only the x-component of field is signifi-
cant. For values of g near 1, this mechanism tends to keep the
electric field at the dust layer constant by reducing the net space
charge as the current density increases.
II. Positive ions are generated in proportion to the threshold current
density, once the total current density exceeds the critical level:
P+ = (2)
where y is an adjustable dimensionless parameter. This mechanism
corresponds to a source-limited generation of ions. It can be seen
to have a built-in hysteresis, because the presence of the positive
ions increases the total current density by a fixed amount at the
threshold.
III. Positive ions are generated in proportion to the total current density,
once the total current density exceeds the critical level:
Si /E b , j >j ,
Jx x ' Jx—Jth
P+ = (3)
76
-------
where 6 is an adjustable dimensionless parameter. This is similar
to mechanism I for current densities far above the threshold, but
should exhibit a hysteresis near the threshold current density.
IV. The positive ions are generated in proportion to the square of the
total current density:
where e is an adjustable dimensionless parameter. This mechanism
could represent a source sensitive to heating effects. In this
particular formulation, there is not a sharp onset threshold, but
the quadratic increase in p is sufficiently rapid to have the appear-
ance of threshold. In practice, the increase of p is so rapid that
numeric instabilities occur unless a maximum limit is placed on p .
This roughly corresponds to the physical situation in which the
rate of heating is eventually balanced by losses to the surroundings.
In this work, the positive ion current density at the plate was
limited to 25 times j .
V. The positive ions are generated in proportion to the negative ion
current density squared:
2
P+ = K3 /J ,vE b , (5)
+ - th x
where j is the x-component of the negative ion current density and
K is an adjustable dimensionless parameter. This mechanism describes
a situation in which the positive ion current does not directly
contribute to the heating effect, as would be the case if generation
occurred at the surface of the layer.
In all the mechanisms except II, the adjustable parameter affects the ratio
of the positive ion current to the negative ion current, and, in general, values
can be found for which the ratio is nearly 1 over an extended range of currents.
Mechanism II is different in that the adjustable parameter affects only the
size of a step increase in positive ion density, and the positive and negative
current densities will be equal at a single current only.
Consequences of Bipolar Particle Charging
In a unipolar ion field, a spherical conducting particle can be shown to
obey the field charging equation:1
= 3Trr2j(l-q/qs)2, (6)
77
-------
where r is the particle radius, j is the current density (assumed to be uniform
far from the particle), t is time, q is the charge on a particle C and qg is
the saturation charge on the particle, given by:
q = 12rrr2e E, (7)
s o '
where E is the electric field magnitude far from the disturbing effects of the
particle and e is the free space permittivity (f /m) . After a sufficiently long
charging time,°a particle essentially attains the saturation charge and stops
charging. In the Stokes law regime, the drift velocity of the particle in the
direction of the field is proportional to Eq } and therefore depends on the
particle radius and the square of the electric field. At a given particle size,
the velocity is proportional to E .
Where ions of opposite polarity impinge upon the particle, a charging
equation of the following form results:
(8)
where the sign convention chosen assumes that q is negative. The net charge
q stops changing at an equilibrium value given By:
.--,-- 1)/(<
q = q Q
max s
The factor multiplying q we call Q for convenience. Q is less than 1
(even negative) for all non-zero values of j . Recalling that q does not
depend on current density, the factor Q represents the degradation in particle
charge that can be expected under bipolar charging conditions, and the particle
drift velocity is proportional to the quantity E2Q under bipolar charging
conditions.
Effects of Particulate Space Charge
Since back corona effects are often observed in operating precipitators,
it is worthwhile to calculate the additional effects attributable to the
immobile particulate space charge. In general, particulate space charge has
the effect of raising the entire voltage-current characteristic to higher
voltages and increasing the electric field at the plate. A more subtle effect
of the space charge is observed in the detailed calculations; it increases the
current density at the plate directly opposite the corona wires and decreases
it in the region between the wires for the same average current density. This
effect modifies the onset of back corona under the five mechanisms from the
condition of zero particulate space charge.
78
-------
4.0 RESULTS OF THE CALCULATIONS WITH FIVE BACK CORONA MECHANISMS
The ESP calculations were made using the following fixed parameters:
temperature
pressure
corona wire radius
wire-to-plate spacing
%wire-to-wire spacing
roughness factor
mobility
J^ „
space charge
615°K
0.86 atm
1.34 mm
11.43 cm
11.43 cm
0.8
7.46x10""^ m2/V-sec
4.0xlO~5 A/m2
0 or -1x10 5 C/m3
The choice of j , is not critical. Too large a value means that unreal-
istically high current densities would be required to exceed the threshold
while too small a value would obscure details as the threshold is crossed.
The non-zero value of q is a substantial space charge density, but not an
excessive one. The effects of the resistivity of the dust layer are not
considered in the early work presented below.
The average current density was initially set to zero, and then incremented
upward to a maximum value, with the corona wire voltage calculated at each step.
The electrical conditions used as the starting values for a given step were the
final solution values of the preceding step to mimic a real measurement of a
voltage-current curve. After the maximum was reached, the current density was
decremented to observe hysteresis effects, but many mechanisms showed no
hysteresis effects at all. Some calculations could not be carried to
completion because of numerical instabilities which were encountered, generally
in cases with extreme back corona.
The voltage-current characteristics were calculated for wide ranges of
the adjustable parameters for each mechanism. Particular attention was paid
to achieving a near-vertical V-I characteristic and to checking for hysteresis
effects, since these are observed characteristics in back corona situations.
Figures 2 through 6 present the voltage-current density curves calculated
for the five mechanisms. In most instances, near-vertical curves could be
obtained, and in some cases hysteresis was observed. Note that hysteresis was
expected in mechanisms II and III (and was calculated), but it also appeared
in mechanism IV in an unusual fashion. The hysteretic behavior of the current
density at levels approaching the critical threshold does not correspond to
physical reality for the following reasons. The electric field at the corona
wire is held constant at the corona-start value; in the calculation it is the
field required to sustain corona current in the absence of external influences,
once it is initiated. But the presence of the positive ion back corona current
is a significant influence which would allow the corona to be sustained at a
lower value of the electric field. The result would be a fairly smooth decrease
in voltage as the current decreases, whereas, in the present calculation, the
calculated voltage "snaps" back to the starting voltage once the back corona
current ceases. The hysteresis at current densities well above threshold
should be little affected by this phenomenon.
79
-------
The significance of the numerical instabilities which occurred for some
values of adjustable parameters cannot be ascertained. The instabilities
invariably occurred at high levels of positive ion current, where real
physical instabilities might logically be expected due to the decreasing net
ionic space charge. Nonetheless, it seems probable that the instabilities
could be suppressed at least partially by careful choice of model parameters.
It is worthwhile to note the stabilizing effect of a particulate space charge
in all the mechanisms. Higher current densities were achieved without
calculational instabilities when particulate space charge was included than
when it was not.
The particulate space charge also tends to suppress the spread of back
corona laterally between the wires. In Figures 3 and 4, the steps in the
curves are due to the cell structure of the calculation, with each step
corresponding to the formation of positive ions in one or two cells as the
local current density exceeds the threshold. With q=0, the transitions
occur in all the cells over a narrow range of total current densities. The
lateral spread of back corona has been verified experimentally.
Consequences of Back Corona
In the Deutsch-Andersen theory of precipitator collection,1'2 the drift
velocity of the charged particles appears as a multiplier in the argument of
an exponential term in an expression for the collection efficiency, and is,
therefore, especially significant. Recalling that the drift velocity is
roughly proportional to E2Q, we have plotted this quantity for the five
mechanisms in Figures 7 through 11. The terms were evaluated at the plate,
because collection occurs at the plate, and a simple linear average over all
the cells was used.
The figures illustrate that the presence of back corona has a strong effect
by whatever mechanism it occurs and regardless of the presence of a particulate
space charge. (We note here that it would prove difficult to maintain an
highly charged particulate in the presence of a bipolar charging rleid.) Two
mechanisms are worthy of note: in mechanism II on the descending legs of the
hysteresis curves, the E2Q product average actually becomes negative; and in
mechanism III, for y = 0.5, the E2Q product approaches zero rapidly as the
current density increases. Both of these conditions are disastrous for
particle collection. The other mechanisms, particularly in the presence of
space charge, exhibit local E2Q products which are zero or negative, but for
which the overall average is positive. The use of a current density weighted
average in these cases would depress the E2Q value calculated even further.
In terms of operating strategy in a precipitator suffering back corona,
the E2Q product is generally a maximum at current densities which avoid
initiating back corona, i.e. low current densities. Unfortunately, although
E2Q represents the drift velocity of a maximally charged particle, it does not
describe the rate of charging which is a function of current density. If
particle charging were accomplished in a prior section, the E2Q product should
then be maximized. If, on the other hand, charging and collection take place
in the same section, operating at low current densities might not fully charge
the particles in that section resulting in reduced collection efficiency.
80
-------
The E2Q curves also point out the enhanced drift velocity in the presence
of a particulate space charge. Even though back corona degrades the E2Q
product, adding space charge can boost it to a significantly higher level.
5.0 COMPARISON WITH EXPERIMENTAL RESULTS
Laboratory measurements to challenge the validity of these mechanisms
have not yet been performed, but two measurements in laboratory precipitators
which exhibit back corona characteristics are presented here. The first is a
voltage current curve measured on a pilot-scale precipitator5 under controlled
conditions after several hours operation collecting a high resistivity dust.
The geometry factors were:
wire-to-plate: 11.4 cm
wire-to-wire : 3.8 cm
wire radius : 1.59x10 3 m (thickly covered with dust)
plate area : 3.0 m2
temperature : 408°K
Such a configuration would be expected to have a corona starting voltage
of about 45 kV under clean conditions and in excess of 30 kV with a roughness
factor of 0.7. The actual V-I curve (Figure 12) was obtained on an X-Y
recorder measuring the corona wire voltage and collector plate current as the
voltage was slowly increased to a maximum and then decreased to zero. The
particulate space charge was zero (no particulate feed).
The ascending leg of the curve exhibits both increases and decreases in
the voltage as the current rises. More importantly, the X-Y recorder reveals
small variations in voltages that would likely be missed in a point-to-point
type of measurement. The descending leg of the curve is more regular but
still has the small step-like character that the ascending leg has.
Of the five mechanisms, only mechanism II has the potential of such a
pronounced hysteresis with step-like increases and decreases in voltage.
Mechanism II is incomplete, however, in that it postulates only a single
threshold. In order to produce a characteristic similar to Figure 12, multiple
closely spaced thresholds would be required. Conceptually, there could be a
spectrum of thresholds around the average value, or the actual current density
along the length of the wire could vary due to minute physical differences,
making the problem a three-dimensional one instead of the two-dimensional
calculation used.
A composite voltage current characteristic based on mechanism II is
presented in Figure 13. This composite can be thought of as a corona wire
made up of five sections, the corona start voltage for each of the sections
differing by 100 volts. This voltage offset corresponds to corona wire
diameter variations of less than 1 percent per offset, which is reasonable
for a dirty corona wire. The composite V-I curve shown in Figure 13 was
generated from a sum of the V-I characteristics given in Figure 3.
81
-------
Other mechanisms can produce V-I characteristics similar to the ascending
leg of Figure 12 with proper choice of the threshold current density and the
adjustable parameters, particularly mechanism IV. The difficulty is in
giving physical meaning to the parameters which, in these calculations, are
arbitrary. In mechanism II, the adjustable parameter controls only the size
of the increase in current, and can potentially be related to the properties
of the dust layer. It is the superposition of many small steps, near the same
threshold, which can produce the steep increase in current, and it is the sum
of the increases which leads to the hysteresis of the descending leg.
Masuda6 has photographed back corona discharges in a point-plane appara-
tus. In a positive corona mode, isolated patches of back corona are observed
in the dust layer, consistent with a composite mechanism II in which different
regions have different thresholds.
The validity of back corona mechanisms can also be tested by examing the
distribution of current across the plate. A current distribution under a dust
layer as a function of horizontal position (Spencer3) is presented in Figure 14.
The data indicate a much higher current density directly opposite the wire for
dirty plates than for clean plates.
Current distribution for the five mechanisms without a resistive dust
layer were calculated. With the exception of mechanism IV, there was not a
strong tendency to concentrate current opposite the wire as in Spencer's3 data.
Mechanism IV did produce such a current distribution, but the significance of
this was uncertain because of the effects of dust resistivity.
A resistive dust layer was then incorporated into the ESP model. The
local voltage drop (across the dust layer within one cell) is given by the
product of the local current density, dust resistivity, and layer thickness.
Because the current density varies from cell to cell, the surface potential
across the cells is variable, and calculable tangential electric fields exist
at the surface of the dust layer. These fields are used as the boundary
conditions for the electric field calculations in the ESP model. The system
is iterated to achieve internal consistency. By itself, one might expect the
presence of surface charges to tend to equalize the current distribution over
the plate, the electric field realigning as the potential difference is
reduced directly under the wire (due to the high negative surface charge). We
further assume that when the dust breaks down, the voltage drop across the
dust in a given cell remains constant at the breakdown value no matter how
much current flows. This situation is analogous to the operation of gas break-
down devices (i.e. neon lamps).
Using these assumptions, we have obtained current distributions such as
those presented in Figure 15. These behave qualitatively like the experimen-
tal curves, with the higher resistivity material showing a broader area in
back corona at the point of initiation. The calculated V-I curves for these
two cases are shown in Figure 16. These were obtained by using very small
current density increments to pick out the step structure of the curves, and
are not composite curves as in Figure 13. The calculated E2Q curves are
shown together in Figure 17.
82
-------
6.0 CONCLUSIONS
A computer model has been developed which is capable of successfully
simulating some aspects of back corona in electrostatic precipitators. Five
simple mechanisms for the generation of back corona were examined. By com-
paring the calculated V-I curves with experimental curves, one particular
mechanism was singled out for further study. Incorporation of the effects of
dust layer resistivity and interstitial gas breakdown produced current distri-
bution and V-I curves in good qualitative agreement with experimental measure-
ments. The effect of back corona on particle collection was estimated for
each mechanism, and the general trend observed is that collection is continu-
ously degraded as back corona increases.
The computer model has proven to be a versatile calculation tool for
diagnosing conditions inside a precipitator and will continue to be of value
in further investigations.
7.0 ACKNOWLEDGMENT
This work was supported by EPA's Industrial Envrionmental Research Labora-
tory (Research Triangle Park, N.C.) under Grant No. R80-58-9701. J.H. Turner
is the grant administrator.
8.0 REFERENCES
1. Oglesby, S. Jr. and G.B. Nichols. Electrostatic Precipitation. New York,
Marcel Dekker, 1978.
2. White, H.J. Industrial Electrostatic Precipitation. Reading, Mass.,
Addison-Wesley, 1963.
3. Spencer, H.W. III. "Electrostatic Precipitators: Relationship Between
Resistivity, Particle Size, and Sparkover." EPA-600/2-76-144 (1976).
NTIS No. PB 257-130, May 1976.
4. White, H.J. J. Air Poll. Cont. Assn. 24, 314 (1974).
5. Penney, G.W. and R.E. Matick. Trans. Am. Inst. Elec. Engr. 1 79, 91,
(1960).
6. Masuda, S. Inst. Phys. Conf. Ser. No. 27, Ch.3 (1975).
7. Masuda, S. and A. Miquno. Proc. 4th Intl. Clean Air Conf., Paper No. V-
47, Tokyo, May 1977.
83
-------
14
12
10
I
U) 6
S
Q
I *
oc
o
35
o SECTION 1
a, SECTION 2
a SECTION 3
— MODEL
40
45
50 55
VOLTAGE (kV|
60
65
Figure 1. Agreement between a model calculation and an experimental VI curve.
SPACE CHARGE, q = -10'5
0.6
0.7 \ 0.5
- - THRESHOLD CURRENT DENSITY
I 1 1 1 1
26
Figure 2. Voltage-current density curves for mechanism I.
84
-------
SPACE CHARGE, q = -10 5 C/m3
THRESHOLD CURRENT DENSITY
26
Figure 3. Voltage-current density curves for mechanism II.
Hysteresis is indicated by dashed line.
55 ,
Z 3
SPACE CHARGE, q = 0
0.4
SPACE CHARGE, q = -1C)-5 C/m3
0.6
0.5
0.3
0 16
18
20
22
24
26
VOLTAGE IkV)
Figure 4. Voltage-current density curves for mechanism III.
Hysteresis is indicated by dashed line.
85
-------
E 4
UJ
IE 2
DC
O
~l 1 1 1 1~
SPACE CHARGE, q = 0
0.023
SPACE CHARGE, q =
0 16
0.04
THRESHOLD CURRENT DENSITY
-10-5C/m3
,0.023
18
20
22
VOLTAGE (kV)
24
26
Figure 5. Voltage-current density curve for mechanism IV.
Hysteresis is indicated by dashed line.
-i 1 1 r-
SPACE CHARGE, q = 0
5-
4 -
1 -
-i 1 1 1 1 r~
SPACE CHARGE, q = -10 5 C/m3
xO.10
0.05
,0.05
,0.10
- - - - THRESHOLD CURRENT DENSITY - -
-1 1 1 1 I L
16
18
20 22
VOLTAGE (kV|
24
26
Figure 6. Voltage-current density curve for mechanism V.
86
-------
MECHANISM I
MECHANISM II
00
1000
1 2 3
CURRENT DENSITY (A/m2) x 10*
1000
ARROWS INDICATE
INCREASING OR
DECREASING CURRENT
1234
CURRENT DENSITY (A/m2) x 104
Figure 7. E2Q as a function of current density
for mechanism I.
Figure 8. E2Q as a function of current density
for mechanism II.
-------
MECHANISM III
MECHANISM IV
CO
CO
1000
100 -
UJ
1234
CURRENT DENSITY (A/m^l x 10*
1000
100
Sf
UJ
10
e= 0 ,
q = -10-5C/m3
• q = 0 C/m3
ARROWS INDICATE
DIRECTION OF CURRENT
CHANGE
CURRENT DENSITY (A/m*) x
Figure 9 E2Q as a function of current density Figure 10. E2Q as a function of current density
for mechanism III. f°r mechanism IV.
-------
CO
MECHANISM V
1000
100
Cu
1
ui
K
10
K= 0
1234
CURRENT DENSITY (A/m^) x 10*
6.0
5.0
4.0
<
~ 3.0
oc
DC
2.0
1.0
0.0
10 20 30
VOLTAGE (kV)
40
Figure 11. E2Q as a function of current density Figure 12. Voltage-current curve for an ESP
for mechanism V. section in back-corona.
-------
6.
5 -
E 4
\ 3
UJ
Q
I-
3 7
tc. 2
1C
o
MECHANISM II COMPOSITE
q = 0 C/m3
16
ARROWS INDICATE CURRENT
INCREASING OR DECREASING
q = -10-5 C/m3
18
20
VOLTAGE (kV)
22
24
26
Figure 13. Voltage-current density curves with hysteresis for a
composite mechanism II.
3-1010fi -cm
5.6-1012fi -cm
10
LU
Q
\- 10-2
QC
DC
O
10-3
« 38 kV
A 30 kV /
— 28 kV, clean plate /
o 22 kV
— 22 kV, clean plate
10
10-3
10-4
DISTANCE ALONG PLATE (cm)
Figure 14. Current density distributions under a dust layer in a
wire-plane geometry for two dust resistivities.
90
-------
10-3
55
UJ
Q
O
_cm
16.3 kV
16.5kV
Sy
100lzn -cm
-— 15.6 kV
15.7 kV
10
-10-5
10-6
Sy
DISTANCE ALONG PLATE
(Sy = 11.4cm)
Figure 15. Calculated current density distributions with resistive
dust layer incorporated in model (two resistivities).
(Jth = 1-10 4 A/m2, left; Jth = 1.10~5 A/m2, right)
o
X
LU
cc
cc
o
VOLTAGE (kV)
Figure 16. Calculated V-I curves for two resistive dusts.
91
-------
100
a
CM
LU
0
1 2 3
CURRENT DENSITY (A/m2) x 104
Figure 17. Calculated E2Q as a function of current density for
model incorporating resistive dust layer (two resistivities)
92
-------
CHARGE MEASUREMENTS ON INDIVIDUAL
PARTICLES EXITING LABORATORY PRECIPITATORS
Jack R. McDonald
Marlln H. Anderson
Ronald B. Mosley
Southern Research Institute
2000 Ninth Avenue South
Birmingham, Alabama 35205
and
Leslie E. Sparks
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
ABSTRACT
The values of charge on individual particles exiting three different lab-
oratory precipitators have been -measured in an experimental apparatus which
utilizes a Millikan cell. Measurements were obtained for dioctylphthalate (DOP)
droplets and fly ash particles at temperatures from 23 to 343°C. At comparable
voltages and currents for positive and negative corona, the data show that the
ratio of the values of negative to positive charge for diameters in the range
1.2-2.6 ym increases from a value of approximately 1 to a value of approximately
2 as the temperature increases from 37 to 343°C. The predictions of a mathe-
matical model of electrostatic precipitation which employs an ionic charging
theory show good agreement with all the positive charging data but only show
good agreement with the negative charging data at the lower temperatures. The
data and model predictions are consistent with the postulation of free electron
charging at elevated temperatures with negative corona.
INTRODUCTION
In the electrostatic precipitation process, particles suspended in a moving
gas stream are charged as the gas is passed through a corona discharge. The
particles are charged due to collisions with molecular ions and, possibly, free
(unattached) electrons in the case of negative corona discharge. Since the
force which drives a charged particle to a collection electrode is proportional
to the charge on the particle, the mechanisms involved in the particle charging
process and the attainable values of particle charge under various operating
conditions are of fundamental importance with respect to understanding and
utilizing the electrostatic precipitation process.
93
-------
In the positive corona discharge, the electrons created in the avalanche
process near the discharge electrode migrate to the discharge electrode leaving
behind positive molecular ions which migrate into the interelectrode space
toward the collection electrode. In this case, the interelectrode space, ex-
cluding the relatively small region near the discharge electrode where active
gas breakdown occurs, consists of neutral gas molecules and positive molecular
ions. Therefore, for the positive corona case, only positive ions may partici-
pate in the particle charging process.
In the negative corona discharge, the electrons created in the avalanche
process near the discharge electrode migrate into the interelectrode space
toward the collection electrode and attach in a probabilistic manner to neutral
electronegative gas molecules to form negative ions. The positive ions formed
in the avalanche process migrate to the discharge electrode. In this case,
the interelectrode space, excluding the relatively small region near the dis-
charge electrode where active gas breakdown occurs, consists of neutral gas
molecules, negative molecular ions, and, possibly, free electrons. Therefore,
there is the potential for both negative ions and free electrons to participate
in the particle charging process.
Evidence and conjecture of the presence of free electron charging has been
documented in the literature for laboratory charging experiments with small-
scale charging devices at room temperature.1'2 The mechanism by which free
electrons charge particles could be quite different than that for ionic charg-
ing. Intuitively, the extent of free electron charging should depend on the
electrode geometry (especially the spacing between the discharge and collection
electrodes), applied voltage, gas temperature, and gas pressure. At the present,
the mechanisms by which electrons would charge particles have not been described
theoretically, and no experimental data are available that truly isolate elec-
tron charging from ionic charging. In applying fundamental principles to full-
scale precipitators, effects due to free electrons are generally neglected.
The purposes of the work presented here were to develop a technique for
measuring the charge and diameter of individual particles in an electrostatic
precipitator and to make charge and diameter measurements in a typical full-
scale, electrode geometry at various temperatures up to that which is typical
in a hot-side precipitator. The values of charge and diameter of individual
particles exiting three different laboratory precipitators have been measured
with an experimental apparatus which utilizes a Millikan cell. Measurements
were obtained for dioctylphthalate (DOP) droplets and redispered fly ash par-
ticles at temperatures from 23 to 343°C at various typical voltages and current
densities.
It was anticipated that by making the above measurements some questions of
importance could be answered. First, how well do existing ionic charging theo-
ries predict particle charge when applied to particles carried in a turbulent
gas stream through a typical precipitator electrode geometry? The utilization
of these theories is complicated due to the existence of a nonuniform electric
field and current density and possible residence time phenomena due to the tur-
bulent nature of the gas stream. Second, how do positive and negative corona
compare in potential effectiveness for charging particles? Higher voltages
and currents before sparkover may result with one type of corona discharge and
94
-------
may lead to significantly higher values of particle charge. The possibility
of different charging mechanisms in the cases of positive and negative corona
may result in significantly different values of particle charge under similar
operating conditions. Third, do free electrons in the case of negative corona
significantly enhance particle charging at higher temperatures (and/or reduced
pressures)? Recent field data from a hot-side precipitator collecting fly ash
indicate enhanced collection efficiencies for particles with diameters between
1.0 and 3.0 ym.3 These collection efficiencies were predicted by using the
measured operating voltages and current densities in conjunction with an ionic
charging theory and were significantly lower than those measured. In this case,
free electron charging may have been prominent in the electrostatic precipita-
tion process.
MEASUREMENT TECHNIQUE AND APPARATUS
The technique utilized for simultaneously determining the charge and
diameter of a particle is based on the downward and upward motion of a charged
particle in an insulating gas under the influence of a uniform, reversible
electric field. When the gravitational force and the force due to the elec-
tric field act on the charged particle in the same direction, then the charged
particle will experience a net downward force given by
FD = maD = - qE - mg + Fd
= - qE - (p - p )4rra3) g + (6imavn)/(l + AA/a), (1)
p 3. *J L)
where
F = net downward force (nt),
m = mass of the particle (kg),
an = downward acceleration (m/sec2),
q = charge on the particle (C),
E = electric field (V/m),
g = acceleration due to gravity (m/sec2),
F, = drag force acting on the particle (nt),
p = density of the particle (kg/m3),
p = density of the gaseous medium (kg/m3),
3.
a = particle radius (m),
H = viscosity of the gaseous medium (nt-sec/m2),
v = downward velocity of the particle (m/sec),
(1+Afc/a) = Cunningham correction factor,
& = molecular mean free path (m), and
A = 1.257 + 0.400 exp (-1.10 a/A).
Similarly, the force acting on the charged particle due to the electric field
can be made opposite to that of the gravitational force in such a manner that
the charged particle will experience a net upward force given by
95
-------
= mau = qE - mg - Fd
- (p - p )Qra3) g - (6iniav )/(! + AJl/a),
P cL j U
where
F = net upward force (nt) ,
2
a = upward acceleration (m/sec2) ,
v = upward velocity of the particle (m/sec) ,
and all other symbols are as previously defined.
Assuming that the terminal velocity of the particle is reached instanta-
neously, then a = a = 0 and
and
where
S/tD (3)
- S/tu , (4)
S = distance over which the particle moves (m) ,
t_ = time to fall the distance S (sec) , and
t - time to rise the distance S (sec) .
Adding equations (1) and (2) yields
- (fira3)(p - p ) g + [(6nnaS)/(l + M/a)] (•£- - f-) = 0. (5)
J P a CD u
Subtracting equations (.1) and (.2) yields
- 2 qE + [(6imaS)/(l + A£/a)] (~ + ^-> = 0. (6)
D u
Equations (5) and (6) can be solved simultaneously to obtain expressions for a
and q.
Equation (5) can be rewritten in the form
4(pp - pa) g
96
-------
Solving for a yields
a =
V
+ If
AH2 -
-AS, + 1 ^p ~ Pa)g ^ fcD
2
Solving equation (6) for q yields
(8)
q = [(3imaS)/(l + AA/a)] (- + -) , (9)
D u
where
E = V/D,
V = voltage applied across two parallel plates (V), and
D = spacing between the parallel plates (m).
A technique and apparatus have been developed for measuring the diameter
and charge on individual particles which have been treated in an electrostatic
precipitator. The technique consists of extracting a sample gas volume from
the precipitator and directing part of this extracted sample into a modified
Millikan measurement cell. A single particle can then be isolated, and its
upward and downward motion under the influence of a uniform electric field
can be utilized to determine its diameter and charge from equations (8) and
(9).
A drawing of the measurement apparatus with its insertion into a precipi-
tator is shown in Figure 1. The gas sample is extracted through 2-inch diameter
tubing. The tubing can be made of teflon or metal as dictated by the temperature
in the gas stream with equivalent results. This tubing contains two bends and
is electrically grounded through a spiral wire which runs along the inner sur-
face of the tubing. The tubing runs upward from the measurement cell and makes
a right angle bend so that the tubing can enter a sampling port. The second
bend in the tubing is made so that the tubing will be oriented opposite to
the direction of the gas flow and so that the entrance to the tubing can be
located inside the electrified regions although the port is located downstream
from these regions. The tubing leading down to the measurement cell contains
a butterfly valve which can be opened to allow a sample of gas to be drawn
into the measurement cell and can be closed to prevent the sample from being
drawn back into the precipitator which has negative pressure. The closed valve
also prevents air flow disturbances during the measurements. Below this valve
and below the measurement cell there is a connection to a pump. With the
valve open, the pump is operated and particles are drawn into the measurement
cell. When a sufficient number of particles are obtained in the measurement
cell, the pump is turned off with simultaneous closing of the valve. Once
the pump is turned off and the valve is closed, the particles very quickly
cease to have any motion due to gas flow. At this point, only the gravitational
field, viscous drag, and electric field which can be imposed across the parallel
plates of the measurement cell have any influence on the motion of the particles.
97
-------
DISCHARGE ELECTRODES
«Jrf^—
P- \ ^*- FILTER
CONTROL PAD
VOLTS
POWER SUPPLY
Figure 1. Experimental apparatus for measuring the diameter and
charge of particles.
98
-------
In those cases where the particle concentrations are too large for easy
isolation of individual particles, the extracted gas sample can be diluted.
Filtered, outside air can be mixed with the extracted sample by means of a
bleed valve preceded by a filter.
The Millikan measurement cell is cylindrical in shape with a diameter of
3.8 cm and a plate spacing (or approximate height) of 0.5 cm. Gas enters the
cell through a small hole in a conical depression in the top plate. The par-
ticles are illuminated by a high intensity microscope lamp and are viewed
through a telescope attached to the measurement cell. The distance traveled
by the particles is determined by a graticule which is mounted near the focal
plane of the objective lens of the telescope. Measurement of the particle
transit times was done with a stopwatch. Voltages on the order of 3 to 15
volts were applied across the plates to produce the electric fields necessary
to generate the data presented in this paper. The temperature of the gas in
the measurement cell was determined by means of a thermocouple.
In the system described here, particles with diameters down to approxi-
mately 0.4 ym can be observed. Measurements were concerned primarily with
fine particles with diameters approximately between 0.6 ym and 3.0 ym since
(1) they are the most difficult to collect in a precipitator, (2) their behavior
in a precipitator is the most difficult to model, and (3) their escape into the
atmosphere offers the greatest health hazard. Measurements on smaller particles
can be readily obtained by allowing the larger particles to settle out. By
waiting approximately 3 minutes after a new sample is introduced into the
measurement cell, only particles of approximately 1.0 ym or less in diameter
will remain in the field of view. Measurements on the larger particles can be
readily made by choosing those particles which fall the fastest under the in-
fluence of gravity. The measurement system also has the capability of deter-
mining the magnitude and sign of the charge on a particle. Thus, the system
has the potential to be used to analyze the effects on particle charge of back
corona and rapping reentrainment. In addition, particles can be successfully
extracted and analyzed from electrified regions at different locations in a
precipitator.
EXPERIMENTAL PROGRAM, DATA, AND RESULTS
The first objective of the experimental program was to determine the
capability of the technique and apparatus described earlier to make accurate
measurements of the diameter and charge of particles treated in an electro-
static precipitator. In order to achieve this objective, measurements were
performed under essentially idealized conditions in laboratory precipitator A.
This precipitator has one gas passage, four electrical sections which are 0.76 m
(2.5 ft) long, plates which are 38.1 cm (15 in.) high, a 25.4 cm (10 in.) plate-
to-plate spacing, a 12.7 cm (5 in.) wire-to-wire spacing, and a 0.24 cm (0.094
in.) discharge electrode (wire) diameter. The normalized standard deviation of
the gas velocity distribution and the gas sneakage per baffled electrical sec-
tion were both measured to be less than 10%. For the experiments in laboratory
precipitator A, low mass concentrations of dioctylphthalate (DOP) droplets were
generated by an aerosol sprayer and were carried through the precipitator by air
at ambient conditions. Since (1) the particles were spherical, (2) particle
reentrainment could not exist, (3) back corona could not exist, and (4) gas
99
-------
sneakage was minimal, the results of the measurements could be interpreted with
little ambiguity.
Measurements were made at precipitator operating conditions consisting of
negative corona, an average applied voltage of 44.2 kV, an average current
density of 21.5 nA/cm2 (20.0 yA/ft2), and a gas velocity of 1.5 m/sec (4.9 ft/
sec). The data obtained at the precipitator outlet from the measurements on
486 individual particles are shown graphically in Figure 2 and are tabulated
in Table 1. The data on each individual particle were obtained from the average
values of three measurements each of the time required for the particle to
travel downward a distance of 0.129 cm under the influence of an electric field
and of the time required to travel the same distance upward. The bars on the
average data points are one standard deviation and are not representative of
the error in the measurement technique but, instead, they are more representa-
tive of the expected spread in charge on a particle with a given diameter due
to the fact that different particles travel different paths through the precipi-
tator and experience different charging electric field strengths and ion densi-
ties. Also, particle diameters in a narrow band have been grouped together
with a diameter given by the midpoint of the band. The solid curve is a model1*
prediction of charge as a function of particle diameter at the outlet of the
precipitator. The model employs an ionic charging theory.5
Table 1. MEASURED AND PREDICTED VALUES OF PARTICLE
CHARGE AT THE OUTLET OF A LABORATORY PRECIPITATOR
Radius
Size Range
(10-6m)
0.25-0.35
0.35-0.45
0.45-0.55
0.55-0.65
0.65-0.75
0.75-0.85
0.85-0.95
0.95-1.05
1.05-1.15
1.15-1.25
1.25-1.35
1.35-1.45
1.45-1.55
No.
Part
3
35
57
46
68
80
66
56
42
20
5
6
2
Mean
Charge
(C)
1.28xlO~17
1.98xlO~17
2. 88x10" 17
4.54xlO~17
6.71xlO~17
9.10xlO~17
1.21xlO~16
1.43xlO~16
1.64xlO~16
1.82xlO~16
3.32xlO~16
2.61xlO~16
4.30xlO~16
Min
Charge
(C)
4.20xlO-18
5. 22xlO~ 18
1. 96x10" 17
2. 76x10" 17
3.74xlO~17
4. 16x10" 17
6.41xlO~17
6. 08xlO~ 17
8.09xlO~17
1.16xlo"16
2.39xlO~16
1.48xlO~16
3.95xlO~16
Max
Charge
(C)
1. 82x10" 17
3. 82xlO~ 17
6.03xlO~17
1.17xlO"16
1.04xlO~16
1.58xlO~16
1.87xlO~16
2.68xlO~16
2.94xlO~16
3.17xlO~16
4.39xlO~16
4.93xlO~16
4.59xlO~16
Normalized
Standard
Deviation
0.481
0.270
0.265
0.346
0.234
0.271
0.280
0.330
0.306
0.289
0.237
0.438
0.748
Predicted
Charge
(C)
1.61x10"17
2.62x10"17
3.81x10"17
5.16x10"17
6.75x10"17
8.49x10"17
1.05xlO~16
-16
-16
1.26x10
1.48x10
1.75xlO~16
2.02xlO~16
2.31x10"16
2.61xlO~16
100
-------
u
Uj
u
io-«
ui
V
ec
i
10'
17
10-7
Figure 2.
T | I I I I |
— THEORY
O EXPERIMENTAL
I I
10*
PARTICLE RADIUS, m
Theoretical and measured particle charge as a function of
particle radius at the outlet of laboratory precipitator A
at ambient temperature.
101
-------
In a wire-plate geometry, the electric field does not vary significantly
over a major portion of the interelectrode space and it varies rapidly only
over a small high field region near the discharge electrode.6 Thus, one would
expect that the majority of the particles of a given diameter exiting the pre-
cipitator would have essentially the same charge but there would be a spread
in charge lying in a range determined by the minimum and maximum values of the
electric field. These expectations are evidenced in the data in that the nor-
malized standard deviations are small. One might also expect that a certain
percentage (possibly 5-10%) of the smaller particles (less than 1 ym in diameter)
would pass through high electric field regions near the discharge electrodes
and would acquire charges which are significantly higher than those predicted
by the average electric field used in the model. Although values of the elec-
tric field in the high field region should be up to 8.5 times the average value
of electric field used in the model, the ratios of measured charge to predicted
charge did not exceed a value of 2.8. It was also observed that, while it was
relatively easy to find particle charges significantly larger than that pre-
dicted by the model for diameters greater than approximately 1.5 ym, very few
particles were found to have a significantly higher charge than predicted by
the model for diameters less than approximately 1.5 ym. In fact, in the di-
ameter range between 0.5 and 1.7 ym, only 1 particle out of 289 measured had
a charge which was over a factor of 1.6 times the predicted charge. In addition
many samples were observed in which the particles larger than 1.0 ym in diameter
were allowed to settle out in order to determine if highly charged submicron
particles were present. In all these samples, the remaining particles all
moved with essentially the same velocity in an applied electric field. Thus,
they all had approximately the same value of charge. Measurements made on gas
samples taken from the middle of the inlet electrical section also produced the
same results.
The limited data shown in Figure 2 and Table 1 indicate that the measure-
ment technique is quite reliable and that the model predictions are in quite
good agreement with the average measured value of particle charge. The higher
theoretical predictions for particle diameters less than 0.6 ym are inherent
in the approximate theory at the values of electric field which were utilized
in the experiment. The use of an average electric field determined by dividing
the applied voltage by the wire-to-plate spacing and of a particle residence
time determined by dividing the precipitator length by the average gas velocity
appears to be adequate for predicting particle charge. Also, an ionic charging
theory appears to be adequate for describing the data.
The second objective of the experimental program was to make charge and
diameter measurements in a typical full-scale, electrode geometry under both
positive and negative corona at various temperatures up to that which is typical
in a hot-side precipitator. In these experiments, the gas stream entering the
precipitator was laboratory air containing low resistivity, redispersed fly ash
particles. In principle, although sparkover may occur at a different applied
voltage for positive corona than for negative corona in laboratory air at atmos-
pheric pressure near sea level, the clean air, clean plate, voltage-current
curves for the two cases should be nearly the same up to sparkover for typical
full-scale plate spacings since the starting voltages and effective ion mobili-
ties do not differ appreciably. Thus, for low mass loadings of low resistivity
fly ash particles, comparisons of particle charging capabilities of positive
102
-------
and negative corona might be made where essentially the same applied voltages
and currents are utilized in both cases. Any significant differences in par-
ticle charging capabilities under these conditions would indicate different
charging mechanisms for the different types of corona.
At higher temperatures (and/or reduced pressures), the mean-free-paths of
ions and electrons increase. Thus, in the case of negative corona, free elec-
trons can penetrate further into the interelectrode space and possibly can have
an increased effect on voltage-current characteristics and particle charging.
Comparison of voltage-current characteristics and particle charging for both
positive and negative corona at higher temperatures should provide further
insight into the effect, if any, of free electrons. If significant penetra-
tion of free electrons occurs in the interelectrode space, this might be evi-
denced in a larger difference in positive and negative voltage-current charac-
teristics than would be obtained if only ions carried the current. However,
even if differences in voltage-current characteristics can not be firmly
established, differences in particle charging capabilities may still exist.
The effect of temperature on particle charging was examined in laboratory
precipitators B and C. Precipitator B has one gas passage, four electrical
sections which are 91.4 cm (36 in.) long, plates which are 91.4 cm (36 in.) high,
a 25.4 cm (10 in.) plate-to-plate spacing, a 22.9 cm (9 in.) wire-to-wire spacing,
a 0.32 cm (0.125 in.) discharge electrode (wire) diameter. Precipitator C has
one gas passage, four electrical sections which are 1.22 m (4 ft) long, plates
which are 1.22 m (4 ft) high, a 25.4 cm (10 in.) plate-to-plate spacing, a 22.9
cm (9 in.) wire-to-wire spacing and a 0.32 cm (0.125 in.) discharge electrode
(wire) diameter. Measurements at 38°C (100°F) were performed at the outlet
of precipitator B, while measurements at 148°C (300°F), 232°C (450°F), and
343 C (650°F) were performed at the outlet of precipitator C.
Figures 3-5 show typical real time voltage-current traces obtained from
precipitator C for clean air, wires, and plates at three different temperatures.
All these curves were obtained sequentially over a relatively short time period.
Under these conditions and with good electrode alignment, it can be seen that
voltage-current curves which are nearly coincident are obtainable. Figures 6-8
show some typical average voltage-current curves obtained when the air stream
contained fly ash particles and when the wires and plates were somewhat dirty.
The data in Figure 6 represent an average over precipitator B. The data in
Figures 7 and 8 were obtained from precipitator C with curves from all the elec-
trical sections shown in Figure 7 and an average over the entire precipitator
shown in Figure 8. All data with particles in the air stream obtained at 343°C
(650°F) from precipitator C were acquired approximately 4 months prior to that
at 148°C (300°F) and 232°C (450°F). During the later measurement period, the
precipitator was not operating as well as before, and the voltage-current curves
with particles for positive and negative corona were widely separated. Since
the inlet mass loading was nominally on the order of 1.14 g/m3 (0.5 gr/acf)
and the particle size distribution was rather large, the effect of particles
on the voltage-current curves could not be attributed entirely to a parti-
culate space charge effect. The observed behavior of the voltage-current
curves might have been due to deposits of a high resistivity ash on the wires
and plates acquired during other experiments performed just prior to these
measurements. In any event, it was often observed that the addition of particles
103
-------
Ul
K
DC
4.0
3.0
2.0
NOTE: CLEAN WIRES/CLEAN PLATES
NEGATIVE CORONA
PLATE SPACING - 26.4 cm
WIRE SPACING = 22.9 cm
SWIRES/SECTION
0.32 cm DIAMETER WIRES
POSITIVE CORONA
PLATE SPACING -25.45
WIRE SPACING = 22.9 cm
SWIRES/SECTION
0.32 cm DIAMETER WIRES
1.0
10
20
30 40
VOLTAGE, kV
50
60
Figure 3. Clean air, wire, and plate voltage-current
curves for precipitator C at ambient temperature.
NOTE: CLEAN WIRES/CLEAN PLATES
NEGATIVE CORONA—^
PLATE SPACING - 254 cm
WIRE SPACING - 22.9 on
5 WIRES/SECTION
0.32 cm DIAMETER WIRES
POSITIVE CORONA ^
PLATE SPACING -25.4 cm
WIRE SPACING - 22.9 cm
6 WIRES/SECTION
0.32 cm DIAMETER WIRES
30
VOLTAGE. kV
Figure 4. Clean air, wire, and plate voltage-current
curves for precipitator C at 148°C (300°F),
104
-------
3.0
NOTE: CLEAN WIRES/CLEAN PLATES
POSITIVE CORONA -
PLATE SPACING - 25.4 cm
WIDE SPACING - 22.9 en
6 WIRES/SECTION
0.32 cm DIAMETER WIRES
NEGATIVE CORONA -r»
PLATE SPACING -
WIRE SPACING - 22.9 en
SWIRES/SECTION
0.32 cm DIAMETER WIRES
20 30
VOLTAGE. W
Figure 5. Clean air, wire, and plate voltage-current
curves for. precipitator C at 343°C (650°F).
TEMPERATURE 38°C
FEEDER ON
• NEGATIVE CORONA
• POSITIVE CORONA
22 26 30 34
VOLTAGE. kV
Figure 6. Average voltage-current curves with particles
for precipitator B at 38°C (100°F).
105
-------
I ! I I I I Tl 1 ! U rl
TEMPERATURE
148°C
FEEDER ON
10 14 18 22 26 30 34 38 42 46
Figure 7. Voltage-current curves for the different
electrical sections with particles for
precipitator C at 148°C (300°F).
5.5
5.0
4.5
4.0
1 "
15 »
111
£C
§ 2.5
U
2.0
1.5
1.0
0.5
II I II I S 1 M
I I :
TEMPERATURE 343°C -
FEEDER ON
I I ! I I I I I I
10 14 18 22 26 30 34 38 42 46
VOLTAGE, kV
Figure 8. Average voltage-current curves with particles
for precipitator C at 343°C (650°F).
106
-------
caused the positive and negative voltage-current curves to separate to a larger
extent than anticipated.
Figures 9-11 show particle diameter and charge measurements made at the
outlet of precipitator B at 38°C (100°F) with an average gas velocity of approx-
imately 1.5 TO/sec (5.0 ft/sec). Model predictions for the different conditions
are also shown for comparison with the data. The data in Figure 9 were obtained
with a constant average current density of 15 nA/cm2 (13.9 yA/ft2) and average
applied voltages for positive and negative corona of 38.0 kV and 34.5 kV, re-
spectively. The data in Figure 10 were obtained with a constant average applied
voltage of 35.0 kV and average current densities for positive and negative coro-
na of 9 nA/cm2 (8.4 yA/ft2) and 15 nA/cm2 (13.9 yA/ft2), respectively. The data
in Figure 11 were obtained for negative corona only at average applied voltages
of 34.5 kV and 38.0 kV, corresponding to average current densities of 15 nA/cm2
(13.9 yA/ft2) and 36 nA/cm2 (33.4 yA/ft2), respectively. For all the data at
38 C (100°F), the theory and data show good agreement. Since the least amount
of data was acquired for the smallest and largest diameter bands, the agreement
with theory should be expected to be less for these particle diameters.
Figure 12 shows particle diameter and charge measurements made at the out-
let of precipitator C at 148°C (300°F) with an average gas velocity of approxi-
mately 1.4 m/sec (4.5 ft/sec). The average applied voltage and current density
for negative corona were 25.7 kV and 31.0 nA/cm2 (28.8 yA/ft2), respectively.
The average applied voltage and current density for positive corona were 33.0
kV and 18.0 nA/cm2 (16.7 yA/ft2), respectively. During measurements with both
positive and negative corona, it was difficult to maintain constant electrical
operating conditions. Wide fluctuations in applied voltage and current occurred
with intermittent sparking. Thus, these data are not useful for comparing
positive and negative particle charging under comparable conditions. However,
the data give an illustration of the capability of the measurement system to
detect adverse charging conditions.
Figure 13 shows particle diameter and charge measurements made at the out-
let of precipitator C at 232°C (450°F) with an average gas velocity of approxi-
mately 1.4 m/sec (4.5 ft/sec). The average applied voltage and current density
for negative corona were 28.1 kV and 28.0 nA/cm2 (26.0 yA/ft2), respectively.
The average applied voltage and current density for positive corona were 31.5
kV and 17 nA/cm2 (15.8 yA/ft2), respectively. After the measurements at 148°C
(300°F), the wires and plates of the precipitator were cleaned by brushing.
This resulted in more favorable and stable electrical operating conditions.
For the operating conditions during the measurements, the theory predicts
essentially the same charge versus diameter relationship for the positive and
negative corona conditions. While the theory agrees well with the data obtained
with positive corona for all particle diameters, it clearly underpredicts par-
ticle charge for diameters between approximately 1,4 ym and 2.4 ym. In this
diameter range, the ratio of the average measured charge to the predicted charge
ranges from 1.25 to 1.59, increasing with increasing diameter. These data in-
dicate that negative corona is more effective than positive corona in charging
particles with diameters between 1.4 ym and 2.4 ym at 232°C (450°F). Comparison
of these data with the theoretical predictions and the information in Figures
2-5 suggests that the use of a completely ionic charging theory to describe
negative particle charging at 232°C (450°F) is inadequate. Since the only major
107
-------
O 10"
uT
O
oc
16
10-
17
10-7
O NEG. COR. V " 34-35 kV
THEORY LINE 1
D POS. COR. V - 37-39 kV
THEORY LINE 2
10-6
PARTICLE RADIUS, m
Figure 9. Measured positive and negative particle charge versus
measured radius and comparison with theory at 38°C (100°F)
and 15 nA/cm2 (13.9
VOLTAOE: »KV
1 O NEGATIVE COMMA 16 nA/OB>
I O POSITIVE CORONA
ERROR BAflt FOR
NEOATIVE CASE
Mr*
Figure 10. Measured positive and negative particle charge versus
measured radius and comparison with theory at 38°C (100°F)
and 35 kV.
108
-------
0 NEGATIVE CORONA
15 nA/cm2, 34-35 kV
2 Q 36 nA/cm2, 38 kV
10-17
10-6
PARTICLE RADIUS, m
Figure 11. Measured negative particle charge versus measured radius and
comparison with theory at 38°C (100°F) with 15 nA/cm2 (13.9 yA/ft2)
and 36 nA/cm2 (33.4 pA/ft2).
0 NT16
flC
10-
17
r i T
1 o NEGATIVE CORONA,
25.7 kV, 31 nA/cm2
2 a POSITIVE CORONA,
33 kV, 18 nA/cm2
10-7
10-6
PARTICLE RADIUS, m
Figure 12. Measured positive and negative particle charge versus measured
radius and comparison with theory at 148°C (300°F).
109
-------
u
of
O
cc
u
10-
17
. ' ' ' ' " ' IT!
NEGATIVE CORONAO) THEORY T
~ 28.1 kV, 28 nA/cm2
i POSITIVE CORONA (2) THEORY
31.5 kV, 17 nA/cm2
ID''
10"6
PARTICLE RADIUS, m
Figure 13. Measured positive and negative particle charge versus measured
radius and comparison with theory at 232°C (450°F).
CURRENT DCNMTY: X
Q NEGATIVE COHONA V 27-M hV
O P081TIVE CORONA V 26-27 hV
—THEORY
Figure 14. Measured positive and negative particle charge versus measured
radius and comparison with theory at 343°C (650°F) and 30 nA/cm2
(27.9 yA/ft2).
110
-------
physical difference between the positive and negative corona process is the
free electron penetration into the interelectrode space in the negative corona
case, the enhanced particle charge with negative corona at 232°C (450°F) might
be due to increased free electron charging.
Figures 14-16 show particle diameter and charge measurements made at the
outlet of precipitator C at 343°C (650°F) with an average gas velocity of
approximately 1.4 m/sec (4.5 ft/sec). The precipitator had been cleaned thor-
oughly prior to these measurements. During the measurements, the electrical
operating conditions were extremely stable. The data shown in Figure 14 were
obtained for a current density of 30 nA/cm2 (27.9 yA/ft2) for both positive
and negative corona with essentially the same average applied voltages of 26.5
kV and 27.5 kV, respectively. Similar to the results obtained at 232°C (450°F),
the theory agrees well with the data for positive corona but clearly under-
predicts particle charge obtained with negative corona for diameters between
approximately 1.2 ym and 2,6 ym. In this diameter range, the ratio of the
average measured charge to the predicted charge ranges from 1.27 to 1.94,
generally increasing with increasing diameter. These data again suggest the
possibility of free electrons participating in the charging process with
negative corona.
Figures 15 and 16 contain data showing the effect of electrical conditions
on particle charging with positive and negative corona at 343°C (650°F). Again,
the theory agrees well with the data for positive corona but underpredicts
particle charge obtained with negative corona for diameters between approxi-
mately 1.2 iJtm and 2.6 ym. The data show the same trend as the theory in that
larger applied voltages and current densities result in higher values of par-
ticle charge. Also, the ability for negative corona to acquire a significantly
higher voltage and current density prior to sparkover resulted in a further
increase in particle charge as compared to the positive corona.
SUMMARY AND CONCLUSIONS
The apparatus and technique described for measuring particle diameter and
charge are capable of providing reliable data on individual particles charged
in an electrostatic precipitator. For temperatures up to 343°C (650°F),
the measured values of charge acquired in positive corona by particles with di-
ameters in the range between 0.6 ym and 3.0 ym are in good agreement with those
predicted by an ionic charging theory. For temperatures less than 37°C (100°F),
the measured values of charge acquired in negative corona by particles with
diameters in the range between 0.6 ym and 3.0 ym are in good agreement with
those predicted by an ionic charging theory. For temperatures of 232 C (450°F)
and 343°C (650°F), the measured values of charge acquired in negative corona
by particles with diameters between 1.2 ym and 2.6 ym are significantly higher
than those measured for positive corona under similar conditions and those pre-
dicted from a completely ionic charging theory. These data indicate that the
charging mechanisms for positive and negative corona differ as temperature in-
creases. The enhanced values of particle charge for negative corona at elevated
temperatures are consistent with the postulation of a free electron contribution
in the particle charging mechanism. Thus, a general theory for particle charging
in negative corona must contain ionic and electronic charging mechanisms whose
relative contributions are temperature dependent. The data indicate that, even
111
-------
POSITIVE CORONA
0 CURRENT DENSITY
30 nA/cm2 - THEORY LINE 1
O CURRENT DENSITY
15 nA/cm2 - THEORY LINE 2
10-"
10-*
PARTICLE RADIUS, m
Figure 15. Measured positive particle charge versus measured radius and
comparison with theory at 343°C (650°F) with 15 nA/cm2 (13.9
pA/ft2) and 30 nA/cm2 (27.9 yA/ft2).
o
g
O
10
,-17
NEGATIVE CORONA
1 O 30 nA/cm2, 27-28 kV
2 O 85 nA/cm2, 30-31 kV
10'7
10-6
PARTICLE RADIUS, m
Figure 16. Measured negative particle charge versus measured radius and
comparison with theory at 343°C (650°F) with 30 nA/cm2 (27.9
yA/ft2) and 85 nA/cm2 (79.0 yA/ft2).
112
-------
in a full-scale precipitator utilizing negative corona, the effect of free
electrons on the charging process can not be ignored for typical temperatures
between 148°C (300°F) and 371°C (700°F).
ACKNOWLEDGMENTS
This work was supported by the Industrial Environmental Research Labora-
tory, U.S. Environmental Protection Agency, Research Triangle Park, North
Carolina, under Contract No. 68-02-2610, Dr. Leslie E. Sparks, Project Officer.
REFERENCES
1. Penney, G. W., and R. D. Lynch. Measurements of Charge Imparted to Fine
Particles by a Corona Discharge. AIEE, 76:294-299, July, 1957.
2. Pontius, D. H., L. G. Felix, J. R. McDonald, and W. B. Smith. Fine Particle
Charging Development. EPA-600/2-77-173, NTIS No. PB 271-727/AS, U.S. En-
vironmental Protection Agency, Research Triangle Park, North Carolina 1977.
3. Marchant, G. H., Jr. and J. P. Gooch. Performance and Economic Evaluation
of a Hot-Side Electrostatic Precipitator. EPA-600/7-78-214, NTIS No. PB
292-648, U.S. Environmental Protection Agency, Research Triangle Park,
North Carolina 1978.
4. McDonald, J. R. A Mathematical Model of Electrostatic Precipitation (Re-
vision 1): Volume 1. Modeling and Programming. EPA-600/7-78-llla, NTIS
No. PB 284-614, U.S. Environmental Protection Agency, Research Triangle
Park, North Carolina 1978.
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. McDonald, J. R., W. B. Smith, H. W. Spencer, and L. E. Sparks. A Mathe-
matical Model for Calculating Electrical Conditions in Wire-Duct Electro-
static Precipitation Devices. J. Appl. Phys., 48(6) -.2231-2246, 1977.
113
-------
OPTIMIZATION OF COLLECTION EFFICIENCY BY VARYING PLATE SPACING
WITHIN AN ELECTROSTATIC PRECIPITATOR
E. J. Eschbach, Research Assistant
D. E. Stock, Associate Professor
Department of Mechanical Engineering
Washington State University
Pullman, Washington 99164
ABSTRACT
Use of the Deutsch equation as an evaluation tool of electrostatic precipi-
tators has resulted in the restriction of precipitator design to a single plate
spacing and has not enabled optimal use of the space charge arising from the
presence of charged dust particles. Space charge is dependent on dust loading
and plate spacing and also influences the electric field, hence the migration
velocity and, therefore, collection efficiency. Thus a trade-off is indicated
between plate spacing and collection efficiency as a function of dust loading.
A first step examination of this trade-off indicates that varied spacing can be
used in precipitators while maintaining high levels of efficiency and decreasing
the required SCA by 15 to 25%.
114
-------
OPTIMIZATION OF COLLECTION EFFICIENCY BY VARYING PLATE SPACING
WITHIN AN ELECTROSTATIC PRECIPITATOR
INTRODUCTION
Electrostatic precipitators (ESP) can only remain competitive with other
particulate control devices if their initial cost can be reduced while holding
their performance, operating cost and maintenance cost at their present level[l].
Presently, most precipitator manufacturers use a standard plate spacing and size
a precipitator based on a modified form of the Deutsch-Anderson equation. The
effective migration velocity used in the Deutsch-Anderson equation is usually
based on past experience and pilot plant testing with the standard plate spacing.
This experience, however, cannot be used to evaluate the effect of using a dif-
ferent plate spacing in each stage of the precipitator. By restricting the de-
sign of an ESP to a single plate spacing, one is not utilizing the space charge
arising from the charged dust particles to its optimum.
Recent experimental work has shown that wider than usual plate spacing re-
sults in increased efficiency [2] and not the decrease of efficiency as predicted
by the Deutsch-Anderson equation and other models [3]. The potential of plate
spacing as a design parameter cannot be evaluated appropriately by the use of
the Deutsch-Anderson equation or by the use of existing data which applies only
to a standard width precipitator. To examine the effect of plate spacing, one
must analyze the fundamental processes governing precipitator performance.
Precipitator collection efficiency for a given physical geometry, gas flow
field and particle size distribution depends on particle migration velocity, plate
spacing and particle turbulent diffusivity. For constant plate spacing and par-
ticle diffusivity, increasing the migration velocity increases the collection
efficiency and for constant migration velocity and particle diffusivity, increas-
ing the plate spacing decreases collection efficiency. Therefore, if migration
velocity can be increased as plate space is increased, it might be possible to
hold collection efficiency constant and reduce the specific collection area,
and therefore the initial cost of a precipitator. The influence of space charge
<3n the electric field and the electric field's influence on migration velocity
hold the key to understanding how the migration velocity varies with plate
spacing.
The electric field at any point in the precipitator depends on the applied
voltage, the plate spacing, and the space charge due to charged particles, ions
and electrons. In this discussion we will hold the applied voltage divided by
wire to plate spacing constant and neglect the influence of ions and electrons
on the total space charge. Also, to make the discussion simpler, the wire and
plate geometry will be modeled as a plate and plate precipitator with the par-
ticles charged to a saturation charge corresponding to the average applied elec-
tric field (applied voltage divided by wire to plate spacing). Using these
assumptions, the electric field at any point in the precipitator depends on the
applied electric field and the space charge due to the charged particles.
In the initial stage of the precipitator where the dust loading is still
high, the particle space charge will have a considerable effect on the electric
field, but in the last stages where most of the particles have been removed, the
space charge will have very little effect on the electric field. The electric
115
-------
field near the collecting plate is, using our assumption, the applied electric
field plus the integral of the space charge from the high voltage plate to the
location of interest. This means for a stage with high dust loading, wide plate
spacing will give a higher electric field near the collecting plate. And since
migration velocity is directly related to electric field, a stage with high dust
loading and wide spacing will have higher migration velocities than one with the
same loading and narrower plate spacing. All this indicates that it might be
possible to use wide plate spacing in the initial stage or stages without suf-
fering a decrease in overall collection efficiency. Since wider plate spacing
might result in a lower initial cost for the precipitator, the trade-off between
plate spacing and collection efficiency as a function of stage loading needs to
be investigated. As the first step toward determining this trade-off, a numerical
scheme based on the assumption given above was developed and used to quantify
the effect of plate spacing.
CALCULATION SCHEME
A calculation scheme was developed to determine how the collection effici-
ency varied with plate spacing as a function of dust loading. Since this was
the first stage of our investigation, we wanted to include all major phenomena
influencing collection efficiency without making the resulting calculations too
involved numerically or too costly to perform. The goal of this set of calcula-
tions was to determine trends and not a detailed prediction.
The scheme is composed of three interconnected sections: migration velocity
calculation, concentration calculation, and electric field calculation. Each
section is described separately below as well as the scheme used to link the
sections together.
Migration Velocity Calculation
The migration velocity is calculated assuming the velocity of the particle
results from a balance of electrostatic and drag forces. The particle is assumed
to be charged to saturation as set by the applied electric field E . Therefore
the migration velocity is given by
E E E d
o c p
w = *-—
where E = electric field at the particle location
E = applied electric field (applied voltage divided by plate to
wire spacing
e = permitivity of free space
u = gas viscosity
For this investigation the applied electric field, E , was held constant at
.4 MV/m. C
Electric Field Calculation
Although relationships exist to calculate the electric field, current, and
spare charge from first principles, a general numerical scheme has not yet been
116
-------
demonstrated. At this stage in our investigation, the electric field can be
assumed to be that generated by a plate and plate precipitation with the space
charge due only to the charged particles. The space charge due to ions and
electrons will change the computed electric field slightly, but the results will
have the same general trends. Neglecting gradients in electric field in the
flow direction as compared to the transverse direction, the electric field is
given by
E = E + £ dy
c y £ y
o
where
y = transverse coordinate
E = applied electric field (applied voltage divided by plate
spacing)
P = space charge due to the charged particles
£ = permitivity of free space
Particle Concentration Calculation
Particle motion in a precipitator is the result of three transport processes:
longitudinal motion of the gas, transverse motion due to the electric field, and
particle turbulent diffusion. Assuming the particles follow the gas flow in the
longitudinal direction and diffusion effects are only important in the transverse
direction, the particle concentration at a point in the flow field is given by
V(3c/3y) + UOc/Sx) + W(9c/3y) - D^c/ay2) = 0
where
c = particle concentration
U = longitudinal gas velocity
V = transverse gas velocity
W = migration velocity
D = particle turbulent diffusion coefficient
x = longitudinal coordinate
y = transverse coordinate
The boundary conditions at the collection plate and charging plate are
B(3c/3y) = We at the charging plate
3c/3y = 0 at the collection plate assuming no re-entrainment
This set of equations has been used by others ±4, 51 to model ESP operation.
A turbulent particle diffusion coefficient must be specified before the
equation can be solved. In general the diffusion would change with both longitu
dinal and transverse location in the precipitator. Since, at this stage of our
analysis, interest is focused on trends in performance, as a function of plate
spacing, a constant diffusion coefficient of 0.0005 m /sec was used. Also the
transverse velocity was assumed to be zero and the longitudinal velocity was
taken as a constant at all locations.
117
-------
Numerical Solution
The partial differential equation for particle concentration was converted
to finite difference form using up-wind difference techniques for the longitudinal
and migration velocity and central difference for the diffusion term. The result-
ing system of algebraic equations were solved by the Gauss-Seidel method.
The solution scheme is started by assuming the electric field at all loca-
tions in the precipitator is equal to the applied voltage divided by the wire to
plate spacing. Using this electric field, the particle migration velocity at
all locations is calculated and finally the particle concentration is determined
by Gauss-Seidel iteration. Using the calculated concentrations, the electric
field is determined including the effect of the space charge created by the
charged particles. The migration velocity is then recalculated using the new
electric field. The particle concentration equation is then solved again using
the new estimate for particle migration velocity. This cycle is continued until
no change is found in the calculated concentration between cycles. Typically,
seven cycles are required for each particle size.
The dust entering a precipitator typically has a wide size distribution.
This continuous distribution can be divided into a number of bands with each
represented by a single particle size and a fraction of the total loading.
Particle concentrations for each band would be calculated separately but the
space charge used in the electric field calculation must include all particle
sizes. In this work only one particle size was used since we were only inter-
ested in trends.
RESULTS
Calculations were first performed for three dust loadings and a range of wire
to plate spacings. The results, shown in Figure 1, show that for a given plate
spacing, increasing the dust loading increases the collection efficiency. The
figure also shows the efficiency decreases with increasing plate spacing, but the
decrease is very slight for the high loading case. We have not yet investigated
the effect of gas velocity, ion space charge or turbulent particle diffusion on
the relationship between efficiency and plate spacing. Under certain conditions
it might be possible to increase the collection efficiency with increasing plate
spacing. The results shown in Figure 1 are for the first stage at a precipitator
with a seven foot plate length.
To show the effect of non-uniform plate spacing, three precipitator configu-
rations were picked and the overall efficiency of each was determined. Tables 1,
2, and 3 give the geometry of each configuration and the calculated efficiency
for each stage. Figure 2 shows the wider spacing decreased the collection effi-
ciency slightly, but greatly reduced the specific collection area.
The results obtained in this initial investigation should only be taken as
showing the trends to expect with wider plate spacing in the initial stages of
a precipitator. The optimum configuration has not yet been determined. Addi-
tional work needs to be done to include the effect of ion space charge in the
calculations.
118
-------
ACKNOWLEDGEMENTS
This work was supported, in part, by Battelle Memorial Institute through
their University Distribution Program.
REFERENCES
1. Stearns-Rogers Incorporated (Denver CO). Economic Evaluation of Fabric
Filtration versus Electrostatic Precipitation for Ultrahigh Particulate
Collection Efficiency. EPRI Final Report FP-775, Research Project 834-1.
2. Misaka, et al. Hitachi Plant Engineering and Construction Co.(Tokyo).
Electric Field Strength and Collection Efficiency of Electrostatic Precipi-
tators Having Wide Collecting Plate Pitches. CSIRO Conference on Electro-
static Precipitation, August 1978.
3. Eschbach, et al. The Influence of Plate Spacing on the Efficiency of an
Electrostatic Precipitator. Washington State University. 1978 Annual
Meeting PNWIS Air Pollution Control Association, November 1978.
4. Feldman, P.L., Kumar, K.S., Cooperman, G.P. Turbulant Diffusion in Electro-
static Precipitators. AIChE Symposium Series, Atmospheric Emissions and
Energy Source Pollution73:165, 1977.
5. Soo, S.L. A Critical Review on Electrostatic Precipitators. AIChE Symposium
Series, Air Pollution and Its Control. 68:126, 1972.
119
-------
Table 1 GEOMETRY AND EFFICIENCY OF PRECIPITATOR A
20"
Stage
-4-
16"
Wire to
Plate
12"
8"
8"
Dust Load Dust Load
In,
Out
Stage
Efficiency
(kg/m ) (kg/m )
1
2
3
4
5
10"
.2540m
8"
.2032m
6"
. 1524m
4"
. 1016m
4"
. 1016m
4.577(10~2)
2.155(10~3)
8.221(10~4)
3.642(10~4)
1.319(10~4)
2.155(10 3)
8.221(10~4)
3.642(10~4)
1.319(10~4)
5.111(10~5)
95.29%
61.85%
55.70%
63.79%
61.26%
SCA =
208 ft
1000 ft /min
Precipitator Parameters
Overall Efficiency, n = 99.888%
7 ft, 2.134m
plate length:
gas velocity:
gas viscosity:
particle diameter: 1 micron
particulate density: 3000 kg/m
average electric field: .4MV/m
6 ft/s, 1.8288 m/s
2.80 x 10~5 kg/m-s
3
120
-------
Table 2 GEOMETRY AND EFFICIENCY OF PRECIPITATOR B
16
12"
8"
8"
Stage
1
2
3
4
5
Wire to
Plate
8"
.2032m
6"
.1524m
4"
.1016m
4"
. 1016m
4"
.1016m
Dust Load
In
(kg/in )
4.577(10"2)
2.005(10~3)
6.778(10~4)
2.257(10~4)
8.503(10~5)
Dust Load
Out
(kg/mJ)
2.005(10~3)
6.778(10~4)
2.257(10~4)
8.503(10~5)
3.342(10~5)
Stage
Efficiency
95.29%
66.20%
66.69%
62.33%
60.70%
c.rA 208 ft Own-mil T7ffi'r.-ir,rvoi7 n - 00 0977
1000 ft /min
Precipitator Parameters
7 ft, 2.134m
6 ft/s, 1.8288 m/s
,-5
plate length:
gas velocity:
gas viscosity: 2.80 x 10 J kg/m-s
particle diameter: 1 micron
•5
particulate density: 3000 kg/m"
average electric field: .4MV/m
121
-------
Table 3 GEOMETRY AND EFFICIENCY OF PRECIPITATOR C
/
V
* n
1 \
^ /
1 \
\ ' t
f \
< /
f >
s
t
8"
8"
8"
8"
Stage
1
2
3
4
5
Wire to
Plate
4"
.1016m
4"
.1016m
4"
. 1016m
4"
. 1016m
4"
. 1016m
Dust Load
In
(kg/in )
4.577(10~2)
1.419(10~3)
3.985(10~4)
1.429(10~4)
5.517(10~5)
Dust Load
Out ,
(kg/mJ)
1.419(10~3)
3.985(10~4)
1.429(10~4)
5.517(10~5)
2.189(10~5)
Stage
Efficiency
96.90%
71.92%
64.13%
61.39%
60.70%
292 ft
SHA = n-irovall T?f f nr-ion^TT n - QQ 0^9'
1000 ft /min
Precipitator Parameters
plate length:
gas velocity:
gas viscosity:
particle diameter:
particulate density:
' 7 ft, 2.134m
6 ft/s, 1.8288 m/s
2.80 x 10~5 kg/m-s
1 micron
3000 kg/m3
average electric field: .4MV/m
122
-------
100
O
y 90
UJ
O
u.
11.
UJ
30 grain/ft
20 grain/ft3
7 grain/ft"
sou
8
8
10
DUCT HALF WIDTH, inches
Fig.I Efficiency variation with dust loading and duct
half width in the first stage.
123
-------
100
99.9
B
-99.8
o
LU
O
u_
u_
UJ
I I I I I I I I I I I
200 220 240 260 280 300
SCA, sq.ft/IOOOcfm
Fig.2 Efficiency and SCA of the precipitator configurations.
124
-------
INTERACTION BETWEEN ELECTROSTATICS AND FLUID DYNAMICS
IN ELECTROSTATIC PRECIPITATORS
By
Samuel Bernstein
and
Clayton T. Crowe
Paper to be Presented at the EPA
2nd Symposium on the Transfer and Utilization of
Particulate Control Technology
July 23 to 27, 1979
Denver, Colorado
125
-------
INTERACTION BETWEEN ELECTROSTATICS AND FLUID DYNAMICS
IN ELECTROSTATIC PRECIPITATORS
By
Samuel Bernstein, Flow Industries, Kent, Washington
Clayton T. Crowe, Washington State University, Pullman, Washington
ABSTRACT
Significant effort has been devoted by past researchers in electrostatic
precipitators to electrostatics and particle chemistry. Relatively little,
however, has been published on the effect of turbulent flow and its inter-
action with electrostatics and particle dynamics. This paper describes an
experiment and a fundamental model which is focused on the coupled effects
of the fluid dynamics and electrostatics.
A laboratory precipitator flow facility with a laser beam illumination
was used to visualize smoke flow. Detailed photographs of flow patterns
with and without electric field were recorded in a horizontal plane through
the precipitator.
A theoretical model was used to numerically simulate the observed flow
patterns in the experiment. In the model the flow was described by time-
averaged mass and momentum equations with a two-equation turbulence model.
The electric field was described by Maxwell electrostatic equation and the
continuity of the current equation. The system of the coupled partial
differential equations was solved by a finite difference scheme.
The interaction between fluid dynamics and electrostatics is shown to
have a significant role in altering the flow in the precipitator. The
practical implication is that the interaction will aid in particle migration
in some regions, but it will increase reentrainment and reduce particle
migration in others. Comparison between the results from the experiment and
numerical computations shows good qualitative agreement between the two.
Since both the electrostatic and the fluid dynamic flow fields depend on
the electrode geometry, further investigation may lead to improved design of
precipitators.
126
-------
INTERACTION BETWEEN ELECTROSTATICS AND FLUID DYNAMICS
IN ELECTROSTATIC PRECIPITATORS
1. Introduction
Electrostatic precipitators have been well suited to remove particulates
from gas streams in applications requiring high collection efficiency of
micron-size particles, large gas volumes, and low pressure losses. One
common, and ever-increasing, application has been the removal of flyash
particles from the exhaust gases of electric power plants fired by pul-
verized coal. In recent designs, however, precipitator size, and conse-
quently its cost, have been increased considerably to meet strict performance
criteria. This adverse economic trend provides the motivation for research
to develop new precipitator technology. A better understanding of the
fundamental process of electrostatic precipitators will provide guidance to
reduce their size and cost while maintaining their high operating efficiency.
The subject of the investigation presented in this paper is one such
fundamental mechanism in electrostatic precipitators; that is, the inter-
action between fluid dynamics and electrostatics. Although this interaction
appears to be quite strong and a -significant factor in precipitator perfor-
mance, relatively little has been published on the subject to date. The
discussion in the paper includes a brief outline of the state-of-the-art
precipitator performance followed by a description of a laboratory experiment
and a numerical simulation, both of which are focused on the coupling between
fluid dynamics and electrostatics. Finally, some preliminary conclusions
and recommendations for further work are presented.
2. Performance of Electrostatic Precipitator
The classically accepted operating principles of electrostatic precipi-
tators are well documented in the literature (for example, see White (1963) ,
Robinson (1971)2 and Oglesby and Nichols (1970)3). The common industrial
precipitator consists of a row of wire electrodes suspended between vertical
plate electrodes which constitute the duct for passage of the dust-containing
gases. A high voltage applied between the wire and the plate electrodes
generates a corona discharge and a flow of ions towards the duct walls.
Some ions accumulate on the particles, imparting a net charge on them. The
resulting charged particles are accelerated towards the duct walls by
Coulomb forces and slowly migrate onto the plates. The accumulated dust
particles on the plates are then dislodged into collecting hoppers by periodic
rapping.
The above-simplified description of precipitator performance was the
basis for Deutsch's classical analysis (see for example White (1963)1).
The analysis resulted in the formulation of a one-dimensional equation which
expresses the particulate collection efficiency as a function of the migra-
tion velocity, collecting electrode area, and the flue gas volume flow.
Deutsch's analysis also assumes that turbulent mixing is strong enough to
maintain a uniform dust concentration profile across the precipitator duct.
It further assumes that the migration velocity is constant along the wall.
127
-------
The migration velocity is predicted by equating the aerodynamic force
and Coulomb force on the particle. However, the computation of particle
migration velocity with Deutsch equation generally fails to yield a reliable
prediction of collection efficiency. Consequently, precipitator designers
use collection efficiency data and the Deutsch analysis to calculate an
"effective" migration velocity for a precipitator design which then serves
as an empirical factor for design of similar units. Numerous researchers
(such as Robinson (1967)4, McDonald (1978)5, Inyushkin and Averbukh (1969)6,
and Friedlander (1959)7 have developed modifications of the Deutsch efficiency
equation by adjusting the migration velocity to account for such phenomena
as turbulent flow effects, non-uniform electric fields and a distribution of
particle sizes. Such an approach is, however, inadequate for either the
evaluation of new precipitator designs or for the evaluation of existing
precipitator designs for different operating conditions on which the migra-
tion velocity data is based. The use of the above approach can lead, and
has led, to units which fall far below design specifications.
Another predictive model for collection efficiency has been proposed by
Williams and Jackson (1962)7 and Cooperman (1969)8. This model is based on
solution of the field concentration equation for particle concentration in
the precipitator.
The major limitations in past investigations have been the neglect or
the use of overly simplified fluid dynamic models, the lack of description
of the electrostatic field interaction with the fluid dynamic model via the
ionic wind and the effect of the boundaries and electrodes on the electro-
hydrodynamic and particulate field. Consequently, these approaches require
assumptions regarding particle transport laws and cannot adequately predict
the performance of the precipitator. Of course, the complete treatment of
precipitator performance requires the evaluation of numerous additional
factors. These include (1) the effect of back corona discharge and particu-
late resistivity as influenced by gas chemistry and coal properties, (2)
particle reentrainment from the collecting plates and from the hoppers both
during rapping and normal operation, (3) plate vibration during rapping (4)
electrode alignment, and (5) particle charging. It is suggested, however,
that a proper description of the fluid dynamic flow field, the electrostatic
field and their interaction is a necessary fundamental step for a meaningful
evaluation of the device.
The following interactive phenomena including fluid dynamics, electro-
statics and particle dynamics is hypothesized (see Figure 1): the fluid
dynamic field is responsible for particle motion and is controlled primarily
by the inlet conditions and duct geometry. In practical operation the flow
field is turbulent, and particle transport will be influenced by the turbu-
lence energy and characteristic length scales of the field.
The particles are charged by the discharge electrostatic field and are
then subjected to the Coulomb force provided by the electrostatic field. At
high electrostatic field intensity, the fluid dynamic flow field becomes
ionized and develops a pressure gradient proportional to the ion charge
density and the electric field intensity. This pressure gradient generates
fluid flow referred to as ionic or electric wind.
128
-------
In order to qualitatively examine the ineraction between the fluid
dynamic flow field and the electrostatic flow field, both a laboratory
experiment and a numerical simulation was conducted.
3. Experimental Investigation
3.1 Experimental Setup
The experimental investigation was conducted in a laboratory precipi-
tator flow facility. The dimensions of the test section of the facility
were 30 inches wide, 15 inches high and approximately 12 feet long. A bell-
mouth entry section with honeycomb flow straighteners preceded the test
section. The gas flow through the facility was driven by a fan downstream
of the test section. Gas flow rate was controlled by a damper and a
variable ratio pulley which drove the fan. The facility was capable of
generating flow rate in the range of 1 to 20 ft/sec. The gas flow in the
test section was turbulent with uniform spatial distribution of its cross
sectional area. All four sides of the test section were made of plexiglass
with an iron frame. A smoke generator (Bernstein (1978)9) was used to
generate fine smoke from petroleum based oil. In order to visualize one
elevation in the precipitator flow facility, an Argon ion laser with beam
intensity of 5 watts was dispersed into a thin sheet of light by means of a
glass rod. The light scattered by the smoke particles at the beam elevation
were then observed from above. In this illumination technique a two-
dimensional view of one elevation was possible. The method is illustrated
in Figure 2. The glass rod was positioned at several locations in the
facility to obtain observation of fine details of the flow field. As illu-
strated in Figure 2 the flow field was observed through a TV monitor for
safety.
Precipitator plate electrodes with normal flanges of approximately
1 inch width were placed in the facility with plate to plate spacing of
9 inches. The plates were mounted against the top and the bottom of the
test section. Discharge wires with 0.109 inch diameters were placed at the
center of the channel with approximate spacing of 8 inches. Two wires were
placed between each two flagnes, the power supply used was capable of
generating negative corona with a sparkover voltage of approximately 70 KV
at ambient condition for the above plate arrangement.
3.2 Test Results
Qualitative observations of flow patterns were obtained over a range of
flow velocities and electric field intensity. The observations were re-
corded in a video cassette recorder. The results reproduced here were
photographed from a TV monitor screen. Figure 3 is a photograph of the flow
pattern without the electric field. Although the electric field was not
energized, one discharge wire may be seen in the photograph. The flow
direction was from left to right and the glass rod can also be seen at the
left portion of the picture. The average gas flow was 1.5 ft/sec and
regions of high and low light intensity represent regions of high and low
concentrations of smoke particles.
129
-------
Figure 4 is a photograph of the smoke flow with average gas velocity of
1.5 ft/sec, field voltage of 39 KV and current of 1 mill! A/m. Two circular
shadow regions can be observed near each discharge wire. The flow between
the discharge wire and the wall electrode is deflected towards the wall and
does not have the same turbulent eddy structure as the fluid flow alone.
Finally, Figure 5 is a photograph of the instantaneous flow patterns
with field voltage of approximately 50 Kv and current intensity of 0.8 milli
A/m. In reviewing these results two comments should be emphasized. The
first is that this preliminary qualitative description of the precipitator
flow field is only an instantaneous view of the unsteady turbulent field.
The second comment is that in numerous variations of gas flow velocity and
electric field intensity it was not always possible to identify the flow
patterns, especially when the video screening was stopped.
4. Numerical Simulation
4.1 Theoretical Model
The model used to describe the interaction between the fluid dynamics
and electrostatics was described by Bernstein and Crowe (1978)10. A similar
model for cylindrical geometry was used by Crowe and Stock (1974)*-*• to
predict the flow pattern in a wire-tube geometry. An outline of the former
model, however, is included here since the referenced work may not be
readily available in the U.S.
The flow field is described by a two-dimensional model similar to the
illuminated plane as shown in Figure 2. The fluid dynamic model, originally
developed at Imperial College (1973)12 uses the time-averaged conservation
equations of mass and momentum and a two-equation turbulence model. Such a
model contains the usual drawback of a turbulence model; however, it does
represent a state-of-the-art in turbulence modeling and it provides good
flow predictions in flow fields not associated with precipitators. The
electric field model includes Maxwell's electrostatic field equation and the
continuity of current equation. The model contains a system of seven
coupled non-linear partial differential equations as follows:
Fluid Dynamic Model
A- Continuity Equation:
^ + ^ = 0 (1)
dx dy ^
where u and v are the time-averaged velocity components in the stream-
wise, x and lateral y directions respectively.
B. Momentum Equation; Using the Boussinesq approximation to relate
the stress on the fluid to the rate of strain with turbulent flow, the
momentum equation in the x-direction is given by
130
-------
__
3x
puu -
, Jiu
eff 3x
3y
c.
ion
E
x
3u
3x
(2)
where Ueff is the effective viscosity, p is the pressure, p is
the ion charge density and EX is the electric field intensity 10n
in the x-direction. The two terms on the left-hand side represent the
momentum transport in the two coordinate directions by convection and
diffusion. The terms on the right-hand side are the force due to pressure
gradient and the body force due to motion of the ions caused by the electric
fields (effect of the electric wind).
C. Y-Momentum Equation;
be written as
The momentum equation in the y-direction can
— (puv - u —
3x reff 3x
(3)
.
ion
The terms are easily identified by referring to Equation 2.
D. Turbulence Kinetic Energy; The differential equation for
turbulence kinetic energy as developed by Launder and Spaldingl2 is
eff 3k
a, 3x
( i eff 3k
(pvk —- -r-
- pe
(4)
where k is the turbulence energy, a, effective Prandtl number for
turbulent energy transfer by diffusion (usually taken as unity) and £ is
the dissipation rate. The two terms on the left-hand side represent tur-
bulence energy transfer in the x and y directions by convection and
diffusion. The term G is the generation term which is given by
G = y
eff
y/
_3v
3x
(5)
The turbulence energy equation as incorporated here does not include ion
flow on turbulence energy generation or decay. Possible contribution of
this effect is the subject of continuing research.
E. Turbulence Dissipation Rate; The differential equation for
dissipation rate of turbulence energy as suggested by Launder and Spalding
is
131
-------
3x
pue +
eff
y
eff jte
3y
(6)
pve -
C EG
where a£ is the effective Prandtl number for diffusion of energy dis-
sipation. The G! and C2 parameters are empirical constants which are
assigned by comparing the predictions with experimental data in simple jet
flow experiments unrelated to precipitators. As with turbulence energy, no
attempt was made to account for particle or electric wind effects.
The remaining expression for "closure" of gas flow equations is the
Prandtl-Kolmogorov equation for effective viscosity in terms of the turbu-
lence kinetic energy and dissipation rate; namely,
yef£ - P + cu pk /£ (7)
where C^ is another empirical constant assigned a value of 0.09.
Electrostatic Field Model:
The two dependent variables describing the electric field are the
voltage and ion charge density. The electric field strength is by defini-
tion the negative gradient of the voltage. The two equations for the
electric field are Maxwell's equation and the continuity of current flow.
A. Maxwell's Equation;
density is
2
The equation relating voltage and charge
(8)
ions
particles
•_ -I
where V is the voltage, e^ is the dielectric constant and pc is the
charge density. The charge of ions as well as the charge on the particles
affect the electric potential.
It is a reasonable approximation, however, to regard the effect of
charge density as a second-order effect and to use the voltage field obtained
by the solution of Laplace equation. Cooperman (see Robinson (1971)2)
obtained an analytic solution for the voltage distribution in a channel with
wires spaced equidistant along the center of the channel as
In
V
m =
cosh [ir(x - mB)/AJ - cos (fry/A)
cosh [ir(x - mB)/A] + cos (fTv/A)J
m
In
cosh (TITO B/A) - cos(irr /A)
cosh (TITO B/A) + cos(irr /A)
(9)
132
-------
where A is the channel width, B is the distance between wires and rQ
is the wire radius and VQ is the corona discharge voltage. The above
series rapidly converge in several terms.
The electrostatic field intensity is obtained by taking the negative of
the voltage gradient. The resulting expressions are
- mB) /A] - - (10)
cos(y,/A)
~ mB) /A] -cos (yir/A)
_oo
E -.k^ sin(yTt/A) £ coah ^"^^ (11)
cosh rir(x - mB)/A"|- cos (yir/A)
m = -co i- -j
where k, is the corona voltage over the denominator for Equation (9). In
the computation performed in this study for a single discharge electrode,
the electric field solution was approximated using the method of images with
an array of source-sink distribution.
B. Continuity of Current Flow - The continuity of current flow is
expressed as
V.J = 0 = V-(pc bE). (12)
ions
where j is the current flow vector, b is the ion mobility and E is
the electric field intensity. The current flow due to particle motion is
neglected, because the particle mobility is much smaller than ion mobility.
Using the same assumption pertaining to the second-order effect of
charge on the voltage distribution, Cooperman (see Robinson (1971)^) showed
that the charge density may be assumed to be a constant in the channel and
equal to
P j ln (d/ro) (13)
c. 27T b V
ion o
where d(m) is a function of duct width and wire spacing.
The solution of the electrohydrodynamic model was obtained by finite
difference solution in a CDC 7600 computer. Typical run required 15 seconds
of central processor for a flow field described by a grid of (22 x 22)
elements.
4.2 Computed Examples
Four examples of computed flow fields are presented for comparison with
the experimental data. These include (a) turbulent flow without electric
fields, (b) turbulent flow with moderate intensity electric field generated
by a single wire (c) turbulent flow with high intensity electric field
133
-------
generated by a single wire, and (d) turbulent flow with electric field
generated by multiple wires. Although numerical values are used in this
example, the results should be considered qualitative for the purpose of
comparisons with experiments. No careful attempt was made to precisely
match the geometry and operating conditions in both cases. The computed
results are given in Figures 6, 8, 9, 10 for each case discussed above.
Figure 7 includes the electric field distribution for the case of a single
wire and multiple wires. This field calculation neglects the presence of
the flanges on the collecting electrodes.
5. Discussion
The discussion is divided into two cases: the fluid dynamic flow field
and the combined electrohydrodynamic field.
The experimental result of the fluid dynamic flow field, as shown in
Figure 3, illustrates turbulent eddies with high and low smoke concentration
of smoke particles. It is evident that turbulent diffusion is not sufficient
to provide uniform particulate concentration as assumed by Deutsch's analysis.
The numerical computations of the same case reveals a small recirculating
bubble aft of the baffle and a short stagnation zone ahead of the downstream
baffle. This predicted flow pattern was qualitatively confirmed in the
experimental study. In the absence of strong ionic wind effect, significant
reentrainment of particles from the wall electrode may develop due to
scouring of the electrode by the flow.
Inspection of both experimental and computational results for the
combined electrohydrodynamic field reveals a strong coupling in the flow.
During the experiments, the flow field was found to be sensitive to the
relative values of fluid velocity and electric field intensity. Qualitative
agreement is observed between the computed flow field and the experiment.
The observed agreement is particularly good between the computed cases of a
single wire (Figures 8 and 9) and the photograph given in Figure 5. The
preliminary computation with multiple wires, Figure 10, predicts multiple
vortex flow near the walls. The results show horizontal flow patterns,
although a vertical ionic flow as studied by Robinson (1976)-'-3 may also be
present. The clover leaf flow pattern of two vortices near each wire, as
hypothesized by Robinson, were not observed in either experiment or compu-
tation for the given electrode geometry and operating conditions. The
preliminary nature of this investigation precludes any further evaluation
regarding the conditions for existence of specific flow patterns with a
given geometry and operating conditions, and these are the subject of con-
tinuing research.
6. Conclusions and Recommendations
The investigation reported here included a qualitative evaluation of
the interaction between fluid dynamics and electrostatics in electrostatic
precipitators. The evaluation was conducted by both an experimental effort
and a numerical simulation in a horizontal plane through the precipitator.
Several conclusions and recommendations can be drawn as follows:
134
-------
(1) The interaction between fluid dynamics and electrostatics appears
to be quite strong. Significant electric wind flow was observed experi-
mentally and numerically at operating conditions of high electric voltages
and currents. The voltage and current levels at which the electric wind was
observed to significantly alter the flow field were lower for low average
gas flow rates.
(2) Qualitative numerical simulation provided good agreement with
observed flow patterns.
(3) The study documented both experimentally and numerically the
existence of a strong electric wind in a horizontal plane. The experimental
data indicate recirculating flow from the discharge wires to the collecting
electrodes and back to the discharge wires. Such a flow pattern satisfies
the continuity of flow requirement; however, a vertical flow due to the
electric wind, as suggested by Robinson (1976)13} can not be ruled out.
(4) The degree of interaction between the fluid dynamic flow field and
the electrostatic field suggests that it may significantly affect the per-
formance of the precipitator. The electric wind may increase the effective
migration velocity at which particles are being transported to the collecting
electrodes; however, the recirculating flow may contribute to reentrainment.
In order to design an electrode geometry which would provide optimum per-
formance, further investigation of the phenomena is necessary.
(5) Further evaluation is recommended to obtain quantitative data on
the above interaction. Quantitative data should be used to verify the
reliability of the numerical predictions which could then be used for cost-
effective design work.
REFERENCES
1. White, H. J. (1963) "Industrial Electrostatic Precipitation" Addison-
Wesley Publishing Co., Inc. Reading, Massachusetts.
2. Robinson, M. (1971) "Electrostatic Precipitation in Air Pollution
Control" Part I. Ed. W. Strauss, John Wiley & Sons, Inc., New York.
3. Oglesby, S. Jr., and Nichols, G. B. (1970) "A Manual of Electrostatic
Precipitator Technology, Part I: Fundamentals" Southern Research
Institute, Birmingham, Alabama - Available through NTIS PB 196 381,
Part II: Application, August.
4. Robinson, M. (1967) "A Modified Deutch Efficiency Equation for Electro-
static Precipitation" Atmospheric Environment, Pergamon Press, Vol. 1,
pp. 193-204.
5. McDonald, J. R. (1978) "A Mathematical Model of Electrostatic Precipi-
tation Vol. 1 Modeling and Programming" EPA-600/7-78 Ilia, June.
6. Inyushkin, N. V., Averbukh, Y. D. (1962) "Influence of Gas Flow Pressure
Conditions on Dust Collection in an Electrical Field" Soviet J. of
Non Ferrous Metals, pp. 35-38.
135
-------
7. Friedlander, S. K. (1959) "Principles of Gas-Solid Separations in Dry
Systems" Chemical Engineering Progress Symposium Series No. 25, Vol. 55,
pp. 135-149.
8. Williams, J. C., Jackson, R. (1962) "The Motion of Solid Particles in
an Electrostatic Precipitator" Proceedings Interactions Between Fluids
and Particles, London, pp. 282-288, 3rd Congress of European Federation
of Chemical Engineering.
9. Cooperman, P. (1969) "A New Theory of Precipitation Efficiency Central
Electricity Research Laboratories" Leatherhead, G. B. Laboratory Note
No. RL/L/N, 10/69, June.
10. Bernstein, S. and Crowe, C. T. (1978) "Fundamental Model of Fluid
Dynamics Electrostatics and Particle Dynamics for Electrostatic Pre-
cipitators" CSIRO Conference on Electrostatic Precipitation, August 23,
24, Leura N.S.W.
11. Crowe, C. T. and Stock, D. E. (1974) "The Effect of Electrodynamic
Secondary Flow on the Performance of Electrostatic Precipitator"
Proceeding of the 1974 Heat Transfer and Fluid Mechanics Institute,
Stanford University Press, pp. 234-265, June.
12. Gosman, A. D. and Pun, W. M. (1973) "Calculation of Recirculating Flows"
Lecture Notes, Imperial College of Science and Technology.
13. Robinson, M. (1976) "Effects of the Corona Discharge on the Electric-
Wind Convection and Eddy Diffusion in an Electrostatic Precipitator.
Health and Safety Laboratory, Energy Research and Development Administra-
tion, New York, HASL-301, February.
136
-------
GO
Inflow Conditions
Geometry
FLUID DYNAMICS
FIELD
Aerodynamic Force
Electric Wind
— Applied Voltage
— Geometry
Space Charge
ELECTROSTATIC
FIELD
Coulomb Forces
PARTICLE DYNAMICS
Charge
Figure 1. Interactive Phenomena in Electrostatic Precipitators.
-------
n
to
OO
Gas Flow
Figure 2. Experimental Setup in Electrostatic Precipitator Flow Facility.
Collecting
Electrode
-------
Flow
Discharge
Electrodes
Collecting
Electrode
Figure 3. Visualization of Smoke Flow in Wire Plate Precipitator.'
(uave = 1.5 ft/sec Plate to Plate Spacing = 9 inches)
Flow
Discharge
Electrode
-Collecting
Electrode
Figure 4. Visualization of Smoke Flow in Wire Plate Precipitator.
(uave = 1.5 ft/sec, Plate to Plate Spacing = 9 inches, Electric
Field Intensity = 39 KV and J = 0.5 Mill! A/M)
139
-------
Discharge Electrodes
Flow
Collecting
Electrode
Figure 5. Visualization of Smoke Flow in a Wire Plate Precipitator.
(uave = 1.5 ft/sec, Plate to Plate Spacing = 9 inches. Electric
Field Voltage = 50 KV and J = 0.8 Mill! AIM)
140
-------
•0.6m
Scale:
1.00 m/s
Maximum Vector
Flow 0.10m
Figure 6. Computed Gas Flow Velocity Near Precipitator Electrode.
-------
IV)
a. Single Discharge Wire.
Scale
Each Line
200
Scale
Each Line
100KV/m
b. Multiple Discharge Wires.
Figure 7. Distribution of Electric Field Intensity in Wire Plate Precipitator.
-------
0.6m
Scale:
1.01 m/s
Maximum Vector
Flow 0.10m
CO
Discharge Electrode
i
Figure 8. Computed Gas Flow Velocity in Electrostatic Precipitator with a Single Discharge
Wire, Average Voltage of 70 KV and Ion Charge Density of 30 Micro C/m2.
-------
0.6m
1 .68 m/s
Maximum Vector
Flow
m
Figure 9. Computed Gas Flow Velocity in Electrostatic Precipitator with Single Discharge
Wire, Average Voltage of 80 KV and Ion Charge Density of 45 Micro C/m2.
-------
Flow
0.05m
0.6m
Scale:
0.115 m/s
Maximum Vector
v
\ M
1 V
1 \\
f
\
\
F|OW 0.10m
•0.6m
Scale:
.1.13 m/s
Maximum Vector
-0.12m
-0.27 m-
_ .» « I •
Figure 10. Computed Gas Flow Velocity in Electrostatic Precipitator with Multiple Discharge
Wires and Average Voltage of 80 KV Ion Charge Density of 45 Micro C/m^.
-------
PARTICLE TRANSPORT IN ELECTROSTATIC PRECIPITATORS
G. Leonard, M. Mitchner and S. A. Self
High Temperature Gasdynamics Laboratory
Stanford University
Stanford, California 94305
We discuss the transport of particles in a precipitator due to the com-
bined effects of mobility and eddy diffusion in the gas flow.
For an idealized model of a duct precipitator, the convective diffusion
equation is solved analytically for the monodisperse particle concentration
n(x,z) as a sum of normal modes which can represent any entrance profile
n(x = 0,z). These modes decay exponentially in the flow direction, x, with
the lowest (dominant) mode having the slowest decay and, therefore, controlling
the efficiency in long ducts.
The dominant mode yields the fractional efficiency formula
H = l-exp-(wgL/ud)F = l-exp-(wpA/Q)F, which reduces to the Deutsch result when
F = 1. The multiplier F which is a function of PE = (wgd/D) , (a measure
of the relative effects of migration velocity WE and diffusivity D) is
evaluated for different assumed boundary conditions. It appears that F
values significantly greater than unity are possible, particularly for
PE » 1.
Numerical solutions of the convective diffusion equation have also been
computed for a uniform entrance profile n(x = 0,z) = no, which agree with
the analytic results and show the entrance profile decaying to the self-similar
dominant mode profile for (WEA/Q) > 1.
The implication of the results to precipitator performance and design are
discussed. In particular, it is argued that the efficiency can be significantly
higher than the Deutsch value, and that by improving gas flow quality to minimize
turbulence, significant reductions of precipitator length could be achieved.
146
-------
INTRODUCTION
In a precipitator the charged particles are transported to the collector
walls by the combined influences of the Coulomb force due to the electric
field and of particle interaction with the flowing gas .
Consider a uniform duct flow of velocity u in the x direction between
parallel plates at z = 0, z = d across which a uniform field E is main-
tained (Figure 1) , and assume that monodisperse ash of radius a and charge
q enters at x = 0. This is a good representation of the collection stage
of a two-stage precipitator, and a fair representation of a single-stage
precipitator if the plane z = 0 is taken as the center line of the duct
where the high voltage wires are located. If the flow is perfectly laminar,
and Brownian motion is neglected, the particles are convected in the x
direction at velocity u and migrate in the z direction with velocity
WE given by
(D
where u. is the viscosity and C is the Cunningham slip factor. Thus the
particle trajectories are straight lines and all the entering particles would
be collected within a distance (u/Wg)d. Also, if the particle concentration
at the entrance were uniform, i.e., n(0,z) = no say, then the profile is
such that n = n0 for d > z > (wg/u)x and n = 0 otherwise (see Figure 2) .
If account were taken of Brownian motion with a diffusion coefficient
DB
where K is Boltzmann's constant, one would expect a similar result, but with
the sharp discontinuities in the concentration profiles rounded out. However
for typical precipitator conditions, e.g., T = 300 K, a = 1 urn,
y = 1.8 x 10~5 Kg/m-s, we find Dg ~ 10~il m2/s, which is so small compared
with the value of Wgd ~ 10~^ m2/s, that the result is indistinguishable from
that neglecting Brownian diffusion.
In all practical precipitators the flow is turbulent and the transport of
particles is determined by the structure of the fluctuating flow. For flow in
smooth-wall ducts the effect of turbulence is usually modeled in terms of an
effective eddy diffusivity. This approach,, while certainly not exact, often
leads to surprisingly accurate representations of turbulent transport processes
provided care is taken to avoid the generation of large eddies at the duct
entrance.
In the Deutsch model, widely used in the design and evaluation of precipi-
tators, it is assumed that turbulent diffusion and mixing is so dominant that
the concentration profile is maintained uniform in the core flow, and the
particles migrate across the boundary layer with the velocity WE- A simple
147
-------
analysis1 then leads to the following exponential decay of concentration in
the flow direction:
WEX
n(x) = n(0) exp - (--)- (3)
This leads to the Deutsch- Anderson efficiency formula
WL WA
where L is the duct length, A is the plate area and Q the volumetric
gas throughput .
It should be noted that eq. (3) applies only to a single size class of
particles having a given migration velocity WE, so that eq. (4) refers to
the fractional efficiency for a given size class. The overall mass efficiency
must be obtained by appropriate integration over the prevailing distribution.
2
The Deutsch model has been widely used both in computer codes for pre-
dicting performance and for characterizing the performance of precipitators
in terms of an effective migration velocity determined from efficiency
measurements and the use of eq. (4) assuming A and Q are known.
It is often stated in the literature that the effective migration velocity,
so measured, is commonly found to be significantly lower than calculated from
eq. (1), assuming that q and E are known either from theory or measurements.
This leads to the viewpoint that the Deutch efficiency is the theoretical
maximum value that precipitators are capable of attaining.
3
However, comparisons of the measured fractional efficiencies on a number
of power plant precipitators with computations from the SRI model^ show that
in the size range -0.1 - 1 ym, where the efficiency minimum normally lies,
the measured efficiencies (and migration velocities) are substantially higher
than those computed. This is unexplained, and suggests a deficiency in the
Deutsch model. Furthermore, the superior performance of precipitators with
plate spacings significantly greater than the conventional value, 4 is also
inexplicable on the basis of the Deutsch model.
The foregoing observations support the view that the Deutsch model is
deficient and suggest that a more exact treatment of the precipitation
process might lead not only to better performance predictions, but also to
prescriptions for improving the efficiency above the Deutch value (or equiv-
alently making smaller precipitators for the same efficiency) .
A brief review of earlier treatments of the effect of turbulent diffusion
on precipitation has been given by Robinson.
148
-------
MATHEMATICAL MODEL
We assume that for the model of Figure 1, the time-average gas velocity
u is uniform but that there are turbulent fluctuations of gas velocity in the
x and z directions. The particles are assumed to follow the instantaneous
gas velocity, without slip, except for the steady, uniform migration velocity
WE relative to the gas in the z-direction.
Taking the time-average of the instantaneous continuity equation for the
particles yields:
a a dn'u' dn'u'
on . dn , x , z
u + W + - + ~
Here u' u' are the fluctuating components of gas velocity in the x and
z directions and n' is the fluctuating component of the particle concentra-
tion. The overbars represent time-averages of the associated products.
In eq. (5) the first two terms represent the divergences of the particle
fluxes due to steady motion in the x and z directions, while the last
two terms represent the divergences of the fluxes due to turbulent transport
in these directions. We model the latter two terms in the conventional way,
by introducing the concept of an eddy diffusivity D. Thus we put:
n-ui-iife ^V = -D|^. (6)
•'u' =-D -^; n'u' =-„
x dx z 9z
Assuming that D is isotropic and uniform, eqs. (5) and (6) can be combined
to obtain the convective diffusion equation:
2 2
3n 3n .3 n 9
u - + w - D (+
The net particle velocity in the z direction due to both the electric force
and turbulent diffusion, can then be written
D 3n fo\
w = w^ --- r~ . \o)
E n dz
Equation (7) is to be solved for n(x,z) subject to boundary conditions to be
specified at z = 0 and z = d, as well as an entrance profile n(x = 0,z).
Having solved for n(x,z), then eq. (8) leads directly to w(x,z), and hence
to the net particle flux to the wall, nw.
To this point the problem is well posed and, subject to the assumptions
stated, an exact mathematical representation of the idealized model. Also,
in general, analytic solutions of eq. (7) are readily found by the standard
technique of separation of the variables .
149
-------
Examination of eq. (7) shows that the problem involves two dimensionless
parameters, which may be taken as (w«/u) and
w d
"E ' -§- ' (9)
The latter is a measure of the relative strengths of external force and
diffusion on particle transport. The limit P -*• 0 corresponds to the
Deutsch hypothesis of infinitely rapid turbulent mixing. The dimensionless
group PF is commonly called the Peclet number. Here P^ is based on the
electric migration velocity, and will be called the electric Peclet number.
Boundary Conditions
The principal problem in proceeding further relates to the specification
of reasonable boundary conditions at the walls z = 0 and z = d, that are
justifiable on the basis of the physics involved. For simplicity, we assume
that re-entrainment due to bouncing, saltation or other effects is negligible,
so that the collector wall(s) acts as a perfect sink. This would certainly be
true for a wet wall precipitator.
Now the standard boundary condition for the diffusion equation at a bound-
ary acting as a perfect sink, that is used in theories of molecular diffusion,
ambipolar diffusion of charged particles in gas discharges, and neutron diffu-
sion in reactors, is that the concentration is zero (n = 0) at such a wall.
In such cases, the diffusion coefficient is assumed constant up to the wall,
or at least to within one mean free path of the wall. However, in the case of
particle deposition in turbulent duct flows, the turbulent diffusivity falls
to zero through the laminar sublayer, and if one assumes constant D up to
the wall, with the boundary condition n = 0, then the calculated deposition
rate due to turbulence alone is several orders of magnitude higher than that
observed experimentally. "~° Effectively, the gas flow in the near-wall region
acts as a barrier to particle diffusion.
In this circumstance there are two procedural choices. The first is to
solve the convective diffusion equation with a non-constant D(z) which models
the effective diffusivity. This requires numerical solution. The second is
to retain D(z) = constant and use a modified boundary condition based on an
analysis of the near-wall region. This second approach, which we follow here,
has the advantage of allowing analytic solutions .
Thus we assume that D = constant throughout the region up to a distance
z = (d - e) where £ « d, and apply at this position a generalized boundary
condition
c + n - 0 (10)
150
-------
Here, c is a constant with dimensions of length which is determined by an
analysis of the near wall region. Clearly, in the limit c •> 0, we recover
the boundary condition n = 0, while in the limit c » d we obtain the
boundary condition n' •> 0.
In the analysis which follows, we derive particular solutions for the
two limiting cases n = 0 and n1 = 0, and it is clear that all possible
solutions fall between these two limits. The analysis of the near-wall region
given in the appendix shows that for particle deposition from turbulent flows,
the appropriate boundary condition lies close to the latter one, i.e. n' =0.
The boundary condition to be applied at z = 0 is more straightforward.
If z = 0 represents the central plane of a single stage wire-plate precipi-
tator, symmetry requires that the z-directed flux at this point be zero,
which, from eq. (8), yields
- n = 0 at z - 0 . (ID
This condition will also be applicable to a two-stage precipitator with the
plane z = 0 being the electrode which repels charged particles .
Analytic Solution
We assume that in eq. (7) n(x,z) is separable in the form
n(x,z) = X(x) Z(z). (12)
Substitution in (7) and division by XZ then yields
Since the first term is a function of x only, and the second of z only,
they must each be equal to a constant ± K (say) , and we have
(14)
dx
and
?„
dz2
151
-------
where the separation constant K, which has dimensions of inverse time, is to
be determined by the boundary conditions, i.e., it is an eigenvalue problem.
Equations (14) and (15) have solutions of the form
X = X exp ax ; a = -c l + (1 + 4
o 2D [_ - u2
(16)
Z = Z exp 3Z
WE f, . ,n 4DK,l/2~]
= 2DL1±(1"7T J
E
(17)
We seek solutions of (16) with a real, negative, corresponding to exponential
decay in the flow direction. This requires
4DK
> 0 , i.e., K real positive
(18)
u
There are two roots for a, one positive, which we discard, and one negative,
which we use.
Thus (16) yields for the x dependence
r \
n(x)
u r/1 . 4DKNl/2
- (1 + — -) -
(19)
u
For the transverse dependence, we seek solutions of (17) with 3 complex, so
as to yield solutions that can satisfy arbitrary boundary conditions at
z = 0,d. Thus we require
4DK
(20)
which is consistent with (19) and requires K real and > wF/4D.
two roots for 3, and the z dependence has the form:
Then we have
n(z)
WEZ
2D
cos
sin
WEZ
2D
,4DK
*• 2
WE
1/2
1) z
(21)
The imposition of boundary conditions at z = 0,d, will yield a set of eigen-
values K and corresponding eigenfunctions which form a complete orthonormal
152
-------
set of modes. Any given entrance profile n(x = 0,z) can be expanded in terms
of this set, and each component mode will decay exponentially in the x direc-
tion, according to eq. (19) at different rates, so that the profile changes
with x. For large x, the profile will tend to that of the fundamental mode,
i.e., that having the smallest eigenvalue K , since it will decay most slowly.
Defining m
w d
fl = E
m - 2D
4DK
m
a/2
- i -^
PJ
1
4DK
m
a/2
(22)
w_
and
, ( p >y
Vu'V
2 P.
26_ 21I 1/2
- 1
(23)
the general solution to eq. (7) for n(x,z) can be written:
n(x,z) =
m
exp-
w x
(-S-)
ud
m
PE
exp(—
A cos(9 f) +B sin(6 f)
m m d m m d
(24)
Here Am, Bm are constants whose ratio is determined by the boundary conditions
at z = 0, d, and whose magnitude is determined by the entrance condition
n(x =0, z).
It will be recognized that the x-dependence in eq. (24) reduces to that of
the Deutsch model, eq. (3), for F = 1. Thus our solution (24) generalizes the
Deutsch result by representing n as a sum of modes with "Deutsch-like" exponen-
tial x-dependences, modified by the factors Fm which are functions of the two
basic parameters (wp/u) and Pg = w^d/D. The exact forms of these factors, as
well as the transverse profiles of the modes, depend on the boundary conditions
at z = 0,d through the eigenvalues Q .
Application of Boundary Conditions - Particular Solutions
In the following, we evaluate the general solution (24) for two particular
cases. In each case we take the boundary condition at z = 0 as eq. (11).
For the boundary condition at z = d, we take the two limiting cases n = 0,
and n1 = 0 corresponding respectively to taking c = 0 and c -> °° in eq. (10)
These two cases bracket all intermediate boundary conditions corresponding to
intermediate values of c (positive).
Application of the boundary condition at z = 0 leads to
A = -- 6B .
(25)
153
-------
Application of the first boundary condition (n = 0) at z - d yields
A = -B tan 9 • (26)
Hence, from eqs. (25) and (26), the 9 values are given by the roots of
29
tan9 = - ^— •
E
(27)
Application of the second boundary condition (n' = 0) at z = d yields,
in place of (26)
A [1 - ( tan6] + B [tan6 + ( ] = 0
E E
(28)
From eqs. (25) and (28) we have 6 given by the roots of
tan6 =
(29)
instead of (27).
For either boundary condition, the transverse eigenmodes are of the form
n(z) = C
m
f)
(30)
where the C are constants determined by the entrance profile n(x=0,z).
The nature of the roots 9 of the alternative eqs. (27) and (29) is
best seen by inspection of the graphical solutions shown schematically in
Figures 3. In either case the root 9=0 is excluded.
Referring to Figure 3(a) for the first case (n = 0) , it is seen that
the set of roots are bracketed by the inequalities
TT/2
IT ,
3TT/2 < 62 < 2-rr ,
(2m- 1)? < 6 < mTT .
2 m
154
-------
It may be noted that for P£ » 1, they tend to their lower limits.
The first
Also with increasing m, the roots tend to their lower limits.
root 9j_ (for the dominant mode) is shown as a function of P..,
in Figure 4.
Referring to Figure 3(b) for the second case (n? =0), it is seen that
the situation is similar, but complicated by the pole in the right-hand side of
eq. (29) at 9 = (PE/2) , In general, if (2p - l)ir/2 < (?E/2) < (2p + l)u/2,
the first p roots occur for (2m - l)ir/2 < 9m < rnTT, and higher roots (m > p)
occur for (m - l)ir < 9m < (2m - l)-rr/2.
mode) is shown as a function of P in Figure 4.
m ,
The first root
0. (for the dominant
The Dominant Mode
The transverse concentration profiles n(z) for the dominant mode calcu-
lated from eq. (30) are shown for the boundary condition n = 0, in Figure 5
for several values PE. It is seen that when PE is large there are relatively
few particles in the inner region (where in eq. (30) the exponential factor
dominates) and there is a pronounced peak in concentration close to the wall.
From eq. (8) it is clear that the average particle velocity w is smaller than
WE inside the peak, and greater than WE between the peak and the wall. As
PE decreases, the relative concentration at the center increases, the maximum
is broader and moves towards the center. For small values of PE the concentra-
tion approaches a monotonically decreasing function of z.
The transverse concentration profiles n(z) for the dominant mode are
shown for the boundary condition n' = 0 in Figure 6. In contrast to those
discussed above, the profiles increase monotonically to a maximum at the wall.
They show the same general dependence on PE as the previous case and for
small
that
E
w < W
approach the uniform profile assumed in the Deutch model.
E everywhere except at the wall where w = WE.
Note
Of more engineering significance is the dependence of the axial concentra
tion profile, since this determines precipitator efficiency. Of particular
interest is the factor F^, for the dominant mode, multiplying the exponent
in the modified Deutch equation (24) , and its functional dependence on the
parameters (WE/U) and PE for the two cases.
From eq. (23) we have
2 _
1/2
-1
(31)
The functions Fl(pE) for tne alternative boundary conditions are shown in
Figure 7 for the case (WE/U) = 0.1. For conditions of interest F^ is a very
weak function of (WE/U).
In the laminar limit PE •> °° both functions tend to infinity as (PE/4) .
For PE decreasing from large values, down to at least
both functions can be approximated by
PE «
10, where
3,
155
-------
i
;
-1
(PE»D. (32)
For still smaller values of PE (increasing diffusivity) the two functions
diverge.
For the boundary condition n = 0, FI decreases to a minimum of Fi^2.3
at PE ~ 3, and then increases again, eventually approaching the value
FI -»- Tt2/(4 PE) (33)
for 1 » P » (w^/u)2 . (34)
h SL
Equation (33) leads to an efficiency
2
n^ 1 - exp-(^- -^-), (35)
ud
which does not involve WE. In this limit, the boundary condition n = 0 pre-
dicts a (mechanical) efficiency which is very much larger than measured. For
instance, putting D = 150V = 2.2 x 10~3 m2/s, a typical value of momentum
diffusivity for fully developed duct flow in air at Reynolds numbers of interest^
and with d = 0.1 m, L/d =50, we find a mechanical efficiency of n, ~ 75%,
which is unrealistically high in the absence of electrical effects.
In contrast, for the boundary condition n1 = 0, as PE decreases, F..
decreases monotonically to unity; specifically, for condition (34), we have
F + 1. (36)
This is the Deutsch limit corresponding to very large diffusivity which, with
the boundary condition n' = 0 gives a uniform profile (see Figure 6 as Pp
becomes small). In this limit, as WE -*• 0, the mechanical efficiency is pre-
dicted to be zero.
While, in practice, the mechanical efficiency is non-zero, it is very much
smaller than given by eq. (35) for the boundary condition n = 0 and for
precipitator applications the boundary condition n' = 0 is much closer to the
correct one. Solutions for other boundary conditions, for intermediate values
of c in eq. (10) will give results lying between those for the two limiting
cases discussed above. In practice for precipitator calculations, we may use
the results for the boundary condition n1 = 0, being assured that it yields a
(slightly) conservative value for F,. .
156
-------
A most important result of the analysis of the dominant mode, given above,
is that when account is taken of a finite particle diffusivity through the
Peclet number PE = (w£d/D), then the self-similar concentration profile which
is established in ducts with (L/d) » 1, has a minimum at the center and a
maximum at the collector wall. Only in the limit PE -*• 0 (infinite diffusivity)
does one recover the Beutsch result of a uniform profile. The larger is the
Peclet number (the smaller the diffusivity), the steeper is the concentration
profile. This is because migration in the electric field is causing particles
to pile up near the collector and back diffusion has only a weak influence in
counteracting this effect.
At the same time, the efficiency for the dominant mode is given by a
Dautsch-like formula with a factor F, (P£) multiplying the exponent. The
factor F^ tends to unity in the Deutsch limit PE -»• 0 (infinite diffusivity)
as it should. However, for non-zero PE (finite diffusivity) the factor F^
is greater than unity, and increasingly so as PE becomes large (the laminar
limit PE -*• °°) .
Physically, the reason for the increase of efficiency over the Deutsch value
can be seen as follows. For the Deutsch model n(z) is constant and equal to
the mean value "n, while the migration velocity (from eq. (8)) everywhere
equals WE< Thus the flux to the walls is n WE. On the other hand, for non-
zero PE, when the concentration is higher at the wall than in the center,
we have n(d) > n. However, at the wall, with the boundary condition n' = 0,
the migration velocity is still w = WE, so that the flux to the wall is
n(d)wg > n" WE. Thus for a given throughput of ash at a given duct location x,
the piling up of ash near the wall with finite diffusivity causes a correspond-
ing increase in precipitation rate, relative to the Deutsch model for infinite
diffusivity.
Entrance Region Effects - Numerical Solutions
The analysis detailed above, in terms of a normal mode expansion is valu-
able in that the dominant mode gives the limiting efficiency for a precipitator
of large (L/d) for any entrance profile. However, to find a complete solution
for an arbitrary entrance profile n(x = 0,z), this profile must be expanded
in the complete set of normal modes and the total solution found by summing
the contributions of all the modes. The higher modes decay more quickly than
the dominant mode and, as a result, an arbitrary entrance profile will gradually
be transformed into the self-similar profile of the dominant mode, as the flow
progresses down the duct.
As an alternative to this analytic procedure, we have solved the two-
dimensional convective diffusion equation numerically for the case of a uniform
entrance profile
n(0,z) = n (constant) (37)
so as to investigate the entrance region effects.
157
-------
In the analysis above, it was noted that F-j_, and hence the dominant
mode efficiency, is insensitive to the parameter (WE/U) for typical conditions
of interest. It can be shown that this is equivalent to saying that the axial
transport of particles due to turbulent diffusion is negligible compared with
that due to convection by the gas velocity u. Therefore, at this point we
neglect the term in (32n/3x2) in eq. (7), which removes the dependence on
(WE/U).
The resulting equation is solved subject to the boundary conditions
eq. (10) (with c -> °°, i.e. n' =0), eq. (11) and the entrance profile eq. (37).
The problem is solved numerically by a Crank-Nicholson finite difference scheme
in the domain (x = 0,L), (z = 0,d). Once the solution for n(x,z) is found, the
penetration is evaluated from the formula
fd fd
P= un(L,z)dz/ un(0,z)dz, (38)
•'o ' o
and the efficiency is then given by
n(L) = 1 - P . (39)
The problem has two independent dimensionless parameters, namely ?„ = (w^d/D)
and
w L w A
•£--5-- <«>
The latter is the dimensionless precipitator length which appears as the exponent
in the Deutsch-Anderson efficiency formula, eq. (4).
Figure 8 shows the computed efficiency as a function of (wEA/Q) for various
values of PE. On a logarithmic plot the Deutsch equation (PE -> 0) plots as a
straight line through the origin. The solid lines for various finite values
of PE start off asymptotic to the Deutsch line (as they should because the profile
is initially flat) and then bend up, becoming asymptotic to straight lines
having a slope F-^ times the slope of the Deutsch line when the profile becomes
that of the dominant mode. The laminar limit (PE -»• °°) is also shown and shows
f) ->• 100% for a finite length L = ud/wE. Also shown for reference, as broken
lines through the origin, are the efficiencies for the same PE values, which
would apply if the entrance profile were that of the appropriate dominant mode.
It is clear from Figure 8 there is an effective entrance region where the
profile develops from the flat entrance one to that of the dominant mode. The
length of this entrance region increases with PE; for instance for PE ~ 10,
it corresponds approximately to a normalized length (wEA/Q) » 1, or about
L ~ 1 m for typical conditions. Past this entrance region, additional lengths
should yield the full efficiency enhancement due to the multiplier F^ in the
Deutsch exponent.
158
-------
CONCLUSION
In evaluating the practical significance of the analytical and numerical
results derived above, the most important question is what values of turbulent
diffusivity D, and hence P_ can be expected in practice.
For the best case, assuming a smooth wall duct with well-engineered
entrance conditions, so that there are no large-scale eddies and a low level of
turbulence, and assuming that ion wind effects produce a negligible degradation
of the flow quality, then we shall have a developing boundary layer flow in the
front end of the duct which will evolve into a fully developed duct flow for
large x/d. Thus we may take the value of eddy diffusivity established^ for
fully developed duct flow as a standard measure of D that might reasonably be
expected to apply if care is taken to maintain good flow quality.
For fully developed pipe flow in the range of Reynolds numbers of interest,
the eddy diffusivity^ is of the order of 150 V where V = U/p is the kinematic
viscosity. For air at STP, V - 1.5 x 10" 5 m2/s, so that we should have
D ~ 2.25 x 10~3 m^/s. Then taking for typical precipitator conditions the duct
half-width d * 0.1 m, and WE in the range 0.1 - 0.2 m/s, we obtain P£
values in the range 5-10.
Reference to Figure 7 shows that for these conditions F-j_ lies in the
range 2-3. This is a very significant factor multiplying the exponent in
the Deutschefficiency formula. For a given precipitator the fractional pene-
tration should be raised to the power FI relative to that given by the Deutsch
value; alternatively, it should be possible to reduce the specific collector
area by a factor F-^ and retain the same fractional efficiency as predicted
by the leutsch formula.
Reference to Figure 8 for the entrance region shows the penetration
at (wgA/Q) ~ 1.5 reduced by a factor ~3 for Pg ~ 5 and a factor ~6 for
FT? ~ 10 relative to the Deutsch value. The effect is even more dramatic for
larger values of (wFA/Q), corresponding to full-scale precipitators.
It may be noted that for particles of size greater than that corresponding
to the efficiency minimum, field-charging theory yields a value of w£ which
increases with size. Hence Pg increases with size and the increase of
efficiency with size should be more rapid than indicated for a constant P,,
(and F^.
Apart from putting precipitator theory on a much more satisfactory theo-
retical basis, which takes the effect of particle diffusion into account in
a physically satisfying manner, perhaps the most significant conclusion to
be drawn from this work is that the gas flow quality, which has received little
attention in the past, is a crucial factor in precipitator performance.
Moreover, it is indicated that theDeuts-ch efficiency is not the theoretical
maximum that can be expected, but that very much greater efficiencies and
effective migration velocities should be possible if good flow quality can
be engineered.
The fact that efficiencies and effective migration velocities no better
than those predicted from the Deutsch model are often attained in practice
159
-------
indicates that either the flow quality is poor (due to large eddy mixing result-
ing from poor entrance conditions, the presence of baffles or possibly ion wind
effects) or there are other unrecognized major effects degrading the performance.
These include re-entrainment , back-discharge and sneakage, for example. The
question of the efficacy of baffles on the collector plates seems particularly
important. It is usually believed that such aerodynamic structures provide
recirculation zones or regions behind the baffles which protect the precipi-
tated layer from erosion by the main gas flow. However, there seems to be
little concrete evidence for the efficacy of such baffles and it has been
reported10'11 that smooth wall precipitators can have efficiencies as high as
structured ones. This is not unexpected, since such baffles will certainly
introduce large eddies in the near-wall flow and produce a large increase in
the effective turbulent diffusivity. Further investigation of this point is
clearly indicated.
Finally, an interesting deduction from the theory is that in the regime
PE » 1, where (from eq. (32)) 7^ « (P /4) , the efficiency reduces to
»EL
n « i - e*P-<->- <«•>
This shows that the efficiency is independent of duct width (2d) , (other
factors such as the electric field being equal) , and is possibly the explana-
tion of why wide-space precipitators^ work as well as conventional ones , which
is in conflict with the Deutsch model. It also shows that the exponent depends
of the square of w^ rather than the first power, and is inversely proportional
to D in this regime.
ACKNOWLEDGEMENT
This work was supported by the Electric Power Research Institute under
contract number 533-1. The encouragement of Dr. 0. J. Tassicker, Mr. D. Texeira,
Mr. R. Hooper and Mr. P. Gelfan is gratefully recognized.
APPENDIX
To determine the appropriate boundary condition on the particle concentra-
tion n at the collector wall, we model the flow field as composed of two
regions. The main region 1, where D is slowly varying occupies 0 <_ z £ (d-e) ;
the near-wall region 2, where D decreases rapidly, occupies (d-e)< z < d,
where e « d.
We solve the convective diffusion equation (7) in region 2 and require
that both the particle concentration and particle flux match at the interface
z = (d-e):
n = n_ at z = (d - e) , (Al)
1 2
160
-------
. 2
WE nl - Dl IT = WE n2 - °2 IF ' (A2)
If we also require DI = D at the interface, then eq. (A2) reduces to
3n.. 8n9
TT = 17 at z = (d-e) • (A3)
Before solving eq. (7) for region 2, we observe that the z-directed gradients
are much larger than the x-directed ones, and that u -> 0, so we may approximate
eq. (7) for region 2 as
8n „ 9n
»E TF - & ' ° •
-------
Using eq. (A3) we find that
exp
= 0
(A9)
Comparing this with the general boundary condition eq. (10), it is clear that
we have obtained a prescription for evaluating the constant c:
expj ^y
[/
f?' <*" i-
exp f(£M\ ^
T -1
(A10)
If we assume that the diffusion coefficient decays linearly in region 2 from
a value D» at £ = e to a value DR at ? = 0, eq. (A10) can be evaluated
c = d
(All)
Taking as typical values (e/d) -0.1, (w£d/DA) * 10 and
larger, we find c ~ 100 d.
~ 10 or
We conclude from these arguments that c is large compared to d, and
that a reasonable boundary condition for the main region 1 is
=0 at z = (d-e)
(A12)
This boundary condition for turbulent diffusion of particles is quite different
from that for molecular diffusion and results from the rapid decrease in turbu-
lent diffusivity in the near wall region. It is consistent with the Deutsch
formulation and is consistent with the very low values of "mechanical" deposi-
tion of particles from turbulent flows. Of course, to obtain non-zero values
of mechanical deposition rate in the absence of electrical effects (w-g •* 0)
it is essential to allow for a large but finite value of c through a more
detailed study of the dependence of D through the near-wall region. However,
for precipitator applications the limit c •> °°, corresponding to the boundary
condition (A12) is adequate.
162
-------
REFERENCES
1. White, H. J. , Industrial Electrostatic Precipitators, Addison Wesley, 1963,
Ch. 6.
2. McDonald, J. R., "A Mathematical Model of Electrostatic Precipitation,
Vol. I., Modeling and Programming," EPA report 600/7-78-llla (June 1978).
3. Gooch, J. P. and McDonald, J. R., "Mathematical Modeling of Fine Particle
Collection by Electrostatic Precipitation," Symposium on Particulate
Control in Energy Processes, San Francisco, May 11-13 (1976).
4. Takimoto, K., "Wide Spacing E.P. is Available in Clearing Exhaust Gas
from Industrial Sources," EPA Symposium on Transfer and Utilization of
Particulate Control Technology, Denver, July 1978, Session A4.
5. Robinson, M., "Turbulence in Electrostatic Precipitators—A Review of the
Research Literature," Minerals Processing, May 1968, pp. 13-17.
6. Friedlander, S. K., Smoke, Dust and Haze, Wiley, 1977, Ch. 3.
7. Beal, S. K., Nuc. Sci. and Engr., Vol. 40, p. 1 (1970).
8. Davies, C. N., Proc. Roy. Soc., Vol. 289, p. 235 (1966).
9. Kays, W. M., Convective Heat and Mass Transfer, McGraw-Hill, 1966, Ch. 9.
10. Dalmon, J. and Lowe, H. J., "Proc. Int. Symp. on Physics of Electrostatic
Forces and Their Applications," Grenoble, France, 1961.
11. Lowe, H. J., "Reduction of Emissions of Pollutants," Phil. Trans. Roy.
Soc. Lend. A 265, 301-307 (1969).
163
-------
HIGH VOLTAGE
PLATE
r
GAS FLOW
WITH CHARGED
PARTICLES
GAS VELOCITY u
MIGRATION
VELOCITY
W
GROUNDED
COLLECTING
PLATE
WIDTH
— d -
Figure 1. Physical representation of
mathematical model for pre-
cipitation theory.
~I
(o)
IT
(b)
Figure 2. Laminar flow precipitator
(a) particle trajectories
(b) particle concentration
profiles.
Figure 3. Graphical solution of (a) eq. 27 and (b) eq. 29.
164
-------
0.5
Figure 4. Solutions
of eqs. 27 and 29.
n(z) ARBITRARY SCALES
.2 .5 .4 .5 .6 .7 .8 .9 I
(z/d)
Figure i. Concentration profiles n(z) for boundary condition
n(d) = 0 for various values of ?„.
165
-------
n(z) ARBITRARY SCALES
Figure 6.
(z/d)
Concentration profiles n(_z) for boundary conditions
n' (d=0) for various values of P_,.
10
9
8
7
6
5
4
3
2
n(d)=0
n'(d)=0
_i i i
i i
O.I
0.2
0.3
0.4
0.5
Figure 1, The factor F1(1/PE) for the boundary conditions
n(d) = 0 and nf(d) = 0.
166
-------
99.9%
90% -
80% -
70%
60%
50%
.5
1.0
A/Q
1.5
Figure 8. Precipitator efficiency for flat entrance profile
taking into account entrance length effects.
167
-------
THE "HUMAN ELEMENT" - A PROBLEM IN
OPERATING PRECIPITATORS
By:
W,J. Buchanan
American Electric Power Service Corp.
Canton, Ohio 44701
ABSTRACT
We are constantly seeking ways to solve the various problems that make
many electrostatic precipitators unsatisfactory, an effort which is urgently
necessary, but I believe the technical problems are often easier to deal with
than the all too common ones with the "human element" which can be quite
complex and frustrating.
Although people, and the way they do their work, are Call too oftenJ a
problem throughout industry there are some "out-of-the-ordinary" elements
that are of particular concern in operating precipitators, this paper examines
what precipitators require and some ways to improve personnel participation.
168
-------
THE "HUMAN ELEMENT" - A PROBLEM IN OPERATING PRECIPITATORS
INTRODUCTION
In the electric utility business, we have been operating boilers,
turbines, generators, transmission lines and distribution equipment for
decades with the realization that our success hinges upon reliability and
efficiency. If this equipment breaks down, we will not be able to meet our
obligations, and we will be in trouble. Any loss of efficiency means a
financial setback, and that, too, can spell trouble.
About 10 years ago, it became necessary to get serious about reducing
the discharge of ash into the air. Although electrostatic precipitators
had been around for some seventy years, there were a few bugs to be worked
out before these devices could achieve the reliability and efficiency levels
they are capable of obtaining today.
One of the problems in operating precipitators has been the fact that,
in general, people do not understand them. If we are to get the best
performance from precipitators, we must determine why they have been so
misunderstood, and discover how to correct that problem.
KNOW YOUR PRECIPITATOR
Historically, dust collectors have been mechanical devices used
primarily to protect induced draft fans and to incidentally reduce stack
emissions by what seemed, at the time, to be a large amount. They required
a high maintenance effort, and substantial fan capacity.
Since passage of the Clean Air Act, the emphasis has been on reducing
particulate emissions, and we have heightened our efforts with advances
in precipitator technology. Today, the electrostatic precipitator is a very
complex piece of equipment that can achieve efficiencies hardly envisioned
a generation ago — if_ it is properly designed and maintained.
Unless a precipitator is designed to make complete allowance for
specific operating conditions and the ash to be collected, its performance
usually will deteriorate relatively rapidly. Likewise, good internal
electrical conditions must be consistently maintained, and gas flow must
be well distributed.
A major cause for unsatisfactory precipitator performance is ash
fouling of collecting and corona electrodes. Consistently acceptable
performance requires adequate cleanliness — a precipitator that will
perform well when collecting "ideal" ash may fail miserably when subject
to "problem" ash.
169
-------
Proper care of anything or anybody starts with understanding, and
precipitators are no exception. In evaluating a precipitator that is
failing to meet requirements, it is essential to know its potential for
doing the specific job assigned to it. This suggests the importance of
appropriate design, of course, but also hints of the need for monitoring
and record-keeping,
Long-term, continuous monitoring of these big, dirty boxes demands
detailed documentation of their entire operation. Clues to conditions
causing malfunction can be found by analyzing information that is dutifully
kept. Among the needed details are such things as coal and ash analyses,
gas temperatures, unit start-ups, and ambient conditions. It is important
to observe the electrical performance to see if voltage, current, and sparking
are normal. Stack emissions can say a lot about the health of a precipitator.
Rapping losses can be discovered by making a "no-rap" test. All problems
and component failures should be recorded in detail. Thorough internal
inspections must be made when the precipitator is out of service.
The person who is keeping watch on precipitator performance must be able
to distinguish between a process change and the deterioration of equipment.
The most intriguing thing about precipitators is the complex interaction of a
number of factors. Unfortunately, there are many among those who operate,
maintain and manage plants where precipitators are used who are not equally
intrigued. It seems that few of them really enjoy solving the mysteries of
good precipitator operation --- perhaps it is easier to accept the first
explanation for a malfunction that comes to mind than to think up the actual
cause. This attitude leads to looking at the precipitators only when some
very obvious condition makes it mandatory.
THE PEOPLE PROBLEM
Many individuals still have not accepted the idea that electrostatic
precipitators are as important to the operation of a plant as the turbine or
steam generator. They represent a major investment, and add significantly
to operating costs. They were installed to reduce emissions, as required by
law. Their reliable, efficient performance requires alert, intelligent
attention. When a precipitator malfunctions, it is then too late to begin
keeping records and become familiar with the operation of this equipment.
At that point, it becomes a kind of panic situation where nobody knows what
to do, and the person who must do something realizes that he should have
taken a more active interest in the precipitator all along. I am not sure
how many crises need to occur before people begin to follow consistent
monitoring programs.
WHO'S RESPONSIBLE?
The general attitude of everyone in a power plant is influenced most by
the manager. His personal manner and technical abilities will earn him
respect, but unless he makes it clear that he wants the plant to be run
efficiently and monitored closely, the typical employee at any level will do
"only what he has to". The manager needs to emphasize the importance of the
170
-------
precipitator by assigning capable, experienced people to supervise it.
Giving the responsibility to the least experienced worker says just the
opposite.
From the manager to the individual physical employee, perhaps nothing
gets work done better or more quickly than keen interest. It provides the
incentive to think, and solve problems, much more so than money. (People do
not become creative for material rewards.) It makes the clock move quickly.
Every employee in a plant needs to know that the manager really cares
about what he C°r shej is doing. And when an employee knows the boss is
running the show, there will be greater accountability. A person in a
managerial position cannot be expected to know every detail about a task he
has delegated, but he needs to know the right questions to ask and how to tell
when he is getting the right answers. Only then can he really know when
something is slipping and what is being done to make corrections. Only then
can he really show that he really cares about what his workers are doing.
Nearly everyone is interested in anything that makes a job easier,
People may not relish taking a lot of seemingly useless readings, hunting
for grounds and broken wires, and crawling around in dirt looking for
answers to uncertain problems. But they may find that easier and more
interesting if they are given clear explanations of what the electrical
readings mean, told why wires break, and are helped to interpret the
conditions that they find inside a malfunctioning precipitator. Most people,
in other words, would like to do the right things — if they only knew what
they were and understood how to go about them.
LEADING THE WAY
It would help if there were uniform practices at all plants, but there
are individual preferences on how to maintain precipitators and how to
assign responsibilities. In some plants, an overall supervisor coordinates
records, inspections, periodic maintenance, monitoring and trouble-shooting.
In others, there is no one to tie it all together. When each department
looks after a limited area — with little concern for the "big picture" —
this makes it difficult to put all the pieces together and solve a problem.
There needs to be better coordination of the operation, maintenance, and
trouble-shooting aspects of precipitators if plants are ever going to really
get "on top" of the situation.
Someone — whether that is one of the plant people or some "outsider"
--- will have to instruct all appropriate plant employees on what they need
to know about precipitators. There should be written instructions and
sessions on making inspections both in and out of service, on how to take
electrical readings and other data, on how to look for grounds and their
causes, and on how to record other useful information.
When several plants are involved within a company it is advantageous
to set up a central coordinating office to monitor precipitator operation
and maintenance. Improved communication — even between only two competent
individuals assigned to this work can accomplish a lot,
171
-------
Likewise, central-office engineers can do many things to improve the
quality of precipitator operation and maintenance, including the development
of standardized reporting forms, assisting in personnel training, trouble-
shooting, interpreting operating data, and helping with routine inspections.
When a system has several different make precipitators and a variety of
operating conditions, the central-office people are in a position to have a
much broader understanding which can be of particular advantage in a problem-
solving situation.
Still another group exists whose members exert a significant influence
on the health and well-being of precipitators: the manufacturers' field
service engineers. Their job is to make pre-operational inspections to insure
that assembly is correct, to set up controls for proper limits and action,
to energize the transformer rectifiers and discharge system, to tune-up the
entire installation under operating conditions — and to instruct the user's
people.
Many service engineers are able to do a reasonably good job of checking
out the equipment initially and getting it in operation properly, but too
many too often add to later confusion by misleading the operators and main-
tenance personnel with inaccurate or false ideas on fundamentals and problems.
Admittedly, precipitator manufacturers have had a monumental task in
staffing their field positions. In a relatively short time, they have had
to hire and train a great number of individuals. There is keen competition
in the industry with high rates of turn-over. Performance problems, however,
often are misinterpreted by inexperienced field engineers who are unable to
effect proper corrections or communicate effectively with competent technical
back-up people in their home offices. More should be done to assure the
technical competence of this important group.
CONCLUSION
Electrostatic precipitators can make significant contributions toward a
cleaner environment, and are here to stay. But their reliability and
efficiency depends in large measure on proper design and adequate maintenance.
Indifference among those who operate and maintain precipitators must give way
to interest. Management can and should cultivate a better attitude among
those involved in this challenging aspect of electricity generation by
providing both leadership and encouragement by example.
172
-------
ELECTROSTATIC PRECIPITATORS - ELECTRICAL PROBLEMS AND SOLUTIONS
By:
Ronald K. Raymond
Pennsylvania Power $ Light Company
Allentown, Pennsylvania 18101
Efficient collection of flyash by an Electrostatic Precipitator is very
sensitive to electrical operation. Most problems that occur in an
Electrostatic Precipitator quickly cause changes in the normal operating
electrical readings. Broken wires, poor mechanical clearances, and
inadequate rapping are examples of these problems. Any one of these
problems will cause the electrical readings (hence the precipitation
efficiency) to degrade.
This paper tells about recent experience gained at Pennsylvania Power §
Light Co. in analyzing and correcting electrical problems associated with
T-R control circuits, electrical sectionalization, support insulator
failure, wire failure, and T-R set metering. These topics cover the common
areas of electrical operation in an Electrostatic Precipitator.
173
-------
ELECTROSTATIC PRECIPITATORS - ELECTRICAL PROBLEMS AND SOLUTIONS
INTRODUCTION
About 4 years ago PP§L established a team of engineers to study
various factors affecting flyash electrostatic precipitator performance.
Among the various factors studied were:
1. Fuel characteristics
2. Flue gas conditioning
3. Rapping systems
4. High voltage control systems
5. Electrical sectionalizing
6. Mechanical design features
In order to study the effects of these various factors on
electrostatic precipitator performance they were altered in a controlled
manner and the change in precipitator performance recorded. While
particulate emission testing was occassionally used for performance
evaluation, most of the time the opacity monitor data was used for judging
a performance change.
This study program has resulted in PP§L undertaking a program to
modify some of its electrostatic precipitators to improve their collection
ability. This paper deals with the electrical portions of this
investigation.
HIGH VOLTAGE CONTROL SYSTEMS
PP§L has, throughout its system, weighted wire design electrostatic
precipitators supplied by a number of different vendors. All of these
units have wire failures to one degree or another. Two factors are
primarily responsible for the amount of wire failures. They are the high
voltage control system and the design of the wire emitting frame system.
A typical control system is shown in Figure 1. It has three
components, the control element, the control logic, and the TR set. The
primary function of this system is to keep the high voltage as high as
possible while preventing excessive sparking and power arcing. Primarily,
two devices are used for the control element. They are the saturable core
reactor and the silicon control rectifier (thyristor).
The saturable core reactor is basically an inductor whose core
material may be saturated magnetically by passing a DC current through a
control winding. When the core is unsaturated, the coil has a high voltage
drop across it. When it is saturated, it has a small voltage drop across
it. This device limits the peak as well as the RMS voltage applied to the
TR set.
174
-------
The silicon control rectifier (thyristor) is a solid state switch. It
conducts with very little voltage drop when turned on and presents an open
circuit when it is off. As such, this device will not limit the peak
current to a TR set, it will only control the RMS value. To get around
this effect, thyristors are used with a series inductance. This inductor
reduces the amplitude of current peaks and tends to broaden the current
wave shape.
Typical current wave shapes for a saturable core reactor are shown in
Figure 2 and 4. A thyristor wave shape is shown in Figure 3.
The control logic circuit senses the operation of the TR set and
attempts to adjust the primary voltage for stable maximum kV operation. As
Figure 1 shows, the logic can control using one or more signals obtained
from various points in the system. Some control manufacturers claim higher
spark sensing ability on the secondary side of the TR set then the primary.
In practice, at PP§L, no appreciable difference in sensitivity was observed
between control systems of each design.
Two types of operation were observed on PP£L operating control
systems. They are the "spark rate" and the "spark sensing" types of
operation. The first type integrates sparking as it occurred in the
precipitator. This signal is compared to a set point value and the TR
primary voltage is adjusted to keep the spark rate constant. The second
system of control increases TR-set primary voltage until a spark occurs or
a maximum value for primary current has been reached. After a spark
occurs, the primary voltage is removed for several cycles of AC voltage.
As the precipitator recovers from the spark, the primary voltage returns to
a value equal to or slightly less than the value it had before the spark
occurred.
The spark rate (first type) control had to be constantly adjusted in
order to prevent wire failures. The second type of control logic was much
more effective in controlling the TR set automatically with little or no
wire failures. To see the effect of the two logic systems on precipitator
operation, two units at PP§L's Martins Creek plant were chosen. The
control logic on Unit 1 and Unit 2 was monitored for one month. Then the
logic on Unit 1 was changed from the "spark rate" type to the "spark
sensing" type. The readings were again recorded for one month. Averages
for primary voltage and current were calculated for each unit before and
after the change. This data is presented in Table 1.
Since both units burn essentially the same type of coal, any change in
operating conditions due to fuel changes would be indicated by a change in
the Unit 2 average before and after modification. Table 1 shows there was
a slight change in the readings due to fuel changes. However, a very
significant change was observed in the readings for the unit whose control
logic was replaced. Even though the readings were higher, there were no
wire failures due to high voltage arcing. It should be noted that the
wires in these units were in poor mechanical condition and that since these
175
-------
tests they have been replaced. Also, the control systems installed for
these tests have been permanently installed on Unit 1 and 2 at Martins
Creek.
The third element in the control system is the TR set. Other than
failure due to improper operating conditions, no difficulties have been
experienced.
ELECTRICAL SECTIQNALIZING
The Martins Creek Unit 1 electrostatic precipitator was tested to
determine the effects of various electrical sections on Unit operation.
This precipitator has five electrical sections connected in series in the
direction of gas flow. Test procedures were designed to obtain electrical
information on various electrical sections and their importance in the
precipitator dust collection process. The effects of additional TR sets as
well as their position in the precipitator were studied. Electrical
sections were deenergized to simulate various SCA's. This was done in an
effort to apply the results to other units as well as the Martins Creek
Unit 2.
Initial testing at Martins Creek on the Unit 1 precipitator consisted
of taking one section of the Unit out of service. This was done using a
systematic procedure—one section at a time from outlet to inlet—one side
of the precipitator at a time. Voltage, current, and opacity readings were
recorded. Figure 5 shows the layout of the Unit 1 precipitator. The
systematic procedure used, referred to as a double response test, was done
by removing sections in an orderly way such as E., D , C , B , A , then E ,
DB, C , B , A . Note, at any time only one section was out of service.
Any cnanges in precipitator operation were attributed to loss of the
deenergized section.
The data recorded during the sectionalizing study had several
limitations in accuracy and validity. Several limitations apply to the
voltage, current and opacity data.
The current and volt meters on the precipitator electrode control
panel stuck at various points in their measurement range. This made exact
readings difficult. The meters had to be tapped gently to obtain a true
value. Also, the meters used to record voltage and current values are
average reading meters, calibrated for RMS sine wave response. Voltage and
current wave shapes at the output of the silicon control rectifiers are not
sine wave. To determine true RMS values for these readings, a shape factor
must be applied. It is virtually impossible to calculate this factor
because it is different for different operating points on the TR set.
Therefore, the readings obtained show relative increases or decreases - not
absolute levels.
Secondary meter readings, kV and mA, are related to the primary
voltage and current. Tests on units equipped with primary and secondary
176
-------
metering show that the relationship between primary and secondary is not
linear. It appears that no constant factor, such as transformer turns
ratio, can be used to relate the primary to secondary. Thus, in a control
cabinet that does not have secondary metering, such as the outlet cabinet
on the Martins Creek Unit 1 precipitator, it is impossible to calculate
one. Qualitatively, it may be stated that as primary values increase, so
do secondary.
The opacity monitor shows that opacity changes occur when a section of
precipitator is being energized. Rapping spikes and a shift upward in base
line values indicate a section has been deenergized. Opacity data
indicates very rapidly a change caused by a section being removed from
service. This data, while useful for changes in any particular short term
test, should not be used for long-term comparisons. Factors such as
particle size distribution and boiler operation (excess air to name one
important parameter) can affect the opacity without changing particulate
emissions. The monitors presently installed on the test unit exhibit a
long-term stability problem. Stack zero checks during different outages
indicate a random drift.
An example of the opacity data obtained during a typical test is shown
on table 4. The outlet fields, D and E, had three electrical buses
connected to them during this trial, see Figure 6. The other sections were
in groups of two. When either D or E was removed from service, more area
was de-energized than from the others. This fact, as well as the fact that
each of them was an outlet, made their effect on opacity more severe than
the other sections.
Other effects were noted in opacity data. When an outlet field was
de-energized, rapping spikes were more prevalent then when the other fields
were de-energized. In almost all tests there was a general decline of
opacity during the tests. Probably, this was due to air sweeping which
occurs during the de-energized test periods.
During testing of the Unit 1 precipitator there was an increase of the
voltage and current readings from initial to final values. Subsequent
tests showed that the increase was due to its electrode controls. They
lacked the ability to keep power levels at optimum values. Table 2 shows
the results before and after installation of new control logic. This table
shows that the controls were responsible for this change in base line data.
Note the absence of the base line shift in the data taken with the new
control logic installed. As the double response tests progressed, the sum
of individual currents was larger than the current with both were connected
(i.e., A. +• A was greater than A - see figure 5). This result was
observed after the new control logic was installed. Therefore, the control
was eliminated as a possible cause for the effect. Table 3 shows these
results. In all but one case the amount of this increase was about 20
percent or higher with little or no sparking.
177
-------
This increase indicates that increase in current input could be
expected from additional electrical sectionalizing. This idea is not new.
It has been expressed by Harry J. White in his book "Industrial
Electrostatic Precipitation". Opacity measurements made during these tests
show opacity spikes are more pronounced when all outlet fields are removed
from service. A conclusion made from these two results is that improved
precipitator performance can be obtained by increasing outlet
sectionalizing.
If present precipitator configuration consists of two sections
electrically tied together and energized by one TR set, then better
performance can be expected if the sections are electrically separated and
each energized by its own TR set. The most significant change from this
type of sectionalizing will occur when an outlet field is separated.
These results are used in one of PP§L's upgrading plans. As part of
this plan, additional TR sets are being added to the outlet fields.
GENERAL COMMENTS
As previously stated, secondary readings for TR set are difficult to
calculate from the primary current and voltage readings. An observation
made during the test program was that TR sets equipped with secondary kV
and mA meters were more useful in determining precipitator operation than
those without these meters. PP§L is installing secondary meters on most of
its precipitator installations.
Another area of frequent difficulty is the emitting frame support
insulator compartments. These compartments protrude above the roof of the
precipitator. Because of this they provide a potential cold spot. Acid
mist and flyash can collect on the support insulators. This can cause
electrical tracking on the insulators. A heating system is usually
installed on the ventilation system that feeds this compartment. This
heating system is being made larger to improve reliability and capacity.
Also, PP§L is studying insulation methods for these compartments.
One final area of difficulty is alignment of the high voltage emitting
frame. PP§L has added stabilization insulators in the bottom of the
emitting frame to help hold it in alignment. This alignment frame and the
wire to plate spacing has been carefully checked and adjusted to the design
criteria.
CONCLUSIONS
The results presented in this paper have been used to formulate part
of a particulate compliance program at PP§L. Some existing precipitators
are being modified to improve their performance. Others need additional
polution control equipment. Modification of existing equipment has
included replacing the high voltage control systems, adding T-R sets, and
improving the emitting frame stability and alignment. As a monitor,
178
-------
secondary metering has been added to evaluate performance and the condition
of the percipitator during operation.
of the percipitator during operation
These modifications were made on a pulverized coal fired boiler.
Their relative effectiveness on other types of installations has not been
investigated. As an example, it may be said -- in a general sense -- that
at installation where heavy sparking occurs the "spark sensing" control
will generally work better then the "spark rate" control. However if the
precipitator does not spark heavily than the control system is not very
important.
These modifications and additions will not change a precipitators
performance from bad to perfect. However, these changes will improve
performance to some degree and assure the existing precipitator is
operating at its maximum capability.
RKR #378:5
179
-------
PNR
IN.
SENSE INPUT
CONVENTIONAL HIGH VOLTAGE
ELECTRODE CONTROL SYSTEM
FIG. 1
180
-------
0.2 VOLT
CM
2 M SEC/DIV.
SATURABLE CORE REACTOR
TR SECONDARY CURRENT WAVE SHAPE
(3o - 4a)
FIG. 2
1 VOLT
CM
2 M SEC/DIV.
SILICON CONTROL RECTIFIER
(THRISTOR)
TR SECONDARY CURRENT WAVE SHAPE
(5d - 6d)
FIG. 3
181
-------
1 VOLT >
CM
.._
2 M SEC/DIV.
SATURABLE CORE REACTOR
TR SECONDARY CURRENT WAVE SHAPE
(3d - 4d)
FIG. 4
182
-------
GAS FLON
SIDE B
SIDE A
>f\t
GflS
FLOW
B
t t
OPACITY
MONITORS
GAS
FLOW
MARTIN'S CREEK SES
UNIT *1 PRECIPITATOR
FIGURE 5
183
-------
GAS FLON
SIDE B
SIDE A
•B
•A
>B
A
A
GflS
FLOW
B
I I
OPflCITY
MONITORS
GAS
FLOW
MARTIN'S CREEK SES
UNIT *1 PRECIPITATOR
FIGURE 6
184
-------
EFFECTS OF CONTROL MODIFICATIONS
UNIT I
UNIT II
TR SET
A
B
C
D
A
B
C
D
PRIMARY VOLTAGE (VOLTS)
BEFORE AFTER %
175
201
186
193
187
183
177
202
191
240
234
259
189
194
186
219
9%
19.4%
25.8%
34.2%
1.1%
6.0%
5.1%
8-4%
PRIMARY CURRENT (AMPS)
BEFORE AFTER %
54
53
53
61
44
38
46
60
63
82
84
110
36
53
50
63
16.7%
54.7%
58.5%
80.0%
1*.2%
39.5%
8.7%
5.0%
TABLE 1
'E" TR SET DID NOT CHANGE ON EITHER UNIT.
185
-------
BASE LINE SHIFT
START
FINISH
TRIAL I
TRIAL II
TRIAL III
NEW
CONTROL
LOGIC
SECTION
A
B
C
D
E
B
C
D
E
A
B
C
D
E
A
B
C
D
E
VOLTAGE
190
220
240
200
370
240
240
230
380
200
230
250
255
370
190
230
240
295
360
CURRENT
55
50
88
69
85
65
85
87
90
65
65
100
105
130
55
60
66
135
90
VOLTAGE
220
240
260
265
385
250
265
250
375
290
260
280
290
360
190
230
235
290
370
CURRENT
75
70
110
109
100
72
104
100
80
145
100
130
138
130
55
60
65
125
85
TABLE 2
186
-------
PRIMARY CURRENT (I ) &
PRIMARY VOLTAGE (Vp)
DATE
SECTION
11/25/76
12/8/76
1/6/76
NEW
CONTROL
6/18/76
A
B
C
D
E
A
B
C
D
E
A
B
C
D
E
A
B
C
D
E
THE LISTED
SECTION ON
V
_E
175
200
180
210
355
200
230
210
200
370
172
200
180
170
380
190
230
240
295
360
BASE
I
50
42
40
62
70
70
80
76
84
110
48
41
39
37
80
55
60
66
135
90
V
175
200
160
200
330
200
240
200
250
340
200
190
205
190
375
170
210
200
235
370
A SIDE
I
25
39
30
54
35
60
70
55
80
45
48
38
38
35
40
20
21
32
54
50
OUT
V
_2
180
195
200
230
350
225
270
230
220
360
200
225
200
190
360
190
215
220
275
365
F
THE LISTED
SECTION ON
B SIDE
I
_£
38
32
39
64
20
70
85
76
85
70
47
41
38
40
40
45
46
50
105
58
OUT
I
_£.
63
71
69
118
55
130
155
131
165
115
95
79
76
75
80
65
67
82
159
108
SUM OF
SECTION CURRENTS A
SIDE PLUS B SIDE
% OF BASE
126
169
172
190
78
185
194
172
196
105
198
193
195
203
100
118
111
124
117
120
TABLE 3
187
-------
OPACITY CHANGES
SECTION
A
B
C
D
E
A SIDE OUT
OPACITY
A B
20%
20%
24%
55%
6%
7%
8%
17%
B SIDE OUT
OPACITY
A B
10%
10%
10%
16%
10%
10%
10%
60%
BASE LINE OPACITY -
A
B
START
14%
FINISH
10%
6%
DECREASE IN OPACITY FROM START TO FINISH
TABLE 4
188
-------
ELECTRODE CLEANING SYSTEMS:
OPTIMIZING RAPPING ENERGY AND RAPPING CONTROL
Author: Michael Neundorfer
Wm. Neundorfer Co., Inc.
Cleveland, Ohio 44124
189
-------
ELECTRODE CLEANING SYSTEMS:
OPTIMIZING RAPPING ENERGY AND RAPPING CONTROL
Methods for improving Electrostatic Precipitator performance by increasing
electrode excitation level during rapping, and by optimizing rapping con-
trol, are presented in this paper. Design modifications can reduce mechan-
ical impedence to vibration transmission during rapping. These same modi-
fications can eliminate areas of high stress concentration where fatigue
failure often occurs. Rapping system control parameters are presented as
they relate to variations in field collection and shedding rates. Anti-
coincident rapping control for plate rappers minimizes peak opacity levels
during rapping if sufficient "rest times" between raps can be maintained.
When anti-coincidence must be sacrificed due to insufficient rest times on
large installations, suggestions for anti-coincident rapping by section are
presented. Methods for optimizing field rapping repeat rates using available
opacity, ash pull, and precipitator power information are discussed.
Michael Neundorfer
Vftn. Neundorfer Co., Inc.
190
-------
ELECTRODE CLEANING SYSTEMS: OPTIMIZING RAPPING
ENERGY AND RAPPING CONTROL
INTRODUCTION:
traditionally, the functions of dry type industrial electrostatic pre-
cipitators have been described as follows:
1) To charge the dust particles
2) To separate the charged particles from the gas stream by attraction
to grounded collecting electrodes
3) To transfer the separated particles from the collecting electrodes
to external containment with minimal loss
In addition to these three functions, a fourth function, dust agglomera-
tion, will be discussed in this paper. Dust agglomeration is critical for
effective precipitator performance since, without agglomeration, nearly all
of the dust removed from the collecting electrodes would remix with the gasses
and be carried downstream. (Plato, 1969)'. Even with agglomeration, trans-
fer losses (rapping reentrainmenti account for up to 70% of total precipitator
emissions (Gooch, Marchant, 1978)
This paper addresses the relationships between electrode cleaning and
precipitator functions and offers suggestions for physical and operational
changes in electrode cleaning which will improve precipitator efficiency.
CHARGING, SEPARATION AND FOULED ELECTRODES
Electrostatic precipitators operate best with cjaan electrodes. This
is especially true with dust resistivities exceed 10 ohm-cm. Dust build
up on collecting plates degrades precipitator performance in three ways:
1) Given a constant average corona voltage level, corona current density
decreases as the thickness and/or resistivity of the dust on the
plates is increased. As corona current density is reduced, corona
power is reduced. The rate of collection can be related to corona
power as follows:
w = k, Pc/A w = rate of collection
K, = constant
PC = Corona Power
A = Collecting surface area
Therefore, a reduction in corona power reduces the rate of collection
which reduces the probability of capture which, in turn, reduces
overall collection efficienty. In extreme cases, improved rappjng
has increased corona current by more than 6 times. (Neundorfer) .
2) As dust layer thickness and/or resistivity is increased, sparkover
voltage is reduced. As sparkover voltage declines, the precipitator
must be operated at a lower voltage to avoid excessive sparking.
This means a reduction in corona power and, therefore, a reduction
in overall collection efficiency. The reduction of sparkover voltage
results from localized electrical breakdown of the dust layer. All
corona current must pass through the resistive dust layer to ground.
The resulting voltage drop across the dust can reach 20KV and follows
ohms law:
191
-------
V., = Voltage drop across resistive dust
j = average corona current density
Q = dust resistivity
1 = dust layer thickness
Sparkover originates at points on the plates where electrostatic re-
sistance values are lowest. These points of low resistance are thin
spots or discontinuities in the dust layer.
3) Back discharge neutralizes negative ions and charged dust particles.
Back corona discharge on the exposed surface of the collecting elec-
trode dust layer can produce positive ions. These ions are attracted
to and migrate towards the discharge electrode and interfere with
normal particle charging and migration.
These three factors indicate that from an electrical standpoint, it
is desirable to maintain the dust layer on collecting electrodes as
thin and as uniform as possible. Build-up on the discharge electrodes
of low resistivity dust will increase the effective electrode diameter
and therefore increase the corona starting voltage. High resistivity
deposits on wires will decrease the corona current at a given voltage.
Discharge electrodes should always be kept clean.
Figure 1. Fouled discharge electrodes: Flyash application with SO-
conditioning. >
192
-------
Fouled discharge electrodes: Cement kiln: High resistivity dust.
TRANSFER, REENTRAINMENT AND AGGLOMERATION
Transfer of the collected dust from the electrodes to the hoppers
is initiated by rapping. Rapping is the direct or indirect mechanical exci-
tation of electrodes. If rapping is effective in shearing the dust from the
plates, the dust will either: 1) reattach itself to the electrode, 2) fall
into the hopper or 3) reentrain in the gas stream for downstream re-separa-
tion or discharge through the stack.
Reentrainment is the recapture of collected dust by the gasses. It
occurs after dust has been attached to electrodes or to other structures,
such as walls and ceilings. Dust also can be reentrained after it has en-
tered the hoppers. Two conditions are required for reentrainment: (1) a
disturbance which dislodges the dust and (2) gas velocity sufficient to carry
the dislodged dust.
The most obvious disturbance which dislodges dust is rapping which ac-
counts for the majority of reentrainment. Sparking can also dislodge dust
from electrodes as can high gas velocity and the sheer weight of a dust layer.
Studies have found that the percentage of emissions due to rapping re-
entrainment increases with increasing particle size. This may be explained,
in part, by the fact that smaller particles agglomerate into larger ones from
inlet to outlet of the precipitator (Spencer, 1976) In these studies, emis-
sions due to rapping losses of particles less than 0.3 um were not signifi-
cant while rapping losses of particles 2.0 um and larger accounted for up to
90% of those particles emitted (Gooch, Merchant, 1978). The total contribu-
tion to precipitator emissions of rapping losses has been measured in a range
between 18%fland 9Q% (Gooch, Marchant (1978), Spencer (1976), (Nichols, et.
al. (1975). These findings demonstrate the potential for overall efficiency
improvement which is possible by reducing rapping losses.
193
-------
As agglomeration of particles into larger particles or "chunks" is in-
creased, the influence of the gas flow on the dust agglomerates is decreased
resulting in a higher probability that rapped dust will be removed to external
containment. Agglomeration can be enhanced through proper rapping control,
process changes, and gas conditioning.
DUST ADHESION AND TENSILE STRENGTH
In order to understand the amount and type of work required to dislodge
dust from collecting electrodes (plates), the conditions affecting dust layer
tensile strength and adhesion must be examined. These conditions are: parti-
cle dielectric and work function properties, dust deposit density, particle
size, gas humidity, interelectrode current density and field strength.
(Tassicker, 1975) .
Electrical forces of adhesion due to field strength and current density
compact the dust layer and hold the dust to the plates. Electrical adhesion
forces increase proportionally as the current density increases. Electrical
adhesion forces also increase with increasing dust resistivity. If electri-
cal adhesion forces were the main component of dust adhesion, very light rap-
ping, with momentary "power off" in the field being rapped, would remove high
resistivity dust. However, since electrical adhesion forces generally are
not the main component of dust adhesion, success with "power off" rapping is
limited. When it is successful, long periods of continuous rapping with
"power off" are required to make measurable improvement in cleanliness.
Although electrical forces of adhesion tend to compact the dust layer
and hold it against the plates, contact potential adhesion holds the dust
layer together. Although the total particle surface charge equals zero, sur-
face charge (work function) may vary in value around the particle surface.
As particles are collected, they are aligned by the field forces and surface
potential differences from particle to particle cause coloumb force adhesion.
The contact adhesion forces are significant for particle sizes less than S.Oxim
(Penney, 1975). In general, fine particles have much higher adhesive char-
acteristics than larger particles.
While the precipitator field force contributes to the alignment of par-
ticles for contact potential adhesion, other influences, such as rapping and
disturbances of the dust layer by newly deposited dust, rearrange particle
positions in the layer. The rearrangement causes higher layer density and
stronger electrostatic and mechanical adhesion. This is supported by the
observation that dust which remains after the first rap has higher tensile
and shear strengths than newly collected dust (Spencer, 1976).
The relationship between dust layer tensile strength and layer density
has not been developed. However, the relationship between dust layer mass/
unit area of pUte and rapping effectiveness is well established (Gooch,
Marchant, 1978) , (Spencer, 1976) , (Sproull, 1965) . With voltage ap-
plied, Plato found that 90% removal was possible with dust mass/unit area
of 0.5g/cm while removal,efficiency with dust mass/unit of .07g/cm was
only 30% (Sproull, 1965). Plato also found that at values up to about
O.lg/cm the dust formed a cloud during rapping. When the dust mass/unit
area exceeded O.lg/cm it caked when rapped. (Sproull, 1965) Therefore,
194
-------
rapping higher mass/unit area dust layers caused less reentrainment.
If rappers are effective, and if precipitator inlet conditions are con-
stant, plate dust layer mass/unit area at any point along the gas flow can
be approximately expressed as "time between raps". This is true until the
dust layer begins to degrade the processes of charging and separation.
THE MECHANICS OF PLATE RAPPING
Rapping effectiveness for a given dust condition is related to plate
acceleration. Plate acceleration refers mainly to acceleration of the sur-
faces of the plates rather than to the whole plate or group of plates as a
unit. Most of the acceleration is plate surface distortion resulting from
the disturbing force of rapping. Peak acceleration normal to the plate sur-
face is primary in dislodging dust; however, simultaneous sbaar acceleration
contributes to cleaning (Spencer, 1976), (Sproull, 1972). In comparing
shearing acceleration with normal acceleration, Sproull found that it took
twice as much shearing acceleration as normal acceleration to shear a given
quantity of dust (Sproull, 1965).Iy
When plates are rapped the dust shears from itself rather than from the
plate surface leaving a residual layer of dust. In order for dust to break
away from this residual layer in large chunks, some sliding must take place.
For sliding to occur the static coefficient of friction must be overcome;
the lower value kinetic coefficient of friction resists sliding. As sliding
speed increases, the value of the kinetic coefficient of friction becomes
smaller.
Peak plate acceleration resulting from rapping can be measured by attach-
ing low mass piezoelectric accelerometers to different points on the plates.
Measurements taken by single axis transducers measuring normal acceleration
can be most consistently related to rapping efficiences. Sustained minimal
peak normal accelerations on the plates of 10-60 g's (0 to peak) will gener-
ate good dust shearing. If single impact or low frequency cleaning is em-
ployed, higher peak acceleration can be required for adequate cleaning.
DESIGN PARAMETERS AFFECTING EXCITATION
For a given rapper disturbing force, mean peak acceleration of the plate
is affected by the following conditions: plate size, shape and rigidity; the
number of plates per rapper; the method of plate suspension; and the location
and method of transmitting rapping force from the rapper to the plate. The
presence of a dust layer on the plates during rapping will reduce plate ac-
celeration.
As plate designs become longer and wider, it becomes more difficult to
attain uniform excitation of the plates. The difficulties of rapping larger
plates has been overcome to a degree by rapping fewer plates with a rapper.
The area of plate per rapper is a good indicator of potential rapper effect*\
iveness. Most?recent weighted wire installations range from 74.3nr (800 ft '
to nearly 465m (5000 ft ) of plate par rapper. European designs are moge
conservative with from less than 27.9nr (300 ftc) to about 83.6nr (900 ft )
per rapper.
195
-------
Minor field modifications to plate suspension members and rapper anvil
beams can more than double minimum peak plate accelerations. In some cases,
relocation of the rapper with respect to plate beam suspension points has
tripled minimum peak acceleration. The use of a common plate frame suggort/
rapper shaft reduces rapper effectiveness significantly. (Neundorfer) .
Transmitting rapping energy from the rapper or point of impact into the
plates requires mechanical linkage. When this linkage is loose, substantial
energy is lost and destructive wear and fatigue can occur. Tapered inter-
ference fits at joints in the linkage can eliminate these problems.
TYPES OF RAPPERS
There are two basic types of rappers: (1) single impact rappers and (2)
frequency rappers. Single impact rappers set up natural frequency vibration
which is maintained entirely by elastic forces in the structure. Frequency
rappers impart a periodic disturbing force which results in steady state
forced vibrations.
Single impact designs''include mechanical drop hammers, magnetic impulse,
and pneumatic single impulse rappers. When plate free vibration is generated
by single impact types, the frequency depends only on the plate/rapper sys-
tem's physical configuration and the initial loading. The only adjustment
available on single impact type rappers is the variation of impact level of
the initial loading. Some magnetic impact rappers impact more than once per
rapping cycle. The frequency of these rappers (less than 10 cps) is too low
to initiate forced vibration in the plates. Therefore, they are classified
with single impact rappers.
Frequency rappers include electric vibrators, eccentric pneumatic vi-
brators (not widely used), and pneumatic frequency impactors. Frequency
rappers operate at frequencies between 25 and about 90 cps. In a rapper re-
entrainment study at Bull Run, plate acceleration in three axes was measured.
Acceleration was highest in the normal axis with frequency rapping. Using
single«impact rappers, normal acceleration was lowest. (Nichols et. al.,
1975).d{
Electric vibrators operate by drawing an anvil against a stop at a given
frequency with variable force. The displacement and anvil mass are small
relative to other rapper types. Therefore, electric vibrators are seldom
used for difficult dusts.
Pneumatic frequency impactors work like a jackhammer. A piston weighing
between 1526g (3% Ibs.) and 5230g (12 Ibs.) is pneumatically reciprocated in
a cylinder with a relatively long stroke. The piston impacts at the lower
end of the stroke and is air-cushioned at the top. Frequency and amplitude
are varied by changing air pressure. Pneumatic frequency impactors simultan-
eously generate both natural frequency and forced vibrations in the electrodes.
Pneumatic frequency impactors are widely used for salt cake and other diffi-
cult dust. Recently, these rappers have been used to upgrade fly-ash rapping
where high resistivity dusts and/or gas conditioning exist.
At the point of highest peak acceleration (near rapper),the main influence
196
-------
to plate acceleration is the free vibration caused by each impact. At the
points farthest from the rapper, transient motion in the first few cycles is
gone and a lower energy natural frequency is sustained.
DISCHARGE ELECTRODE RAPPING
Discharge electrode rapping is accomplished with both single impact and
frequency rappers. There is little published regarding wire rapping effect-
iveness. This is probably because: (1) discharge electrode rapping contri-
butes little to reentrainment emissions and (2) charging and separation de-
gradation due to discharge electrode fouling is less obvious than degradation
due to plate fouling. Some installations of hot-side fly-ash precipitators
have experienced very difficult wire dust build-up problems. These have been
solved by tightening wire rapper transmission assemblies and using frequency
impact rappers with relatively high energy, short duration (less than 0.4 sec)
bursts. This approach produces a high level of wire excitation for a short
period of time. The high level of excitation assures that all wires on the
frame vibrate. The short duration minimizes chances for wire mechanical fail-
ure.
Discharge electrode frames are supported with a minimum number of steel
pipes or bars in tension. These members are supported by ceramic or porcelain
insulators on the top of the precipitator shell. The insulators are in com-
pression. The best rapping results are achieved when rapper rods are posi-
tioned on the wire support frames independent of the support members. Double
tapered ceramic shafts with tapered steel connectors and a tapered rapper
attachment guarantee tight shaft couplings, minimal mechanical impedence to
energy transmission from the rapper to the frame, and the best resistance to
electrical tracking available. This system will function at well above the
highest precipitator operating temperatures.
197
-------
Original wire rapper
insulating shaft.
Bolted line-post in-
sulator. Note mis-
alignment at anvil.
Replacement of origin-
al shafting with
threaded/socket coup-
ler and double-tapered
ceramic shaft.
Wire rapper shafts us-
ing 2 double tapered
ceramic shafts each
for electrical isola-
tion.
DISCHARGE ELEC.TRODE RAPPER INDUCED MECHANICAL FAILURE
Discharge electrode failure is the most frequent and/or severe precip-
itator maintenance problem. The cause of discharge electrode failure is most
often arcing; corrosion is second in causing failure; and mechanical failure
is the least frequent cause of discharge electrode failure. (Engelbrecht,
1976). '
It is important to be able to differentiate between failure caused by
arcing, corrosion and fatigue. It is also important to record wire failures
by wire location, type of failure, and point of failure along the wire. With
this information, patterns relating to discharge electrode failures can be
observed.
Most mechanical failures of wires can be traced to over-rapping. Many
wire rapper vibrating systems operate for several seconds per cycle. If the
dust has not broken from the electrode within one second, it generally will
not be sheared by that level of excitation. Long periods of continuous wire
rapping can cause wire breakage.
Wire frame support insulator and rapper insulating shaft breakage are
probably more common than entirely mechanically induced wire breakage. When
the coupling at the ends of the insulating shafts become loose, misalignment
and chatter cause measurable rapping energy loss and can cause insulating
shaft breakage. The solution to this problem is to periodically tighten
198
-------
couplings or replace bolted and spring loaded coupled shafts with double-
tapered ceramic shafts.
Wire frame support Insulators are often broken by surface contamination
and subsequent tracking. When rapping causes support Insulator failure, it
1s the result of uneven distribution of the frame weight on the insulator,
or, more commonly, inadequate clearance where the rapper shaft penetrates
the insulator cap causing rapping energy to transmit through the insulator.
Rapping through the wire frame support shaft can also cause support insula-
tor breakage.
Inadequate or uneven tensioning of the wire will reduce rapping effect-
iveness and cause excessive wear at suspension points. This wear problem in
the wire support frame at the wire button can be caused by spit arcing as
the wire bounces and rotates with respect to the frame.
Seals at points where rapping shafts penetrate precipitator shells and
penthouse roofs should be gas and water tight. However, they must not sig-
nificantly dampen rapping energy or transfer it to the shell. Packing glands
are not acceptable since they absorb too much rapping energy. Steel guide
pipes with less than 1/8" radial clearance can corrode and bind the rapping
shaft. Leaky seals will allow corrosion of the guide pipe and shaft. This
leakage can cause contamination of discharge electrode insulating shafts by
water, gas, condensables, and dust, leading to electrical tracking from lower
conducting assemblies to the precipitator shell.
To summarize, when the following conditions are met, maximum discharge
electrode rapping will result:
1) Shaft connection tight from the rapper to the frame (preferably all
tapered).
2) Rapper shafting well aligned and not binding.
Wires adequately and evenly tensioned.
Rapper shafting independent of frame supports.
Rapper shafting located on the frame for best action.
Short bursts (0.5 sec or less) of frequency impact rapping at an
intensity and repeat rate required to maintain clean wires.
MECHANICAL FAILURES OF COLLECTING ELECTRODES
Collecting system mechanical failure occurs in the form of stress cracks,
stress fracture, broken bolts and broken welds. These fractures occur at
points of high stress concentration such as rapper shaft attachment to the
anvil beam, plate attachment to the anvil beam, and rapper shaft connections.
The reduction of stress concentration at these points of failure can often be
accomplished with relatively minor field modifications.
The highest stress concentration in most plate rapping systems is at the
base of the rapper shaft. At this point the total rapping energy is?trans-
mitted through as little as 20.6 cm (3.2 = in ) of area and 40.6 cnr (6.3
Inches) of weld. Since the lower end of the shaft and the anvil beam are
not mated, machined surfaces, the actual area of energy transmission is much
199
-------
less than 20 cm2 (3.2 1»2) and will be concentrated in the weld. Welded gus-
sets from rapper shaft to beam can distribute rapping imposed stresses.
The best method of eliminating shaft/anvil beam failure is to install a
cast steel tapered socket on the beam and use a tapered rapper shaft. The
cast steel socket receives rapping energy through an interference fit tapered
connection with the shaft and delivers this energy over an area 6 to 8 times
higher than a butt-end connection. Weld length is increased 3 to 4 times.
This approach reduces stress concentration at the shaft base and in the anvil
beam since the rapper loading is transferred to the beam over a broader area.
Tapered rapper shaft connections of 5 cm (2 in) to 7.6 cm (3in) of coupling
length length guarantee shaft alignment, and minimize stress concentration
through high contact area. The important result of this stress reducing de-
sign is that over stressing is less likely to occur.
Fatigue failure results when metals are subjected to high stress levels
for repeated loadings. If any component is loaded repeatedly with stresses
above its fatigue limit, it will fail after a number of cycles. The number
of cycles before failure depends on the stress level. If rapping induced
stress levels in all components of the plate assembly are below the fatigue
limit for the materials of construction, fatigue failure will not occur re-
gardless of the number of cycles. S-N curves indicate that if fatigue has
not occurred by about 10 cycles, the system has not been stressed above the
fatigue limit. In the case of the frequency type vibrators and rappers, 10
cycles will occur in 5 to 6 hours of continuous rapping.
If stress failure has occurred at some point in the collecting plate
system, the proper fix is not necessarily the obvious one of reducing rap-
ping intensity. Effective rapping energy (measured as plate acceleration)
will often be increased by reducing stress concentrations at failure points.
In one case, replacement of the shaft butt weld with a tapered base
soc.ketpincreased energy transmission and reduced stress levels (Neundorfer,
1977). Vertical zero to peak acceleration of over 1000 g's were measured
on one precipitator 22.8 cm (9 in) from the rapper shaft termination on the
anvil beam. On the same plate system, there was no measurable normal excita-
tion at the bottom of the plate. The anvil beams were stiffened, the plates
to anvil beam connection improved, and tapered sockets and rapper attachments
were installed. These improvements increased the normal acceleration at the
bottom of the plates to about 15 g's (0 to peak). This improvement included
bolting a tapered base to the existing single impact rappers but no rapper
change. When tapered mount pneumatic freguency rappers were installed, the
vertical anvil beam acceleration 22.8 cm (9 in) from the shaft was reduced
from more than 1000 g's in the original case to less than 300 g's and the
normal.acceleration at the bottom of the plate averaged 28 g's (Neundorfer,
z
200
-------
See scope traces below:
Vertical acceleration at anvil
beam: magnetic impulse rapper.
1000 g's (0-peak) 20 ms/DIV
lOv/DIV.
Vertical acceleration at same
point on anvil beam: pneu-
matic frequency rapper.
293 g's (0-peak^
5 ms/DIV. 2v/DTV.
operating pressure 45 psi
Normal acceleration at bottom
of plate: magnetic impulse
rapper. 0.0 g's (0-peak)
lOms/DIV .02V/DIV
Normal acceleration at bottom
of plate: pneumatic frequency
rapper. 28 g's (0 - peak)
5 ms/DIV .2V/DIV
Operating pressure 35 psi
The improved plate and wire cleaning raised the corona current from a maxr
mum of 150 ma with the original non-tapered, non-reinforced system to current
limit (600 ma) with the improvements. Outlet grain loading was measurably
reduced as would be expected.
IMPROVEMENT OF RAPPING MECHANISMS
Effective rapping energy cannot be transmitted through air. High area,
tight connections between the rapper, rapper shaft, rapper shaft anvil beam,
201
-------
and anvil beam to platewill maximize the effect of available rapping energy.
In one installation, the acceleration (0-peak) normal to the plate at the
bottom of a top rapped plate was increased by 45% by applying a 1-1/2 long
weld between the plate and its support beam at the top of the plate. This
weld increased contact area.
Areas of high stress concentration in rapping systems can result in
mechanical failure, but also cause high impedence to the transmission of
rapping energy. Designing these areas out of the system using tapered conn-
ections and broad based anvil sockets solves a two-pronged problem.
A stress analysis of a plate or wire assembly would be very costly, if
it were possible. The use of low mass accelerometers to measure accelera-
tions at various points on both the electrode frames and electrodes generates
Information useful in design modifications to existing systems which will en-
hance rapping effectiveness.
CONTROL
Precipitator rapper control plays a crucial part in attaining high pre-
cipitator efficiencies and low stack opacity. Thirty years ago continuous
rapping of precipitator electrodes was introduced. This improvement was the
beginning of an operational refinement with high potential for reducing pre-
cipitator emission. Rapping reentrainment losses, which can account for up
to 90% of all precipitator emissions, can be reduced by as much as 85% through
optimization of rapper control (Sproull, 1972).
The following general principles of precipitator operation apply to most
installations:
1) Plate rapping effectiveness increases with increased mass/unit area
of dust (reentrainment is reduced).
2) Mass/unit area of dust can be expressed as a "time between raps" for
any point in the precipitator along the gas flow as long as dust
build-up does not affect charging and separation and as long as rap-
ping is effective when it occurs.
3) This time between raps is inversely related to distance from the in-
let of the precipitator for a given dust mass/unit area.
4) Plate response to a given rapping disturbance is reduced as dust
build-up increases.
5) Simultaneous plate rapping increases the magnitude of instantaneous
rapping loss. This may not necessarily increase overall emissions,
but can increase the level of opacity.
6) Minimum rapper "on times" which result in good cleaning should be
used.
Precipitator collecting electrode rapper control is normally designed so
that rapping fields correspond to electrical fields. Each rapping field re-
peat time (the time it takes for a given rapper in that field to repeat after
it has rapped) should be adjustable independent of the other field repeat
times. The reason for independent field repeat time control is that in a
multifield precipitator the mass build-up rate of dust from inlet to outlet
field can vary by one to two orders of magnitude. Since rapping effectiveness
is related to mass/unit area of dust on the electrodes, the field rapping
202
-------
repeat rate must be separately adjustable from inlet to outlet.
Unfortunately, field repeat rates cannot be set strictly according to
field build-up rates. One reason is that migration velocity increases with
particle size. Therefore, the precipitator is also a classifier, collecting
larger particles in the inlet fields, and finer particles in the outlet fields,
This classification affects dust tensile strength, agglomeration, and adhes-
ion, and, therefore, affects the relationship between mass/unit area and re-
peat time. The extent of particle classification is reduced to some degree
by the fact that larger particles are reentrained in much higher percentages
than smaller particles; this results in passage of some of the larger par-
ticles from inlet fields to outlet fields through rapping reentrainment in
the inlet fields.
Field repeat rates must be established for each installation. In most
installations, processes, fuels and precipitator operating conditions may re-
quire the development of several sets of field rapping rates. Automatic
field repeat feedback systems are being investigated but feedback parameters
available are indirect and delayed. This lack of isolated direct feedback
makes automatic repeat rate control a difficult problem.
Most precipitators are set up with much too frequent field repeat times.
In almost every case, when field repeat times are studied in relationship to
opacity and overall efficiency, repeat rates are found to be2too frequent.
In studies by Gooch, Spencer, Sproull, Nichols and Schwartz, field repeat
rates were changed from minutes to hours with very positive results.
In addition to field repeat rate control, plate rapper control should
include anticoincidence. This feature allows only one plate rapper to oper-
ate within a time block. This time block should be greater than or equal to
the time it takes for a rapping disturbance to clear the precipitator outlet.
Anticoincidence rapping minimizes the magnitude of opacity spikes resulting
from rapping disturbance.
Anticoincidence rapping can be applied to multiple precipitators shar-
ing a common stack. Rapping only one plate rapper per stack at a time is
helpful in meeting opacity regulations.
On some large precipitators and multiple precipitator installations,
the number of plate rappers per stack may be as high as 700. In these cases,
it is impossible to maintain a sufficient time block between raps, proper
field repeat times, and total anticoincidence. When this condition exists,
total anticoincidence is sacrificed in favor of anticoincidence in as many
outlet fields as possible while maintaining a sufficient time between raps
and field repeat times. Anticoincidence in outlet fields will have the
greatest effect on minimizing rapping opacity spike magnitude. Opacity in-
creases during rapping can last for more than 6 seconds; therefore, the time
block per rapper is set at more than 8 seconds.
203
-------
RAPPER CONTROL DESIGN
Rapper controls should offer field repeat rate adjustment, rapper on
time adjustment, (or intensity or both) and rapper rest time adjustment for
each field. Anticoincidence should be available. Totally solid state digi-
tal systems are desirable since they can offer broad program flexibility,
with relatively few components. Time values are set with labeled switches
with linear resolution.
Digital solid state rapper controls are very reliable if properly de-
signed. Electrical line noise should be filtered and relays should be pho-
toisolated. Rapper energization indicators and rapper failure indication
are desirable features.
METHODS OF TUNING RAPPER SYSTEMS
Optimizing precipitator rapping with normally available feedback informa-
tion requires documentation, patience, and some degree of imagination. The
feedback information is obtained from:
1) Transmissometers (optical density)
2) Automatic voltage control console readings
3) Hopper evacuation system vacuum charts or other indications of rela-
tive amounts of dust discharged from each hopper.
4) Visual observation of electrodes being rapped during precipitator
operation. (This can be done in some cases using access covers with
pyrex viewing windows and high intensity lights).
5) Plate fouling observed during internal inspections with the precip-
itator shut down.
6) Recording of outlet grain loading and particle size distribution
(this is generally not available).
Transmissometers, which measure optical density of flue gas, can be used
to determine the degree of rapping reentrainment and the duration of the dis-
turbance after the rap. Optical density relates linearly to dust concentra-
tion for a given type of dust. The relationship between optical density and
dust concentration is available or can be developed for specific dusts.
Optical density increases resulting from rapping of the outlet fields
and one or two fields directly upstream of the outlet fields, can be used
directly to evaluate rapper intensity and field repeat times. Inlet field
rapping disturbance often blends into the background opacity level. There-
fore, inlet field repeat times and rapping intensity are adjusted to reduce
opacity background levels.
Console readings are closely monitored during rapping optimization.
Precipitator power should be maintained as high as possible between raps. A
decrease in precipitator power due to electrode fouling will also be ob-
served as an increase in optical density. When this occurs field repeat
rates and/or rapper intensity in the field(s) affected should be increased.
If information can be developed indicating the amount of dust being col-
lected from each field, rapping can be adjusted so that actual field collec-
tion efficiencies approach theoretical values.
204
-------
Actual viewing of the electrodes during rapping while the precipitator
is operating will reveal the amount of agglomeration of the dust and the
size of dust chunks broken away by rapping. Corona quehch resulting from
high dust concentration during rapping may also be observed.
Precipitator inspections during outages will reveal the thickness and
smoothness of the dust build up on electrodes. Direct manual rapping of
electrodes or wash down during an outage and its effect on operation after
start-up can demonstrate whether improved on-line electrode cleaning will
improve performance. Highly irregular dust build up on the plates and back
corona craters indicate special problems associated with high resistivity ash.
If outlet grain loadings and particle size distribution can be taken re-
liabily and continously during a rapping system "tune-up" the information de-
rived will most directly help to establish an optimum rapping program.
205
-------
REFERENCES:
1. Plato, H. Rapping of Collecting Plates in Electrostatic Precipitators.
Staub-Reinhalt Luft (in English) 29: 22-30 (1969)
2. Gooch, J.P., G. H. Marchant, Jr. Electrostatic Precipitator Rapping
Reentrainment and Computer Model Studies. Prepared for EPRI by SRI.
EPRI FP-792. 3: (1978)
3. UN Published work done by Mm. Neundorfer & Co., Inc. Two Eastern Util-
ities (1976)
4. Spencer, H.W. III. Rapping Reentrainment in a nearly full Scale Pilot
Electrostatic Precipitator. Prepared for U.S. EPA by S.R.I. EPA-600/
2-76-140: 76 (1976)
5. Gooch and Marchant, op. cit.
6. Ibid
7. Spencer, op. cit.
8. Nichols, G.B. S.W. Spencer, and J.D. McLain. Rapping Reentrainment Study,
S.R.I. November (1975)
9. Tassicker, O.J. Some Aspects of Electrostatic Precipitator Research in
Australia. Symposium on Electrostatic Precipitors for Control of Parti-
cles. P.B. 240440:101-129 (1975)
10. Penney. G.W. Adhesive Behavior of Dust in Electrostatic Precipitation.
Symposium on Electrostatic Precipitators For The Control of Fine Particles
EPA-650/2-75-016: 65 (1975)
11. Spencer, op. cit.
12. Gooch and Marchant, op. cit.
13. Spencer, op. cit.
14. Sproull, W.T. Fundamentals of Electrode Rapping in Industrial Electrical
Precipitators. J.A.P.C.A. 15/2: 50-55 (1965)
15. Ibid
16. Ibid
17. Spencer, op. cit.
18. Sproull, W.T. Minimizing Rapping Loss in Precipitators at 2000 Megawatt
Coal Fired Power Station J.A.P.C.A. 22/3: 181-186 (1972)
19. Ibid
206
-------
20. Unpublished work done by Wm. Neundorfer & Co., Inc. for a midwestern util-
ity. (1977)
21. Nichols et al, op. cit.
22. Englebrecht, H.L. Electrostatic Precipitator Inspection and Maintenance.
Plant Engineering pp 193-196 April (1976)
23. Neundorfer, .1977, op. cit.
24. Neundorfer, 1976, op. cit.
25. Sproull, 1972, op. cit.
26. White, H.J. Industrial Electrostatic Precipitation. Addison-Wesley Pub-
lishing Co., Inc. 1963.
27. Schwartz, L.B. Effect of Rapping Frequency on the Efficiency of an Elec-
trostatic Precipitator at a Municipal Inunerator. Proceedings of the
Fourth Annual Environmental Engineering and Science Conference Mar. 3-4,
1974, U. of Louisville.
207
-------
COMPOSITION OF PARTICULATES
SOME EFFECTS ON PKECIPITATOR OPERATION
By:
Jack D. Roehr
Wahlco, Inc.
Santa Ana, California 92704
Precipitator operators and designers often review spectrographic
analyses when attempting to resolve problems with installations
or designing future installations. These analyses can be mis-
leading. Those responsible for operation and design must have a
knowledge of the compounds of the elements that can make up the
particulate under study as well as the process that produced
these compounds and the operating conditions existing at the
point of particulate formation. Several of the most common
elements and compounds are discussed with a review of their
effects on precipitator operation.
208
-------
COMPOSITION OF PARTICUIATES
SOME EFFECTS ON PRECIPITATOR OPERATION
When this planet Earth was formed it was made up of various and sundry
elements. These elements - whether they be carbon, copper, zinc, calcium,
or whatever - were subsequently discovered by man and he set about finding
ways and means of extracting these elements for his use. It was soon learned
that in no two places on the surface of the Earth were these elements found
in the same abundance or in the same ratios or in the same mixtures making
up compounds of the elements.
Because of these differences and also because of technological advances
and/or the findings of a multitude of inventions, we have almost as many
different methods of processing ores to extract the wanted elements or of
making use of some compounds such as the burning of coal as there are indi-
vidual industrial plants on this Earth.
This is the root of the problem when we are concerned with the operation
of an electrostatic precipitator being used or to be used to clean the gas
evolved from one of these processes. In any case we are concerned with the
capture of particulates made up of compounds of various elements when we
very seldom, if ever, know what the compounds are.
For purposes of expediency as well as economy, chemical analyses of
particulate are usually obtained by means of a spectrograph. In some in-
stances no analyses will be made on the dust or particulate from an installa-
tion unless the vendor is in trouble. If the equipment meets performance
guarantees, that point is proven and the test people pack up their gear and
leave. Nothing beyond that is learned. For particulate analyses, each
element reported in a spectrographic analysis is usually stated as the oxide
of the element, i.e., iron is reported as Fe2O3 and calcium as CaO when, in
fact, the elements could be in the form of a sulfide or a sulfate or some
other compound.
It is these results that are fed into computers over the world in attempts
to derive an equation for calculating the drift velocity or effective migra-
tion velocity in the Deutsch equation. Although several of these have been
published, the authors usually admit that each has its limitations and draw-
backs. As an indication of the problems that can be had please refer to
Table 1. Here are the coal and ash analyses from a pulverized coal-fired
boiler and precipitator system.1 In this case there were two parallel pre-
cipitators. The ash analyses for each was, for all practical purposes, iden-
tical. One precipitator exceeded the design performance while the other
failed by a wide margin. In fact on the basis of the tested performance
the precipitator would have had to be doubled in size to meet design con-
ditions, even more if you are a believer in the square root rule or double
power function version of the Deutsch equation.
209
-------
They subsequently learned that due to some unknown factor in the boiler
system the 803 content in the gas to precipitator IB was about 30 ppm and
the SOa in the gas to the other unit was about 1 or 2 ppm. After conditioning
the gas with SO3 the performance of the precipitator exceeded the design by
a considerable amount.
The purpose of this discussion is not to reveal the latest in black magic
on how to arrive at proper precipitator size or prognostication on the antici-
pated performance of an existing or planned installation. Rather it is to
discuss operating problems that can be experienced when certain elements or
compounds are present in the particulates or gases to be cleaned with a
precipitator. Also, we are not restricting this to a single application such
as a pulverized coal-fired boiler.
ALUMINUM
This is found in such applications as coal-fired boilers, cement kilns,
fluid catalytic cracking units and in the primary aluminum industry. Gener-
ally it is in the form of aluminates and silico-aluminates of high electrical
resistivity.
For successful precipitator operation the gas must be conditioned to
attain surface conductivity. Water vapor is most commonly used, either by
water evaporation in conditioning chambers or as a natural consequence in
the process such as in a wet process cement kiln; ammonia, in addition to
water vapor, has been found to be a good conditioner especially for precipi-
tators on fluid catalytic cracking units, and sulfur trioxide is the pre-
ferred agent for fly ash. Certain alumina containing clays can be particu-
larly troublesome. For example, a precipitator on a coal-fired boiler in
Alabama gave operating difficulties when they were burning coal from a
particular mine. There was very little difference in the spectrographic
analyses of the ash. However, it was learned that in the coal ash from the
one mine that gave the trouble the aluminum and silica in a portion of the
ash were present as an especially fine kaolin clay dust. This fraction of
the ash resisted all attempts to be rapped from the collecting plates
leaving a thin film of very high resistivity dust.
ARSENIC
This is only mentioned here for those in the non-ferrous metallurgical
industry that are concerned with a furnacing operation with a charge con-
taining arsenic. Arsenic will be readily volatilized off in the furnace as
the oxide and will condense out of the gas as the temperature decreases.
Gas cleaning is usually done in two stages, the first over 500°F to get the
bulk of the dust and pass the volatile arsenic oxides and the second below
270°F to catch the condensed oxides. The gas must be shock cooled so that
the arsenic will be condensed out ahead of a cold precipitator. Any arsenic
that condenses out in the precipitator can give real trouble, the most com-
mon of which is the growth of needle-like crystals on the discharge elec-
trode with decreased electrical clearance and low voltage operation. In the
210
-------
temperature range of 350-450°P the condensing arsenic compounds seem "plastic"
and sticky resulting in sure problems in the precipitator. Very low
operating temperature can also be troublesome as acid dew point is reached
with the 803 formed in the furnacing operation.
CALCIUM
This is usually found in the dust to the precipitator, depending on the
process involved, as calcium oxide, calcium carbonate or calcium sulfate.
Any of these can be detrimental to precipitator operation as they have an
inherent high electrical resistivity. The first "hot" precipitator in the
industry, on a cement kiln in 1912, had high voltage support insulators made
up from a stack of marble slabs (a form of limestone).
When calcium compounds are the predominate constituent in the dust to be
collected you must follow the resistivity curves in your precipitator opera-
tion and have either high moisture content or very low or very high operating
temperature. For example - resistivity problems on iron ore sintering
machines can be directly related to the lime content of the charge. High
lime, self-fluxing plus sinter mixes are the worst in this regard.
Calcium sulfate can form in a lime-gas reaction in the system. If per-
mitted to form while the dust is in the precipitator it will result in a
very hard dust cake on both the discharge electrodes and collecting plates.
Where the flues are long or large enough to permit formation of the sulfate
either in the process or in the flue system ahead of the precipitator, the
resultant dust will be essentially free-flowing after it is collected.
IRON
If this is in the form of the dry oxide the electric resistivity will
be high and suitable precautions must be taken to have operating temperatures
on either side of the peak of the curve and sufficient moisture must be
present for conditioning, particularly on the high temperature end of the
scale.
Precipitators have been used for collecting the fumes from the oxygen
blown steel-making facilities, both for building ventilation and for vessel
discharge gases. The fume is essentially iron oxide and can readily be
collected at ambient conditions for the first case and with the use of water
evaporation conditioning for the second case.
Furnacing operations that have sulfur in the charge such as with a pyrite
roaster will result in a dust that will consist of a high percentage of iron
sulfates. The iron oxide in the gas serves as a catalyst for the formation
of 303 which then reacts with the oxides to form the sulfate. Iron sulfate
itself is not a difficult dust to precipitate and handle if it is in that
form as it enters the precipitator.
211
-------
If sufficient time has not elapsed between the roasting operation and the
precipitator so that the formation of the sulfate occurs after the dust is
collected, it will invariably result in build-up on the discharge electrodes
and collecting plates of hard clumps that can be of rock-like consistency,
impossible for the rappers to remove.
No two cases are ever alike and solutions are difficult to find. The
process is time and temperature dependent. For example, a large clump or
nodule taken from a discharge electrode can be analyzed for iron sulfate.
You will find that the stratum that was closest to the discharge electrode
will be very high in sulfate - say 25% - with correspondingly less sulfate
content as you approach the outer boundary of the clump or the "newest"
portion. Also, as the sulfate content increases, so does the hardness.
The best solution to this sulfating problem is to change the operating
temperature. The reason for this is not clearly understood. An increase
in operating temperature from 700°F to 900°F has corrected a situation,
while in others going to 1100° was required. In still other cases a lower
temperature has been the answer.
LEAD AND ZINC
These two have been lumped together for the purpose of this discussion
because the end results, as far as precipitator operation is concerned, are
so similar.
Most precipitator applications are on the gases from sintering machines
or roasting operations. The differences in operations and conditions are
as numerous as there are plants in operation. No two are alike.
Sinter machine practice with older techniques gave fumes of very high
resistivity and made external flue gas conditioning a necessity. A general
rule of thumb was to evaporate sufficient water to result in a relative
humidity of 60%. Consolidated Mining Company, Trail, B.C., used to condition
the gas to achieve a moisture content of 1% in the dust. The precipitator
worked satisfactorily at this condition. Less moisture than that resulted
in spontaneous combustion of the dust in the hoppers and more than that re-
sulted in a sticky dust that could not be rapped from the collecting plates.
In the earlier days when the multiple hearth furnace was common in roasting
zinc concentrates, precipitator operation on these gases was no real problem.
This was because of the relatively low roasting temperatures with sufficient
time to permit the formation of zinc sulfate. With the advent of more modern
roasting machinery such as fluid bed reactors it was a different situation.
Here there is a very high temperature and a very short residence time with-
in the roaster, and the dust to the precipitator is a real fume made up of
condensed zinc oxide which had volatilized as zinc in the furnace and then
oxidized to create a sub-micron particle most difficult to collect in a
precipitator.
212
-------
This dust not only has high resistivity but it is very light, not unlike
feathers. Some plants have had to resort to closed damper-power off rapping
to achieve successful operation. Others have abandoned the precipitator
entirely and have resorted to high temperature fabric filters.
Most copper smelters have had to increase the dust collection efficiency
of their plants because of air pollution restrictions. Since they return
the precipitator catch to the system, because of the high copper content,
they are, if they have appreciable zinc or lead content in their concen-
trate charge, adding to the zinc or lead content of the dust. This is a
vicious cycle since the only way to get rid of these "impurities" is via
the stack or by discarding the dust. If they don't discard the dust the zinc
or lead content continues to increase and with it a degradation of precipita-
tor performance. A typical example of this is a copper smelter in Canada,
where the recirculating dust has resulted in very high concentrations of
lead and zinc from their converter gas. Lead content varies from 18 to 40%
and zinc content varies from 6 to 20%, dependent on the stage of the blow
for the various converters. In situ resistivity measurements showed that
the resistivity was a function of zinc content varying from 10 at 8% to
over lO*2 at 20% at the operating temperature of 500 - 600°F.
In a study to determine what could be done to upgrade precipitator per-
formance a pilot precipitator was installed. The only thing that would
reduce opacity at the stack was a combination of moisture and 303 injection.
The addition of about 3% by volume water vapor and 50 ppm SO-j reduced opacity
to, essentially, zero.
MAGNESIUM
High magnesium content dusts are most commonly found in plants that
extract the element from brines or sea water and precipitators are used on
the magnesium hydroxide calciners. Precipitation is ordinarily easy due to
high water content of the gases, although rapping can be difficult due to
the light, fluffy nature of the collected material. One exception is in
plants that do not have a fresh water wash for the precipitate and wind up
with a high chloride content which is volatilized off in the kiln resulting
in a very fine fume making precipitation more difficult.
MOLYDENUM
For this element we find precipitators on moly-sulfide roasters.
After difficulties with the earliest installation it was established that
success hinges on the temperature of operation. Above 700°F the dust is
sticky and plastic, soon resulting in a wall-to-wall mess. Although the
SO2 content of the roaster is not very high, the S03 formation is quite high
resulting in a very high dew point of approximately 400°F below which tem-
perature corrosion is a real problem. These factors limit the successful
operating temperature to a fairly narrow band. However, it can and has been
done.
213
-------
PHOSPHOROUS
This element is included here as a matter of interest since it has been
pegged as one of the culprits in the failure of a precipitator to perform
on a coal-fired boiler not too far from Denver.
Precipitators have been used for elemental phosphorous furnaces for
years. This is one of the worst if not the worst duties for a precipitator
due to rapid and extreme temperature variations. However, these precipita-
tors are not really for the purpose of collecting phosphorous as the tem-
perature is high enough that all phosphorous, theoretically, is in the vapor
state and the precipitator is to collect the particulate, mostly silica,
before the phosphorous is condensed out down the line in the process.
It has been reported by the operators at the above mentioned utility
that 1-4% ?2C>5 in the ash resulted in impairment of precipitator operation.
This could be the result of vaporization of the phosphorous in the boiler
with subsequent condensation into one of several forms that can be very
sticky and build up on discharge electrodes.
A plant in Idaho has a precipitator on a phosphorous rock calcining
kiln and with very good success. The dust has a P20s content of about 15%.
The dust is very fine making a venturi scrubber, which they tested before
purchasing a precipitator, out of the question because of the very high
pressure drop requirement. Their furnacing operation is only in the range
of 1600°F and this may be the reason for the difference in the resultant
dust in comparison to the coal-fired boiler operation.
POTASSIUM AND SODIUM
These two are also lumped into the same discussion. They are both in
the alkali group of metals and both have comparable results concerning
precipitator operation. Either can do the following, they can act as condi-
tioners to enhance precipitator performance or they can be the chief cause
of impaired precipitator performance. The real key to this is in the process
that caused the dust in the first place.
Sodium content of the ash has been an indication of the failure or suc-
cess of precipitators on coal-fired boilers being fired with low sulfur
coal, with the trend being better operation with higher sodium content.
As examples we have Utah Power & Light's Huntington and Naughton stations.
At Huntington when they had 5-6% sodium in the ash their precipitators
operated in the high 99%+ range with drift velocity in the neighborhood of
8 cm/sec. At Naughton where the sodium was down in the 1% or lower range
the precipitator on one boiler operated in the 3-4 cm/sec, range and they
had to use 803 flue gas conditioning to achieve satisfactory precipitator
performance.
214
-------
Sodium has also been used with varying degrees of success as an additive.
This certainly is not new. Thirty-five years ago a saturated solution of
salt water was added to the gas from a dry process dolomite kiln. This per-
mitted them to operate their precipitators at 550°F with good performance.
Without the salt addition they had to operate at 700°F for good performance.
Salt has been added to the furnace charge in various non-ferrous metallur-
gical operations. This was not for the purpose of enhancing precipitator
performance but to aid in volatilizing elements such as cadmium or lead.
A secondary result was an aid in precipitator performance. It is believed
that the chloride compounds formed had an affinity for water vapor and this
lowered the resistivity. The sodium may have had a bearing also.
In some applications, particularly cement kilns, the potassium and
sodium actually downgrade the precipitator performance. The precipitator
will actually be selective in that the cement making constituents, i.e.,
CaO, A12C>3, Fe2O3, MgO and SiC>2 will be precipitated out in the first fields
and the 1^0 and Na2O will be caught in subsequent fields with the highest
potassium and sodium content being in the stack effluent. This is due to
the method by which the soda and potash fumes are made. The kiln temperature
in the burning zone is high enough to volatilize off the alkali. In fact
this is what they want - to get the alkali out of the clinker. This alkali
subsequently condenses out as a fine fume and downgrades precipitator per-
formance, because of the fine particle size. On cooling, they may go through
a plastic condition in the 600 - 650°F range and should this occur in the
precipitator you are in trouble from build-up. You must get the temperature
down to at least 550°F and this should be accomplished by water evaporation.
With kilns that practice insufflation, i.e., kiln feed charged via the
burners, the situation is aggravated because of the extreme temperature
where all the alkalies will be driven off as a vapor.
In cement plants that have gypsum in the feed or are firing with high
sulfur oil sulfates of both potassium and sodium can be formed. These
must be cooled so that condensation occurs in the flues well ahead of the
precipitator.
SULFUR
This element can mean success or failure with a precipitator on a pro-
cess. If in the elemental .form, as. caught in the precipitator, it means
trouble. (This is very rare but it has happened in special cases.) It
melts at about 300°F and is a solid a few degrees below that temperature,
and a super-viscous liquid a few degrees above. It has a high resistivity.
It can quickly plug up a precipitator and/or result in build-up and short
circuiting.
When oxidized to 303 it can condition the gas and dust so as to greatly
enhance precipitator performance.
215
-------
If the SO3 goes to the formation of sulfate compounds it can result in
build-up problems as discussed above.
GENERAL
There are other elements or their compounds that precipitator operators
or designers will be confronted with. However, the above discussion should
cover the most prevalent.
CONCLUSION
When designing precipitators for a new plant or when attempting to
determine the reasons for the poor performance of an existing installation
an analysis of the particulate will give you only a partial answer. You
must consider the environment at the place in the process where the particu-
late is being formed, as well as subsequent to this point in the process,
so that you will have a better understanding of the exact chemical and physi-
cal nature of the particulate under consideration at the time it enters
the precipitator.
216
-------
TABLE 1
COAL ANALYSES
Hydrogen 3.4%
Carbon 51.0
Water 15.0
Sulfur 1.6
Ash 21.6
ASH ANALYSES
Unit A Unit B
Si02 48.9% 49.6%
A1203 25.1 26.1
Fe2O3 8.6 8.2
CaO 4.8 4.4
MgO 1.3 1.4
TiO2 0.7 0.7
K2O 5.4 5.6
P205
NaOH 2.6 2.4
SC-3 1.1 1.4
PRECIPITATOR EFFICIENCY
Design 98.5% 98.5%
Test 89.9 98.9
Test W/S03 99.2
217
-------
Literature Cited:
1. Busby and Darby (May, 1963) Journal of The Institute of Fuel
218
-------
INCREASING PRECIPITATOR RELIABIILTY BY PROPER LOGGING AND
INTERPRETATION OF OPERATIONAL PARAMETERS -
AN OPERATOR'S GUIDE
By:
Peter P. Bibbo - Product Manager, Precipitators
Peter Aa - Product Analyst
Research-Cottrell, Inc.
Utility Division
Somerville, New Jersey 08876
Maintaining efficient and reliable electrostatic precipitator operation
is dependent to a large extent on the operator's knowledge of the precipitation
process and on the type, amounts and use of operational data and records. Ana-
lysis of properly recorded information can heighten the operator's knowledge of
precipitator behavior and, through his actions, precipitator performance and
reliability can be enhanced.
This paper discusses the type of data needed for an operator's evaluation
of precipitator performance, where and how to obtain this data, and how to in-
terpret the collected data.
The procedures contained in this paper are the recommendations of a manu-
facturer's field service personnel, who have had extensive experience in
working with and training operators in the proper operation and care of pre-
cipitators. With the more intimate knowledge and experience that only comes
with daily plant level responsibility for precipitator performance, it is ex-
pected that operators who read this paper will tailor the detailed recommen-
dations to better suit their own situation.
219
-------
INCREASING PRECIPITATOR RELIABILITY BY PROPER LOGGING AND INTERPRETATION OF
OPERATIONAL PARAMETERS - AN OPERATOR'S GUIDE
BASIS OF PRECIPITATOR PERFORMANCE
Since this paper deals with the collection and interpretation of opera-
tional data which could be useful in maintaining optimum precipitator per-
formance, a review of the fundamental operating parameters which control pre-
cipitator performance is an appropriate starting point.
Simply stated, the operating conditions at the inlet of the precipitator
plus the operating conditions inside the precipitator determine exactly what
conditions result at the outlet of the precipitator. In spite of this simple
cause and effect relationship, different theories and opinions abound which
attempt to explain the precise relationships among the many parameters that
control precipitator behavior. Fortunately for the operator, the parameters
to which he can measure and adjust are not complex, and can.be effectively
controlled to maintain precipitator performance at the maximum possible levels.
Operating conditions at the inlet of the precipitator are completely
dependent upon the fuel being fired, and by the boiler design and its opera-
tion. If a gas conditioning system is installed, these inlet conditions may
be altered somewhat. The inlet conditions important to the operator are shown
in Table 1. The conditions inside the precipitator are mainly a function of
the precipitator design. To the extent that the manufacturer has followed
good design practices and provided the means for adjustment, this is where the
operator has the greatest control over the precipitator performance level.
Important precipitator conditions are shown on Table II.
MAINTAINING EFFICIENCY AND RELIABILITY
The availability performance of electrostatic precipitators is typically
among the best for any mechanical equipment in a power plant. Availabilities
over 90 percent are not uncommon.
However, the progress of environmental legislation in this country indi-
cates that availability levels at or above 90 percent will have to become the
ordinary rule rather than a proud exception. It has been and still is the
position of Research-Cottrell that the key to maintaining reliable precipitator
performance is through a working preventive maintenance program, regardless of
individual precipitator design features or manufacturer's claimed reliability.
The data logging and interpretations recommended in this paper augment
any precipitator preventive maintenance program. In fact, some of the recom-
mended data logging activities are identical to preventive maintenance steps
outlined by Research-Cottrell in earlier literature. An example of recom-
mended preventive maintenance steps for precipitators is shown in Table III.
220
-------
DATA TO BE RECORDED FOR INLET CONDITIONS
Since the precipitator operator typically has little control over boiler
operation, there may be a natural reluctance toward obtaining useful boiler
data at the appropriate times. However, at all times, practically all the
conditions that the precipitator must treat are being produced by the boiler.
Speaking from our own experience, we can say that there have been many cases
where proper boiler data collection has revealed causes for fluctuations in
precipitator performance, and operational changes were made to favor precipi-
tator performance, without detriment to the performance of the boiler.
Table IV lists the data source documents that can be used to obtain all
the essential data indicated in Table I for precipitator inlet conditions.
Boiler Data
Frequently, we find that too much boiler data is recorded when its only
purpose is to support analysis of precipitator behavior.
Figure 1 shows a recommended boiler data sheet. The heading should be
completely filled out each time a new sheet is used. Too often, this detail
is overlooked, causing considerable confusion. This commentary pertains to
all data sheets. Data taken on separate days should not be entered on the
same sheet. Information on boiler operation should be recorded at least once
per day, at representative load conditions.
Coal Analysis
Many different report formats are used, but they all contain essentially
the same information. Common deficiencies in these reports are failure to
report the exact dates the samples were taken and the coal source identification.
To accurately identify the coal, the mine name, seam, and geographical location
must be logged. To be most meaningful and accurate, the analyses should at
least be reported "as received" (A.R.).
Ash Analyses
Similarly, ash mineral analysis report formats vary widely although they
generally contain the same information. Dates and coal source should also be
logged, and since ash composition is influenced by the combustion process, a
notation should be made indicating whether the sample was fly ash or coal ash.
Coal ash is the residue obtained by completely combusting a coal sample in a
laboratory. Fly ash should be considered as a sample obtained downstream of
the boiler economizer section. Ten minerals plus LOI (loss on ignition, which
is not the same as carbon content) are the constituents that are always of
interest to precipitator performance.
221
-------
Many plants perform composite coal proximate analyses once per week. The
operator should obtain a copy of these and any other coal and ash analytical
results that are available, and keep them in the precipitator file.
Precipitator Performance Test Results
Although this data source is normally obtained only once per year at most,
it is the only source that contains the most comprehensive set of related data
useful in evaluating precipitator performance, and usually it is the only docu-
ment that contains information the operator cannot readily obtain himself, such
as actual dust concentration, particle size distribution, ash resistivity, and
so on. Gas flow uniformity is usually measured only once and may not be mea-
sured at all. Since uniformity is determined mostly by the geometry of the
actual installation (and to a much lesser extent operating conditions such as
boiler load), and cannot be easily adjusted, it is of little interest to the
operator.
DATA TO BE RECORDED FOR PRECIPITATOR INTERNAL CONDITIONS
A properly designed precipitator should be capable of efficiently treating
all the conditions delivered to its inlet by the boiler. Therefore, the operator
should find the greatest control he has over performance and reliability will
be by the actions he takes on the precipitator.
The important operating variables that lead to maximum precipitator per-
formance and reliability aren't many, but it is not possible for an operator
to recall the essential information without a systematic procedure of data
logging. Table V lists the data source documents that can be used to record
all the data necessary to understand and maintain maximum precipitator per-
formance and reliability.
Electrical Energization Data
The entire precipitation process depends fundamentally on the quality of
electrical energization, and power readings are sensitive to changes in coal
quality and boiler operation. For these reasons, power data should be taken
frequently. A form used by Research-Cottrell for recording these readings is
shown in Figure 2, but any format which requires the same information would
be useful.
Voltage-Current Curves
Again, speaking from practical experience, it is difficult to imagine any
data that is as useful as voltage-current curves in understanding precipitator
behavior, yet they seem to be used infrequently at the plant level. While
more time-consuming to obtain than electrical readings, voltage-current data
need not be tafcen as frequently as power data, and curves do not have to be
generated for every T/R set for the data to be useful. The data can be recorded
on the form shown in Figure 3, but it is best interpreted when plotted on a
curve, as shown in Figure 4.
222
-------
Energization data should be taken at least once per day, or at any time
when a major change in boiler operation occurs. Voltage-current curves should
be recorded at least once on all T/R sets to reflect normal operating condi-
tions for each major coal source fired in the boiler. Thereafter, voltage-
current curves need only be taken for a fraction of the T/R sets to serve as
a check. Plants may wish to record these electrical data more frequently, but
experience will probably show it is not necessary.
Electrode Cleaning Equipment Data
Next to energization data, information on electrode cleaning system
operation is of most use to the operator. Electrode cleaning system data
important to precipitator operation appears in Table VI.
Few operators log the operation of electrode cleaning equipment. Yet
this system's operation has a large effect on the efficiency and reliability
of the precipitator.
Since electrode cleaning system operation is infrequently changed, it is
not necessary to record this data on a frequent basis. However, any changes
that are made to the operation should be logged, the reason for the change
noted, and the effect on precipitator operation recorded.
Ash Removal System Operation
The most prevalent cause of poor precipitator availability is electrical
section grounds caused by full hopper conditions. An ash removal system
operation log can be an important aid in analyzing and understanding precipi-
tator reliability. Dates and times of all full hopper conditions or suspected
full hopper conditions should be entered in the log.
Since the actual ash pulling cycle is frequently controlled automatically
on a demand basis, it should only be necessary to record dates and time of all
full hopper conditions or suspected full hopper conditions. To identify dust
maldistribution in the precipitator, the operator may wish to record the time
required to pull each hopper.
Internal Component Inspection Data
The internal condition of the precipitator is a major controlling factor
in precipitator operation and reliability. During any outage of sufficient
length, a precipitator internal inspection should be performed. Major items
to be inspected during an internal inspection are listed in Table VII.
Research-Cottrell has developed an internal inspection log, shown in
Figure 5 that simplifies the logging of internal conditions. During an internal
inspection the inspector simply codes his findings on the log, and converts the
codes into a verbal inspection report after the inspection is complete. Addi-
tional items can be added to the code as necessary.
223
-------
Complete inspections and reports should be made with each annual outage
and on any other occasion when time permits. Internal inspections are par-
ticularly important during unscheduled outages if localized poor electrical
performance has occurred.
Replacement Parts Log
The recording of part descriptions, failure dates and replacement dates
will show whether component failures are on a reoccurring or random basis.
There are many components that have a direct effect on associated components,
and without replacement records the direct cause of component failure may
never be detected.
HOW TO INTERPRET THE DATA
In order to maintain optimum precipitator performance, we believe the
operator must first know what the design or expected precipitator performance
level is. He should then be able to compare this to the present performance
level, and finally investigate and correct the apparent cause for malperfor-
mance if it exists. Thus, we also believe the operator should have a basic
understanding of what precipitator performance depends upon. This paper does
not pretent to present all the knowledge of precipitation theory in a few
pages of print, for that would result in a gross oversimplification. However,
by at least recognizing some of the more important aspects of precipitator
behavior, the operator will be able to do his job better.
Precipitation Fundamentals
Precipitator performance level is totally dependent on three general
quantities: the amount of energized bus sections, the total gas volume
passing through the precipitator, and the sum of all other system parameters
such as ash composition, particle size distribution, precipitator voltage
and current levels, and all the other inlet and internal conditions. The
baseline is the level of performance that results when all these conditions
are considered "normal".
Deviation from expected performance cannot be detected unless the base-
line performance is known. Since the baseline in reality will be a broad
line, information that the operator records that is indicative of "normal"
performance level will contribute to the definition of baseline performance.
Likewise, recorded information that is not indicative of "normal" per-
formance gives the operator insight as to what factors may be adversely
affecting precipitator performance. It is convenient to divide the data
into two categories: operational factors and physical factors.
Operational Factors
Boiler data, coal and ash analyses, electrical readings and voltage-
current curves, rapper operation, and ash removal system data are all examples
of operational factors. All operational factors are related to one another
in some way, and they all make a contribution to precipitator performance.
224
-------
For example, gas volume, one of the three precipitation funda-
mentals, can be estimated from simple boiler data. A better estimate
of gas volume can be obtained by a coal analysis. Comparison of the
estimated gas volume with "normal" gas volume can indicate a cause
for deviation in precipitator performance.
Likewise, comparison of electrical readings can supply insight
into precipitator behavior. For instance, changes in voltage levels
can indicate changes in coal or ash properties, internal dust build-up,
gas temperature excursions, and so on.
The relationships among all the operational factors can become
complex. Figure 6 is.a simplified guide to understanding some of the
most important relationships that normally exist. This guide will not
satisfactorily answer every question that can arise, but it should help
the operator avoid misconceptions and assist him in defining potential
problems.
Physical Factors
Precipitator performance and reliability can be affected by actual
or imminent equipment faults which are completely independent of the
operational factors. These can also be complex. Figure 7 is a fault
tree ^ for typical precipitator hardware. The operator can use this tree
to quickly ascertain real or potential equipment failures that can cause
deviations from "normal" performance.
CONCLUSIONS
Logging important precipitator operating and physical conditions
will give the operator useful insight into the factors that contribute
to changes in precipitator performance levels. A written record of the
cause and effect relationships that have occurred in the past, along
with a fundamental knowledge of precipitator behavior, will aid the
operator in identifying and correcting potential failure causes in the
future. Hopefully, the information and guidelines presented in this
paper will be helpful in enhancing precipitator performance and reliability
by making the operator more effective in what he does. In this latter
regard, we make three final comments
First, we recommend that all pertinent data collected on precipi-
tators be maintained separately from records for other plant equipment.
Also data should be filed by date. This is to ensure that information is
readily available for analysis and cross-checks without the need to
collate various files or sub-files.
225
-------
Second, wherever possible, a single person or group at the
plant level should be made responsible and held accountable for proper
record keeping and for precipitator performance and reliability.
The reason for this is that the precipitation process depends upon
the interaction of process and equipment variables that are best
understood and controlled by those with hands-on experience.
Finally, operators should recognize that the procedures and
recommendations contained in this paper can readily be incorporated
into a preventive maintenance program; that the information is useful
in identifying real problem areas that require corrective maintenance;
and that the sum total of all the corrective and preventive maintenance
actions the operator performs will ultimately lead to optimum precipi-
tator performance and.reliability.
(1)
The Fault Tree shown in Figure 7 is a partial tree. A complete
tree can be obtained by writing to the authors.
226
-------
Table -1. FUNDAMENTAL ESP OPERATION CONDITIONS
PRECIPITATOR INLET
GAS VOLUME, VELOCITY & UNIFORMITY
GAS TEMPERATURE
GAS HUMIDITY
TYPE AND RANK OF FUEL FIRED
DUST CONCENTRATION
DUST SIZE DISTRIBUTION, RESISTIVITY, AND
OTHER PROPERTIES
CARBON CARRYOVER & OTHER BOILER OPERATING
CONDITIONS
227
-------
Table 2. FUNDAMENTAL ESP OPERATING CONDITIONS
PRECIPITATOR INTERNAL
ENERGIZATION & SECTIONALIZATION
INTER-ELECTRODE ALIGNMENT
DISCHARGE & COLLECTING ELECTRODE RAPPING
& CLEANLINESS
GAS FLOW DISTRIBUTION
AMBIENT AIR INLEAKAGE
HOPPER ASH LEVEL
228
-------
Table 3. PM CHECKLIST FOR A TYPICAL FLYASH PRECIPITATOR
DAILY
TAKE AND RECORD ELECTRICAL READINGS AND TRANS-
MIS SOMETER DATA
CHECK OPERATION OF HOPPERS AND ASH REMOVAL SYSTEM
EXAMINE CONTROL ROOM VENTILATION SYSTEM
INVESTIGATE CAUSE OF ABNORMAL ARCING IN TR
ENCLOSURES AND BUS DUCT
WEEKLY
CHECK RAPPER AND VIBRATOR OPERATION
CHECK AND CLEAN AIR FILTERS
INSPECT CONTROL SET INTERIORS
MONTHLY
CHECK OPERATION OF STANDBY TOP HOUSING PRESSURIZING
FAN AND THERMOSTAT
CHECK OPERATION OF HOPPER HEATERS
CHECK HOPPER LEVEL ALARM OPERATION
QUARTERLY
CHECK AND CLEAN RAPPER AND VIBRATOR SWITCH CONTACTS
CHECK TRANSMISSOMETER CALIBRATION
SEMIANNUAL
CLEAN AND LUBRICATE ACCESS DOOR DOG BOLTS AND
HINGES
CLEAN AND LUBRICATE INTERLOCK COVERS
CLEAN AND LUBRICATE TEST CONNECTIONS
CHECK EXTERIOR FOR VISUAL SIGNS OF DETERIORATION,
AND ABNORMAL VIBRATION, NOISE, LEAKS
229
-------
Table 3. PM CHECKLIST FOR A TYPICAL FLYASH PRECIPITATOR (CONT'D)
ANNUAL
CONDUCT INTERNAL INSPECTION
CLEAN TOP HOUSING OR INSULATOR COMPARTMENT AND
ALL ELECTRICAL INSULATING SURFACES
EXAMINE AND CLEAN ALL CONTACTORS AND INSPECT
TIGHTNESS OF ALL ELECTRICAL CONNECTIONS
CLEAN AND INSPECT ALL GASKETED CONNECTIONS
CHECK AND ADJUST OPERATION OF SWITCHGEAR
CHECK AND TIGHTEN RAPPER INSULATOR CONNECTIONS
SITUATIONAL
RECORD AIR LOAD AND GAS LOAD READINGS DURING AND
AFTER EACH OUTAGE
CLEAN AND CHECK INTERIOR OF CONTROL SETS DURING
EACH OUTAGE OF MORE THAN 72 HOURS
CLEAN ALL INTERNAL BUSHINGS DURING OUTAGES OF
MORE THAN 5 DAYS
INSPECT CONDITION OF ALL GROUNDING DEVICES DURING
EACH OUTAGE OVER 72 HOURS
CLEAR ALL SHORTS DURING EACH OUTAGE
CHECK ALL ALARMS, INTERLOCKS AND ALL OTHER SAFETY
DEVICES DURING EACH OUTAGE
230
-------
Table 4. DATA SOURCE DOCUMENTS FOR PRECIPITATOR
INLET CONDITIONS
BOILER DATA SHEET
COAL ANALYSIS - PROXIMATE (A.R.)
ULTIMATE (A.R.)
ASH MINERAL ANALYSIS
PRECIPITATOR PERFORMANCE TEST REPORTS
MANUFACTURER'S GAS FLOW DISTRIBUTION
TEST RESULTS
231
-------
Table 5. DATA SOURCE DOCUMENTS FOR PRECIPITATOR
INTERNAL CONDITIONS
ELECTRICAL ENERGIZATION READINGS
VOLTAGE - CURRENT CURVES & WAVEFORM DATA
RAPPER OPERATION DATA
ASH REMOVAL SYSTEM OPERATION
INTERNAL INSPECTION REPORTS
REPLACEMENT PART LOG
232
-------
Table 6. ELECTRODE CLEANING SYSTEM DATA
TIMING CYCLE OF CLEANING SYSTEM
INTENSITY OF CLEANING DEVICES
SEQUENCE OF CLEANING DEVICES
233
-------
Table 7. MAJOR INTERNAL INSPECTION ITEMS
AMOUNT OF PARTICULATE BUILDUP ON COLLECTING
AND DISCHARGE ELECTRODE SURFACES
ALIGNMENT BETWEEN COLLECTING ELECTRODES
BROKEN OR SEVERELY WORN DISCHARGE ELECTRODES
WARPED OR BOWED COLLECTING ELECTRODES
MISSING DISCHARGE ELECTRODES
CLINKER FORMATION
CLEANLINESS OF SUPPORT AND ISOLATION INSULATORS
AND SIGNS OF TRACKING
EXTREMELY CLEAN OR DIRTY AREAS VS. NORMAL AREAS
234
-------
RESEARCH-COTTREUL. INC.
BOILER READINGS
Customer ___^
Location
Tima
Megawatts
Staam Flow, lOOW/nr
Stttm fna, la Jtvjt cnirorn*)
Stsam Tamo.' P, tat st»g»
-------
RESEARCH-COTTRELL, INC.
Sat No. .
Rapper lift.
PREC1P1TATOR ELECTRICAL READINGS
lnh N"
C^trrftnar ( ....__
Dura
~^O Nil Air nr rta* 1 rtart
i- i ^
HW or FW .
. amps or MA®.
B^^ amps _
, >ec; AC undervoitaoa crip.
.volts
AUTO OR
MANUAL
I 0
ISO I 200 ! 250 I 300
750 I :QOO I 1500
s sec, xv.
S SSC. XVj
SPK HATE
I
Set No.
Sapper lift-
et Size .
DC overcurrwit trip .
HW or FW
nps or MA j>
Vibrators — arnos
. sec. duration.
. sec: AC undarvoitaga trip.
. volts
AUTO OR
MANUAL
ISSC
S?K a ATS
0 I 50 i 100 i 150 I 200 2SO i 300 I SOO I 750 I 'OOP I 1500
J1PSI. i
i S 3SC. XV i ]
S SSC. xv, ;
i i 1 1 : ! ' i ! i
! 1 1 ! 1 ! i ! i i
1 i 1 1 i i ' i | ;
S«t No.
lift,
Sat Size
OC ovarcurrafit trip .
. HW or FW .
. amps or MA 5.
Vibrators __^ amns _^_ SBC. duration.
.sec; AC jnaervoltags crip.
. voits
AUTO OS
MANUAL
15. !»«!.
I I
JJ SSC.
|SPK 3AT5
SO [ IQQ : ISO ! 200 '• 250 : 300 ! 500 | 75O I IQOO
jl?RI. !
S5SC. KVi i
| s SSC. XV2 !
I 1 i ! ! : '
i I ! 1 ! '• 1 1 ! i
1 > ! ! ! i i ! ; '
I
Sat No
Rapper lift.
Sat Siza
OC ovwcurrent trio .
. HW or FW _
. arnojor MA §_
Vibrators __ imos
duration.
. sac: AC undttrvoltags trip-
AUTO OS
MANUAL
so i irjo ; !H ; 200 j 2so ' :cxi ' =cxj
750 I 1000 I 1500
I S 3EC.
-------
Research Cottrel!
ELECDRED PROGRAM - PRECIPITATOR ELECTRICAL DATA
ro
to
JOB NO.
T
LOCATION
TTTTY
0011 ES N
TT
-
—
-
-
o
SET
NO.
._.
-•
-
-
--
-
-
-
-
~
-
-
—
IT
TT
CUSTOUEK
HW
FW
__
-
-
6
;
-
-
/
i
apt
Tl
PRI
VOLTS
-
-
8
-
_.
ti
-
to
o
[ITT
-
-
-
11
SEC
MA
--
-
-
-
-
-
-
-
_
-•
-
-
-
IT
TIT
It
PR)
AMPS
-
-
_
-
-
-
-
--
_.
-
-
CSI
-
-
TITTTT
1
TmTTTTTITTT
NO.
nr
—
SEC
KV1
-
-
-
—
...
-
-
-
-
-
-
r
•
SEC
KVZ
-
-
._
-
-
*TE
r
SPK/
MIN
--
-
-
-
--
1 1
-
-
-
t
SFCS/
TR
-
-
-
-
-
_.
—
-
-
irm"
UE
IR.
;iP
-
-
-
—
rrm
rrrr
rrriTr
-
COMMENTS
--
-
-
—
—
-
-
-
-
-
MOO
T
-
-
-
ITIT
mnrr
E. AUTO/ U AN
-
-
-
—
-
-
-
-
-
-
-
-
-
Name
Oepl.
—
Tele. No
-
-
-
-
—
-
-
-
-
-
-
-
&5
bb
57
&a
-
ia
-
ut
-
bt
-
-
-
-
w
-
ti'i
M
-
Figure 3. Precipitator electrical data
-------
JOS NO.
TB NO. _
DATE _
C3 HW
CjFW
1500
1400
1300
1200
1100
1000
< 900
300
LU
IT
X
g
700
IT
0
< 600
0.
o
S 500
0.
400
300
200
100
VOLTAGE - CURRENT CHARACTERISTICS
READ O LEFT a RIGHT
0 UL
-H-
TIME
GAS TEMP -
OPACITIES:
RAPPING
j i ill
20 30 40
PRECIP1TATOR VOLTAGE. KiLOVOLTS (KV)
50
T-LB
300
S
8
-" 0
60
IAMPS (MA
ATOR CURRENT.
Figure 4. Voltage-current characteristics
238
-------
OQ
C
PLANT NAME
PRECIPITATOR INSPECTION LOG
REASON FOR INSPECTION
PRECIPITATOR »
DATE
INSPECTED BY
LOCATION OF INSPECTION
I
GAS
PLOW
PRECIPITATOR LAYOUT REFERENCE
ro
CO
UJ
H
(D
O
H-
rt
So
rt
O
ri
fD
o
r+
H-
O
INSPECTION CODE
1 KINKED DISCHARGE ELECTRODE
2 WARPED DISCHARGE ELECTRODE FRAME
3 BROKEN DISCHARGE ELECTRODE FRAME
4 BROKEN DISCHARGE ELECTRODE
5 BOWED COLLECTING ELECTRODE
6 HEAVY DUST BUILD-UP
7 HEAVY CORROSION
8 RAPPER ALIGNMENT OFF-CENTER
9 IMPROPER COLLECTING ELECTRODE SPACING
10 IMPROPER DISCHARGE ELECTRODE TO
COLLECTING ELECTRODE SPACING
GAS FLOW
w
w
D
«
3 5
2 4
11 13 15 17 19 21 23 25 27 29 30 32 34 36 38 40 42 44 u
10 12 14 16 18 20 22 24 26 28 30 31 33 35 37 39 4l 43 45 «J
COLLECTING ELECTRODES Q
-------
ro
->
o
c
i-i
(B
I
o
n>
rr
H-
O
g
I-1
Hh
O
rt
O
PARAMETER CHANGE-
CAUSE
-INLET CONDIT10NS-
Gas Volume
Gas Temperature
Gas Humidity
Type of Fuel
Dual Concentration
Dust Resistivity
Duflt Particle Sice
Carbon Carryover
-INTERNAL CONDITIONS-
Energlzatlon - Voltage
EnergiEation - Current
Collecting Plate Cleanliness
Discharge Electrode Cleanliness
Flow Distribution
Air Inleakage
Hopper Asti Level
Spark Rate
SOURCE
DOCUMENTS
Boiler Data
Coal Analysis
Boiler Data
Boiler Data
Coal Analysis
Test Report
Coal Analysis
Ash Analysis
Coal Analysis
Boiler Data
Ash Analysis
Test Report
Coal Analysis
Tt*Ht Report
Aah Analysis
Electrical Data
V-l Curves
Electrical Data
V-I Curves
Rapper Log
V-I Curves
Electrical Data
Rapper Log
Test Report
Boiler Data
Electrical Readings
Hopper Log
Electrical Data
CHANCES - NOTES
* Increase
- Decrease
+ Increase
- Decrease
+ Better Rank
— Poorer Rank
* Increaae
- Decrease
* Increase
- Decrease
+ Larger 8 Lie
- Smnl UT Slie
+ Increase
- Decrease
* Increase
- Decrease
+ Increase
- Decrease
+ Cleaner
- Dirtier
+ Cleaner
- Dirtier
* More Unlforo
- Less Uniform
+ Mom Leakage
- Lcab Leakage
+ High Level Potential
- Low Level
+ Incrunsu
• DccruiiHH
EFFECT - CHANGED PARAMETERS
i
+
*
E
N
N
N
N
N
N
H
H
N
+
N
N
GAS TEMPERATURE
N
*
F.
H
N
H
N
N
N
H
H
N
+
N
N
[ GAS HUMIDITY
N
*
t
H
H
N
H
N
H
II
H
N
I
N
H
b.
O
W
a.
N
H
H
H
N
N
N
N
N
H
N
II
N
H
»
DUST CONCENTRATION
N
+
I
»
H
H
+
H
N
H
N
II
N
H
N
1 DUST RESISTIVITY
N
+
+
+
N
N
*
N
H
N
N
H
I
N
H
DUST PARTICLE SIZE
N
N
N
t
H
N
+
N
N
II
1!
N
N
H
II
CARBON CARRYOVER
N
N
N
H
H
N
H
N
H
N
N
H
N
H
N
ESERG. - VOLTAGE
N
+
+
(N)
+
-
+
+
+
E
*
•
N
+
*
+
1 ENtRG. - CURRENT
N
+
+
+
«)
+
+
+
*
E
+
+
H
+
+
+
PLATE CLEANLINESS
*
N
*
•r
+
+
*
+
E
E
N
N
H
N
+
1 D.E. CLEANLINESS
+
N
*
+
+
+
-
+
E
E
N
H
H
H
«
1 FLOU DISTRIBUTION
N
H
N
N
N
N
N
H
N
H
N
N
4
N
N
AIR tHLEAKACE
-
+
I
N
N
N
N
N
N
N
N
N
N
»
N
s
at
a.
i
-
N
*
N
-
N
I
»
E
E
N
N
N
N
N
u
-
-
-
+
-
+
+
E
*
+
+
*•
+
*
*•
I PERFORMANCE - OPACITY
*
(N)
E
4
+
*
* •
»
E)
E)
+
*
t-
*•
>
-
Aasumptlona ft Keyi
1, At typicsl full load
2. Cold side ESP
3. Changes are from "normal"
4. Symbols: Opacity: * " In«««
- " Decreaae
H i No Direct Effect
I : Insignificant Effect
E : Either improves or degrades
( ) i Hay occur
-------
ro
1
Highest Order Event
Precipitator Fault Tree
1*1 Or
-n
(D
ENIIVIIM CwfuiMHtl
TtMOUtjiMMl j
a* I
( ) Lowest Order Even!
Hotii UM
OiiUlbuliuii
H 1X1 or Mui. al
ChiHiitMi Fwliii
Fail
4
B(*K
DlCI.H
FwMi
If M(MH
MlMM
F,ul
CwillfNI,
D«W
if.
MI4MI Ol
.LlM*
rfui*
-------
ELECTROSTATIC PRECIPITATORS
START-UP, LOW LOAD, CYCLING
AND MAINTENANCE CONSIDERATIONS
By:
F. A. Wybenga and R. J. Batyko
Babcock & Wilcox Company
Fossil Power Generation Division
Barberton, Ohio 44203
ABSTRACT
Particulate control devices for large fossil fired steam generators, now pre-
dominantly coal fired, are being designed for increasingly higher efficiencies
to meet current and proposed new regulations. Heavy emphasis is placed on
minimizing occurrences when emission limits are exceeded during all phases
of operation or for reasons of malfunction of the control device.
These requirements make it desirable, if not necessary, to energize electro-
static precipitators during the early stages of boiler start-up. This also
calls for planned, regular maintenance.
This paper discusses precipitator design and operating considerations that
will allow energization during initial coal firing of precipitator equipment.
With start-up and shut-down becoming of even greater importance on boiler units
designed for cycling service, the paper also addresses items to be considered
in design of precipitators, precipitator accessory equipment and the total
boiler system to facilitate cycling operation and extended low load operation.
242
-------
ELECTROSTATIC PRECIPITATORS
START-UP, LOW LOAD, CYCLING
AND MAINTENANCE CONSIDERATIONS
INTRODUCTION
With increasingly stringent limits being placed on particulate emissions
from fossil fired steam generating equipment, and the need for reporting
instances when emissions exceed established limits, additional emphasis has
to be placed on maintaining clear stacks over the full range of operating
conditions. In this paper, we will be discussing electrostatic precipitators
as applied to coal fired utility boilers, particularly what is being done
and what can be done to allow precipitators to perform reliably over a full
range of operating conditions. In addition to the normal start-ups, these
operating conditions may include, either initially or in the future, frequent
start-ups and shut-downs for load cycling and extended low load operation.
All will likely result in low flue gas temperatures in back-end particulate
control systems, considered undesirable for obvious reasons.
There are limits to what can be done to particulate control devices,
whether precipitators or bag filterhouses, to make them suitable for low gas
temperature operating conditions. For this reason, more emphasis has to be
given to the "back-end" equipment in a power plant when the total boiler system
is planned and designed. If precipitators are expected to handle whatever
is produced by the power production process, poor performance may be expected
if there are equipment incompatibilities. Poor performance and resulting
dirty stacks are no longer acceptable. Consequently, the particulate control
devices as well as all other pollution control equipment have to be given
equal consideration in the total plant design.
The pollution control problems facing the utility industry are numerous
and complex. Considering the varied needs of the utility companies, the
variety in available fuels, local environmental regulations and economic
restrictions, the solutions will be different for each installation. In this
paper, we do not claim to solve these problems but is is our intent to frankly
state what precipitators can do, based on our experience, and also address
what they cannot be expected to do. For the latter case, we will be giving
some food for thought to the users, their consulting engineers, and the
equipment suppliers so that, together, solutions can be found and incorporated
in the total plant system.
While it may appear from some of our statements that we are taking the
total industry to task for past omissions, this is not the case. What has been
done in the past was done for good reasons, and in general, the objectives
strived for were achieved. We now find ourselves in different situations and
in a different environment which may call for a different approach from that
followed in the past.
243
-------
BOILER "BACK-END" SYSTEM DESIGN PRACTICES
Before we discuss electrostatic precipitators and their operation, it may
be well to review normal design practice for the boiler "back-end" system i.e.,
all equipment located downstream of the boiler economizer. Refer to figure 1.
Boiler
i
i
Gas
Air
1
OeJS
Bypass
Air
Heater
1
"1
Electrostatic
Precipitator
WW
Air Bypass
F.D. Fan
To Stack
or
I.D. S02 Removal
Fan System
Steam
Coils
Figure 1 Boiler "back-end" system.
Exit Gas Temperature
When preparing utility boiler specifications, it is customary for the
owner/user or his Consulting Engineer to select and specify the air heater
exit gas temperature at full load operating conditions. This value is
selected on the basis of the fuels to be burned and overall economics. With
rapidly rising fuel costs and 40 F gas temperature being roughly equal to 1%
in boiler efficiency, the lowest possible exit gas temperature is of course
preferred. Typically, exit gas temperatures are in the 260 -300 F range for
coal fired units.
Air Heaters
Air heaters are selected and sized by the boiler manufacturer to produce
the specified exit gas temperature and the air outlet temperatures required
for the fuel preparation and burning equipment. With air heater size estab-
lished, a review is made of air heater material requirements and corresponding
minimum cold end average temperatures. For regenerative air heaters with
corrosion resistant low-alloy steel in the cold end elements, minimum average
metal temperature recommendations are shown in Figure 2. The minimum temper-
ature limits are a function of the flue gas acid dewpoint, hence the increase
244
-------
as sulfur content in the fuel increases. By maintaining these minimum temper-
ature limits, air heater corrosion can be kept to a minimum.
Min. avg. cold
end temp.
°F
1
240-
220-
200-
180-
160-
140.
Oil Firing
/ Curves for low alloy
x steel cold end elements
Bit. Coal Firing
0123456
% Sulphur in fuel
Figure 2 Air heater minimum average
cold end temperature guide.
Air Heater Cold End Temperature Control
Controlling cold end average temperatures over a normal range of boiler
loads requires either an air by-pass or auxiliary heating at the air heater
air inlet. Typically, either steam, hot water, or glycol heaters are installed
at the air heater air inlet for this purpose. They are sized to handle not
only a load range but also the anticipated range of ambient air temperatures.
Electrostatic Precipitators
Precipitators are normally purchased for full load operating conditions.
This is certainly the correct approach if collection efficiency under normal
operating conditions, at or near full load, is the main consideration. How-
ever, as new units are added to the system, the older installations are now
frequently called on for cycling duty or, if not suitable for cycling service,
kept at extremely low loads for extended periods. It can be expected that
units purchased today for base loading will at some time in the future also
be used in cycling service.
In a recent survey of the utility industry conducted by Babcock & Wilcox
Company, we asked what operating and maintenance problems had been experienced
on boiler back-end equipment used for cycling or low loads. All those respon-
dents who had any appreciable experience with this type of operation reported
problems that can be broadly described as "dew point" problems. These include
problems with air heater pluggage and corrosion, precipitator fouling with
resulting poor performance, ash handling system pluggage and corrosion of
flue expansion joints and stacks. Mostrespondents expressed the opinion that
the solution to these problems is to maintain gas temperatures above the acid
245
-------
dewpoint under all operating conditions. Steam coils were the most frequently
offered suggestion, followed by air heater gas or air by-pass systems pr even
an economizer by-pass. Better materials, suitable for below dewpoint operation,
were suggested by 20% of the respondents.
The results of this survey reinforce our opinion that start-up, cycling,
and low load operation, as they affect precipitators and other back-end equip-
ment, should be given full consideration when new systems are designed and
equipment is selected. For units not initially slated for this type of service,
economics may well dictate that those units be designed with limited initial
flexibility. Even so, major equipment, not readily modified, should be
selected with the future in mind.
START-UP OPERATIONS
For precipitators, the preferred solution to minimizing dewpoint and
moisture problems has traditionally been to delay energizing the precipitator
until gas temperatures are above 200 F. Even though the equipment still has
to pass through the dewpoint transition, delayed energization will minimize
the deposition of potentially wet and sticky materials. Unfortunately, federal
and/or local regulations may no longer permit this mode of operation if it
results in particulate emissions above allowable limits. This then points to
the desirability of energizing electrostatic precipitators as soon as coal
firing is begun. We see no reason why this cannot be done with properly
designed equipment.
Pre-Start-up
In preparation for precipitator energization upon first coal firing, our
experience indicates that certain pre-start-up procedures should be followed.
For any given installation, precipitator operating instructions manuals should
be consulted for specific recommendations.
Prior to boiler start-up, preferably eight hours or longer, all rapping
systems on the precipitator should be placed in operation. Also, all heating
systems on the hoppers and insulators, if provided, should be energized. This
will allow hoppers and insulators to heat up and stay dry through subsequent
start-up operation. Seal or purge air fans, if provided, should also be
started early.
Hopper ash removal systems should be ready to operate and preferably in
service to remove all accumulated ash.
Boiler Purge
Regardless of all other considerations, we consider it of utmost importance
that the precipitator power supplies be kept de-energized whenever explosive
mixtures may be present within the boiler system setting. While documented cases
of precipitator gas side explosions are rare, sparking from any source, whether
fuel ignitors or precipitators, should not be allowed during any boiler purging
operation or whenever the boiler master fuel trip prevents firing. We, there-
fore, recommend that all precipitator power supplies be interlocked with the
master fuel trip and firing permissive.
246
-------
Initial Firing
Typical utility practice is to use oil or gas ignitors or start-up burners
for the initial firing of the boiler. As long as these fuels are burned
exclusively and good combustion conditions are maintained, there should be no
need for precipitators to be energized for stack appearance reasons. It would,
therefore, be our recommendation to keep precipitators de-energized while
maintaining good combustion with the auxiliary fuel, both for stack appearance
reasons and protection against air heater deposits and the possibility of
fires.
Coal Firing
Exactly when coal firing can be initiated will be a function of the indiv-
idual boiler unit, its burner design, coal characteristics, and the ability to
obtain adequate combustion air temperatures to sustain good combustion, with
or without auxiliary fuel support. Cost and availability of auxiliary fuel
and load carrying capacity on auxiliary gas or oil will also have to be con-
sidered. Assuming that coal firing will begin at or near unit synchronization
and that precipitators are to be energized at this time, gas temperatures will
still be quite low at the air heater outlet as shown in Figure 3 for a typical
coal fired unit. The curve shows exit gas temperature rise above entering air
temperature.
100
Exit gas
temp, above
air temp.°F
60
25% Load,
Synchronize
Roll Turbine
Figure 3
Time from light-off
Exit gas temperature rise above air
inlet temperature during initial
operation.
Exit gas temperature being directly related to air heater air inlet
temperature then points to the desirability of having high air preheating
capability, preferably steam or water coils. There are other means of accel-
erating the air heater gas outlet temperature rise such as air or gas by-
passing or even completely stopping the air heater rotation. All of these
will also have the undesirable effect of drastically reducing combustion air
temperatures. Air heater gas by-passing will also reduce air heater cold
end temperatures. Inlet coils are, therefore, the preferred solution.
247
-------
Figure 3 also points out the desirability of shortening the boiler warm-
up phase and, once synchronized, raising unit load as quickly as possible.
This will minimize the length of time during which the precipitator has to be
exposed to low gas temperatures.
With generally lower acid dewpoints with coal firing as compared to oil
firing, and the inherent neutralizing effect of coal ash, anything that can
be done to minimize the auxiliary oil firing percentage will also be helpful
to the precipitator.
Precipitator Design Considerations
Continuing on the assumption that the precipitator will be energized
upon initial coal firing, the design has to be capable of handling this
operation. It has been our experience that properly designed precipitator
equipment can readily handle this situation. Essential design features are
discussed below.
Hopper and Insulator Heating Systems. An adequate heating system on all
ash hoppers is a must for early start-up capability, preheated hoppers will
minimize flue gas condensation problems in the critical lower portion of the
hoppers and assist with ash removal.
A heating system on all electrical insulators exposed to flue gas and
fly-ash is equally important. This heating could either be direct or in
conjunction with the seal/purge air system used on insulator components.
Insulators should preferably be kept dry and clean, although dry ash or dust
deposits, except with extremely low resistivity dust, will probably not cause
any problems.
Insulator Materials, Location, and Arrangement, when possible, electrical
insulators should be kept out of the dirty gas stream. The most troublesome
arrangement is to have flue gas on one side or surface with colder ambient
conditions on the other side. Condensation and resulting damp deposits are
now most likely to occur, unless adequate heating is provided on the colder
side.
Support insulator systems frequently used on the B&W precipitator con-
sist of the primary support insulators completely outside the gas stream as
shown in Figure 4. Even with this arrangement, a sealing (shroud) insulator
is required around the hanger rod. By externally heating the shroud insul-
ator, deflecting flue gas away from the insulator and introducing some purge
248
-------
air, this insulator can be kept clean and dry.
Sealing
Insulator
A
Electric
Heating
Figure 4 Support insulator arrangement.
Concerning insulator materials, we are of the opinion that only the best
available materials should be specified. This is undoubtedly alumina. Besides
having excellent dielectric properties, it is extremely strong and can better
withstand temperature variations and even tracking and its potential damaging
effects.
Rapping Intensity. Of primary importance to precipitator operation at
low gas temperatures is heavy rapping. Most European design precipitators
have rapping systems that were developed for difficult to remove, high resis-
tivity, ash. This also makes them better suited to also remove caked on
deposits collected during low temperature operation.
Our experience indicates that heavy rapping can adequately remove start-
up deposits. While actual removal may not take place until the deposits have
dried, subsequent precipitator operation and performance are not significantly
affected. Depending on the frequency of start-ups, time required at low
temperatures and sulfur content of the fuel, there is still the possibility
that hardened, difficult to remove deposits could accumulate to the point
where optimum performance can no longer be maintained. Water washing may
then have to be used to restore performance. We have not had occasion to
recommend water washing on any of our utility precipitator installations but
are familiar with a German installation that water washes periodically. This
unit typically has 20 start-up cycles yearly and maintains precipitators
energized throughout start-up.
249
-------
Heavy rapping impact on collecting plates is as much a function of rapper
design as it is of collecting plate support and attachment. Transverse
acceleration (perpendicular to the plate surface) is most effective in removing
deposits. A rigid means of attachment for each plate will cause most of the
available rapping force to translate into a transverse cleaning component.
On designs where individual plates are loosely supported, much of the avail-
able rapping force will go into swinging the entire curtain. The resulting
decrease in transverse acceleration values has been demonstrated in tests on
collecting curtains. As shown in Figure 5, impact values decreased 40% when
plate fixation bolts were loosened.
Collecting plate
height - Ft
50
40
30
20
10
Plate Fixation
Bolts Loose
Plate Fixation
Bolts Torqued
50 100 150
Plate rapping impact (transverse)-g
200
Figure 5 Effect of plate fixation on rapping
impact.
While transverse acceleration values are of primary importance, the
acceleration component within the plane of the plate (parallel) can also be
very beneficial if plates are rolled with shear planes and ridges to
facilitate removal of deposits.
The overall importance of adequate rapping impact on collecting surfaces
is now generally recognized and specifications frequently call for minimum
accelerations of 50g to lOOg. Not always specified is the direction in which
it is to be measured, transverse or parallel to the surface. How meaningful
actual reported impact data is to the user is an unresolved question. For
example, it has been demonstrated that a lower weight accelerometer will give
250
-------
higher readings, all other conditions being the same. This is an area where
some industry standards would be helpful so that test results using different
procedures and instrumentation can be quantified and comparisons made.
High impact, low amplitude rapping has given excellent cleaning results
on our units now in service. Falling hammer type rapping systems and rigidly
fixed collecting plates can combine to give these results. Regardless of
how specified, the important consideration has to be the ability of the rapping
system to keep collecting plates and electrodes clean, especially for start-up
operations.
For discharge electrodes, rapping impact may not be as important as on
plates, nevertheless, it should be adequate to keep electrodes reasonably
clean for optimum power input. Rigid discharge frames with tightly strung
electrodes and individual falling hammer type rappers have excellent cleaning
characteristics. Sharp edge, sheared discharge strips will also facilitate
cleaning.
Precipitator Operating Procedures
One possible option available to the operator to minimize harmful effects
from start-up operations on the precipitator is to isolate certain chambers
or casings, if adequate dampers have been provided. This type of operation
will expose only selected portions of the precipitator to start-up conditions.
Another approach commonly used is to allow gas to pass through the
entire precipitator while energizing only a single field. The initial field
to be energized should be the last field since it will collect the least
amount of ash during normal operation and any remaining ineffectiveness due
to start-up deposits will have a minimum effect on subsequent operation.
On the other hand, there are some who believe that the leading field
should be the first one energized, since subsequent operation with coarser
ash will help to scour off remaining start-up deposits. Whichever approach
is followed, partial energization can be beneficial. It does, however,
require more operator involvement especially if remote start-up capability
or added automation has not been provided.
Summarizing, we firmly hold to the opinion that routine boiler start-
ups can be accomplished with precipitators in operation whenever coal is
fired. This is based on proven capability and experience with properly
designed equipment.
LOW LOAD OPERATION
As covered in the earlier section of this paper dealing with back-end
system design practices, if there is a need for extended low load operation,
the total system has to be specified and designed accordingly. Current practice
calls for predicting boiler performance down to 25% of full load. This does
not necessarily insure, however, that exit gas temperatures are high enough
for satisfactory precipitator performance down to these low loads.
251
-------
Figure 6 shows a plot of gas and air temperatures for a typical coal
fired unit over a wide load range. Steam coils are provided for air heater
cold end average temperature control. For this example, the coils would be
in operation whenever inlet temperature is decreased below 80 degrees F
and/or unit load drops from full load in order to maintain an average cold
end temperature of 180 degrees F, typical for 3% sulfur coal.
Temperature-°F
300-
200
100
Minimum for
3% Sulphur
No
Control
AH Air Inlet
With control of
avg. AH cold end
temp, at 180°F.
25
50
Boiler load-9
75
100
Figure 6 Exit gas temperature vs. load.
Much has been written on acid dewpoint as related to fuel sulfur content
and at what temperatures moisture and corrosion become troublesome. With 3%
sulfur coal, a reasonable minimum gas temperature is on the order of 250 F.
Referring back to Figure 5, this temperature level for our typical unit is
reached at 50% load with steam coils in service. Operating for extended
periods at even lower loads can now lead to severe fouling problems,
corrosion, and ash handling problems.
System Design Solutions
The most obvious solution is to provide additional inlet coil air
preheating capacity which will in turn increase exit gas temperature. Gas
temperatures have to be increased if the back-end equipment is expected to
operate successfully at these low loads.
Precipitator Design Solutions
Design features such as hopper and insulator heating systems and high
rapping intensity as found necessary for early start-up capability are equally
needed on units expected to operate at low loads. Even so, these will not
allow operation below the dewpoint for extended periods.
With operating temperatures approaching dewpoint, the main concern now
becomes corrosion. Metal surfaces exposed to gas on one side and air on the
outside are likely to operate considerably below gas temperature and conden-
sation will occur. While this may not be detrimental to operation, it can be
extremely damaging over the long term. Possible solutions are given below:
252
-------
Corrosion Resistant Materials. Low alloy steels have proven improved
resistance to acid dewpoint attack. Using these materials for the precip-
itator shell and hoppers can be of definite advantage. We question the value
of using similar materials for precipitator internals, if this choice is made
on the basis of extended low load operation only. If gas temperatures are
kept high enough to keep the ash dry, as it should be, then corrosion of
precipitator internals should be no problem either.
Insulation. Thermal insulation on back-end equipment is normally kept
to a minimum since heat losses no longer affect system efficiency. If low
load operation at marginal temperatures is being considered, insulation
thicknesses should probably be specified at 3 or 4 inches, double layer prefer-
ably, rather than the 1 or 2 inches now normally used to provide personnel
protection. There should be an emphasis on good workmanship, thermal barriers
and no voids around penetrations thereby eliminating cold spots on all casing
surfaces.
CYCLING OPERATION
By cycling operation, we refer to frequent nightly and weekly start-up
and shut-down. As applied to boiler equipment, a cycling unit would typically
be designed for 200 full start-up and shut-down cycles over the life of the
unit, five nightly cycles per week with boiler pressure decaying down to
500 psig and one weekend cycle per week with boiler pressure decaying down to
100 psig. For 30 years of plant life, this could add up to 9,560 cycles.
The normal start-ups, 200 or over the 30 year life time or 7 per year,
can be handled with properly designed equipment as already described. To
handle potentially six start-ups per week, even though they are hot restarts,
is more of a problem. There are few coal fired units operating in this manner
and operating experience is limited. To be sure, "dewpoint" problems could be
severe unless system design is carefully considered.
Precipitator Design Considerations
Design features as required for early start-up or low load operation
should be incorporated for cycling service. Also to be included are corrosion
resistant collecting plates and discharge electrodes primarily for out of
service corrosion reasons.
Whereas seal air/purge air fans for insulator compartments are normally
not required on negative pressure units, such fans should be called for on
cycling units. During outage periods with induced draft fans shut down, fly-
ash will drift into insulator compartments and eventually fill them. A
continuous flow, even though small, of heated air is needed to keep these
areas purged.
253
-------
System Design Considerations
isolation. One definite requirement for cycling equipment subject to
frequent hot restarts is a need to maintain residual heat. Good thermal
insulation and tight isolation dampers at system inlet and outlet would be
important.
Gas Temperature Control. Controlling exit gas temperatures at
reasonable levels for all operating conditions should be investigated, Later
vintage cycling boilers are now frequently sold with turbine steam by-pass
systems. Although these systems as designed result in minimum quantities
of available excess steam, some steam should be made available to supply a
large air heater steam coil.
Hot Side Precipitators. The other item that should be given further
consideration is the use of hot side precipitators for cycling duty. Normal
pressure decay in a cycling boiler is such that temperatures in the vicinity
of the economizer may still be above 300 F even for a 32 hour weekend shut-
down. This could well make the choice of hot side precipitators attractive,
but not for reasons of collectability of ash as was promoted several years
ago. The undesirable aspects of hot side precipitators such as higher gas
volumes and system temperature loss would have to be weighed against the
advantage of having the precipitator located in a more favorable gas
temperature zone.
MAINTENANCE
How well equipment operates is highly dependent upon the amount of mainte-
nance performed, especially preventive maintenance. With precipitators not a
part of the main process of generating power, past industry practice has been
to provide only a minimal amount of maintenance. With the current legislative
environment and the absolute necessity of meeting EPA emission limits or paying
fines or reducing boiler load, the need for maintaining this equipment in good
operating condition is a must.
Qualified and trained personnel should be assigned to inspect the electro-
static precipitator and also be given the responsibility to schedule and oversee
required repair work and a preventive maintenance program. External inspections
of the precipitator should be carried out on a routine basis with internal
inspections whenever the unit is out of service. The precipitator vendor
normally supplies operating and maintenance instructions. Using these instruc-
tions and with the vendors' help, an inspection and preventive maintenance
schedule should be developed. If the schedule is followed and adequate
maintenance performed, cases of malfunctioning equipment causing excessive
emissions can be kept to an absolute minimum.
254
-------
For personnel training, many utilities now have vidio tape equipment and
many vendors have the ability to provide both operating and maintenance train-
ing tapes. B&W can provide s-uch training tapes for our precipitators. With
the normal turnover in operator and maintenance personnel, vidio tapes can be
of tremendous value in bringing new people quickly up to date on maintenance
requirements.
Another important consideration is keeping good records of both operating
and maintenance data. Daily recording of current and voltage characteristics
on power supplies should be obtained and tracked. This will detect changes in
operating conditions and can point to specific areas to be inspected during
subsequent outages,
CONCLUSION
Long term reliability of electrostatic precipitators and continued good
performance over a wide range of operating conditions are now a matter of
necessity. Precipitators have to be selected with design features and accessory
equipment to allow operation at less than ideal operating conditions. On
installations being designed for cycling service or extended low load operation,
it is essential that full consideration be given to the design and operation of
back-end equipment.
Once decisions have been made on what special features are needed in the
equipment to be selected, these should be carefully specified to potential
vendors so all will be bidding to the same needs. The air pollution control
industry is an extremely competitive business with many qualified bidders.
Few vendors are willing to incorporate more expensive options, whether they be
better materials, added features or systems, to provide additional long term
reliability unless such options are either specified or will be given favorable
evaluation. The situation in the boiler industry is no different. If, for
example, during the planning stages for a new plant, it is deemed necessary
to maintain a high exit gas temperature over the load range, this will have
to be specified to the boiler vendors. A mere statement in equipment speci-
fications requiring that the "equipment shall be suitable for a wide range
of loads with any fuel(s) specified" will most likely not produce the desired
end result.
Once purchased and installed, electrostatic precipitators have to be
given equal consideration in the total plant maintenance effort. Precipitators
are now a vital part of the total system and have to be treated accordingly.
255
-------
ELECTROSTATIC PRECIPITATOR EMISSION AND OPACITY
PERFORMANCE CONTROL THRU RAPPER STRATEGY
By:
William T. Langan
Jon H. Oscarson
Envirotech Corporation
Buell Emission Control Division
Lebanon, Pennsylvania 17042
Scott Hassett
Utah Power and Light Corporation
Salt Lake City, Utah 84110
ABSTRACT
The opacity and emission performance of two "identical" cold-side
precipitators at a power plant utilizing a Western coal source was
investigated through an extensive test program. The test program included
establishment of outlet emission versus opacity correlations for the two
"identical" units, correlation of rapper strategy (frequency and amplitude)
with opacity, and correlation of power-off rapping strategy with opacity.
Analysis of the test program data demonstrates that the opacity-
emission correlations from the two "identical" units can be significantly
different. The potential reasons for the differences between the two units
are identified and examined in light of the test program data.
Strong correlation of opacity with rapping strategy was developed from
the test program. Results demonstrate that an automatic power-off rapping
technique can significantly enhance precipitator performance on ash from
Western coal.
256
-------
ELECTROSTATIC PRECIPITATOR EMISSION AND OPACITY
PERFORMANCE CONTROL THRU RAPPER STRATEGY
INTRODUCTION
The Utah Power and Light Corporation has two "identical" cold-side
electrostatic precipitators at the Huntington Station in Huntington, Utah.
Utah Power and Light Corporation required these electrostatic precipitators
to be designed to a higher performance level than the then existing EPA
regulatory of 0.1 Ib/lO^ BTU. Utah Power and Light Corporation demanded
this higher than required performance to ensure that the scenic Utah
surroundings would not be degraded by pollutants. To meet this goal, Buell/
Envirotech designed the precipitator units to achieve a 99.5$ collection
efficiency, which corresponds to an emission rate Of 0.05 lb/10° BTU.
Each of the precipitator units was designed to collect fly ash
resulting from the firing of the same low sulfur coal.
The coal is mined from the Hiawatha seam of Deer Creek mine in Emery
County, Utah.
Not only are the station's two precipitators duplicate units, but
the other equipment in the system from the boiler to the precipitator is
also identical. Utah Power and Light Corporation ordered duplicate equipment
to ensure that the second system would achieve the same high level of
performance/operation as the initial system.
The Unit #1 electrostatic precipitator was started up in June, 1977,
and the Unit #2 electrostatic precipitator (see Figure //I) was started
up in August, 1974. The precipitator efficiency has been measured at
99.6$.
PROGRAM DESCRIPTION
The present program was initiated to determine the opacity resulting
from normal operation of the system and to assess the influence of the
collection system rapping strategy upon the opacity.
To obtain these data, performance tests were conducted on one-half
(chamber) of each precipitator. The opacity measurement was obtained by
a portable Lear Siegler instrument installed in the outlet ductwork. The
emissions were obtained by use of EPA-17 test method at a test port
located slightly upstream of the opacity measurement.
The emission-opacity relationship is a complex function of many
factors. These factors include particle size distribution, moisture
content of flue gas, and representative light path of measurement.
Correlation of these factors has resulted in the following opacity
relationship (Reference l).
257
-------
In (I/Io) -
Where w is outlet grain loading
I/Io is light transmittance
L is light path length
P is average particle density
K is extinction coefficient, a function of the
particle size for a given refractive index.
For this study, an emission-opacity correlation was experimentally
determined for each unit. There are several approaches that may be
practically employed to obtain such a relationship. These approaches
include: reduction of the precipitator's effective specific collection
area by de-energizing various electrical sections, variation of the
precipitator's effective specific collection areas by variations of the
boiler load, and alteration of the collecting system rapping strategy.
Of these possible approaches, the alteration of collecting system
rapping strategy was employed to obtain emission-opacity correlation for
each unit.
This approach was appropriate for the Huntington Station since the
fly ash exhibited adhesive characteristics. This characteristic principally
resulted from the level of sodium in the fly ash. With modification of the
collecting plate rapper system strategy, additional fly ash is allowed to
accumulate on the collecting plates.
This results in a reduction of the effective collection field applied
voltage caused by the voltage drop through the dust layer (Reference 2).
This phenomenon is illustrated in Figure #2. The practical implementation
of this method indicated that the change of one steady state level fly ash
layer to another was achieved within one day after changing the particular
rapping approach. This is illustrated in Figure #3 which presents the
opacity level versus time for two different rapping approaches. The opacity
corresponding to each particular rapping strategy was measured for one week
before changing to a different rapping strategy.
During the emission-opacity correlation tests, coal samples, boiler
conditions, fly ash samples and opacity measurements were all monitored.
Two measurements were taken for each of four different collecting system
rapper conditions for both the Unit #1 and the Unit #2 precipitator.
EMISSION-OPACITY CORRELATION
The results of the emission-opacity correlation tests for Huntington
#1 are presented in Figure #4. The measured opacity values shown in this
figure correspond to the outlet duct location, rather than the stack, since
only one-half of the precipitator was tested. If the unit were performing
identically in both chambers, the stack opacity would be roughly 1.7 times
258
-------
the duct opacity, and the stack emissions would be twice the duct emissions.
Because the opacity meter was calibrated on clean outside air, the
data have been fitted with a special case of the method of least squares
that requires the curve to pass through the origin. The resulting corre-
lation (Figure #4) is a good fit through the data with little scatter.
The mean coal and ash analyses for the samples taken during the
Huntington #1 emission-opacity tests are shown in Table I. These data
indicate that there were no major variations in the coal or ash properties
during the test series. The particle size distribution and the resistivity-
temperature dependence of representative fly ash sample obtained during
the test program are presented in Tables II and III, respectively. The
results of the laboratory bulk electrical resistivity measurements demon-
strate that range of resistivities are within acceptable values for
effective precipitator performance.
The results of the emission-opacity correlation tests for Huntington
#2 are presented in Figure #5. The curve fitting approach utilized was
the same technique employed for Huntington #1. However, the resulting
correlation is not the same quality of that for Unit #1, since most of
the emissions and opacities are quite low. The mean coal and ash analysis
for the emission-opacity correlation tests for Huntington #2 are presented
in Table IV. Examination of the.se data indicate there were no major
changes in either coal or fly ash characteristics during the test program.
The measured particle size distribution for Huntington #2 is given in
Table V, along with the loss on ignition and specific gravity values. The
measured electrical resistivity from a representative fly ash sample for
Huntington #2 is shown in Table VI.
Comparison of resulting emission-opacity correlations from Unit #1
and Unit #2 show a large difference between the two units. Examination of
the differences between the coal and fly ash samples for Unit #1 (Table I)
and Unit #2 (Table IV) does not reveal any characteristic difference that
would adequately explain the differences between the units.
A comparison of particle size distributions between the two units
(Figure #6) indicate that Unit #2 exhibits significantly higher particle
sizes than Unit #1. These particular size distribution measurements were
taken at the inlet of the precipitator in an attempt to resolve the
difference in emission-opacity correlation that apparently exists between
the two units. The fact that Unit #2 has a higher particle size than Unit
#1 qualitatively explains the fact that Unit #2 has a lower opacity than Unit
#1 for a given emission level. Particle size influence upon the opacity is
larger than one would estimate. This anomaly may result from an insufficient
number of data measurements at higher opacity values for Unit #2.
Examination of the hardware at the two "identical units" revealed that
mill differences developed between the two units due to operational condi-
tions. This situation explains the different particle size distribution
259
-------
between the two units.
EMISSION-POWER RESULTS
The correlation of emissions versus precipitator power was derived
from the present test program. The data and correlation for the emission-
corona power for Huntington #1 and #2 are presented in Figures #7 and
#8, respectively. These data show an asymptotic approach to limiting
emission as power is increased. This indicates that, although power may
be added, it is not useful in terms of enhanced collection efficiency.
That is, the increase in power (through current, not increased voltage)
does not effectively improve precipitator performance.
This was verified through modeling of the precipitator performance
through use of the Southern Research Institute Mathematical Model
(Reference 3). Analysis confirmed that increased corona power delivered
through increased current, not voltage, may not significantly enhance
precipitator collection efficiency.
RAPPING STRATEGY
Since the resulting emissions were found to be a function of the
rapping approach, the technique of power-off rapping was applied to
assess the potential performance enhancement for a given rapper intensity
level.
The technique of power-off rapping is to de-energize an electrical
section and rap the collecting plates. This approach has the potential
advantage of enhanced fly ash removal from the plates, since the electrical
pinning force is removed during the rapping sequence and minimum fly ash
is collected on the plates since the applied collection field is removed.
This approach is only attractive with highly sectionalized electrostatic
precipitators.
The power-off rapping technique was applied to Unit #1. Comparison
of the emissions indicate the power-off rapping technique obtains the
same opacity with rapper levels that are only 60% of those required for
power-on rapping. Test results from the power-off rapping technique are
presented in Figure #9.
SUMMARY
The opacity-emission correlation from two "identical" cold-side
electrostatic precipitators operating on ash from low-sulfur Western
coal were found to exhibit significant differences. The particle size
distribution difference between the units was found to contribute to the
opacity-emission differences. This was found to result from mill
difference between the two units. Additionally, correlation was found
to be sensitive to the opacity levels upon which the correlation was
derived.
260
-------
The emission-corona power correlation demonstrates asymptotic
emission limit with increasing power. Data demonstrates power levels
may be achieved after which additional power will not significantly
enhance precipitator performance.
The technique of automatic power-off rapping has indicated sig-
nificant enhancement of precipitator performance on low-sulfur coal.
Rapping intensity, using the power-off rapper technique, can be only
of the normal rapping level to achieve equivalent performance.
261
-------
TABLE I
COAL AND ASH ANALYSES FOR EMISSION-OPACITY
CORRELATIONS AT HUNTINGTON NO. 1
Moisture
Ash
Sulfur
Si02
Al20o
Ti02
Fe203
CaO
MgO
Na20
K20
P205
so3
Base/Acid
Silica/Alumina
Mean
2.50
13.57
0.51
60.48
18.66
0.81
3.31
5.23
1.33
2.50
1.34
0.24
4.33
0.18
3.24
Standard Deviation
0.10
0.50
0.02
0.88
0.51
0.11
0.13
0.68
0.23
0.18
0.08
0.02
0.16
0.01
0.10
262
-------
TABLE II
UTAH POWER AND LIGHT COMPANY, HUNTINGTON #1
SUMMARY OF PARTICLE SIZE DISTRIBUTION,
LOSS ON IGNITION AND SPECIFIC GRAVITY
Hopper Sample
+149
+ 45
+ 40
+ 35
+ 30
+ 25
+ 20
+ 15
+ 10
+ 5
- 5
Specific
Gravity
Loss on
Ignition
y
y
y
y
y
y
y
y
y
y
y
Test #4
0.43*
6.1
7.1
8.2
9.8
11.8
15.8
21.8
37.0
58.5
a. 5
2.2076
1.56*
Test #5
0.34$
4.6
5.2
6.2
7.6
9.0
11.7
17.0
29.9
50.0
50.0
2.3679
1.74*
Test #6
0.36$
4.9
5.6
6.2
7.3
8.8
11.7
17.2
31.0
50.1
49.9
2.1952
1.72$
Test #7
0.26$
8.8
9.9
11.1
13.0
15.5
19.0
25.5
40.2
57.2
42.8
2.1285
1.38$
Test #8
0.45$
8.3
9.4
10.6
12.2
14.7
18.8
25.0
40.0
62.2
37.8
2.1277
1.45$
263
-------
TABLE III
UTAH POWER AND LIGHT COMPANY, HUNTINGTON #1
SUMMARY OF BULK ELECTRICAL RESISTIVITY
Temperature (°F) 280 300 340
Test #4
Resistivity (OHM-CM) 9.00 x 109 6.79 x 109 3-35 x 109
Breakdown Voltage (VOLTS/CM) 1.70 x 10^ 1.30 x 10 1.40 x 104
Moisture (%} 6.12 6.10 6.10
Test #5
Resistivity (OHM-CM) 1.00 x 1010 7.50 x 109 6.08 x 109
Breakdown Voltage (VOLTS/CM) 1.50 x 10^ 1.60 x 10^ 1.40 x 10^
Moisture (%} 6.16 6.13 5.91
Test #6
Resistivity (OHM-CM) 9.53 x 109 6.18 x 109 3.29 x 109
Breakdown Voltage (VOLTS/CM) 1.50 x 10^ 1.60 x 10^ 1.30 x 10^
Moisture (%} 6.21 6.26 6.08
Test #7
Resistivity (OHM-CM) 1.48 x 1010 8.45 x 109 6.57 x 109
Breakdown Voltage (VOLTS/CM) 1.70 x 104 1.60 x 104 1.50 x 10^
Moisture (%} 6.08 6.22 5.90
Test #8
Resistivity (OHM-CM) 1.46 x 1010 8.77 x 109 4.92 x 109
Breakdown Voltage (VOLTS/CM) 1.50 x 10^ 1.80 x 10^ 1.60 x 109
Moisture (%} 6.10 6.22 6.12
264
-------
TABLE IV
COAL AND ASH ANALYSES FOR EMISSION-OPACITY
CORRELATIONS AT HUNTINGTON NO. 2
Moisture
Ash
Sulfur
Si02
A1203
T102
Fe203
CaO
MgO
Na20
K20
303
Base/Acid
Silica/Alumina
Mean
1.59
11.86
0.49
59.03
19.55
1.08
3.62
6.86
1.62
2.98
1.19
0.30
4.75
0.21
3.02
Standard Deviation
0.11
1.10
0.02
2.07
0.39
0.13
0.08
0.73
0.73
0.29
0.14
0.03
0.54
0.02
0.10
265
-------
TABLE V
UTAH POWER AND LIGHT COMPANY, HUNTINGTON #2
SUMMARY OF MICRO PARTICLE DISTRIBUTION,
SPECIFIC GRAVITY AND LOSS ON IGNITION
Inlet Thimble Sample
Test #
+149 y
+ 45 n
+ 40 y
+ 35 y
+ 30 y
+ 25 y
+ 20 y
+ 15 y
+ 10 y
+ 5 y
- 5y
2
34
36
40
43
48
53
61
70
83
16
1
.89$
.2
.8
.0
.5
.0
.8
.0
.5
.7
.3
2
2.92$
32.1
34.1
37.0
40.2
44.5
50.0
56.2
67.5
79.0
21.0
3
3.63$
34.3
36.2
38.5
41.6
45.5
50.1
56.5
66.5
77.2
22.8
4
3.31$
33.5
35.8
38.2
41.5
44.8
49.0
55.9
65.9
77.2
22.8
3.
33.
35.
37.
40.
43.
48.
55.
65.
76.
23.
5
83$
2
0
0
0
9
8
0
5
5
5
3.
31.
33.
36.
39.
43.
47.
53.
63.
79.
20.
6
52$
9
9
5
5
0
0
5
5
5
5
7
5.86$
50.2
52.5
55.0
58.0
61.5
65.5
70.5
78.0
86.4
13.6
8
; 2.49$
35.5
37.9
40.5
44.0
47.5
53.5
59.5
68.8
78.8
21.2
* 2.1704 2.1595 2.1708 2.1109 2.2566 2.0864 2.1865 2.1346
** 2.18$ 2.34$ 2.70$ 2.84$ 3.09$ 2.90$ 2.40$ 2.45$
^Specific Gravity
**Loss on Ignition
266
-------
TABLE VI
UTAH POWER AND LIGHT COMPANY, HUNTINGTON #2
SUMMARY OF BULK ELECTRICAL RESISTIVITY
Temperature (°F) 250 280 310
Test #1
Resistivity (OHM-CM) 5.32 x 109 5.06 x 109 2.12 x 109
Breakdown Voltage (VOLTS/CM) 1.30 x 10^ 1.20 x 10^ 1.20 x 10^
Moisture (%) 6.92 7.02 6.64
Test #2
Resistivity (OHM-CM) 5.84 x 109 5.82 x 109 5.15 x 109
Breakdown Voltage (VOLTS/CM) 1.40 x 10^ 1.30 x lO* 1.30 x 10^
Moisture (%} 7.06 6.78 6.56
Test #3
Resistivity (OHM-CM) 8.04 x 109 6.37 x 109 5.87 x 109
Breakdown Voltage (VOLTS/CM) 1.40 x 10^ 1.20 x 10^ 1.20 x 10^
Moisture (%) 7.06 6.94 6.58
Test #4
Resistivity (OHM-CM) 7.93 x 109 5.72 x 109 5.32 x 109
Breakdown Voltage (VOLTS/CM) 1.40 x 10 1.20 x 10^ 1.20 x 10^
Moisture (%} 7.04 6.92 6.64
Test #5
Resistivity (OHM-CM) 7,52 x 109 6.47 x 109 4.70 x 109
Breakdown Voltage (VOLTS/CM) 1.10 x 10^ 1.20 x 10^ 1.20 x lO4'
Moisture (%} 6.80 6.90 7.10
267
-------
TABLE VI (CONT'D)
UTAH POWER AND LIGHT COMPANY, HUNTINGTON #2
SUMMARY OF BULK ELECTRICAL RESISTIVITY
Temperature (°F) 250 280 310
Test #6
Resistivity (OHM-CM) 1.11 x 1010 7.54 x 10 5-94 x 109
Breakdown Voltage (VOLTS/CM) 1.20 x 104" 1.40 x 10* 1.40 x 10*
Moisture (%) 6.86 6.92 7.27
Test #7
Resistivity (OHM-CM) 1.49 x 1010 5.97 x 109 5=32 x 109
Breakdown Voltage (VOLTS/CM) 1.20 x 10* 1.20 x 10* 1.10 x 10*
Moisture (%) 6.84 6.99 6.86
Test #8
Resistivity (OHM-CM) 9.85 x 109 9.16 x 109 5.67 x 109
Breakdown Voltage (VOLTS/CM) 1.40 x 10* 1.40 x 10* 1.40 x 10*
Moisture (%} 6.74 7.24 7.08
268
-------
References
1. Ensor, D. S. and Pilat, M. J. Calculation of Smoke Plume Opacity
from Particulate Air Pollutant Properties. APCA Journal, Volume
21, No. 8, August 1971.
2. White, H. J<, Industrial Electrostatic Precipitation. Addison-
Wesley, Reading, Massachusetts, 1963.
3. McDonald, J. R. A Mathematical Model of Electrostatic Precipitation:
Revision 1, Volume I, Modeling and Programming. EPA-600/7-78-llla,
U. S. Environmental Protection Agency, Research Triangle Park,
North Carolina, June, 1978.
269
-------
Figure 1. Huntington Station Unit #2 on-line
270
-------
'DUST LAYER
EMITTING,
ELECTRODE
.COLLECTING
PLATE
SURFACE
•Vo
vt
-vd
Vt = Vg + Vd
= Vg + RJd
where Vt is total voltage drop
Vg is effective voltage drop across the gap
Vd is voltage drop across dust layer
R is resistivity of dust layer
d is dust layer tickness
J is current density
Fig. 2 - SCHEMATIC REPRESENTATION OF VOLTAGE
DROP THROUGH DUST LAYER.
271
-------
STRATEGY NO. 1
60
50
< 30
Q_
O
20
10
0
TIME-30MIN. INCREMENTS
STRATEGY NO. 2
60
50
40
>-
B30
<
a.
O
20
10
n
iwv^vii
altfantt *»u»
uv^
kAJtfvil.il
T" !(f*V*
... JL i«!LJL
L.-'Ju1 • •
'V.*l"«y.> Mf
KiLtLJ • . ''
rVH«»w
UV»>K'"" J'ltl
TIME-30MIN. INCREMENTS
Fig. 3 REPRESENTATIVE OPACITY FOR DIFFERENT RAPPER STRATEGIES
272
-------
Q.
IN3 O
w h-
o
20
18
16
14
12
10
8
6
4
0
0
Fig. 4 -
.01
.05
.06
.02 .03 .04
DUCT EMISSION GR/DSCF
HUNTINGTON NO. 1 — OPACITY VS. OUTLET CONCENTRATION
-------
0
0
.02
DUCT
Fig. 5 — HUNTINGTON NO. 2
.05
.06
.03 .04
GR/DSCF
OPACITY VS. OUTLET CONCENTRATION
-------
100
90
80
70
60
50
40
30
LU
Q_
LU
o
20
15
10
f
J-SITU MEASUREMENTS
© HUNTINGTON NO. 2 - 2 TESTS
A HUNTINGTON NO. 1 - 2 TESTS
f
1.5 2 2.5 3
Fig. 6 - PARTICLE SIZES
5 6 7 8 9 10
275
-------
ro
100
80
CO
CO
I 60
CO
CO
CO
LJJ
40
20
0
0
07
I I
100 120 140 160 180 200
PRECIPITATOR POWER-(KW)
220 240
Fig. 7 — HUNTINGTON NO. 1 — EMISSIONS VS. POWER
-------
ro
250
200
CO
en
— 150
CO
CO
CO
LLJ
100
50
140
160
180 200 220 240
PRECIPITATOR POWER (KW)
260
280
Fig. 8 — HUNTINGTON NO. 2 — EMISSIONS VS. POWER
-------
60
50
t 4°
< 30
0.
0
20
10
n
^y*V*^
•*/»4v>*»Wj>'
iH^sw^V"
MA^VV/MM
W^W*"**!
TO'-fV
i i
Wf**"1""**11
TIME-30MIN. INCREMENTS
Fig. 9 REPRESENTATIVE OPACITY FOR POWER-OFF RAPPING
278
-------
RAPPING SYSTEMS FOR
COLLECTING SURFACES IN AN
ELECTROSTATIC PRECIPITATOR
By:
Heinz L. Engelbrecht
Air Pollution Control Division
Wheelabrator-Frye Inc.
Pittsburgh, Pennsylvania
ABSTRACT
Dust build-ups on adjacent surfaces is often an unwanted effect in indus-
trial processes handling pulverized materials. In the case of electrostatic
precipitation, build-up of dust on the collecting surface is the purpose of
the process. Removal of these build-ups is sometimes difficult, but necessary
to maintain the continuity of the process. Dry-process methods for dust re-
moval include rapping by impact.
It is the intent of this paper to provide information on the general sub-
ject of dust removal from plates, on rapping systems used in electrostatic
precipitators, and on measurements and order of magnitude of rapping forces.
Rapping acceleration test results obtained in the Laboratory are expanded
to cover full-scale collecting surface plates.
Effects of plate design, plate height and length, method of support,
rapping, hammer weight and lift are analyzed.
279
-------
INTRODUCTION
Handling of pulverized materials is common to many industries and pro-
cesses; a few examples may be manufacturing of cement, processing of foods,
and separation and recovery of materials. Sometimes, these materials cause
build-ups on adjacent surfaces as an unwanted effect, but in the case of an
electrostatic precipitator, build-up of dust on the collecting surface plates
is the explicit intent.
The electrostatic precipitation process includes the phases of particle
charging, precipitation and removal. All of these phases are affected by the
residual layer of particles on either the collecting surface plates or the
discharge wires. Thus, periodical removal of these build-ups is necessary to
maintain the continuity of the precipitation process. The removal of the pre-
cipitated dust requires consideration of the precipitation process and preven-
tion of reentrainment.
Reentrainment of particles occurs constantly in an electrostatic precipi-
tator. All particles entering the electrical field are charged, but not all
charged particles are precipitated. Of those which are, a certain number is
reentrained by saltation, by erosion, by electrical wind, and/or electrical
arcing. More reentrainment occurs during dust removal. Additional reentrain-
ment of particles can originate in the hopper area caused by hopper boil-up.
All reentrainment will add to the fraction of particles which escape the elec-
trical field without ever being precipitated. This total makes up the outlet
dust residual of the precipitator.
A quantitative analysis of the various fractions of the outlet dust re-
sidual has not been attempted, but it has been shown that rapping reentrain-
ment is the major contributor to the outlet residual.
Therefore, the removal process is important for the continuity of the
precipitation process, requiring a carefully selected equilibrium between clean-
ing of the collecting surface plates and prevention of particle reentrainment.
Dry-process methods for removal of dust from collecting surface plates
depend on mechanical devices using impact forces generated by single impact
hammers or multiple impact vibrators. It is the intent of this paper to des-
cribe a rapping system, which uses rotating, single-impact hammers to achieve
dust removal from the collecting surface plates.
COLLECTING SURFACE PLATES
Collecting surface plates consist of thin metal plates (16 to 20 Ga.)
spaced evenly at centerline distances of 8 to 20 in. to form gas passages, in
which the discharge or emitting electrodes are centered. Curtains with a
height of up to 50 ft. and a length of up to 12 ft. in direction of gas flow
can be made out of single stiffened sheets or individual roll-formed strips
18 to 24 in. wide.
Basic requirements on the design of a collecting surface are: rigidity,
light weight, quiescent zones and good vibrational characteristics. These are
280
-------
met by a roll-formed strip, which includes three pockets and means to inter-
lock with adjacent strips to form a continuous curtain (Figure 1).
The curtain is supported at the top by two hangers per strip and con-
nected at the bottom by means of a continuous bar and anvil assembly
(Figure 2).
RAPPING SYSTEM
Rappers are of the impulse or vibrator-type; one producing a single im-
pact, the second, a continuous vibration of a certain duration. The impulse
type rapping system described in this paper uses a slowly rotating shaft,
mounted cross-wise to the gas flow inside of the precipitator, with hammers
attached in a staggered fashion around its periphery. Each hammer is in line
with a rapper bar, which connects the lower ends of a number of collecting
surface strips making up one collecting surface curtain (Figure 3). As this
shaft, which is normally located at the outlet end of each electrical field,
rotates, one hammer after the other reaches a vertical position with the
hammer head standing up. At this point, the force of gravity takes over, and
the hammer head rotates approximately 180 degrees around a pin in its attach-
ment and hits the anvil at the end of the rapper bar. This impact is intro-
duced into the collecting surface curtain and produces definite plate move-
ments and acceleration in two axis, normal to the plate and in direction of
the impact, of which the acceleration normal to the plate is of greater im-
portance in the removal process of the particles.
MEASUREMENTS OF PLATE ACCELERATIONS
Plate accelerations provide data for comparisons of various plate de-
signs, sizes, and rapping systems. Accelerations normal to the surface of
the plate are measured in multiples of g, with one g = 32.2 ft/sec^. These
are measured at selected points using an electro-mechanical pick-up system,
with a piezo-electric effect to visualize and/or measure the acceleration
of the collecting surface curtain. The voltage is proportional to the ac-
celeration of the object.
The electrical signal from the pick-up is amplified in a voltage ampli-
fier and fed to an impulse-type sound level indicator which allows the dis-
play of the effective values (RMS) or the maximum values by using a holding
circuitry.
The measurements of the plate accelerations are made at a suitable num-
ber of test points on the collecting surface curtain. The number of test
points (horizontally and vertically) has to be determined in such a manner
as to give sufficient data to evaluate the expected acceleration pattern.
The rapping impact is produced with a standard rapping hammer by manually
releasing the hammer from the upper lift point. The average acceleration
values are calculated for each collecting surface strip and also for each
elevation and plotted graphically. As a lower limit for suitable removal
of the accumulated dust layer, a minimum acceleration of lOOg is considered
acceptable for most electrostatic precipitator applications.
281
-------
Up to six individual measuring points are used in each collecting surface
strip in each test elevation (Figure 4). Normally, testing is done in three
different elevations: top, center, and bottom. The number of strips to be
tested depends on the total number of strips in the collecting surface curtain.
Three to five measurements are taken at each of the designated test points,
and the following average test results calculated for each of the test points:
a) arithmetic average of the test points in one location
b) arithmetic average of the test averages in one of the
tested elevations of one strip. This average is used
for the acceleration profile.
c) arithmetic average of the acceleration in one collecting
surface strip
d) total arithmetic average of the rapping acceleration in
one collecting surface curtain
This total average represents a mean rapping acceleration for a specific
collecting surface curtain and can further be used to calculate a rapping pro-
file based on percentages of the mean rapping acceleration. An example for
test results obtained at the test points of the collecting surface plate shown
in Figure 4 is given in Table 1; graphical presentations of the acceleration
profile are made for absolute values of g, and for percentages of the mean
rapping acceleration at each test elevation (Figure 5).
INFLUENCES ON RAPPING INTENSITY
Influences on the rapping intensity in an electrostatic precipitator,
not necessarily in sequence of their importance, are:
- Plate design
- Curtain length
- Curtain height
- Plate connection to rapper bar
- Plate support
- Location of impact point
- Hammer weight
- Hammer lift
Tests have been performed with various plate designs, methods of connect-
ing plates to rapper bars, curtain heights and lengths, and hammer weights.
The lift of the hammer was always taken at approximately 180°, which repre-
sents the maximum lift inherent with this rapper design.
PLATE DESIGN
Basic tests were conducted using a so-called C-type collecting surface
plate. Comparative tests showed that a CS-type plate has a higher level of
rapping intensity than the C-type. For example, the absolute acceleration
for the C-type plate was 188g in the far corner diagonally opposite the point
of impact, whereas the CS-type showed 400g at the same location (Figure 6).
282
-------
In addition, the acceleration of the far points represented 51% of the
average acceleration of the C-type and 78% of the average acceleration of the
CS-type plate. This shows a lower dampening effect of the CS-type curtain.
(Figure 7) .
CURTAIN LENGTH
As it can be expected, there is a definite influence of the curtain
length in the acceleration of the collecting surface plates. Comparing cur-
tains with 4.5, 7.5, and 12 ft. length, the maximum accelerations at the point
of impact are 1076 to 1160g, reduced to 560, to 400g respectively at the far
points (Figure 8). These tests were performed using a standard CS-type col-
lecting surface curtain with a field height of 20 ft. The mean accelerations
in g are 747g, 575g, and 513g respectively. Local accelerations expressed in
percent of the mean accelerations are shown on Figure 9. The acceleration
profile remains almost constant for the first 4.5 ft. from the point of impact
regardless of total curtain length.
CURTAIN HEIGHT
Tests with a curtain height of 50 ft. revealed a substantial accelera-
tion at the point of impact, which rapidly decreased to approximately mid-
height and then remained almost constant for the upper half of the collecting
surface curtain. The acceleration profile in g and the percentage of the av-
erage acceleration are shown in Figure 10.
METHOD OF PLATE SUPPORT
The only method of plate support further investigated in this respect was
characterized by a loose connection between plate and support beam. All other
methods using a rigid connection were ruled out because of their inherent in-
ability to provide for graduated expansion of the collecting surface curtain
during start-up and the dampening effect of the rigid connection on the accel-
eration of the upper third of the collecting surface curtain.
METHOD OF CONNECTING PLATE TO RAPPER BAR
Comparative tests between a system using a "rigid" connection between the
plate and the rapper bar and one using a "loose" connection (Figure 11) re-
vealed that the rigid connection will achieve a higher average acceleration
(412g compared to lOlg), and also provide for a higher acceleration in the
upper far corner (296g compared to 89g).
LOCATION OF IMPACT POINT
Locating the rapper hammer impact point normal to the surface of the col-
lecting plate results in a considerable lower average plate acceleration, as
well as a drastically reduced acceleration in the upper far corner (Figure 12).
Therefore, the hammer impact point is always parallel to the axis of the
plate.
283
-------
Locating the rapper bar at approximately the mid-point of the collecting
surface curtain and maintaining the direction of the hammer impact, the local
acceleration at the point of impact is extremely high (Figure 13), but the dis-
tribution of the plate acceleration is rather non-uniform. Therefore, besides
design considerations, such as accessibility, etc., it is best to maintain the
point of impact at the bottom of the collecting surface curtain.
HAMMER WEIGHT
A comparison of the rapping accelerations obtained with two different
hammer weights is shown on Figure 14. The average accelerations follow the
ratio of the hammer weights, i.e., 1.58 to 1.00
HAMMER LIFT
It is inherent in this design that the hammer lift approximates 180 de-
grees, therefore, no tests at other lift angles were made, although it can
be expected that the average acceleration is also a linear function of the
hammer lift.
AVERAGE PLATE ACCELERATION
An analysis of the results of numerous rapping tests revealed a mathe-
matical relationship between the average rapping acceleration of a collecting
surface curtain and the curtain dimensions and the hammer weight.
This relationship can be described as follows:
with b = average rapping acceleration (g)
W = hammer weight (kg)
L = curtain length (m)
H = curtain height (m)
c, d, K = constants
INFLUENCES ON PRECIPITATOR EFFICIENCY
Periodical removal of the precipitated dust from the collecting surface
plates is necessary to maintain the continuity of the precipitation process.
The preferred method of dust removal in dry-process precipitators is by plate
acceleration using a suitable rapping system. The rapping hammer impact on
the collecting surface curtain can be varied by freqency and intensity.
284
-------
The influence of rapping frequency on the performance of the electrosta-
tic precipitator is not an easily predictable parameter, but depends on
process variables, such as particle shape and size, electrical field and spe-
cific dust resistivity (all of which vary with the location in the precipita-
tor) , and general variables, such as gas velocity, gas temperature, collecting
surface plate design, field height, and field length. Thus, rapping frequency
becomes one of the parameters to be established during precipitator start-up
or fine-tuning.
The influence of rapping intensity is used to counteract electrical, me-
chanical and chemical forces opposing the removal of the precipitated dust,
and is, thus, more germane to the source of the dust; i.e., the process itself.
Requirements for rapping intensity are more predictable and can be handled by
design of the hammer; i.e., weight of the hammer head and/or length of the
hammer handle (shaft).
Rapping intensity expressed as mean rapping acceleration can be calcu-
lated. For a typical CS-type collecting surface plate with the following
data:
Length: 2.4m K = 197
Height: 6.0m c = .387
Hammerweight: 6.2 Kg d = .227
the average plate acceleration calculates to 436g (see Table 1, Figures 4
and 5). The results of a series of calculations of average rapping intensi-
ties for various field lengths; i.e., various number of strips in comparison
with measured average rapping accelerations is shown on Figure 15.
Rapping frequency is controlled by adjustable electric timers, which can
be set to provide either or two typical modes of operation: rapping of one
single curtain followed by a short pause or rapping of several or all curtains
followed by a long pause. In either mode, an individual plate will be rapped
approximately the same number of times during a long time span.
Two dust samples (fine and coarse particulates) were investigated in a
pilot precipitator and the penetration noted as a function of the rapping
cycle. In the case of the sample consisting of fine dust, an optimum per-
formance; i.e., a minimum of penetration was observed at a rapping cycle of
160 minutes, whereas in the case of the coarse dust sample, the optimum was ob-
served at a rapping cycle of 110 minutes. The penetration levels with con-
tinuous rapping were 40 and 10 percent respectively (Figure 16).
Rapping optimization tests at an electrostatic precipitator serving a
municipal incinerator were made, resulting in a change from a stack opacity
reading characterized by heavy spiking every 10 minutes (rapping cycle 7/8
minutes on and 9-1/8 minutes off) to an opacity reading under 10 percent
with practically no spikes (calibration spikes show every 4 hours). The
latter was taken with a rapping cycle of 1 minute on and 20 minutes off
(Figure 17) .
285
-------
SUMMARY
Rapping systems for collecting surface curtains in electrostatic precipi-
tators have long been considered mechanical devices designed for ease of in-
stallation, reliability of operation, and low cost. The recent emphasis on
high collecting efficiencies, even for fine dust particles, requires a new
look at the design and operation of these systems. Fine tuning of the rapping
systems in multi-field electrostatic precipitators will strive to prevent rap-
ping losses, which could make up for 60 to 80 percent of the outlet residual.
Here, the work on the outlet field is of prime importance. Real time instru-
mentation, such as opacity meters and oscilloscopes, have to be used to gain
immediate results from changes made during the fine tuning process. Needless
to say that this is an ongoing continuous process and any changes in the oper-
ation of the source of the gas can result in changed operating conditions for
the precipitator and, thus, in a need for adjustments of the rapping system.
ACKNOWLEDGEMENT
Information and data for this paper were provided by LURGI Umwelt und
Chemotechnik GmbH, Frankfurt, Germany.
286
-------
COLLECTING SURFACE • CHS TYPE
DISCHARGE ELECTRODE • ISODYN
\
PLATE / WIRE GEOMETRY
FIGURE 1
PLATE SUPPORT
RAPPER BAR
FIGURE 2
287
-------
HAMMER
PLATE RAPPER
BEARING WITH BUSHING
FIGURE 3
PLATE SUPPORT
TEST POINTS '"
TEST POINTS
STRIP #
RAPPING
HAMMER U d
ELEVATION 3
ELEVATION 2
ELEVATION 1
RAPPER BAR
FIGURE 4
288
-------
STRIP
3 •
ELEVATION
2 •
ELEVATION
1 •
STRIP
AVERAGE
TOTAL
AVERAGE
TABLE 1: ACCE
(ALL VALUE
STRIP 1
410 470 430 26C 630 450
445
450 450 440 310 560 430
440
1280 1220 1240 780 1960 950
1240
710
* ARITHMET
LERATION MEASUREM
S IN MULTIPLES OF I
STRIPS
290 260 330 150 370 210
270
320 310 330 220 310 240
290
480 460 750 390 740 280
515
360
44S
1C AVERAGE OF THREE TESTS
ENTS
91
STRIPS
250 200 230 130 240 190
205
270 220 230 135 310 220
230
320 310 400 220 770 320
380
275
TABLE 1
87
50 50 47
'•'LJIllllLJ
RAPPING ACCELERATIONS IN 191
RAPPING ACCELERATIONS
IN % OF AVERAGE
IbAVG. • 437 9|
RAPPING ACCELERATIONS FOR TEST
COLLECTING SURFACE
FIGURE 5
289
-------
CS - TYPE PLATE
C • TYPE PLATE
550
RAPPING ACCELERATIONS IN 19)
COMPARISON BETWEEN CS • TYPE AND C • TYPE
COLLECTING SURFACE PLATE
FIGURE 6
CS • TYPE PLATE
C • TYPE PLATE
107
RAPPING ACCELERATIONS IN 1%)
OF AVERAGE ACCELERATION
COMPARISON BETWEEN CS • TYPE AND C • TYPE
COLLECTING SURFACE PLATE
FIGURE 7
290
-------
686
560
574
454
550
400
FIELD
LENGTH
3 STRIPS
14.S FT.I
S STRIPS
IT.» FT.)
« STRIPS
112.0 FT.I
COMPARISON OF RAPPING ACCELERATIONS
OF VARIOUS FIELD LENGTHS IN 191
FIGURE 8
92 75 100
79 107
FIELD 3 STRIPS
LENGTH (4.9 FT.I
5 STRIPS
17.5 FT.I
8 STRIPS
(12.0 FT.I
COMPARISON OF RAPPING ACCELERATIONS
OF VARIOUS FIELD LENGTHS IN 1%)
FIGURE 9
291
-------
95
57
42.5
RAPPING ACCELERATION
IN 19 I
RAPPING ACCELERATION
IN 1*1 OF AVERAGE
RAPPING ACCELERATIONS OF SO FT. / 12 STRIP
C-TYPE COLLECTING SURFACE PLATE
FIGURE 10
296
100
•LOOSE* CONNECTION
•RIGID* CONNECTION
BETWEEN COLLECTING SURFACE
PLATES AND RAPPER BARS
RAPPING ACCELERATIONS
FIGURE 11
292
-------
520
307,
188
480
RAPPING IMPACT
NORMAL TO THE SURFACE
OF THE PLATE
RAPPING IMPACT
PARALLEL TO THE SURFACE
OF THE PLATE
RAPPING ACCELERATIONS IN (9)
FIGURE 12
RAPPING IMPACT
NEAR THE CENTER
OF THE PLATE
RAPPING IMPACT
AT THE BOTTOM
OF THE PLATE
RAPPING ACCELERATIONS IN (91
FIGURE 13
293
-------
HAMMER WEIGHT
30 IBS.
HAMMER WEIGHT
19 LBS.
RAPPING ACCELERATIONS IN (91
FIGURE 14
AVERAGE
RAPPING ACCELERATION
200 —
iiiir*
0 2 4 6 B 10
NUMBER OF COLLECTING SURFACE STRIPS
COMPARISON BETWEEN CALCULATED
AND
TESTED RAPPING ACCELERATIONS
FIGURE 15
294
-------
PENETRATION
CONTINUOUS RAPPING
10
I I I I I I
90 110 120 150 175 180 210 MIN.
RAPPING TIME
RAPPING OPTIMIZATION
FIGURE 16
------~'""'
TIME
SHORT RAPPING CYCLE
-740-
i
TIME
LONG RAPPING CYCLE
RAPPING CYCLE AND OPACITY
FIGURE 17
295
-------
LOW POWER PRECIPITATION
A Logical Solution To Collection Problems Experienced
With High Resistivity Particulate.
By:
John H. Umberger
36 Main Street
Holmdel, N.J. 07733
Cold-side precipitation offers capital and energy savings
over hot-side collection of high resistivity particulate. The
major limitations of cold-side collection are premature back
corona and tenacious particulate accumulations on collecting
surfaces and discharge electrodes.
Consistent with theory, reduction in corona current does
much to alleviate these limitations. This report describes
field-proven methods to accurately identify back corona,
adjust power supply for maximum precipitation, and to adjust
rapping to minimize accumulations and reentrainment.
Although the presentation is geared for operators and
manufacturers representatives, it contains useful feedback for
Research and Development personnel.
296
-------
LOW POWER PRECIPITATION
I. INTRODUCTION
Typically Precipitators are tuned to a spark-limited equilibrium.
This report describes another equilibrium to tune to, one which better
satisfies all conditions required for efficient, reliable collection of
moderately high resistivity particulate. It discusses energization and
rapping tuning methods whereby conventional cold-side wire-plate precipitators
can be tuned to obtain reliable omega's in the range of 9.0 cm/second,
when collecting particulate with in situ resistivity in the 1011 ohm-cm
range. These tuning methods were developed in the field on cold-side
ESP's collecting fly-ash from the firing of low sulfur, low sodium oxide
southeastern coals. Abbreviated versions of these methods were applied
on ESP's collecting fly-ash from western low-sulfur coals with enough
success to merit further work with this system on all cold-side high
resistivity applications. Success of any tuning method was determined
with the help of the optical transmissometer. All units tuned employed
single impact drop-hammer rappers which could impact with energies of up
to 16 ft-lb. Area of collecting surface was no more than 5400 square
feet per rapper. Specific collection area (SCA) for the precipitators
tuned ranged from 0.55 sec/cm to 0.72 sec/cm. (275 to 360 ft2/1000
ACFM).
II. TUNING THE ELECTRICS
A. Background
Electrical power is used to charge the particulate and then assist
in its motion towards the collecting surface. The electric field
provides the force for charging the particles and the current density
supplies the charge carrying ions and electrons. The electric field
forces these charges onto the particles. Current flowing in the precipitator
not only supplies charge, it also generates a force which holds the
particulate to the collecting surface. As resistivity increases this
holding force also increases for a given current density. Both collecting
surface buildup tenacity and positive corona discharge increase with
increasing current density. Therefore, it seems reasonable to reduce
operating current to solve both problems.
B. Back Corona Detection
The first step in correct tuning is to locate the energization
point at which back corona starts. The method is simple and reliable,
and is accomplished by monitoring the raw precipitator voltage wave
form. The equipment needed to detect back corona is a single trace
oscilloscope of at least one megaohm input impedance, a voltage divider
with a dividing ratio selected to assure a balance between safety and
S/N ratio, sufficient RG59u coaxial cable to transfer the voltage waveform
from the power supply to the power supply control, a functional secondary
averaging current meter, and a reliable control element with sufficient
297
-------
stability to limit secondary current at any point between corona onset
and sparking. No averaging or integrating elements are allowed between
the power supply and the oscilloscope in the KV monitoring circuit.
However, the "averaging" corona current meter supplied with the vendors
controls is sufficient for obtaining characteristic curves. Back corona
detection is also possible using the peak corona current obtained from a
second trace on the oscilloscope, but the shape of the characteristic
curve is different.
After installing or modifying all circuit elements as shown in
Figure la the transformer-rectifier (T/R) being measured should be
energized and adjusted to an operating point below sparking. The waveform
viewed on the oscilloscope should appear the same as shown in Figure Ib.
If not, a check should be made on all wiring for errors or damaged
elements.
Attenuate the T/R to just above the threshold of corona, that is the
point where secondary current just begins to flow in the precipitator.
To detect back corona slowly increase the current, noting the deformation
of the waveform. During any half cycle the voltage rises to a peak
value (KVp) and the decays to a minimum value (KVm). Watch the relationship
of KVm and the average precipitator current (ma). As you slowly increase
the current you will note a value of current at which KVm stops increasing,
where dKVm/dI=0. This is the point of the onset of back corona. If you
run the current at higher levels, for instance on the outlet field only,
you can get as much as a 12% increase in opacity without sparking.
Figure 2 shows the graphical location of back corona.
KV minimum can take any position along a continuous line within the
shaded area after the dKVm/dI=0 point. Once in a while, opacity will
not deteriorate at current levels above the dKVm/dI=0 point. One fact
always holds! Precipitator performance will not be any better than
it is at this point!!!
For in situ resistivity of 1010 to 1011 ohm-cm the back corona glow
can barely be seen at this point. For very high resistivity (10
to 10 ohm-cm) particulate the back corona glow is obvious.
C. Half-Wave and Full-Wave Energization
Figure 2 also compares the characteristic curves of both half-wave (H-W)
and full-wave (F-W) energization for equivalent precipitation conditions.
The differences are:
1) For equal average current the average KV is roughly the same.
2) The average current at which back corona occurs in full-wave energization
is roughly twice the average current at which back corona occurs in half-wave.
298
-------
FIGURE 1
HIGH VOLTAGE MONITORING
Precipitator
Rl
X MEG
Oscilloscope
RG59
R2
X
a) Monitoring Circuit
KVmin- E=£H
KV peak
--.*——! i '—tn—~. ~.~.~. rj
b) Typical Voltage Waveform
299
-------
FIGURE 2
KV, Voltage
Figure 2 : Characteristic V-I Curve showing Back Corona (B/C)
300
-------
3) The peak voltage at back corona onset is roughly the same for both
F-W and H-W energization.
4) The percentage increase in average corona current from the point of
back corona to the initial point of sparking can be as great as:
a) 100 to 200% for F-W
b) 10 to 20% for H-W
The more effective the rapping, the greater this percentage increase can
be.
5) KVm is much lower and KVp is much higher at the same average current
for H-W.
All these observations are explainable and consistent with theory.
Most workers in this field agree that back corona occurs when a
critical field strength (Eds) is reached. This field strength equals p
times a critical current density (jds)- To avoid back corona we can
move the mountain to Mohammed — that is, use gas conditioning to lower
particulate resistivity — or we can bring Mohammed to the mountain —
that is, lower precipitator current to lower the electric field in the
dust layer. I chose the latter and successfully expanded the useful
range and reliability of every cold-side precipitator on which I lowered
the current density to the onset of back corona point.
I find H-W energization superior to F-W for three reasons:
1) 1/3 to 1/4 the power consumption of F-W at back corona (B/C)
point
2) Slightly lower opacity than F-W at B/C point
3) Longer term stability of tuned-in performance
There are two other advantages to the low power strategy which
alone justify the small effort required to detect back corona and to
keep transmissometers running. The first is obvious, the virtual
elimination of wire discharge electrode failure. Keep in mind that the
precipitator is now operating well below the threshold of sparking, and
sparking is the major cause of D.E. failure. Corrosion another cause of
failure, is generally not a problem for low sulfur coals. The qualifier
"virtual" is used because hoppers can still be overfilled and even at
low power levels it is possible that a D.E. can be burnt in two. This
cause of failure can be avoided by a reliable undervoltage trip. I
recommend that undervoltage be sensed on the secondary of the T/R unit.
301
-------
I have also discovered that particulate accumulations on the discharge
electrodes are frequently caused by back corona. During numerous
internal inspections of collecting surface buildups, I got used to
seeing wire buildups ranging from 5mm total diameter (2.77 mm wire
diameter + buildup) to 20 mm total diameter. Generally the population
and size of accumulations increased from inlet to outlet of any given
chamber. After retuning the same precipitators to the threshold of back
corona the internals were re-inspected to monitor collecting surface
buildups. During these inspections I found that the same wires were
metal clean. I concluded that this type of wire buildup was the collection
of positively charged particulate from back corona sites on the collecting
surfaces.
To review, the electrical tuning methods are simple:
1) Identify back corona by method specified and reduce
secondary current to this point by current limit.
2) Energize precipitator in double half-wave mode to get maximum
ratio of KVp/Iave.
For low power collection or 10^ ohm-cm particulate and half-wave
energization you will find KVa = 21., KVp -40 to 50, and jave~ 6 na/cm2
(5 ma/1000 ft2). You will also find that total power consumption will
be 5 to 10% the power consumed when the precipitator operates in a FW spark
limited mode. Thus, the label low power precipitation.
III. TUNING COLLECTING SURFACE RAPPING
Poor electrical stability and excessive re-entrainment are largely
due to improperly applied collecting surface rapping. Deterioration of
precipitator performance due to ineffective rapping is most critical
when collecting high resistivity particulate.
Owen Tassicker (1972)^- did a great deal of study on the mechanics
of collecting surface rapping and, like the earlier researchers Ruckelschausen
(1957) and Sproull (1965) ^, concluded that normal to the collection
surface accelerations are the predominant means of particulate removal.
After much study within and outside of the precipitation field he derived
his formulation for the tensile strength (P) of a particulate layer on a
collecting surface in the influence of an electric field.
(Equation 2)
P = A1/d2 + A2/d2 + 105 fo J2
302
-------
P = Dust layer tensile strength (gm/cm^)
A;L = Van der Waal constant
A£ = Capillary constant
d = Particle diameter (cm)
EQ = Permittivity of free space (F/m)
J = Current density (A/crn^)
p = Layer resistivity (ohm-cm)
p = Gas resistivity (ohm-cm)
O
e = Relative permittivity of dust layer (solids
A and voids) (no units)
Figure 2.1 a & b (taken from Tassicker) gives a reasonable illustration
of the electrical parameters involved.
This is primarily a D.C. equation, since the mode of energization I
promote has greater A.C. components, capacitances of the circuit in
Figure 2.1 could have relevant components which may alter the validity
of the equation. The relevance of the electrical term of this equation
has been observed by many rapping researchers and in ways can be drawn
from field experience. Tassicker presented a table which I reproduce
here.
TABLE 1
E8
£A
J
Pg
p£
P
(KV/CM)
(pa/crn^)
(ohm- cm)
(ohm- cm)
(gm wt/cnr)
Case 1
(low )
5
4
0.05
1011
109
-0.011
(-repulsive)
Case 2
(med )
5
4
0.05
ion
5xl010
0
Case 3
(high )
5
4
0.05
SxlO11
+1.1
(+at tractive)
303
-------
-Va
H.T.
DISCHARGE
ELECTRODE
u
OUST
GAS
Eg
c\
~-' 4-
POTENTIAL
^1 <•$ 2
-------
Table 1 helps explain the rapping problems experienced with both
high resistivity and low resistivity. We know that we can rap low lO^
ohm-cm particulates without too much trouble. Usually we do not have
many problems with heavy accumulations unless the precipitator serves
cycling duty. In this case, the capillary term in equation 2 probably
takes prominence.
Generally, we can effectively dislodge 10^ ohm-cm particulate with
a light to medium impact. This light rap provides enough force to
overcome the Van der Waals and capillary forces. For simplicity's sake
we will assume that these components of particulate layer strength are
unaffected by changes in resistivity and electrical conditions.
The last term of Equation 2 provides the justification for reducing
current densities to the low levels suggested. The electrical component
of tensile strength increases with the square of the electrical field in
the dust. From Figure 2 it is evident that as current density increases
the electrical component of particle tensile strength grows larger at an
increasing rate. If we double the current density, we will more than
double the electrical component of the tensile strength. When running
in a spark limited mode, current densities typically varied between 20
and 50 na/cm^. After correct electrical and rapping tuning techniques
were developed, the same precipitators were operative (artificially
current limited) at approximately 5 na/cm „ In effect the electrical
component of the fly-ash strength had been reduced considerably. It
seemed logical to reduce current density and now that it has given
positive results, it seems even more logical.
The mechanical aspects of rapping and dust strength should also be
considered. As the earlier researchers noted, a residual layer remains
on the collecting surface after a rap. Ruckelshausen defined this
residual as basic layer. Rapping efficiency, as defined by Sproull, is
the percentage of the particulate layer removed by a single blow or
total layer thickness minus basic layer thickness, all divided by total
layer thickness. Sproull found that rapping efficiency increased with
an increase in layer thickness. Or, for a given rapping efficiency, the
thicker the buildup the lighter the rap needed to dislodge it.
The conventional American design precipitator has the disadvantage
of having a relatively unresponsive collecting surface structure.
Generally, too many collecting surfaces are cleaned by one rapper.
Also, it is probable that too much rapping energy is absorbed by the
collecting surface supports of the top-rapped precipitator. As particulate
resistivity gets higher the demands on the rapping system become greater,
and the marginal abilities of the top-rapped precipitator become obvious.
The precipitators on which the low power techniques were developed could
be considered worse than average in rapping responsiveness. Yet, the
methods to be prescribed are quite effective.
305
-------
With Sproull's findings in mind, it seems logical that for any
given collecting system there would exist a thickness of particulate
accumulation for which the rapping accelerations available could clean
with a reasonable efficiency. To assure reasonable rapping efficiencies,
the lower the acceleration available, the thicker the accumulation would
have to be. Taking complete advantage of Sproull's findings requires
that collecting surfaces should be rapped only when a certain minimum
particulate layer thickness is reached, and not before. Since each
successive collecting fields in a precipitator collect particulate at
slower rates (in thickness or mass per unit time), rapping should be
less. f.r.equent-at outlet .ward, fields^relative, to. inlet ward^fields. To
find the relative layer growth rates the Deutsch equation can be used.
Consider the following example:
A precipitator having an SCA of 0.8 sec/cm (400 ft2/ 1000 ACFM) is
designed to collect high resistivity particulate ( P = SxlO- ohm-cm)
with an efficiency of 99.5%. The inlet dust concentration is 5.72xlO~
gm/cc (2.5 gr/ACF). The precipitator has four equal collecting,fields
in the direction of gas flow. Particulate density is 1.1 gm/cmj
(70#/ft3).
Using the Deutsch equation, we get anojof 6.7 cm/second. Each of
the four fields should collect 73.4% of the fly-ash they receive.
Allowing for fly-ash concentration and density we get the following
growth rates: first field, 0.675 mm/hr; second field, 0.179 mm/hr;
third field, 0.048 mm/hr; last field, 0.013 mm/hr.
If we use a constant thickness system we would choose rapping
periods for any precipitator field in direct proportion to its particulate
growth rate. In this example, the period for one rap for each collecting
surface would be; (inlet) T , 3.78 T, 14.1 T, 52.OT (outlet). We now
know the ratio of buildup rates and all rapping should be timed in
accordance.
The next task is to ascertain the absolute timing. This is a
tricky business. Incorrect rapping adjustments may take three days to
four weeks to deteriorate performance on a 300 to 500 SCA precipitator,
and as long as five months on a 1000 SCA precipitator. The best known
way to find T would be to monitor inlet field opacity. T should be set to
the longest time possible before opacity deteriorates. Remember, the
thicker the ash deposit, the easier it is to dislodge. Transmissometer
monitoring at the outlet of the precipitator is not really suitable for
direct determination ofTbecause ef f ective a) f or each collection field
will vary from position to position and in time during the performance
degeneration period.
306
-------
Experience shows that once rapping is timed as recommended, precipitator
sensitivity to particulate source transient behavior is drastically
reduced. Reasonable success has been.realized using T in the range of
0.75 to 1.50 hours. For the 400 SCA precipitator rapping times would
be:
Field 1 (inlet) each rapper is fired once every 0.7 to 1.5 hours
Field 2 each rapper is fired once every 2.8 to 5.6 hours
Field 3 each rapper is fired once every 10.4 to 20.8 hours
Field 4 (outlet) each rapper is fired once every 38.5 to 76.9 hours
To many these intervals man seem too long, but keep in mind that
these timings allow only 0.5 to 1.0 mm average accumulation of particulate
on the collection surfaces for this particular precipitator.
If current densities are constant from inlet to outlet and the
constant thickness system is applied to rapping timing, there is no
logical reason to say the intensity of rapping should not be the same
throughout the precipitator. However, in practice, field representatives
and operators generally reduce rapping intensity at the outlet of a
precipitator in an effort to reduce rapping puffs. From my experience
this is a bad practice. The effort to reduce short-term emissions
results not only in higher long-term emissions due to ineffective collecting
surface cleaning, but can actually aggravate the problem the reduction
was applied to eliminate.
75
§50
25
Figure 4 says it best:
FIGURE 4
Comparison of Outlet Field Rapping Puffs
75
t(sec)
Case A: Light rap = 4 ft-lb
(everywhere) T = 0.033 hr.
t(sec)
Case B: Heavy rap = 16 ft-lb
(outlet) T = 8 hr.
307
-------
Figure 4 demonstrates some differences between very light-frequent
rapping and very heavy-infrequent rapping. Although the peak opacity is
higher for heavy rapping the total emissions may be less. The secondary
puffs for Case "B" are believed to be due to hopper boilup. This does
not happen for Case "A", implying that little or no particulate is going
into the hopper, but instead totally re-entering the gas stream.
Considering the conditions, one could conclude that the light-
frequent rapping dislodges the fine and/or most recent accumulations
from the collecting surface before they have a chance to agglomerate.
These observations reinforce the recent findings of Nichols (June 1978) .
When the low power strategy is applied to solve collecting surface
cleaning problems, other advantages will be realized. The major one is:
when operating at low current densities, the electric field strength in
the gas zone stays the same or increases with increases in particulate
layer thickness on the collecting surfaces. In comparison, the typical
high current density approach encourages a degradation of electric field
strength in the gas interface as particulate layer thickness increases.
This fact alone would discourage the use of a slow rapping system. One
would have to clean collecting surfaces at accumulation thicknesses
approaching the diameter of the particles themselves, a virtual impossibility.
One concern, not mentioned before, is the need for uniform current
distribution at the collecting surface. Figure 5 taken from Tassicker
(1972)-* shows the increase in J uniformity with decrease in average
current. The analogy between the correlation of Figure 5 and precipitators
tuned to low power should be made with caution since the dust layer and
space charge in an actual precipitator will alter the analogy. Both
dust layer and space charge will increase the ratio of peak KV to average
current. This would suggest greater uniformity for low power tuning
than is shown in Dr. Tassicker's data since actual current densities are
one to two orders of magnitude less. In addition the mere presense of
the dust layer will alter current density uniformity although it is not
obvious which way. However, the pattern of back corona craters found on
collecting surfaces does tend to agree with Dr. Tassicker*s profiles of
a two wire system. The craters exist only on the collecting surfaces
directly across from the discharge electrodes, and on the edges of
collecting surface gas flow baffles.
Figure 6a compares stack opacity for a precipitator tuned in a
conventional manner (F-W, Jave = 30 na/cm2, KVp = 62 KV; T = 0.033 hours
everywhere; outlet rapping energy = 4 ft-lb.) To the same precipitator
tuned to low power criteria (H-W, jave = 5 to 10 na/cm2; KV = 40 KV; T
= 8 hours on outlet; rapping energy anywhere = 16 ft-lb.).
Figure 6b compares long-term performance of the same precipitator
operating in F-W spark limited mode to one operating in the prescribed
low power method. This plot helps to explain why low power tuning is
308
-------
FIGURE 5
6.0
o
O
cn
C
0)
O
a;
5
RADIUS FROM ELECTRODE (cm)
Figure 5; Comparison of Electric Field and current density distribution
along the anode for 15 and 20 KV applied voltages.
(afterTassicker, 1972 5)
309
-------
FIGURE 6
100 -
0 -
Spark Limited: j = 40 na/cm3 ,
0.033 hr.
_ _ T. i_..j-£ _l_- i.J l:j. --J—:^
ry
*Vfc*
min.
Low Power: j = 8 na/cm3 , Ta 8.0 hr.
a) Stabilized: Comparing Spark Limited, light and frequent
Rapping-to-Low Power, Heavy and infrequent rapping. (See Text)
100 -
Spark Limited
Low Power
b) From Metal Clean Startup; (See Text)
310
-------
not obvious. The vendor start-up engineer sees much better performance
on a metal clean ESP with high powered full-wave energization than with
low powered half-wave energization. At start-up, high powered energization
may give opacities of 2 to 6% (possible efficiency of 99.8+%) where low
power would give opacities of 10 to 20% (possible efficiency of '99.0 to
99.5).
As particulate accumulates on collecting surfaces the electrical
equilibrium changes. For spark limited precipitation the combination of
back corona and excessive voltage drop across the dust layer degenerate
precipitator performance to unacceptable levels, that is, opacities in
the 30 to 60% range (efficiency ranging from 85 to 98.5%). The rate of
deterioration depends on particulate resistivity (the higher the p the
faster the deterioration), the precipitator SCA (the larger the precipitator,
the slower the deterioration), and the effectiveness of collecting
surface geometry and responsiveness (ratio of acceleration/energy input),
collecting system suspension, and as suggested here, electrical input.
For low power precipitation the accumulation of particulate actually
improves the behavior of the precipitator. When started-up, stack
opacity will be greater. As particulate accumulates on the collecting
surface and equilibrium is reached, performance will improve slightly.
The initial relatively poor low power performance is still much better
than the level of deterioration which spark limited operation will
yield.
After proper application ot the recommended low power techniques,
it will become obvious that average current density of 5 - 10 na/cm^ and
peak KV's of 40 to 50 can be more than sufficient to obtain overall
omega's greater than 9 cm/sec for 10^ ohm-cm particulate. It will
become obvious that increases in corona current can cause more harm than
good.
The methods and philosophies prescribed here are not without limitation.
High 10 ohm-cm and higher resistivities could require precipitators
which can support KVp/jave, at least one order of magnitude higher
(strangly enough this ratio comes out in the units of resistivity, ohm-
cm: for p =10 ohm-cm , KVp/jave = 10 for p = 10^2 ohm-cm, Ep/jave =
10*--% etc.) and rapping accelerations one order of magnitude higher (if
minimum sufficient normal acceleration for 10^ ohm-cm particulate is 20
g's for 1Q12 ohm-cm you may need 200 g's minimum normal acceleration).
Like most precipitators, the low power precipitator is sensitive to
cycling. The only solution that is possible with present technology is
a massive (100 g's or greater) power-off rap before the gas temperature
drops below 230°F. But, if you forget to do it once when coming-off
line, while collecting high resistivity particulate, there are only two
cures 1) thorough, and I mean metal clean, water-wash or 2) a three to
five day constant firing of high sulfur coal to condition the baked-on
particulate.
311
-------
The rapping adjustments are also simple: 1) Rap any collecting
field once and only when the particulate layer thickness reaches a
predetermined minimum thickness. Precipitators of different manufacture
will have different critical minimum thickness. Determining the best
timing for the best performance will take weeks, not minutes or hours.
2) Rap the surface with energies high enough to get sharp spikes on your
transmissometer readout. Rap all fields with equal energy. I prefer
the drop weight rapper to any other rapper. Although the results are
preliminary, I have found that the vibrator-type of rapper tends to
stimulate basic layer growth and generate undesirable electrical instabilities.
IV. CLOSING REMARKS
The methods described will give positive results. The closer one
adheres to these methods, the more sucessful these results will be. If
one chooses to diverge from any of the recommended adjustments, I
advise that only one adjustment be changed at a time. The success or
failure of any change (power or rapping) can only be reasonably judged
after completion of at least two weeks of operation.
In closing power strongly affects rapping and rapping strongly
affects power.
312
-------
REFERENCES
1) Tassicker, Owen, "Aspects of Forces on Charged Particles in
Electrostatic Precipitators, "Doctoral Thesis, University of New
South Wales, July 1972, p.p. 3742, 4594 (available in EPRI reprint).
2) Ruckelshausen, Kurt, "Removal of Powdered Deposits from Technically
Smooth Metal Surfaces," Doctoral Dissertation, Technischen
Hochschule, Stuttgart, No. DK Giese-Druk, Kg Offenbach/M, 1957.
3) Sproull, Wayne T., "Fundamentals of Electrode Rapping in Industrial
Precipitators, "Journal of the Air Pollution Control Association,
Feb. 1965, No. 2, p.p. 50-55.
4) Nichols, Grady B., "Rapping Reentrainment in a Near Full Scale Pilot
Electrostatic Precipitator,"Southern Research Institute for the
Environmental Protection Agency, EPA-600/7-78-112, June 1978,
p. 3.
5) Tassicker, Owen, Ibid., 1972 - p.p. 156-157.
313
-------
HIGH INTENSITY IONIZER TECHNOLOGY
APPLIED TO RETROFIT ELECTROSTATIC PRECIPITATORS
By
C. M. Chang and A. I. Rimensberger
UNION CARBIDE CORPORATION
Linde Division
Tonawanda, New York 14150
.ABSTRACT
This paper presents the results of a study concerning the
application of High Intensity lonization technology to a flyash
electrostatic precipitator.
Described in detail are the adopted method of performing
functional design for ionizer-precipitator systems, selected re-
sults of ionizer-precipitator flow model tests, and the proposed
physical retrofit. Also presented is the projected performance
of the ionizer-precipitator system, derived from an empirical
model which takes into account the phenomena of space charge and
gas velocity distribution.
314
-------
HIGH INTENSITY IONIZER TECHNOLOGY
APPLIED TO RETROFIT ELECTROSTATIC PRECIPITATORS
INTRODUCTION
The High Intensity Ionizer (HII) is a novel device capable of charging
particulates contained in flue gases to levels significantly higher than what
can be achieved in conventional electrostatic precipitator (ESP) systems,
Schwab, et al (1976)1 and Tassicker and Schwab (1977)2. By using the High In-
tensity Ionizer as a pre-charger, greater overall migration velocities result
in the precipitators. Of particular value is the use of HII to enhance the
ESP's collection efficiency for fine particulates, i.e. particles with diameter
less than three microns, which determine the opacity of the process gas leaving
the system.
This paper presents the results of a study concerning the application of
High Intensity Ionizer technology to a flyash electrostatic precipitator. It
includes detailed description of the proposed physical retrofit, results of
flow model tests, and the projected performance of the precipitator retrofitted
by High Intensity Ionizers.
DESCRIPTION OF BASE ELECTROSTATIC PRECIPITATOR
Figure 1 shows the arrangement of the base electrostatic precipitator to
be retrofitted by High Intensity Ionizers. It has 80 parallel ducts. There
are three mechanical fields and five electrical sections in the gas flow direc-
tion. The base conditions of this precipitator, as shown in Table 1, are de-
rived from data provided by the customer. The upgrade requirement is such that
the emission rate should not exceed 60.2 g/109 J (0.14 Ib/mm BTU) which corres-
ponds to 0.11 grams/m (0.048 grains/ACF) at the maximum load of 150 MY.
FUNCTIONAL DESIGN OF IONIZER-PRECIPITATOR SYSTEMS
In predicting the performance of an electrostatic precipitator retrofitted
by High Intensity Ionizers, use has been made of the following resources;
A. All available HII-ESP pilot test data, including those
conducted at TVA John Sevier steam plant, Huang, et al,
(1978)3, which have been systematically correlated.
B. Results of an in-house computer study on three-dimen-
sional space-charge distribution in HII-ESP .systems,
as affected by the design of the transition zone be-
tween HII and ESP, as well as the prevalent space
charge density therein.
315
-------
C. Results of an empirical correlation for optimal
HII placement in an ionizer-precipitator system.
D. EPA-SRI Computer Program, McDonald (1978)4, for
electrostatic precipitators, modified to use for
ionizer-precipitator systems.
E. Comments of Union Carbide consultants with recog-
nized expertise in precipitation technology.
The empirical correlation of optimal HII placement in an ionizer-
precipitator system takes into account the fact that as the single HII array is
moved downstream from the inlet of a precipitator with a fixed overall length,
the collection area of the precipitator section downstream of HII becomes
smaller, resulting in having a smaller specific collection area of the precipi-
tator being augmented by the enhanced charging capabilities of the HII array.
Furthermore, the total charge imparted to the entrained particulates by the HII
array decreases due to the reduced concentration and smaller mean particle size
of the particulates processed by the HII array, as the HII array is moved down-
stream. On the other hand, ionizer performance is improved as the grain loading
of the gas entering the HII array decreases. This allows the HII to operate
more closely to its maximum design capabilities in terms of achievable corona
power density and specific corona current. The space-charge density in the
transition zone between the HII array and the downstream precipitator also de-
creases, so that the chances of discharge and arcing, which tend to neutralize
the high charge produced by HII, are significantly reduced. An additional
effect favoring a downstream location is that the HII is applying all its
charging potential to smaller particles which are typically more difficult to
collect. Thus, factors affecting the optimization of ionizer-precipitator
systems include the collection efficiency of the base electrostatic precipitator
without HII, inlet grain loading, particle resistivity, prevailing ambient pres-
sure, purge air flow as a fraction of total gas flow, and any change in precipi-
tator flow distribution due to the installation of HII.
Figure 2 shows that the maximum performance of a typical HII-ESP system is
to be obtained by placing HII between fields and at about eleven feet down-
stream from the leading edge of the first collecting field of the precipitator.
The EPA-SRI computer program for ESP performance simulation, McDonald (1978) ,
developed by Southern Research Institute, was modified to allow electrostatic
charge to exceed the saturation level permitted by the local electrical field.
Charge density level achieved at a given location within the precipitator is
carefully investigated to monitor the amount of particle charge added by HII.
A preliminary empirical correlation of electrostatic charge produced by HII was
developed based on limited experimental data available to date. Additional work
is required to develop a more complete understanding of HII charging capability.
The maximum charge density permissible in a given transition zone design was
assessed.
In order to make space for the installation of the HII array, three feet of
collecting plate length of the base electrostatic precipitator must be removed.
Our results show that by placing the HII array at a location 12 feet downstream
316
-------
from the leading edge of the first collecting field of the base precipitator,
a system performance of 99.361 could be expected.
As shown in Figure 3, the mass loading decreases rapidly in the region
downstream from the High Intensity Ionizer whereas the charge density increases
first before it decreases rapidly thereafter.
Figure 4 illustrates the dependency of the predicted system performance on
the standard deviation of velocity distribution. The fractional collection
efficiency of the HII-ESP system computed by the modified EPA-SRI program is
illustrated in Figure 5.
FLOW MODEL STUDY
The objective of this experimental program was to determine the gas flow
distribution downstream of HII, which results from a change in the position of
the inlet turning vanes and the installation of discharge electrode system, HII
array, and a velocity distribution device downstream of HII. The mechanical
arrangement of a 1/10 scale flow model tester is shown in Figure 6. The labora-
tory assembly of this unit is shown in Figure 7. Figure 8 illustrates the
schematic diagram of the velocity profile recording circuitry.
The hot wire anemometer used is the Data Matrix Model 800-LV with a Model
U-18 probe. An automatic traverse device was developed to produce predeter-
mined vertical and horizontal motions of the hot wire anemometer, whose signal
is sequenced and timed by the data logger. A teletype records on tape typically
520 to 1040 data points per test, which are fed into a data processor where the
standard deviation of flow distribution is computed. A Calcom plotter is also
connected to produce three-dimensional velocity distribution plots. Figure 9
illustrates the modification of turning vanes implemented during the test
series. Sample results of the flow model tests are included in Table 2. Fi-
gure 10 illustrates a typical chart recorder plot of the velocity distribution
downstream of HII, whereas Figure 11 depicts a three-dimensional velocity dis-
tribution plot.
MECHANICAL SYSTEM DESIGN
The HIT system, as shown in Figure 12, consists of a purged bulkhead
resting on support beams, with an array of installed HII throats and a high
voltage discharge system. Each throat consists of a bellmouth, diffuser, purge
rings and an exit cone (see Figure 13). A supply of clean, heated gas is re-
quired to purge the rings in the HII throats, in order to effectively prevent
back corona therein. The discharge electrode system consists of a mast and
electrode assembly which is suspended from the precipitator roof, supported by
insulators and stabilized at the bottom. A velocity distribution device is
located downstream of the bulkhead. A commercially available high voltage
power supply and control system is used.
317
-------
To install the HII system in an existing electrostatic precipitator pre-
sents two fundamental design challenges, namely, physically locating the HII
and determining a source and system for the purge gas supply. The present
design was for a side-loaded precipitator with HII located downstream of the
inlet and between the fields. Figure 14 presents the proposed arrangement.
To provide space for the HII system in this existing precipitator, three feet
of collecting plate length from the second field had to be removed and use was
made of the internal catwalk space between the fields. In a side-loaded pre-
cipitator, temporary reinforcement of the side walls and roof will be required
before cutting an opening for the installation of the bulkhead. The bulkhead
assembly could then be slid into the precipitator on its support beams. The
discharge electrode system and distribution device could also be installed
from the same location (see Figure 15). High voltage power supply for the HII,
e.g. transformer/rectifiers, is located on the roof of the precipitator. In
some cases, a relocation of the transformer/rectifier sets for the existing pre-
cipitator may be required to prevent interference. Insulator housings and the
high voltage feed to the discharge electrode system are located on the precipi-
tator roof.
Reliable operation of the HII system requires a continuous purging of the
HII anodes and a source of heat to temper the purge gas, as required. Several
potential sources of heat are available in most applications: blow-off steam,
air coming off the preheater, electric heaters, waste heat, e.g. boiler-house
air, or clean flue gas. Since an ample supply of boiler-house air was avail-
able, the arrangement as shown in Figures 16 and 17 was selected. This arrange-
ment of the purge gas system consists of: spool piece with filters, inlet
nozzle into the bulkhead, duct work, fan, heat exchanger and electric heater as
a back-up system.
Based on a comprehensive evaluation of many design options, the design
described above represents the most efficient and economical choice available
for this site-specific application.
SUMMARY AND CONCLUSIONS
In defining the performance of a flyash electrostatic precipitator retro-
fitted by an HII, the following program steps have been taken:
A. Assess performance of the base electrostatic
precipitator, White (1963)5.
B. Define upgrade requirements.
C. Define flow velocity distribution in HII-ESP
by scaled model tests.
D. Define the most economical design choice and
perform mechanical system design.
E. Predict the performance of ESP retrofitted by HII.
318
-------
These steps have been carried out in a systematic manner. As a result of
this study, we have proposed to retrofit a flyash electrostatic precipitator
by installing an H1I array at a location 12 feet downstream from the leading
edge of the first collecting field of the precipitator, removing three feet of
collecting plate length from the second field, and installing a suitable velo-
city distribution device downstream of HII. Based on the flow model test re-
sults and available HII-ESP pilot test data, the performance predicted for this
HII-ESP system is well above the level required to meet the emission rate of
60.2 g/109 J (0.14 Ib/mm BTU).
ACKNOWLEDGEMENTS
The authors wish to acknowledge the able assistance of Dr. Roger Brown,
in performing the computation using the EPA-SRI Computer Program for electro-
static precipitators; Mr. W. C. Fong, in supervising and implementing the
flow model tests; Dr. Ravi Bansal, in designing the flow model test data
processing system; Mr. A. B. Stewart, in designing and constructing the auto-
matic traverse device used for the flow model tests; and Mr. E. L. Tytka, in
instrumentation and data reduction.
We also appreciate many useful discussions with Dr. Harry J. White,
Scientific Consultant, Carmel, California.
REFERENCES
[1] Schwab, J., et al. Development Program for an lonizer-Precipitator
Fine Particle Dust Collection System as Applied to Coal=Fired Utility
Steam Generators. Final Report EPRI FP-291, Volume I § II, October 1976.
[2] Tassicker, O.J. and Schwab, J. High Intensity Ionizer for Improved
ESP Performance EPRI Journal, 56-61, June/July 1977.
[3] Huang, C. M., et al. Pilot Evaluation of High-Intensity Ionizer for
Improving Electrostatic Precipitator Efficiency. Paper #73B, AIChE 85th
National Meeting, Philadelphia, Pa. June 4 - 8, 1978.
[4] McDonald, J. R. A Mathematical Model of Electrostatic Precipitation.
Volumes I § II, EPA-600/7-78-lllb. June 1978.
[5] White, H. J. Industrial Electrostatic Precipitation Addison-Wesley.
1963.
319
-------
40 DUCTS
GAS
FLOW
525.00 AC FM
AT
320°F
40 DUCTS
12'-
12'-
•33'+4' —
Figure 1, Arrangement of base ESP.
0 .2 .4 .8 .B
FRACTION OF COLLECTING PLATE LENGTH UPSTREAM OF Mil (-1
Figure 2. Sample results of performance
prediction for an HII-ESP system.
320 7
-------
a
o
s 10-'
10-*-
- GRAIN LOADING
O • 625,000 ACFM
G| - 3.0GR/ACF
T • 320"F
V • 97.52X
iw - 0.20 FPSI
0 3 E 9 12 IS 18 21 24 27 30 33
LENGTH OF COLLECTING PLATE" FROM INLET 1FTI
Figure 3. Sample results of performance prediction
for" an HII-ESP system.
-0.17lbj/mmBTU •
0.14 lb!/mm BTU
Q - 525.000 ACFM
G[ » 3.0 GRAINS/ACF
0 0.1 0.2 0.3 0.4 0.5
NORMALIZED STANDARD DEVIATION OF VELOCITY DISTRIBUTION. S(-)
Figure 4. Summary of predicted HII-ESP
performance results.
321
-------
80,0-
70,0-
60.D -
50.0-
40.0-
10.0-
NOTE: 12 FT OF EFFECTIVE COLLECTING PLATE
LENGTH BEFORE HII.
1 2 4 6 8 10 20 40 60 100
PARTICLE DIAMETER [ ,u ml
Figure 5. Fractional collectional
efficiency of HII-ESP system.
Figure 6. Schematic diagram of flow model tester.
322
-------
Figure 7. Flow model tester assembly.
323
-------
AIR
HII
-v
\
s
II
][
\
ESP
Figure 8. Schematic diagram of HII-ESP flow model
velocity profile recording circuit.
GAS FLOW
GAS DISTRIBUTION
BAFFLE
2°
Figure 9. Modification of turning vanes,
324
-------
TEST NO. 1
DUCT NO. 10
DUCT NO. 12
DUCT NO. 18
Figure 10. Velocity distribution.
325
-------
35.
CO
ro
cr>
H-
OP
CD
O
o
Cu
O'
rt
H-
O
O
rt
S-AXIS(IN)
TEST NO. 3
-------
INSULATOR
INSULATOR COMPARTMENT
DISCHARGE ELECTRODE ASSEMBLY
GAS FLOW
COLLECTING
FIELD
I
,'. p
:•:
v
:•; x-x- •
c** c
TiV>, , '.'J^
SiSivXv
•.'.•.•.'.V.V,
fff •Ksl\-'
•'.<•'. x-!{xj
Xv v^Ii^;
:::::XxXx:
*•"•*»*•*•*•"•*•*.
x:|:::vx|:;:
:$:Sx?S
?:$:?:?:^;
*t.
\ m m-
J«.
~— ,
/
COLLECTING
FIELD
II
LE
^-~-GAS DISTRIBUTION DEVICE
--—BULKHEAD
x-SUPPORT STEEL
A
LEGEND
Figure 12. High intensity ionizer system.
327
-------
BELLMOUTH
DISCHRAGE
ELECTRODE
ASSEMBLY
PURGE RINGS
EX IT CONE
Figure 13. Throat assembly.
328
-------
\
Figure 14. High intensity ionizer system.
329
-------
Figure 15. Construction alterations.
330
-------
HEATER ,- HEAT EXCHANGER
LEGEND
:X:*1 HII EQUIPMENT
Figure 16. High intensity ionizer system arrangement (elevation)
331
-------
ELECTRIC"!
HEATER i /-FAN
HEAT
EXCHANGER
' s- TRANSFORMER/RECTIFIER
HEAT
EXCHANGER
INSULATOR HOUSING
I-x HII EQUIPMENT
ELECTRIC
HEATER
BOILER HOUSE
Figure 17. High intensity ionizer system arrangement (plan view).
332
-------
TABLE 1. DESIGN BASIS FOR HII RETROFIT
LOAD
FLOW RATE
TEMPERATURE
INLET GRAIN LOADING
MIGRATION VELOCITY
COLLECTION EFFICIENCY
OUTLET GRAIN LOADING
RESISTIVITY
MASS MEAN DIAMETER
PARTICLE SIZE
STANDARD DEVIATION
SULFUR CONTENT
ASH CONTENT
MOISTURE
HEATING VALUE
150MW
525,000 ACFM
320° F
3.0 GR/ACF (4.32 GR/SCF)
£ 0.21 FT/SEC
97.52% (PRESENT ESP)
0.0672 GR/ACF (PRESENT ESP)
= I011fl — cm
24 n m
2.4
1.1 TO 1.3%
14.88%
7.52%
= 11.625 BTU/LB
TABLE 2. SAMPLE RESULTS OF FLOW MODEL TESTS
.._.I7
GAS
FLOW
1 1
1 1 s-
1 ' '
1 1 ' V^-
!!v.
FUN NO,
VELOCITY DISTRIBUTION DEVICE NO.
HOTWIRE PROBE POSITION
FACE VELOCITY IFT/SECI
APAO '"H301
af-AB ""HjOl
APHII * DISTRIBUTION DEVICE ("H2OI
APCD l"H2°'
S(-)
1
1
E
4.M
3
1.12
0.42
11
0.67
2
1
f
4J9
Z.40
0.9
0.35
1.1
0.«5
3
3
E
S.S2
3.90
1.44
0.64
1.80
O.M7
333
-------
BOXER-CHARGER - A HOVEL CHARGING DEVICE
FOR
HIGH RESISTIVITY DUSTS
By:
Senichi Masuda and Hajime Nakatani
Department of Electrical Engineering
Faculty of Engineering, University of Tokyo
7-3-1, Hongo, Bunkyo-ku, Tokyo, Japan
Akira Mizuno
Ishikawajima Harima Heavy Industries Ltd.
Technical Research Center
3-2-16, Toyosu, Koto-ku, Tokyo, Japan 135
ABSTRACT
A novel charging device has been developed for effective precharging of
high resistivity dusts to be collected in electrostatic precipitators, bag-
filters or other electrostatically augmented dust collectors. Charging is
performed by mono-polar ions in an ac field by bombardment from both sides of
dust particles, so that this device is called "BOXER-CHARGER". The charging
current density is one order of magnitude higher than that in the conventional
dc corona current, so that its charging speed is very high, allowing its insta-
llation inside the inlet-duct where gas speed is higher than 10 m/s. The de-
vice consists of parallel electrode assembles facing each other, between which
an ac main voltage is applied. They are excited alternately by high frequency
voltage sources at the negative period of the main frequency to produce a planer
plasma ion source which supplies negative ions to the inter-electrode space.
The charge accumulated on the dust deposit on the electrode assemply is complet-
ely eliminated by this plasma in each exciting period, so that no back disch-
arge occurs even at the resistivity beyond 10^-5 ohm-cm. The characteristics
and performance of this charger are decribed with its application data in a
model precipitator in laboratory.
334
-------
BOXER-CHAEGER - A NOVEL CHARGING DEVICE FOR HIGH RESISTIVITY DUSTS
1. INTRODUCTION
The development of high performance pre-chargers capable of charging dust
particles, especially with high resistivity, to a theoretical limit in a short
time has become a matter of increased concern. Such pre-chargers can be used
not only for retrofit purposes, but also in combination with the advanced pre-
cipitator systems, such as the pulse charging type precipitator. They can be
also used in all of the electrically augmented dust collectors. The High Int-
ensity Charger developed by Air Pollution Systems Inc. and being tested in the
EPRI Test Facilities in Alapahoe Power Station represents one of such develop-
mental efforts. Another effort is being made by the Southern Research Instit-
ute to develop the tri-electrode type charger having a grid electrode near the
collecting electrode.
The authorsand his co-workers have developed another novel type of charg-
ing device called "BOXER-CHARGER" in which charging is made by bombardment of
unipolar ions in an ac field. This device consists of the parallel planar el-
ectrode assemblies facing each other, between which an ac main voltage is appl-
ied to produce an ac charging field. In synchronization to this voltage is
applied a high frequency exciting voltage to each one. of the electrode assembl-
ies alternately to produce a plasma on its surface when it becomes negative in
polarity. The plasma emits negative ions to the charging space, so that dust
particles coming into this space are bombarded by the negative ions from both
sides alternately. The charging current density by these negative ions is ca.
one order of magnitude higher than that obtainable in the conventional dc coro-
na charging. As a result its charging speed is very high, allowing its insta-
llation inside an inlet-duct of a collector. The charged dust particles under-
go an oscillatory motion, so that they are mostly fed out of this charger with-
out being collected on the electrode assemblies. The dust deposits occuring
on the electrode assemblies in a small amount do not cause back discharge, be-
cause the charge accumulation due to oncoming ions is quickly neutralized by the
plasma to appear in the next exciting period. This results in a large advant-
age of back discharge free operation in this device even at a very high dust re-
sistivity beyond 10 ohm-cm.
In this paper are described the basic construction and charging characteri-
stics of the BOXER-CHARGER with its application data in a model precipitator in
the author's laboratory.
2. PRINCIPLE AND CONSTRUCTION
Figure 1 illustrates the basic construction of the BOXER-CHARGER which is
the modification of its prototype (Masuda (19T8)1) made applicable in the dusty
gas at elevated temperatures. In Figure 1 (a), three planar electrode assemb-
lies are arranged parallel to each other in the gas flow direction. Each ass-
embly consists of a number of parallel discharge electrodes as illustrated in
Figure 2, connected every two units to form two groups which are insulated from
each other. When a dc or ac exciting voltage is applied between the two groups
of the discharge electrodes, corona discharge occurs to form a planar plasma ion
source along the electrode assembly. The main ac voltage of a sinusoidal or
square wave form at a low frequency of 50 - 500 Hz is applied between the adja-
335
-------
EXCITING MAIN
VOLTAGE(A) VOLTAGE
T ? ? T
y laaJ
EXCITING
VOLTAGE(B)
cent electrode assemblies (A) - (B) and (.B) -. (A1)to produce the uniform ac ch-
arging fields therebetween. When an electrode assembly takes a predetermined
polarity, negative in this case, its is supplied with the exciting voltage,
which is a high frequency ac voltage at 1.5 - 20 kHz in this case, to produce
the planar plasma ion source over this assembly. Thus, the electrode assemb-
lies (A), (B) and (A1) are alter-
nately excited as shown in Figure
1 (b) during the negative half
period of the main voltage. In
the next positive half period,
the excitation is interrupted so
that no positive ions are supplied
to the charging spaces. The ex-
citation has to be stopped slight-
ly ahead of the polarity change to
allow the extinction of residual
plasma capable of providing posi-
tive ions, so that the exciting
period Tc is made shorter than the
half period of the main voltage.
The negative ions are extracted
by the main field from the planar
ion sources to travel across the
charging spaces in the alternat-
ing directions, and bombard the
dust particles coming into the
charging spaces from both sides
alternately to impart them negat-
ive charge. The residual ions
arrive at the opposite electrode
assemblies to be absorbed there.
(a) Electrode Assemblies.
(A) , (A")
, , /Tc : Charging Period \
\Te : Elimination Period/
(b) Voltage Applied to (A), (A1) and (B).
Figure 1 Construction of BOXER CHARGER-
If, however, the opposite assemb-
lies are contaminated by high res-
istivity dust, the oncoming ions
are accumulated on the surface of the dust deposit to raise its surface poten-
tial. But this accumulated charge is effectively eliminated in the next half
period by the plasmas to appear on the opposite electrode assemblies now in the
exciting cycle. Hence, under the proper design of current density, J, and ex-
citing period, Tc, the maximum surface potential of the dust deposit can be kept
below its breakdown threshold, so that no back discharge takes place. In case
the elimination action is insufficient in the excitation period, an additional
excitation for elimination has to be made for a period Te, as shown in Figure 1
(b), at which the main voltage must be kept zero to avoid the emission of posi-
tive ions. The possibility of this back discharge free operation represents
one of the largest advantages inherent to the BOXER-CHARGER. The second advant-
age is the uniformity of charging field in space which contributes to the well
defined theoretical charge to be imparted to dust particles. The third advan-
tage is that the particles, even if highly charged, undergo only an oscillatory
motion without being collected onto the electrodes as in the case of cLc corona
charging, so that they can be supplied at the charger outlet with little loss.
The forth advantage is that, owing to the bi-directional charging, the particles
having an extremely high resistivity can be much better charged than in the case
of the uni-directional charging by dc corona (Masuda and Washizu (1979)2).
336
-------
••0.2
Insulator
(pylex-glass)
Metal
(Al tape)
t-1.2
(a) Proto-type
(b) Assynmetrical
type
't-1.2
(c) Symnetrical type I
The fifth advantage is the high field intensity beyond 10 kV/cm obtainable in
air at NTP under medium or low dust resistivity condition, which provides &
high saturation charge to the dust particles. Finally, the sixth advantage is
a very high ionic current density, J, more than one order of magnitude higher
than that obtainable in dc corona charging, which results in a very high char-
ging speed as to allow its installation in the inside of an inlet duct of a
precipitator or other dust collector. Figure 2 illustrates the different con-
structions of the electrode assem-
blies tested in the present resea-
rch, (a) represents the proto-
type electrode assembly consisting
of two series of strip electrodes
attached on both surfaces of an
insulator plate in zigzag relat-
ionship. The exciting voltage is
applied between the two series to
cause silent glow corona on the
periferies of each strip electrode.
The insulator plate serves as a
holder of the strip discharge el-
ectrodes and avoides the occurence
of sparking which can cause igni-
tion of flammable gas or powders.
This represents a large advantage
in its applications in electrost-
atic painting or powder coating,
although its ion emissibity is
rather poor. (b) represents an
assymmetrical construction consis-
ting of strip discharge electrodes
and non-corona rod electrodes arr-
anged alternately. This constru-
ction allows the use of a dc bias voltage to be inserted in series to the exci-
ting voltage, which hampers the occurence of corona in the non-exciting period.
As a result, the applicable main field intensity can be raised, so that the sat-
uration charge to be imparted to dust particles can be increased. On the other
hand, its charge elimination ability is rather poor, so that it cannot be used
in the high resistivity dusts. (c) represents the construction having the lar-
gest teeth length, which provides the highest ion emisibity but lowest applica-
ble main field strength. Hence, its maximum saturation charge is rather limit-
ed, whereas its charge elimination ability is very high so that it is suitable
in high resistivity dusts. (e) represents the construction having the smallest
teeth length which provides a high saturation charge and poor charge elimination
ability, so that it is suitable to medium or low resistivity dusts. (d) is the
intermediate construction having the chopped teeth, which can be used in a wide
dust resistivity range. The detailed characteristics of these electrode assem-
blies are to be described later.
104>
(d) Symnetrical type II (e) Synmetrical type III
Figure 2 Construction of electrode
assemblies-
3. THEORETICAL CHARACTERISTICS
The performance of BOXER^CHARGER has to be considered from two different
aspects; one from its charging ability in terms of the maximum charge and char-
ging speed, the other from its charge elimination ability to avoid back dischar-
337
-------
3,1 Charging characteristics
When a square wave main voltage is used, the quantity of charge imparted
to a spherical particle by BOXER-CHARGER can be expressed by Pauthenier's equa-
tion (Pauthenier and Moreau-Hanot (1932)3);
Q = 15 kV/cm
11 kV/cm
=10 kV/cm
>10 kV/cm
lU kV/cm
Est max
(T=150°C)
15 kV/cm
10 kV/cm
6.5kV/cm
=10 kV/cm
12 kV/cm
tabulated in Table 1. The effect of teeth length as described in the preceding
section is clearly indicated. The effect of the bias voltage, V^, is also tes-
ted in the assymmetrical configuration (b) in Figure 2 where the saw teeth ele-
ctrode (c) is used instead of the strip electrode. When V-^ = 0, the magnitude
of Est max is 10 kV/cm, which can be raised to 13 kV/cm at V^ = 8 kV. This in-
crease is resulted by the dc bias voltage inserted in series to the exciting
voltage to lower the concentration of field at the teeth edges. It should be
noted that these values of E , measured in the dc tests does not immediately
c: T" Tf| Q ~y
provide the magnitude of E^x in the BOXER-CHARGER expressed in terms of avera-
ge field intensity, Ema^. = Vmax/d. In BOXER-CHARGER the field intensity at the
electrode assemblies at rest (without excitation) becomes maximum owing to the
space-charge field, so that the magnitude of Emax as above is always more or
less smaller than that of E
In the ideal case when the space-charge li-
._ „„ _ — _i__ - _.__.. — j- _
mited current is prevailing, Equation (5) to be decribed later provides
(.2/3) E,
Jst max'
338
-------
3.1.2 Charging time constant. Since the magnitude of E must be kept as high
as possible to raise the saturation charge, Qc», the only possible way to low-
er the charging time constant, T, to increase the charging speed is to raise
the charging current density, J, up to its maximum value, Jmax, obtainable at
E = Emax- Jmax is limited by either of the two factors. When the planar
ion source can produce a sufficiently high plasma density to provide mono-polar
ion emission more than needed for the space-charge limited current density,
Jsp, the magnitude of J will be limited by the value of Jsp at E = Ejnax, i»e.
Jsp max- Otherweise, depletion occurs at the ion source, so that Jmax is
limited by the ion emitting ability of the source at E = Emax and becomes lower
than JSp max- The duty factor of the corona excitation, DF, is also another
factor.
We derive the general expression of the charging current density, J, at
first. The potential U, local field intensity E and charging current density
JSp inside the charging space under the space-charge limited condition can be
derived by solving the Poisson's equation and Ohm's law, considering the corr-
esponding boundary conditions. The solutions are:
U = V0 (1 - x/d)3/2 (10
E = - (3/2) E0 (1 - x/d)1/2 (5)
Jsp = (9/8) e0y E02/d (6)
where Vo = main voltage, Eo = average field intensity, x = distance from the
electrode assembly at rest, and d = gap between the two adjacent electrode
assemblies. The charging current density can be expressed in terms of JSp as
follows:
J = K-DF-Jgp (T)
= K.DF-(9/8)e0p E02/d (8)
where K = coefficient of ion emission ability < 1.
In order to examine the validity of the relationship as described by Equa-
tions (7) and (8), experiments are made using the apparatus shown in Figure 3,
the the exciting voltage applied. The magnitude of J is affected by E0, the
exciting voltage, Yex, its frequency, fex, gas composition and its temperature,
T, and electrode geometry- The measurements are made in air at normal press-
ure and different temperatures.
Figure U represents the effect of Vex on J in the electrode assembly (c)
measured at DF = 1 and a frequency fex = 20 kHz with different values of EQi J
rises linearly with Vex until sparking occurs between the two adjacent disch-
arge electrodes, and no saturation occurs. This indicates that even this type
of the electrode assembly •having the highest ion emission ability does not pro-
vide a sufficient ion density so as to produce the space-charge limited current.
Figure 5 indicates the effect of Eo on J in the electrode assembly (c) mea-
sured at DF = 1 and Vex just below the sparking threshold with the differ-
ent values, of fex, The dotted curve (l). represents the space-charge limited cu-
rrent, and the curve (.5) the game characteristics £n the. assembly (.a). It can
be seen that the electrode assembly (c) can produce a quite satisfactory curr-
339
-------
3456
Exciting Voltage V
(kV)
Effect of Vex on J at
Figure h
de-tests.((c); d=5.5cm; DF=1;
20kHz; air at T=20°C)
(1) Space charge limited current
(2)
(3)
ex
20 kHz
2 kHz
.electrode(c) /
/
(1)
/
(2)
01234 56
Main Field Strength E0(kV/cm)
Figure ^ Effect of Eo on J at de-tests-
((a) and (c); d=5,5cm; DF=1; Vex Just below
the sparking threshold; air at T=20°C)
ent density J = 0.76 Jsp at Eo = 6 kV/cm with fe3= 20kHz and J = 0.57 Jsp at the
same field intensity even with the lowest frequency fex=l kHz. Whereas the as-
sembly (a) having an insulator plate produces a much lower current density J =
0.10 Jsp at the same field intensity with fex = 9 kHz. J rises proportionally
with the square of Eo, and the effect of fex is not substantial, allowing the
use of a low exciting frequency, such as fex = 1 kHz.
The validity of Equations (7) and (8) are further tested by plotting (Jxd)
against Eo at d = 35 and 60 mm and fex =1.5 and 20 kHz, using the assembly (c).
The results obtained are shown in Figure 6, which indicates the rise of (Jxd)
to be approximately proportional to Eo^, as expected from Equation (8). The
magnitude of J are further measured with changing the duty factor in the range
of DF = 0.1 - 1.0. The exciting voltage is periodically interrupted for this
purpose. It is confirmed that J changes proportionally with DF.
From the foregoing test results, it seems pertinent to express J by Equat-
ions (7) and (8), taking K as a constant representing the ion emission ability
of each electrode assembly. The values of K for each assembly measured under
different conditions are given in Table 2.
Table 2 EMISSION ABILITY K
electrode
(a)
(b)
(c)
(d)
(e)
fex=20 kHz
0.1 - 0.2
0.8 - 1.0
0.8 - 1.0
0.8 - 0.9
0.8 - 1.0
fex=1.5 kHz
-
-
0.6 - 0.65
0.6
-
Finally, the effect of T on J is mea-
sured. The results obtained are shown in
Figure 7- Since the ion mobility, y, is
proportional to the absolute temperature,
J is expected to rise proportionally with
the temperature (see Equation (8)). The
measured results agree approximately with
this assumption.
340
-------
0.4
0.3
0.2
0.01
1 2 3 4 5 7 10
Main Field Strength E0(kV/cm)
Figure 6 (J x d) vs. Eo-
20 °C
123456
Main Field Strength E0(kV/cn)
Figure 7 Effect of Temperature on J
j vex Just below 8kV; air; 209C) ((c); Vex Just below spark voltage; d=5.5
cm; fex=2 kHz; DF=1; air).
3.2 Charge elimination characteristics and conditions for avoiding back corona.
The charge elimination characteristics is largely dependent on the geomet-
ry of the electrode assembly, as previously described. But, it also depends
on the exciting frequency, fex. The use of a low frequency results in a bett-
er charge elimination effect, because a longer time is available for the ions
of opposite polarity to arrive at the surface charge accumulated on the deposit
through Coulombic transport and diffusion. On the other hand, the ac corona,
appearing in a form of point glow at high frequency (20 kHz), tends to turn in-
to streamer corona which causes the reduction in sparking threshold of the exc-
iting voltage, Vex- This, however, does not produce much disadvantage in rel-
ation to J, as indicated in Figure 5.
The back discharge can be avoided when either of the following conditions
are fulfiled:
(i) Charging period ends before the breakdown of the deposit layer on the
opposite electrode assembly at rest occurs, and all the accumulated charge
on the layer is completely discharged by the succeeding elimination period:
Ed = (l/es e )-J-Tc-S § Eds
(9)
where Ed = field strength in the desposit, Tc = charging period, S = space fac-
tor of the electrode assembly, and E
-------
ectrode assembly at res,t within a period when the main voltage appears in its
half cycle, Ta:
ra-yE(t)- dt £ d
o
(10)
When the main voltage is a square wave with no additional elimination period
to be made at zero voltage, Equation (.10) "becomes:
y E • (l/2f) i d
(11)
K CHARGING PERFORMANCE
The charging performance of BOXER-CHARGER is tested using a steel ball
and calcium carbonate powders.
U.I Charging performance test using a steel ball.
A steel ball with 3.0 mm in diameter is hung by a nylon thread (diameter:
100 ym) into the charging space of a BOXER-CHARGER for a sufficiently long ti-
me, and its saturation charge is measured with a Faraday-cage- Tests are made
on the effects of the charging polarity, electrode geometry, charging field in-
tensity, charging position, and the additional charge elimination action.
U. 1.1 Maximum saturation charge. At first the comparison is made of the char-
ging performance between the positive and negative charging polarities, using
the electrode assemblies (c). The polarity can be easily altered by revers-
ing the phase of excitation. Figure 8 shows the results obtained. The theo-
retical saturation charge according to Equation (l) is also depicted. At pos-
itive charging, the saturation charge, Qm, rises with the main field intensity,
E0, approximately along the theoretical line up to Eo = 5 kV/cm, at which, how-
ever, it drops very sharply. This is because the streamers develop from the
teeths of the positive discharge electrodes being excited into the charging sp-
ace to produce bi-polar ions along their channels and decrease the saturation
charge. At negative charging, on the other hand, Qoo rises proportionally with
Eo along the theoretical line up to a much higher level limited by sparking of
the main field, because the development of streamers are greately hampered in
this case. The main voltage frequency used is f = 50 Hz with DF = 1.
Figure 9 shows the comparison of the charging performance between the el-
ctrode assemblies (c) and (e). The magnitudes of Q^ in (c) and (e) do not di-
ffer from each other and agree very well with the theoretical line up to their
sparking thresholds. The sparking threshold of E0, which represents the maxi-
mum applicable main field intensity, Emax (average values as described), is mu-
ch higher in (e) than in (Q) in accordance with the difference in Egt max given
in Table 1. The magnitude of Emax an<3- "the ration -CEst..max/Emax) are given in
. . , . Table 3 for each electrode
Table 3 COMPARISON OF Emax BETWEEN (c) and (e) assemMy together its calcu-
lated value of charging time
constant, T, at Emax• It
can be seen that this ratio
is not much different from
its theoretical value (3/2)
and, thus , Q^,, from the
electrode
(c)
(e)
curve
(i)
(ii)
Emax
6.5kV/cm
8 kV/cm
Est max /Emax
i.U
1.75
T
3.5 ms
IK 5 ms
described above. This enables the estimation of
value of Est max measured from the deftest.
U.I.2 .Augmentation of charge elimination. In the case when the surface of
342
-------
o-
£
4J
CO
Negative
Charging
Positive
Charging
Space
/ Streamer
Main Field Strength E0(kV/cm)
Figure 8 Saturation charge of a steel
ball in the positive and negative char-
ging plotted against Eo.((c); d=83mm;
DF=1; fex=20kHz; f=50Hz; 2a=3mm; 20°C)
(ii)
s
'
8,
h
0 L»
Theoretical
i)
(i) electrode assembly (c)
Emax =6.5 kV/cm
(ii) electrode assembly (e)
=8.5 kV/cm
[a) wave form # 1 (Tc=5ms; Te=0)
(b) wave form # 2 (Tc=Hms
~1 ' ' ' ' (c) wave form # 3 (Tc=2ms; Te=Ums)
4567
Hain Field Strength E0(kV/cm) Figure 10 Wave form of main vol-
Figure 9 Saturation charge of a steel tage used for augmentation of cha-
ball in electrode assemblies (c) and
(e). (negative charging; DF=1; fex=20
kHz; f=50Hz; d=65mm; air at 20°C).
rge elimination ability (frequency
of main voltage f 50 Hz).
343
-------
the electrode assemblis are covered with extremely high resistive layers, the
charge elimination ability of BOXERS-CHARGE at its ordinary operation mode with
DF = 1 becomes insufficient so as to cause back discharge. In order to over-
come this difficulty, the following methods of augmenting charge elimination
can be used:
(a) A part of the half period of the main voltage is made zero, and an additio-
nal excitation for charge elimination is made for a period Te during this
zero voltage interval under zero main field condition (additional eliminat-
ion method).
(b) The excitation for charging is limited to a part of the interval when the
charging main voltage appears (reduced exciting period method). This meth-
od is equivalent to the use of a higher main voltage frequency, The excit-
ing period in this case is denoted as Tc.
These methods are illustrated in Figure 1 (b) and Figure 10 (a) - (c).
Tests are made on these augmentation methods by measuring QK. of the steel
ball under an extremely high resistivity condition where the polyethylene sheets
with resistivity beyond 1Q15 ohm-cm and many pinholes (diameter =0.3 mm; spac-
ing = 5 mm) are attached on all of the discharge electrodes of the electrode
assemblies (d) facing to.each other. Te is varied from zero to U ms whereas Tc
from 2 to 5 ms. The results obtained are compared with the blanc case with cl-
ean electrodes.
The solid curves in Figure 11 show the results obtained using the main vol-
voltages of the wave form # 1 -
# 3 in Figure 10, measured at
different position in the charg-
ing space with fex =1.5 and 20
kHz. The data for the clean
electrodes are also indicated by
the dotted curves. It can be
seen that Q^, rises in the vici-
nity of the electrodes when no
additional elimination is made
(Te = 0), possibly due to the
space charge field. Whereas,
when the additional elimination
is made with Te = ^ ms, Qoo at
the center region is apprecia-
bly lowered and it drops in
the near electrode region. This
tendency becomes more pronounc-
ed whan the lower frequency f =
1.5 kHz is used. These pheno-
mena seems to be caused by the
plasma produced in the additio-
nal elimination period, Te. The
positive ions .from the plasma
will be extracted by the opposi-
tely charged suspended particle
to reduce its saturation charge, Q^, The charge reduction will be more effect-
ive in the near electrode region, and when a lower frequency f = 1.5 kHz is used.
o
o
0?
0)
be
C
O
Center
ll
M
h
Electrode
T w
I a
0
Center
Electrode
024
Distance from Center (cm)
024
Distance from Center (cm)
(a) £„
1.5 kHz
(b)
f =20 kHz
ex
Solid curve: Covered Electrode
Dotted curve: Clean Electrode
Figure 11 Effect of charging position on Q,,,
(E0=3.8kV/cm; steel ball; 2a=3mm; air at 20°C)
344
-------
o
T-l
I
Cf
00
M
A
4J
(0
3 -
3456
Main Field Strength Efl(kV/cm)
(a) fex =1.5 kHz
s 4
I
c-8
3 -
2 -
3 4 5
Main Field Strength E0(kV/cm)
(b) fex = 20 kHz
Figure 12 Effect of augumented charge elimination for avoiding Lack discharge
(d = 80 ram; air at 20 °C; main voltage frequency = 50 Hz).
The visual observation of the coronas shows that they take a form of diffused
streamers bridging across the neighbouring discharge electrodes at the lower
frequency fex=1.5 kHz, whereas they remain to be the stable spot-like glows at
a higher frequency fex = 20 kHz. It is evident that the diffused streamers
can provide the positive ions more effectively to cause the lowering in Qco.
It should be added that a severe back discharge starts to occur at Eo = 3 kV/cm
when the ordinary wave form with DF = 1 is used.
Figure 12 indicates the effect of the methods of augmenting charge elimi-
nation using the wave forms # 1 - # 3 in Figure 10 in avoiding back discharge.
The wave form # 1 represents the case where only the exciting period is reduced
from 10 ms (f = 50 Hz) to Tc = 5 ms. Tc is further reduced in # 2 and # 3 to
4 and 2 ms respectively, whereas the additional elimination is made at Te = k
ms in these wave forms. The magnitude of Q» is measured at fex =1.5 and 20
kHz under the covered and clean electrode conditions at different positions in
the charging space, and its average values are plotted against E0. Again the
solid curves represent the results obtained with the covered electrodes, and
the dotted curves those with the clean electrodes. The theoretical line is also
depicted. It can be seen that the maximum value of Qpo can be greately raised
by using the wave forms # 1 - # 3 which increase the back discharge starting
threshold of Eo from 3 kV in the ordinary wave form with DF = 1 up to U, H.7
and 6 kV respectively. The solid curves rise parallel to and near the theore-
tical line, and they begin to saturate at the threshold field intensities as
described above. If compared at the same magnitude of Eo, #1 with the least
augmentation provides the highest Qoo. But, the largest value of the maximum
Qoo can be obtained by # 3 with the highest augmentation. The dotted curves
for the clean electrodes are also parallel to the theoretical line, and no sat-
345
-------
Feeder
Suction Type
Faraday Cage
Electrostatic
Precipitator
uration appears in the tested range of E , The wave form # 1 provides the
highest value of Qoo, whereas # 3 the lowest value, probably due to the balance
between the effects of different charging time, Tc and eliminating time Te. Qoo
lies above its theoretical value except for the case using # 3 in (a). This
phenomenon is also attributable to the effect of space-charge which produces
in every half cycle of the main voltage the field intensity higher than its
average value Eo at any local point in the charging space. The comparison
between the figures (a) and (b) leads to the conclusion that the use of a low-
er exciting frequency fex =1.5 kHz, much more preferable in all the respects,
is acceptable because the magnitude of the maximum Qoo obtainable at this fre-
quency is almost the same as that obtainable at a much higher frequency fex =
20 kHz, so far as a suitable augmentation measure be taken in charge elimina-
tion. However, it should be noted that the use of any augmentation method
as previosly described causes inevitably the rise of charging time constant,
which requires the increased length of BOXER-CHARGER.
U.2 Charging performance Tests using powder samples.
The charging performance of
a BOXER-CHARGER for powder sampl-
es is tested in a race track sys-
tem as shown in Figure 13 under
the ordinary and back discharge
conditions. Calcium carbonate
powders are used for this purpo-
se, as they provide an ordinary
resistivity p
-------
i.o r
of
^
0)
m
u
u o
•H
••H
O
&
w
1.0
,5 -
Point
A
B
C
D
vg (m/s)
16.3
7.5
4.9
1.7
t/T
0.9
1.8
2.5
8.1
0.5
Dotted line: Q/Q^ = t/(t + i)
012345678
Normalized Charging Time ( t/T )
Figure lU Normalized specific charge of
powder as a function of normalized charg-
ing time ((d); #1; d=120mm; Eo=5kV; T=
21ms; fex=1.5kHz; f=50Hz; calcium carbon-
ate powder; Pc[=9xlo9ohm-cm; air at 20°C).
faO
01234 5
Main Field Strength EQ(kV/cm)
Figure 15 Specific charge of pow-
der at room temperature ((d); #1; d
=120mm; fex=1.5kHz; f=50Hz; calcium
carbonate; pa=9xlo9ohm-cm; air at
20 °C).
it. 2.1 Tests at room temperature. The wave form # 1 is used throughout these
tests. At first, the dependence of powder specific charge on the charging
time, t, is examined, by changing the duct air velocity, Vg. In this case, t
= L/vg. The operating condition of the BOXER-CHARGER is kept constant, so th-
at its charging time constant, calculated from Equation (8) and Table 2, beco-
mes constant at T = 21 ms (see Figure lU). The measured values of the specif-
ic charge are normalized by its value at vg = 1.7 m/s, which corresponds to t/T
= 8.1 to produce 89 % of the its saturation value if Equation (l) holds in this
case. Then, these normalized values are plotted against the charging time no-
rmalized by the charging time constant T = 21 ms.in Figure lU. The point D in
the figure represents the reference value used for normalization. The dotted
curve represents the theoretical curve according to Equation (l). A very good
agreement between the two curves provide a support to the time increase in pow-
der specific charge to follow the theoretical relation (l). Thus, the speci-
fic charge is expected to arrive approximately at its saturation value at t =
3 T.
Then, the powder specific charge Q/M measured at a very low gas velocity
vg = 1.7 m/s (t/T = 8.1 at E0 = 5 kV/cm) and different Eo is plotted against Eo
in Figure 15. The measured values, are expected to represent approximately the
saturation values, and agree very well with the theoretical curve for 20 ym pa-
347
-------
rticle size. The theoretical curve for 30 ym particles is also depicted. As
the average particle size of the powder sample used is ca, 25 pm, the results
in Figure 15 lead to the conclusion that the BOXER-CHARGER can provide the the-
oretical saturation charge according to Equation (2), so far as no back disch-
arge occurs.
U.2.2 Tests at 150 °C. The specific charge of the powder samples is further
measured at 150 °C at which its resistivity becomes as high as pa = 8 x 10l3
ohm-cm to cause a severe back discharge in the ordinary dc corona The three
wave forms # 1, # 2 and # 3 in Figure 10 are used in this case as the augment-
ing means of charge elimination. The measurements are made again at vg = 1.7
m/s using fex =1.5 kHz and f = 50
Hz. The results obtained are plo-
tted in Figure 16. The normalized
charging time, t/T, lies in the ra-
nge of 3 - 8 as shown in Table k,
so that the measured specific char-
ges approximately represent their
saturation values before back dis-
charge begins to take place.
Table
NORMALIZED CHARGING TIME
wave
form
# 1
# 2
# 3
EO (kV/cm)
3
6
5
(2.5)
h
(8)
1
3.3
5
(11)
(8)
k
6
(13)
(10)
5
012345
Main Field Strength (kV/cm)
Figure l6 Specific charge of calcium
carbonate powder at 150 °C ((d); d=120
mm; fex=l-5kHz; f=50Hz; pd=8xlOl3ohm-
cm; air)
It can be seen that the specific ch-
arges rise at first proportionally
with E0 in all of the wave forms.
They arrive at their respective pe-
aks, and then drop sharply because
of back discharge initiation. Their
values in the normal regions are sl-
ightly lower than the theoretically
expected values. The wave form # 3
having the highest augmentation pro-
vides the highest peak value of the
specific charge and the highest thr-
eshold of E0 for back discharge ini-
tiation. It should, however, be noted that the wave form # 1 with the lowest
augmentation can provide a fairly high peak value of the specific charge, only
17 % lower than that given by # 3, although the threshold value of Eo for back
discharge occurence is substantially lowered in # 1 wave form. It is conclud-
ed from Figure 16 that the BOXER-CHARGER can provide approximately the theoreti-
cal specific charge even to the extremely high resistivity powders up to a cert-
ain threshold field intensity of Eo in the range of 3 - 6 kV/cm if the appropri-
ate augmenting wave form for charge, elimination be used. The corresponding
maximum specific charge for the present coarse particles with 25 urn average dia-
meter is ca. 1 pC/g . In case of much amaller particle size, this magnitude is
348
-------
expected to rise in the inversely prpportiona.! relationship with its size.
5. PRECIPITATION TEST
Preliminary tests are made to examine the effect of the BOXER-CHARGER to
enhance the collection performance of a precipitator located at its downstream
using the race track system shown in Figure 13. Four units of a tri-electrode
precipitator are connected in series as a collecting section, as shown in Figu-
re IT. A dc high voltage of 27 kV is applied to the third electrodes, ~and a
variable voltage is applied between the third and discharge electrodes to pro-
vide different current densities, JEp, in the collecting electrodes. Tests are
made at a room temperature T = 30 °C with the gas velocities vg = 7-5 m/s in the
BOXER-CHARGER section and vg = 1 m/s in the precipitator section. The dimen-
dc HV 1
dc HV 2
c
c
£ 500mm-J*
CAD
""' ° IT ° T ° 1 T
D UA
3*1 o 1 J o JL o 1 JL
LT\
C\J
B> ?1
o
,,
i
h
\
/
A: discharge electrode B: collecting electrode ~
C: third electrode D: shielding electrode E: perforated plate
Figure 17 Tri-electrode precipitators used as a collecting section in the
race track system.
sions of the BOXER-CHARGER and its operating conditions are the same as in the
preceding section. Calcium carbonate powders are used as the test sample which
does not cause back discharge under the present room temperature. The magnit-
ude of E0 in BOXER-CHARGER is kept fairly low at Eo ^ k kV/cm. The mass conc-
entration of dust at the outlet of the collecting section, Co (mg/Nm3), is mea-
sured with a continuous dust monitor "KONITEST" based on tribo-electrification.
The mass concentration of dust at the inlet, Ci, is kept at 2 g/Nm3. The
KONITEST is calibrated with a glass filter for the uncharged calcium cabonate
powders used in the present tests. It shows the blanc emission level Ci = 890
mg/Nm3 when both the BOXER-CHARGER and the collecting section are separated from
the voltage sources. Since mechanical collection also occurs in the collecting
section, this blanc emission level is used as the reference emission level to
indicate only the electrical effects, and the mass penetration of dust, Co (mg/
Wm3), under different operating conditions is expressed in a percent ratio as
100 (C0/890).
Figure 18 shows the results obtained at different values of Eo for differ-
ent precipitator current densities JEp. The penetration lowers with increas-
ing E0, and its reduction becomes 30 - 50 % of the initial penetration for
each value of JEP at Eo = h kV/cm. The results at Eo =• h kV/cm are replotted
against JEP in Figure 19, which clearly shows a substantial decrease in penetr-
ation caused by the BOXER-CHARGER at each precipitator current level. The co-
llection efficiencies shown in Figures l8 and 19 are not the overall efficienc-
ies, but those resulted by the electrical- effects based on the reference emiss-
ion level as described above, i.e. (l - penetration). It should be noted that
349
-------
99
^98
£97
c
«3
O)
IBO
±70
o
JEP =
0.039
2
3
5 ^
101
20 |
30 J
50 °-
100
99
C98
£97
c
180
±70
o
050
with
2
3
5 _
10 ^
o
20 2
30 |
100
0 1 2 3 A 5
Eo (kV/cm)
Figure 18 Dust penetration vs. Eo.
(air at 30 °C)
0.1
JEP (mA/m2)
0.2
Figure 19
Dust penetration vs.
(E0 = ^ kV/cm; air at 30°C)
"KONITEST" monitor produces an error reading for the already charged particles
because of its principle, and that this error in this case is expected to be an
over reading. This means that the penetration decrease in the charged conditi-
ons may be higher than that indicated in Figures 18 and 19. This rough meas-
uring means is used primarily for the purpose of quick examination at the pres-
ent stage of tests, and the more exact measurements are being undertaken.
It can be concluded from the present tests that the BOXER-CHARGER evident-
ly enhances the collection performance of the succeeding precipitation section.
6. CONCLUSION
The performance of BOXER-CHARGER as a precharging means is tested in det-
ail, and the following conclusions are derived:
(l) BOXER-CHARGER has a sufficiently high charging performance to be used as a
precharger of an electrostatic precipitator. The magnitude of the average
operating field intensity in its charging space can be made as high as 8 kV
/cm in air at room temperature and normal pressure. Its operating current
density amounts to 60 - 90 % of the space-charge limited current in a para-
llel plane system, which is one order of magnitude higher than the current
density obtainable in the ordinary dc corona system. As a result, its ch-
arging time constant becomes very low at ca. 5 ms. This produces a high
charging speed so as to enable the use of the BOXER-CHARGER inside the inlet
duct of a precipitator.
(2) BOXER-CHARGER can provide a sufficiently large quantity of charge even to
the extremely high resistivity powders with more than 10^3 ohm-cm resisti-
vity, without causing back discharge. It can be operated at the main
350
-------
field strength of 6 kV/cm at 150 CC, eyen when its electrodes are complet-
ely covered with the very high resistivity powder deposit with 8 x 10^3
ohm-cm resistivity. Back discharge is avoided automatically by the inhe-
rent charging mechanism of BOXER-CHARGER , such that the plasmas produced
on the electrode assemblies for charging purpose eliminate at the same time
the charge accumulated on the high resistivity deposit covering the elect-
rodes in each charging cycle.
(3) This charge elimination ability can be augmented easily by the use of an
additional excitation of the electrode assemblies at zero main field and
the reduced excitation period.
(k) The use of a low exciting frequency fex =1.5 kHz and a low main voltage
frequency f = 50 Hz is acceptable, which enables a large cost reduction in
the power pack of BOXER-CHARGER.
(5) The use of a BOXER-CHARGER enhances the collection performance of a precir-
pitator installed at its downstream.
(6) BOXER-CHARGER can be installed not only inside-the inlet duct of a precipi-
tator, but also within its casing in front- of each collecting field, because
its length in the gas flow direction can be made very small when the gas
velocity is very low in the range of 1 m/s.
From these inherent advantages of BOXER-CHARGER, it is expected that it can be
used also for the retrofit purpose of an existing precipitator or other kind of
dust collector to increase its overall collection performance,
ACKNOWLEDGEMENTS
The authors are indebted to their co-workers, Mr. K, Akutsu and 'Mr, M.
Washizu for their great help and valuable discussions.
REFERENCES
1, Masuda, S'« 5 M. Washizu, A, Mizuno, a.nd K'. Akutsu, BOXER-CHARGER - A Hovel
Charging Device for High Resistivity Powders. Record of IEEE/IAS 1978
Annual Meeting (Tronto, Canada), p. 16 (October, 1978).
2. Masuda, S. and M. Washizu. Ionic Charging of A Very High Resistivity
Spherical Particle. Jounal of Electrostatics, 6 (1979) p. 51.
3. Pauthenier, M.M. and M. Moreau-Hanot. La Charge des Particules Spherique
dans un Champ Ionise. J. Pysique Radium, 3 (1932) p. 590.
351
-------
PRECIPITATOR ENERGIZATION UTILIZING
AN ENERGY CONSERVING PULSE GENERATOR
H. Hoegii Peter sen
F.L. Smidth & Co.
77, Vigerslev Alle, DK-2500 Valby
Copenhagen, Denmark
Pre"ben Lausen
High Voltage Laboratory, Building 329
Technical University of Denmark
100, Lundtoftevej, DK-2800 Lyngby
Copenhagen, Denmark
ABSTRACT
The performance of a conventional two-electrode type precipitator col-
lecting high resistivity dust can be improved by pulse energization. Pulses
of a suitable duration and repetition rate superimposed on the DC voltage
permit higher peak-voltage without sparkover, improve particle charging and
current distribution, and allow independent regulation of the precipitator
voltage and current. A considerable quantity of energy, however, is re-
quired for each pulse to charge the precipitator to the pulse voltage level.
For high pulse voltages only a minor part of this energy is necessary for
the discharge current in the precipitator. For reasons of economy, the
recovery of the energy stored in the precipitator capacitance during each
pulse is therefore extremely important for pulse energization of large
precipitators. An energy conserving pulse generator with pulse transform-
er, a pulse initiating switch element, and a feed-back diode for the
energy recovery is described. The design and the instrumentation of a pilot
precipitator specially developed for the comparison of different kinds
of precipitator energization are outlined. The ability of the applied pulse
energization system for controlling the corona discharge current indepen-
dently of the precipitator voltage is demonstrated. Further, results from
field tests showing the performance of the system under operating condi-
tions are presented.
352
-------
PRECIPITATOR ENERGIZATION UTILIZING AN ENERGY CONSERVING PULSE GENERATOR
INTRODUCTION
Pulse energization for improvement of the performance of a precipitator
was first investigated by H.J. White about 30 years ago. It has since been
further examined by a number of investigators in U.S.A., Europe, Australia,
and Japan.
The advantages claimed for pulse energization in comparison with conven-
tional DC-energization are:
Higher peak voltage without excessive sparking, and therefore improved
particle charging in accordance with the classical theory for particle
charging.
- More effective extinguishing of sparks and better suppression of in-
sipient back corona.
By variation of the pulse repetition frequency and pulse amplitude the
discharge current can be controlled independently of precipitator volt-
age. This allows reduction of the discharge current to the back corona
threshold limit for a high resistivity dust without reducing the preci-
pitator voltage.
- With short duration pulses the corona discharge takes place well above
the corona onset level for constant DC voltage and is. suppressed during
the remaining part of the pulse by space charges. This results in a
more uniformly distributed corona discharge along the discharge elec-
trode.
- Corona discharges from short duration pulses are less influenced by
variations in gas and dust conditions. This improves the internal cur-
rent distribution of a separately energized field.
- Corona discharges are obtainable from surfaces with larger diameter
curvatures. This permits the use of large diameter discharge wires, or
rigid type discharge electrodes with comparatively short and blunt
tips, reducing the risk of discharge electrode failures.
- Permits a higher power input and thereby improves precipitator perfor-
mance.
- Increases particle migration velocity, particularly for high resistivi-
ty dusts, permitting reduction of the collection area for new instal-
lations or improvement of the efficiency of existing installations
without increase of the collection area.
Utilization of pulse energization on a commercial scale, however, has not
been realized until now, mainly because of problems connected with energy re-
quirements and pulse generator design. A pulse energization system solving
these problems is described here. Further, results from investigations carried
out to evaluate the practical value of the system are presented.
353
-------
PULSE ENERGIZATION PRINCIPLES
Two-Electrode Systems
In the pulse energization development work carried out by White and Hall
around 1950 a high-power pulse generator, supplying high-voltage pulses m the
order of 100 microseconds' duration to a conventional two-electrode system
precipitator through a hold-off diode, was used. The electric energy stored
in the capacitance of the precipitator at the crest value of the pulse main-
tained the corona current during the period between the pulses. The pulse volt-
age therefore had a fast rise and a slow decay and somewhat resembled a saw-
tooth wave shape in form. The development work was brought to full scale field
test with a prototype pulse generator. The lack of reliable high-power switch
elements, however, then prevented commercial development.
The later advent of high-power electronic switch elements created renewed
interest in pulse energization. The Environmental Protection Agency .in 19&9
thus sponsored a pulse energization development and field test program. How-
ever, only the first phase of this program, which was based on the earlier
work of White and Hall, was completed.
One of the problems in pulse energization is the amount of energy required
for repetitive charging of a precipitator to a high pulse voltage level. An in-
teresting solution of this problem has been described by Milde^. High voltage
pulses of sufficiently short duration to charge only a small portion of the pre-
cipitator at any given instant of time are superimposed on an operating DC vol-
tage. The very narrow pulses propagate as travelling waves along series-connect-
ed discharge wires or discharge wire systems, sequentially charging these. Hereby
is obtained an energy saving in comparison with pulse energization systems
where the total precipitator capacitance is charged simultaneously and the
stored energy is dissipated between pulses either in the precipitator , or in
a pulse forming network 5. jn the energy conserving principle described here,
the saving is obtained by returning the energy stored in the precipitator to
the pulse generator.
Three-Electrode-Systems
Liithi introduced a pulse system with an auxiliary electrode in the vici-
nity of the discharge electrode. With this third electrode it was possible to
create a uniform field with a high sparkover level between third electrode and
collection electrode, and therefore high charging and collecting field
strengths in the main field. The corona-creating pulse field was produced be-
tween the discharge electrode and the third electrode making the corona dis-
charge almost independent of the main field. The corona current could be re-
gulated by the pulse voltage as well as by the pulse repetition frequency.
With this system Liithi obtained independent regulation of the current density
and the charging and collecting field strengths in laboratory precipitator ex-
periments, achieving more favourable electrical precipitator energization with
high resistivity dusts. Pulse widths from 1 to 50 us were used, and the high
ion concentration in these short duration corona discharge pulses was shown to
cause sufficient expansion of the ion cloud to ensure a reasonably good current
distribution on the collection electrode in spite of the uniform main field.
According to Liithi, the capacity of the high resistivity dust layer on the
354
-------
collection electrode smoothed the pulse current flow through the dust layer
sufficiently to avoid back ionization.
Masuda et al.f improved the three-electrode system by introducing a DC
bias-voltage between the discharge electrode and the third electrode. The bias-
voltage maintains the potential of the discharge electrode in the pulseless
period at a level ensuring that no corona discharge occurs between pulses, also
with the gap between discharge electrode and third electrode increased to a
practical level of 10 cm. The bias-voltage also enables the use of half or
full wave AC-voltage. Large scale installations using-this latter mode of ener-
gization are today in operation in Japan."
To improve the distribution of the current density in the dust layer on
the collection electrode in the three-electrode system Penney and Gelfand
have proposed a special third electrode design in combination with a lowering
of the potential of the third electrode during the corona pulse.
GENERAL PRINCIPLE
The system described here is used with conventional two-electrode type
precipitators. Short duration high voltage pulses are repetitively superimposed
on an operating DC voltage, refer Figure 1. The pulse duration is within the
range 50 - 200 ys, and pulse repetition frequencies from 25 to kOO pulses per
second are used.
-100
aT-80
o>
ro
0-60
lm
o
£ -40
'a.
'5
2-20
o_
Pulse repetition frequency 200 pps
Corona onset
0.5 5
Time, ms
5.5
Figure 1. Voltage wave form for pulse energization.
It is characteristic of the operation of the system that the DC voltage
is maintained s.lightly below the corona onset voltage. The object is
to extinguish the corona discharge completely after each pulse. This allows
full control of the discharge current by means of the pulse repetition fre-
quency as demonstrated later.
355
-------
Further, the extinction of the corona discharge after each pulse combined
with suitably long intervals between the pulses allows the DC field to remove
the ion space charge from the interelectrode spacing, before the next pulse is
applied, permitting high pulse peak voltages without sparking.
100
o
o>
m
1 -50
o
•**
eg
'a.
'5
o>
Voltage
100
0 3
O
+*
s
'a.
'o
£
Q.
100
50
Time,
100
Figure 2, Ideal wave forms of precipitator current and voltage.
Figure 2 idealized shows the precipitator pulse current and voltage.
During the first half of the pulse time a negative current flows from the
later described pulse generator to the precipitator, charging the precipitator
capacitance from the operating DC voltage level to the pulse voltage peak
level. For high pulse voltages only a minor part of this charge is emitted as
corona discharge during the pulse in order to maintain a collection plate cur-
rent density of 0.1 - 0.2 mA/m2. The remaining, and major part of the charge
is returned to the generator as a reversed current flow during the second half
of the pulse time, thus bringing the precipitator voltage back to the operat-
ing DC voltage level.
356
-------
ENERGY CONSERVING PULSE GENERATOR
DC supply
II
Charger
II
-DC
Switch
.Jp Precipitator
Pulse transformer
Figure 3. Energy conserving pulse generator principle.
10,11
A basis diagram of the energy conserving pulse generator system ' is
shown in Figure 3. The DC operating voltage is maintained by a DC supply with
a blocking inductor L™ preventing the pulse voltage from entering the supply.
A coupling capacitor CQ blocks the DC operating voltage from the pulse trans-
former. The pulse circuit includes a charger supplying a storage capacitor C^,
a thyristor switch, a feed-back diode, and a series inductance Lg. The storage
capacitor G^5 the series inductance Lg together with the pulse transformer
leakage inductance, the coupling capacitor GC, and the precipitator capacitance
CY forms a series oscillatory circuit.
The storage capacitor Cr> is charged to a controlled DC level by the
charger. The thyristor switch is turned on, and the precipitator represented
by the capacitance C-p is charged to the maximum pulse voltage level by the
first half-period of the oscillatory current. Because of the series oscilla-
tion, the energy supplied to Gp is returned to CR through the feed-back diode
by the current in the second half-period. During this interval, the thyristor
switch is turned off and the current in the pulse circuit is blocked until the
next ignition of the thyristor switch. The energy is stored in Gp during the
interval between the pulses and is used for the next pulse.
Typically, an industrial precipitator has a capacitance of kO - 50 pF per
square meter collection area. Charging of this capacitance to a high pulse
voltage level a large number of times per second without energy recovery would
require a considerable quantity of energy. Depending on pulse voltage level
and pulse repetition frequency, the charging energy might amount to several
times the energy absorbed by the precipitator as useful corona discharge ener-
gy. For reasons of economy, recovery of the charging energy is consequently
extremely advantageous in connection with pulse energization of large precipi-
tators.
357
-------
CURRENT-VOLTAGE RELATIONSHIP
The electrical parameters, as described by the corona current density at
the collection electrode surface and the precipitator voltage, determine the
precipitator performance.
For conventional DC energization, the relationship between the parameters
mentioned above is usually presented as a current-voltage curve, showing ave-
rage value of current density versus average value of precipitator voltage.
For pulse energization the current-voltage curves can be presented as ave-
rage value of current density versus peak value of superimposed pulse voltage
Up for various pulse repetition frequencies. For such a set of curves, the
operating DC voltage level UDC and the pulse width are fixed. Figure U shows
an example of such current-voltage curves obtained with a special pulse dis-
charge electrode.
These current-voltage curves have been obtained under operation conditions
with hot dustladen gas and with the DC voltage level set slightly below the
corona onset voltage.
E
E
« 0.1
c
0>
•o
c
fc
3
O
0.01
DC voltage 30 kV
Pulse width 60 \is
PRF = 400 pps
Gas temperature 250°C
15
20 25
Pulse voltage, kV
30
Figure k. Current density versus pulse voltage with pulse repeti-
tion frequency PRF as parameter.
358
-------
Alternatively, the same set of current-voltage curves can be presented as
current density versus pulse repetition frequency for various pulse voltages
as shown in Figure 5.
DC voltage 30 kV
Pulse width 60 us
Gas temperature 250°C
O
100 200 300
Pulse repetition frequency, pps
400
Figure 5.
Current density versus pulse repetition frequency with
pulse voltage U as parameter.
The curves show that, within wide limits, the precipitator current can
be regulated independently of the precipitator voltage by variation of the
pulse repetition frequency. A practical consequence of this is that it becomes
possible to adjust the current to the limit for back ionization without re-
ducing the precipitator voltage. This means more favourable electrical ener-
gization for high resistivity dusts than obtainable with conventional DC-ener-
gization, where the current cannot be regulated independently of the precipi-
tator voltage.
DISCHARGE ELECTRODES FOR PULSE ENERGIZATION
Discharge electrodes for pulse energization must primarily have a geometry
which, in combination with the collection electrodes used, will ensure suitable
current-voltage curves for this particular mode of energization.
Further, the electrode geometry must produce a sufficiently high collect-
ing field strength close to the surface of the collection electrode with the
operating DC voltage maintained just below the corona onset voltage. Finally,
the electrode geometry must ensure that the collecting field close to as well
as the current distribution on the surface of the collection electrode are
reasonably uniform.
359
-------
These pulse discharge electrode requirements favour the use of rigid type
electrodes with comparatively short and blunt tips. Further, such designs re-
duce the risk of discharge electrode failure and deterioration.
DOUBLE PIPE TEST PRECIPITATOR
A mobile, double pipe test precipitator in which the operation conditions
of two parallel pipe precipitators can be kept identical during slipstream test-
ing has been used for studying the differences between DC and pulse energiza-
tion as well as the effect on precipitator performance of the different pulse
energization parameters.
Grav. dust meas.
Cooler
rT^LJ
D-
Stack
0
Opacity meter
Slipstream
Figure.6. Schematic diagram of double pipe test -precipitator.
Figure 6 shows a schematic diagram of the double pipe test precipitator.
The gas slipstream from the main gas duct is fed through a conditioner and a
heater to obtain the desired values of humidity and temperature. A carefully
designed, adjustable splitter together with two independently controlled fans
ensures an equal division of dust and gas.between the two parallel precipita-
tors. The temperature of the two precipitators is independently controlled by
heating elements on the pipes.
360
-------
The diameter of the precipitators is 250 mm and their effective length is
3.5 m. The main design criteria have been a collection efficiency at DC opera-
tion of about 99 %, a gas velocity of 0.5 -* 2 m/s, an inlet dust concentration
of maximum 50 g/Nm3, and an operating temperature of maximum UOO °C.
The test precipitator has controls for the operation parameters and equip-
ment for measurements of mechanical and electrical characteristics. The col-
lection efficiency is determined gravimetrically "by measurement of the outlet
dust concentration together with weighing of the dust collected in the hoppers.
Opacity meters are inserted in the tubes leading from the precipitators to the
stacks.
For measurement of the precipitator voltage a voltage divider is used, de-
signed for a peak voltage of .120 kV and a frequency range from DC to 20 MHz..
The current to the discharge electrode is measured at high potential by means
of a shunt and a battery operated oscilloscope. The shunt is placed directly at
the discharge electrode in the hot gas in order to eliminate the effect of
leakage resistances and stray capacitances, refer the cross section of the pipe
precipitator shown in Figure 7.
Insulator
Insulator
Shunt
Heating tape
Discharge
electrode
Weight
AI203 stabilizer
Figure 7. Cross section of pipe precipitator.
361
-------
Further, the current flowing to the collection electrode part of the pipe
precipitator can be measured separately at low potential. Also possible cur-
rents from the rod supporting the lower end of the discharge electrode, from
the insulator housing, and from the hopper can be measured separately.
The total installation, including the high voltage power supplies and the
control and measuring equipment, is placed in a 30 feet standard container for
easy transport.
PILOT PRECIPITATOR FIELD TESTS
For a period of about one year and a half, the double pipe test precipi-
tator was installed at a lime burning rotary kiln. The precipitator here de-
dusted a slipstream with a gas volume of about twice 200 m3/h, taken from the
350 °C hot exit gases from the 290 t/2h h lime kiln.
The fuel used was a mixture of coal and oil, typically 60 % and hd %, re-
spectively, related to calorific values.
Typical pilot precipitator gas temperature was about 250 C in order to
obtain a high resistivity level, water content about 15 % moisture by volume,
and dust concentration 25 - ^0 g/Nm3. Particle size median was 17 um, and
dust resistivity was 1011 - 1013 ohm-cm at 250 °C, refer Figures 8 and 9 show-
ing dust particle size distribution and dust resistivity, respectively.
99
10 100
Particle size, um
1000
Figure 8. Particle size distribution of dust from lime kiln.
362
-------
10
14
10
13
E
o
I
o
l012
I 1011
w
to
cc
10
109
10
~ 15%
Moisture by volume
Oil and
coal firing
- Oil firing
100 200 300
Temperature, °C
400
Figure 9. Resistivity of dust from lime kiln.
For the pulse energization tests the DC voltage level was usually set just
below the corona onset in order to avoid corona discharge between pulses. DC
corona discharges between pulses resulted in poor current control and deterio-
ration of precipitator performance.
A great number of comparison tests were performed with various pulse ener-
gization parameters and with conventional DC operation as reference. The dis-
charge electrode used in the DC reference precipitator was a 3 mm diameter wire.
The comparison tests were performed simultaneously and with the same corona dis-
charge current density (typically 0..1 mA/m2 or 0.2 mA/m2) in the two pipes. It
was generally not possible to produce sparkovers in the precipitators with these
current densities.
The precipitator efficiency and the w, migration velocity were determined
from measurements of the inlet and outlet aust concentrations and the gas
volume flow rate. The w concept is useful, since the necessary precipitator
volume is inversely proportional to w .
An improvement factor was defined as the ratio between the wk values for
pulse and for DC energization found for the same corona discharge current.
As seen from the following table, the improvement factor was strongly depend-
ent on the degree of back ionization as judged from resistivity measurements
and the corresponding current-voltage curve for DC energization.
363
-------
OPERATION CONDITION
Without back ionization
~ 10^ ohm* cm
Moderate back ionization
~ 1012 ohm -cm
Severe back ionization
~ 1 0 -1 3 ohm • cm
wfc IMPROVEMENT FACTOR
1.2
1.6
> 2
PILOT PRECIPITATOR LABORATORY TESTS
During another period of about one year, the double pipe test precipita-
tor was used in a laboratory set-up comprising a gas burner, a dust feeder,
and a steam generator. The dust used vas from the earlier mentioned lime kiln,
with particle size distribution and resistivity as shown in Figures 8 and 9-
The purpose of the laboratory experiments was to examine thoroughly the
influence on the precipitator performance of pulse voltage level, DC voltage
level, pulse repetition frequency, and pulse width, etc. The tests revealed
that the pulse height is the most important factor influencing the dust loss.
This is illustrated in Figure 10, which shows the relationship between collec-
tion efficiency and precipitator peak voltage for constant DC voltage and for
various current densities.
97.5
~ 98.0
v
o
= 98.5
c
o
•5 99.0
a>
"5
o
99.5
Gas temperature 250 C
0.01 mA/m2
0.025 mA/m2
0.050 mA/m2
0.09 mA/m2
35 40 45
Precipitator peak voltage UDC + UP, kV
Figure 10. Collection efficiency versus precipitator peak voltage
with current density as parameter.
364
-------
The laboratory experiments further confirmed that the precipitator per-
formance is less influenced by high resistivity dust with pulse energization
than with DC energization. Increasing the dust resistivity to a high level by
stopping the steam injection reduced the Wv value for the DC energized preci-
pitator by a factor of about 3, while the w^ value for the pulse energized
precipitator was only slightly reduced.
FULL SCALE TESTS
Full scale tests have been carried out for about one year on a precipita-
tor dedusting the earlier mentioned lime kiln. The total collection area of the
•precipitator is about 1^00 m2, the duct width is 250 mm, and the discharge
electrodes are of the conventional 2.T mm diameter helical type. The gas velo-
city is about 0.6 m/s, giving a treatment time of 6 seconds.
The gas temperature in the precipitator is about 350 °C, the water content
about 15 % by volume, and the dust load about 50 g/Nm3. The dust resistivity
varies from about 1010 to 1012 ohm-cm at 350 °C depending on raw materials and
kiln operation, refer Figure 9-
The precipitator can be energized from either a pulse generator or a con-
ventional DC power supply (single phase, full-wave raw rectified DC).
The sparking rate with this DC energization is iri the order of 60 sparks
per minute at an operating voltage of about 30 kV. With the pulse energization,
the sparking rate is drastically reduced to a value of about one spark every
3 minutes for a sum of operating DC voltage and pulse voltage of about 60 kV.
The installed pulse generator which was originally designed for a smaller pre-
cipitator cannot give sufficient pulse voltage to operate the precipitator at
higher sparking rate where better use might be made of the possibilities of
the pulse energization principle.
The full scale tests performed under these conditions have given w^ im-
provement factors of 1.3 - 1.U thus confirming the wk improvement tendencies
found with the double pipe test precipitator.
A prototype of an automatic control, specially designed for pulse ener-
gization has been in operation at the above mentioned installation for about
four months.
The automatic control equipment regulates the DC voltage level, pulse
height, and pulse repetition frequency on the basis of criteria related to sup-
pression of corona discharges between the pulses, the sparking rate, and the
current density, respectively.
Further, the automatic control includes fast precipitator voltage recove-
ry after sparkover, as necessary in order to fully utilize the pulse energiza-
tion at high sparking rate. This is illustrated in Figure 11, which shows pre-
cipitator voltage recovery after sparkover. It is noticed that the operating
DC voltage as well as the pulse voltage is recovered after approximately 20 ms.
365
-------
Time, 10 ms/div
Figure 11. Precipitator voltage recovery after sparkover.
At present, experiments are "being performed with the purpose of refining
the control criteria and the control equipment. Further, endurance tests to
check the reliability of the equipment are being carried out.
CONCLUSION
Short duration high voltage pulses are superimposed on an operating DC
voltage by means of a pulse generator capable of conserving a predominant part
of the energy stored in the precipitator capacitance at the crest of the pulse
voltage, and not used by the corona discharge. The energy conservation is
especially advantageous with high pulse voltages, high pulse repetition fre-
quencies, and large precipitators. A pulse generator, based on pulse trans-
former and thyristor technology available to-day, has been developed, and has
been in operation for about one year at a full scale precipitator installation.
Further, a mobile double pipe test precipitator with equipment for measurements
of mechanical and electrical characteristics has been built in order to study
the differences between DC and pulse energization. The pilot precipitator
tests, as well as full scale tests, have shown that by maintaining the DC
operating voltage just below the limit for corona discharges between pulses
it is possible to control the discharge current by means of the pulse repeti-
tion frequency, and it is further possible to operate at high precipitator
peak voltages with a low sparkover rate. Regulation of the precipitator current
independent of the precipitator voltage means more favourable electrical ener-
gization than obtainable with conventional DC operation, particularly for ap-
plications with high resistivity dusts.
366
-------
Determination of precipitator efficiency and migration velocity w^., per-
formed on the double pipe test precipitator as well as on the full scale pre-
cipitator, shows substantial w-^ improvements as compared to conventional DC
energization for high resistivity dusts.
The w^ improvement factors show that the pulse energization system can be
used successfully to improve the performance of existing precipitator instal-
lations having conventional discharge electrodes and operating with high resi-
stivity dusts.
The test results further show that considerable reduction of the total
collection area is obtainable for new high resistivity application precipita-
tors utilizing specially designed pulse discharge electrodes.
Finally, the current control capabilities of this pulse energization system
make it well suited for energization of the collection stage of a two-stage
precipitator where effective control of a small discharge current for limita-
tion of reentrainment is desirable.
For all these applications the energy conserving pulse generator system
will be advantageous, particularly for large precipitators.
REFERENCES
1. White, H.J. A Pulse Method for Supplying High-Voltage Power
for Electrostatic Precipitation. Trans. AIEE, Nov., 1952.
p. 326-329.
2. Belco Pollution Control Corporation. Technical Description of
the Belco Pulsed Power Supply. Environmental Protection Agency
Contract CPA 22-69-1^3. 18 March, 1970.
3. Milde, H.I. Reduced Power Input for Improved Electrostatic
Precipitation Systems. US Patent ^,133,6^9. Jan. 9, 1979.
^. Feldman, P.L. and H.I. Milde. Pulsed Energization for Enhanced
Electrostatic Precipitation in High-Resistivity Applications.
Symposium on the Transfer and Utilization of Particulate Control
Technology. Denver, Colorado. July, 1978. Conference Records
Vol. 1, p. 253-27^-
367
-------
5. Fuchs, H. Electronic Dust Separator System. US Patent 3,981,695.
Sept. 21, 1976.
6. Liithi, J.E. Grundlagen zur elektrostatischen Abscheidung von
hochohmigen Stauben. Diss. Nr. 392U. ETHZ, Zurich. 1967.
7. Masuda, S., I. Doi, M. Aoyama, and A. Shibuya. Bias-Controlled
Pulse Charging System for Electrostatic Precipitator. Staub-
Reinhalt. Luft 36. No. 1, January 1976. p. 19-26.
8. Masuda, S. Novel Electrode Construction for Pulse Charging.
Symposium on the Transfer and Utilization of Particulate Control
Technology. Denver, Colorado. July 1978. Conference Records
Vol. 1, p. 2*11-251.
9. Penney, G.¥. and P.C. Gelfand. The Trielectrode Electrostatic
Precipitator for Collecting High Resistivity Dust. J. Air
Poll. Control Assoc. 28, No. 1. January 1978. p.53~55-
10. Kide, L. Electrostatic Precipitator Arrangements. US Patent
14,052,177. Oct. U, 1977.
11. Lausen, P., H. Henriksen, and H. Hoegh Petersen. Energy con-
serving Pulse Energization of Precipitators. To be presented
at IEEE/IAS 1979 Annual Meeting, Cleveland, Ohio. October 1-lt,
1979-
368
-------
PRECHARGER COLLECTION SYSTEM - DESIGN FROM THE LABORATORY
THROUGH FIELD DEMONSTRATION
Morey A. Nunn
Lodge-Cottrell-Dresser
Houston, Texas 77002
Duane H. Pontius
Southern Research Institute
Birmingham, Alabama 35205
J. H. Abbott and L. E. Sparks
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
Abstract
Southern Research Institute with the support of the EPA has
developed a two-stage precipitator which has shown encouraging
results in the laboratory. The two-stage ESP consists of a pre-
charger for the first stage and a conventional ESP operating at
low current densities. Lodge-Cottrell's assistance was enlisted
to design and supply a 1060 m3/min precharger collector system.
The design was to be based on the functional parameters arrived
at from the laboratory tests. In addition to the basic require-
ments, the system was designed utilizing features that could be
readily incorporated in systems treating gas volumes of commercial
size.
369
-------
PRECHARGER COLLECTION SYSTEM - DESIGN FROM THE LABORATORY
THROUGH FIELD DEMONSTRATION
INTRODUCTION
Southern Research Institute, with the support of the partic-
ulate Technology Branch, Industrial Environmental Research Lab-
oratory, Research Triangle Park, U.S. Environmental Protection
Agency, has been investigating a two-stage approach to a solution
of the problem of precipitating high resistivity particulate
materials. In a two-stage ESP, particle charging occurs princi-
pally in the first stage or precharger. Particle collection occurs
in the second stage, or collector, which would be designed to
operate at a relatively low current density and high electric
field strength. Thus it is in the precharger that back-corona poses
the greatest difficulty.
The precharger design for suppression of back-corona is based
on a simple wire-plate electrode geometry, with the addition of
perforated screen electrodes parallel to the plates, as illustrated
in Figure 1 (Patent Pending - U.S. Serial No. 882,673, March 3,
1978). A high voltage applied to the corona wire will bring about
corona conduction. In general, part of the current will go to the
plate and part to the screen, depending on the relative potentials
at each. If the plate is grounded the screen is set at a potential
having the same polarity as that of the corona wire; ions moving
toward the plate will tend to be repelled away from the screen.
If the magnitude of the potential at the screen is great enough,
the screen current may go to zero, while a considerable corona
current passes through the openings in the screen proceeding from
the corona wire to the plate.1
After successful confirmation of the theory in the laboratory,
the Southern Research Institute decided to solicit the assistance
of a precipitator manufacturer to design, fabricate, and erect a
pilot precharger and precipitator. Lodge-Cottrell was awarded the
contract for this work in October 1978.
DESIGN OBJECTIVES
The conventional design philosophy was modified in response
to two general considerations. First, the unusual nature of the
precharger requires special treatment due to the clearances asso-
ciated with two nested, high voltage electrode systems, and second,
provisions for adjustments and modifications are desirable because
of the developmental aspects of the program. An inflexible design
would allow no room for correcting nonoptimum parameters. In
addition to the basic requirements, the system was to be designed
utilizing features that could be readily incorporated in systems
treating gas volumes encountered in commercial operation. For
370
-------
example, the use of insulating spacers between the screen and
plate electrodes in the original prototype were found to be
impractical, and the techniques used for electrical isolation
of the three separate electrode systems were not readily adapt-
able to scaling up. A second generation device designed by the
cooperative efforts of Southern Research Institute and Lodge-
Cottrell incorporated a more rugged design philosophy along with
many of the electrical suspension features planned for the pilot
demonstration unit. This device performed at least as well as
the original prototype, thus confirming the overall design strategy.
DESIGN CONCEPTS
Design of the pilot pr-charger began by utilizing design
concepts evolved from Lodge-Cottrell's precipitators. Examples
of this include:
1) "Lead thru" insulators with the support insulators
located on the roof out of the flue gas.
2) Supports for the collectors, screens, and discharge
electrodes isolated from the casing with springs.
3) Methods for guiding and adjusting the alignment to
obtain the initial electrical clearances and maintain
them.
The first problem encountered when'reviewing the required
equipment was available space. The earlier 28 nr/min units had
not required the support system necessary for the larger unit.
The necessity of maintaining electrical clearances and the pro-
visions for erecting the unit grew in complexity.
This led to the choice of stacking the discharge and screen
electrode support frames, separated by insulators. A final arrange-
ment of the hardware inside the unit, which maintained the necessary
electrical clearances and could be erected efficiently, was developed
after extensive layout work (Reference Figure 2). The arrangement
drawings were reviewed with the personnel involved in constructing
the smaller unit and it was decided to provide additional adjustments,
The next design problem was in maintaining electrical clear-
ances between the screen and the collector plate. The collectors
are a nominal 2.16 m in width and 3.12 m in height. The height
to width ratio was of concern and a rib was added to the plate
section to increase its rigidity. The screen to plate clearance
was increased by 6.4 cm because of fabrication and erection toler-
ances.
The rappers used with the precharger are electromagnetic
vibrators. They were chosen because of the ease with which the
intensity can be varied. The main concern was effective rapping
371
-------
of the screen. It is difficult to transmit vibration through
the cross section of the screen and it was important to keep
the screen as clean as possible without reentrainment of the
dust. Spring diameters and heights were selected on the basis
of availability, so that the spring rate could be easily changed.
As with any type of electrostatic precipitator, good gas
distribution and elimination of gas sneakage is important. A
1/2-scale model of the precipitator and precharger was built and
gas surveys were performed (Figure 3). Baffles were added to
confine the flue gases within the treatment areas and gas dis-
tribution devices were installed in the nozzle to obtain even
distribution at the inlet of the precharger.
CURRENT STATUS - CONCLUSION
The precharger concept has been tested with regard to elec-
trical characterization, particle charging effectiveness, and its
effect on the collection efficiency of a two-stage system with
various electrode configurations in the downstream collector stage.
These tests have been carried out on laboratory pilot scale systems,
both at Southern Research Institute and at the U.S. Environmental
Protection Agency's Industrial Environmental Research Laboratory,
Research Triangle Park. The devices used in these tests were
designed to handle nominally 42 m3/min, and redispersed fly ash
was used under controlled temperature conditions where the resist-
ivity was maintained in the range of 10ai - 5xl012ohm cm. Some
of the results of the tests were reported in another paper presented
at this conference. For tests under actual flue gas conditions,
another 42-84 m3/min unit has been fabricated for use in the field
at a coal-burning power plant that produces a high resistivity
fly ash. The small pilot scale field evaluation is scheduled to
take place later this summer.
The pilot testing program carried out thus far has provided
valuable information concerning both the feasibility of the concept
and the nature of the problem areas to be considered in the design
of the larger pilot demonstration unit.
A site has been selected for installing and testing the large
pilot demonstration unit. The principal requirement of the test
site was that the fly ash resistivity be high enough to represent
a significant challenge for the device - on the order of 101l -
10:i2ohm cm. The design work is approximately 90% completed at
this time, and the fabrication phase will begin within the next
several months.
The ultimate goal of this program is to demonstrate that
the precharger concept under development can perform under real-
istic field conditions, and to determine to what extent benefits
may accrue from this new technology in the form of improved pre-
cipitator collection efficiency and/or reduced cost of pollution
372
-------
control equipment. If proven feasible, the precharger could be
useful, not only as part of new, integrated two-stage ESP system,
but also because of its small size along the direction of gas
flow, as a retrofit to existing precipitators where dust resist-
ivity is significant problem.
Reference
Pontius, D. H., and L. E. Sparks. A novel Device for Charging
High Resistivity Dust. APCA Journal, Vol. 28, No. 7, July, 1978.
373
-------
TOWARDS A MICROSCOPIC THEORY OF ELECTROSTATIC PRECIPITATION
BY
C.G. Noll, T. Yamamoto
United McGill Corporation
Groveport, Ohio 43125
ABSTRACT
The statistical mechanics of particles at the surface of a dust deposit
are presented as a step towards a microscopic theory of electrostatic
precipitation. The paper consists of three major divisions: (I) Field
induced entrainment, (II) Particle condensation at the surface, and (III)
Hydrodynamic limits on precipitation.
In Part I, the entrainment of dust from surfaces by electrostatic forces
was considered as a simple Markov process. A choice of x~1 for the
dependence of the adhesive potential (van der Waals particle attraction) led
to an exponential emission rate in the Kramers-Chandrasekhar approximation.
This result suggested a simple experimental procedure for determining the
adhesive properties of the dust. In addition it was found that both thermal
Brownian and particle-deposit collisions contribute to the entrainment.
In Part II, the precipitation problem was discussed in terms of a
two-dimensional Ising model. The Markovian assumption demanded that the
efficiency of the EP geometry be divided into a product of factors depending
on particle migration and condensation in the two-dimensional surface. From
this perspective a new collection criteria was established and the importance
of particle collisions was emphasized.
Finally, in Part III the role of bulk fluid flows in particle condensation
and migration was addressed. The superficial gas flow and secondary flow were
found to introduce anisotropy into the Ising lattice and also to enhance the
mean particle energy at the surface. The Navier Stokes Equation was
numerically integrated for the case of a wire-plate type corona. Plots of the
stream function and gas velocity profile were presented and these suggested
the existance of a lower limit on the shear stress that acts on surface
particles.
374
-------
TOWARDS A MICROSCOPIC THEORY OF ELECTROSTATIC PRECIPITATION
INTRODUCTION
A property common to all electrodeposition is the growth of a deposit.
Hence, whether we are concerned with solid or liquid deposition, a theory of
the growth process is extremely desireable.
The complexity of the usual problem, makes it necessary to formulate
models that attempt to describe the growth process at different levels.
Macroscopic models are in abundance and these can be used to predict the
number of particles or thickness of a deposit at time t. Several
representative models are listed in the References. (Cooperman (1977)1,
Gothard (1977)^, Robinson (1967)3) Should an adequate model of this type
be formulated, it then becomes necessary to construct models which deal with
the physical and chemical properties of the particulate that might be
responsible for generating the observed collection. Models of this type might
be called microscopic.
Most of the work on electrostatic deposition has been primarily concerned
with the development of deterministic models of the macroscopic type. The
emphasis . has been directed towards particle transport in the gas stream and
such parameters as "effective migration velocity" and diffusion coefficient
have been introduced. In this approach a functional equation (differential or
integral equation, etc.) for the number of particles in the layer at time t,
say n(t), is derived on the basis of some postulated mechanism. (The
postulated mechanism may or may not reflect the internal processes responsible
for changes in layer thickness). This deterministic equation, together with
boundary conditions, is then solved to obtain n(t). The operable results are
expressed as an efficiency. The major problem with such treatments is that
they are deterministic. One obtains the result that the dust layer thickness
at a given time t will always be the same if the initial conditions are not
changed. Since macroscopic models do not take into consideration the large
number of random or chance factors that can influence the growth of the
deposit, a multitude of empirical constants must be introduced.
It is the contention of the authors that an approach to the microscopic
model must be made through an understanding of the statistical mechanics and
the surface physics of the particles at the dust layer/gas stream interface.
The combined process of particulate deposition and adhesion is known to be
aided by an electric field (Penney and Klingler (1962)^). However,
adhesively bound particles are also known to be entrained from a surface by an
electric field (Zimon (1969)^). Thus, a ^knowledge -of the steady state
behavior of the deposit, conditioned by these competing mechanisms, is crucial
to the formulation of the microscopic model.
375
-------
Part I of this paper considers the entrainment problem as a simple Markov
process; that is, the two particle mechanics at the interface are
deterministic between impulse events and, consequently, the second order
(phase space) transition probability density satisfies the Smoluchowski -
Chapman - Kolmogorov equation. This first order approach leads to diffusion
equations for gross particle transport and a local exponential
(Deutsch-Anderson) type collection equation. However, such a model is
inadequate, since the real problem must include both deposition and
entrainment. The combined process requires the deposition to be described by
at least a third order Markov process. The present simplified case, though,
gives needed insight into the entrainment problem.
A model-surface/particle - binding potential is proposed, which contains
both adhesive and electrostatic terms. The subject particle localization is
then discussed in terms of the Kramers-Chandrasekhar solution of the
generalized Fokker-Planck equation. Such a model suggests an analogy between
electrostatic precipitation and chemical kinetics. The model is also
consistant with trends in coagulation theory (Ruckenstein (1978)6).
In Part II the precipitation problem is discussed in terms of a two -
dimensional Ising model. The Markovian assumption demands that the efficiency
of an EP geometry be divided into a product of factors depending on particle
migration and condensation in the two-dimensional surface. From this
perspective a new collection criterion is establishe'd' and the importance of
particle collisions is emphasized.
Finally, Part III summarizes some work done by one of the authors on the
electrohydrodynamics of electrostatic pr*ecipltators. This material is
included to link the surface transport .phenomena to bulk fluid flows. The
superficial gas flow and secondary flow; introduce anisotropy into the Ising
lattice and also enhance the mean part,icle energies at the surface.
The purpose of the electrohydrddynamic study was to experimentally and
theoretically investigate riot only the induced secondary flow but also the
interaction of the secondary flow with the primary flow. Numerical solutions
for the electric field strength with a space change density distribution were
obtained through an extension of Leutert and Bohlen's (1972)7 numerical
methods. The principle objectives of the theoretical study were (1) To
calculate the secondary flow velocity distribution for the single wire and the
two wire configurations by integrating the Navier Stokes equation for the case
of two-dimensional incompressible flow, and (2) To calculate the flow
interactions with the primary flow for the single wire and the two wire
configurations, for various magnitudes of flow.
The uniqueness of the present work lies in the division of various
particle surface interactions into low frequency effects, which determine the
binding potential and particle migration, and stochastic accelerations, which
statistically determine entrainment. From this perspective the fundamental
concern lies in the adhesion of dust to the surface.
376
-------
PART I. DUST ENTRAINMENT
Statement of Problem
When an external field is applied to a moderately conductive (a > 10~^
J2~^ m~1) dust layer, particles in the surface of the layer experience a
competition between adhesive and electrostatic forces. This competition can
lead to an entrainment of the dust.
The particle-dust layer interaction 4>(r*, t) consists of adhesive and
external electrostatic terms. A differential equation, subject to this
potential, can be defined to describe the motion of the particle.
|f « -fit + A(t) + Ktf.t) (1)
In (1) the quantities r* and If are independent random variables which
describe the position and velocity jjf the particle in phase space, K(r*,t) is
the gradient of the potential, and A(t) is the driving term. The relaxation
time 3 is included to account for viscous energy losses: 3 =
where r\ is the viscosity of the medium and (p,d) is the
(bulk density, diameter) of the dust.
For the problem to be tractable two additional assumptions must be made:
(1) It must be assumed that the particle exhibits Brownian motion, subject to
binding in the layer. Such an assumption is plausible since the gas stream
and surface of the dust layer are in thermal equilibrium. However, in the
case where particles collide with the surface and other disturbances exist,
the spectral density must be shown to be constant and frequency independent
within the response range of the particle layer system. (2} It must be
assumed that the potential (j»(r*rtj, which binds the particle, remains
continuous during the time interval in which the Brownian motion process is to
be described. This assumption demands that the impacting particles do not
adhere at the site of interest. In the ensemble sense, this restriction is
not crucial, because the adherence of impacting particles leads to new or
replacement sites. However, a removal of this restriction would require the
deposition/ entrainment process to be described as a third order Markov
process.
Under the present assumptions, the solution of (1) can be written as a
second order conditional probability density W(r*,i!,t \ ?o,it0) to) , a
quantity which obeys the generalized Liouville (Fokker-Planck) Equation.
• i^gradjW = /3diva(WiI) -
where D is determined by the spectral density of the source,
377
-------
An approximate, quasi-steady state solution to (2) was obtained by Kramers
for the case of atomic scale particle emission past a barrier of height
4>(xc) (Kramers (1940)8, Chandrasekhar (1943)9). In terms of an
emission rate, his efforts gave
where - , «= , and
c
Although Kramers' interest was directed towards problems of chemical kinetics,
a generalization of the thermal spectrum to include other driving components
and a generalization of the analysis to macroscopic particles are trivial.
The important point is this: The emission of initially bound particles
depends solely on properties of the binding potential and its modification by
an external field. Since the adhesive potential is short range, the potential
need only be modeled within the diffusion boundary layer thickness <5. This
is the interaction force boundary layer approximation employed in coagulation
theory. &
Application to van der Waals Binding
The adhesive component of the potential (j)(r*, t) can be modeled (crudely,
but adequately (Zimon (1969)^)) by a macroscopic van der Waals attractive
term and by a macroscopic Born type repulsive term. In the present case it is
assumed that the minimum in the adhesive well is located at XA and that XA
and WA are not* appreciably affected by an external field. Thus, for
the following potential can be written:
m
where the Born repulsive term has been ignored and A is an effective Hamacker
constant (A-10-19 j). The particle's charge in the presence of an
external static field E is determined by the Maxwell relation (Cho (1964)11).
where c is the permitivity of the surrounding medium. For the potential
defined by (4) and (5):
378
-------
and
(6o)
Equations (5) and (6) indicate that the barrier height (xc) decreases
linearly with increasing field and that we ~ E3.
Thermally Induced Entrainment. If wc»6/2 and only thermal
Brownian forces drive the particles, then (3) reduces to
tH./SATZ (?)
Thus, at constant temperature the particle emission rate increases
exponentially with increases in field and at constant field the particle
emission rate increases with increasing temperature. Field criteria for the
application of (7) are estabished from the assumption on WQ and the fact
that the potential well is not to be saturated by the external field,
These criteria are met for most dusts encountered in electrostatic deposition
problems.
Particle Induced Entrainment (Sticking and Scattering Events Ignored).
Collision incidents at the surface of the deposit obey Poisson statistics,
that is, if an ensemble of intervals of duration T is averaged and if the
average number of collisions per second at a site is X, then the
probability of k events at the site in the internal T is given by the Poisson
distribution:
(9)
379
-------
Not all collisions need to be direct "hits" since the close contacts of the
bound particles transfer energy to the site of interest. The spectral density
for a collision-type source is written
where f^(v) is the Fourier transform of the pulse train. The approximate
result in (10) was determined for Poisson distributed rectangular pulses with
duration T and kinetic nnor-^y £. The bound particles respond only to
spectral components near WA, and thus, the primary concern is directed
towards processes for which T<2Tttjj1. Such processes require direct
impact of particle and surface; this is shown as follows.
In analogy with (4) the potential's action on the incident particle must be
(ID
where a(x) is the fraction of the Maxwell charge carried by the incident
particle (-12moj1 and one must retreat all the way
back to ( 1 ) .
Incidently, it is not sufficient to assume that dust resistivity is the
determining factor in entrainment. The important parameter is the incident
particle's kinetic energy and the distance the particle lies from the
potential barrier when charge reversal occurs. Again, the adhesive potential
will determine the -deposition.
380
-------
PART II. PARTICLE CONDENSATION
The Assumption of a 2-D Canonical Ensemble
At the basis of the generalized Liouville (Fokker-Planck) Equation is the
assumption that the surface of the layer can be described by a canonical
ensemble. The Boltzmann factor in the particle emission rate (3) is just a
consequence of this assumption. If the assumption of a canonical ensemble is
valid, a whole host of mathematical and physical techniques can be applied to
the precipitation process. It is therefore an imperative that we consider
the particle-site lattice in greater detail.
Suppose that the sites have equivalent statistical geometry (identical
spherical particles) and that they form a surface of thickness equal to the
interaction range XQ. The mean kinetic energy of particles in this ensemble
is determined by (1) the thermal energy of the bulk (kT), (2) the kinetic
energy of the incident particles, and (3) the energy supplied by structural
vibration and rapjdng. The steady-state (or mean) energy of the surface is
then defined as E. Of course, the layer may grow or deplete but the surface
will remain quasi-static in energy. Note the limitation imposed by the simple
Markovian approximation.
The energy of each site may change with time but should take on a distinct
set of values. The discrete set of levels may be made arbitrarily large, and
thus, their number inconsequential so long as E is maintained . If the
allowed states are numbered !,'••, r, then the number of distinct ways of
configuring a particles such that a-j of them are in state r=1, 32 are in
state r=2, etc, is given by
no,!
or
r r
InT * In a! - I In a.
r r
(13)
Since the surface contains many paricles and many states, Stirling's
approximation may be applied and
InF £j a In a - £arlnar
r
The extremum of this quantity, subject to the constraints on particle number
and mean energy, may be determined by the method of Lagrange multipliers.
£(lnor+ a' + /3'Er)Sa = 0 (15)
r ' ' r
381
-------
where Er is the energy of the rth site and (a', 3) are the Lagrange
multipliers. With a proper choice of (ot',3), all the a(j~r-1 • It is thus
proposed that the surface of the layer behaves as canonical ensemble, raised
to a temperature
9 « T + Xf/£k (18)
Note that the relaxation rate 3 is included to account for energy transfer
from the surface - only thermal and collision related energy processes are
included in (18).
A dilute particle fluid model, subjected to other energy sources besides
thermal, is quite familiar in physics. In particular, much progress has been
made in understanding the NMR spin-lattice relaxation process in solids by
employing a spin-temperature hypothesis (Goldman (1970)13). Since the
Larmor frequencies for the electron and nuclear systems differ by many orders
of magnitude, rf radiation can be used to separately alter the nuclear and
electron "temperatures". The particle case, though, has reduced
dimensionality and expected condensation. (Note that the hyperfine
interaction (|>~r~3).
The central concern for precipitation lies in an analogy between site
occupancy and the long range order of two-state magnetic systems. Deposition
is then defined in terms of the long range order or correlation between
sites. The interaction parameter which determines occupancy is the well depth
m<}>(xc) and the range of the interaction is XQ.
The 2-D Ising Model Approximation
In 1952 Lee and Yang^ sought to describe gas condensation in terms of
the Ising model, a model developed to explain magnetic phase transitions. The
present phase transition has two characteristics which make the
two-dimensional Ising model an attractive choice: (1) The lattice, unto which
condensation occurs, is static and (2) The condensation problem is purely two
382
-------
dimensional. The only "easy" two-dimensional problem is the Onsager (square,
equivalent site) lattice with constant aniaotropy (McCoy and Wu (1973)1^).
In this case the long range order or occupancy correlation is written:
\
(19)
where mtt and m^ are the trapping energies along and transverse to the gas
flow.
The collection efficiency in terms of (13) is simply
> <20>
'"
Consequently, the overall efficiency of an EP, which collects monodisperse
particles, must be the product of ns and an efficiency dependent on
particle migration
(21)
The quanity rim ~ (1-exp(-t/T)) where t is the residence time for the
dust in the EP and t=H/bE. In the expression for T, H is the spacing
between collection electrodes and b is the particle mobility. We have ignored
both neutralizing collisions in the gas stream (density correction) and
secondary flow processes.
Equation (19) is significant, because it introduces a critical temperature
determined by the following relation
• I (22)
C
Above this effective temperature, dust collection cannot occur.
PART III. FLUID TRANSPORT IN THE EHD FIELD
The efficiency product (15) was written to suggest a statistical
independence of gas stream transport and surface condensation. This
assumption is consistant with the simple Markovian approximation, and thus, it
is sufficient to demonstrate the strength of the interaction and then to
include the coupling phenomenologically as a change in the adhesive potential
for surface particles.
Recently, T. Yamamoto succeeded in numerically intregrating the Navier
Stokes Equation in the presence of a uniform gas flow normal to a wire-duct
corona geometry. The calculations were extensive, so only an outline and
example results will be presented here. The details will be published
elsewhere.
Under normal EP flow conditions, the external body force due to the space
charge and field strength is not affected by the secondary flow, since the ion
velocity is approximately two orders of magnitude greater than the secondary
flow velocity. Consequently, the flow problem is simplified by a separation
383
-------
of the problem into a solution of Poisson's equation followed by a solution of
the Wavier Stokes equations, modified by body force terms due to the electric
field.
The Poisson and continuity equations were solved simultaneously, subject
to experimentally determined boundary conditions on the electrode potentials
and current density. A finite difference procedure was chosen along with the
successive-over-relaxation method (SOR) to assure faster convergence. The
equations were solved self-consistantly. When the calculated and measured
current densities at the collection surface became matched, the space charge
density became determined.
Next, the two-dimensional Wavier Stokes equation, suitably modified by the
body force terms due to the electric field, was transformed into a vorticity
equation and then solved along with the Poisson equation for the stream
function. Although the calculation scheme permitted a specification of the
incident flow conditions, a uniform initial flow was assumed. This permitted
a reflection symmetry about the plane of the wire. Also, a condition of
non-slip was demanded at the wall.
The vorticity equation was solved using finite difference approximations
with special attention given to the non-linear convection terms. Numerically
induced oscillations were experienced, when using the central difference
approximation for cases of high current density and low kinematic viscosity.
In the explicit finite difference approach to the vorticity equation,
stability was obtained by using the up-wind difference method in the
convection terms.
The computational cost of the solution depended strongly on the efficiency
of solving the problem in the finite difference form. Because of this, the
Fast Direct Solution (FDS) method was employed for solving the Poisson
equation for the stream function. A detailed algorithm for the FDS is
described by Nakamura (1977)16.
The steady state velocity field and stream function were calculated for
the one and two wire in duct geometries. In the two wire case, the
dependences of these fields on gas flow is exhibited in Figures. 1-3, along
with the appropriate initial conditions. At zero gas flow the results are in
qualitative agreement with the circulation reported by Yabe, Mori, and
Hijikata (1978)1?, except that the center for circulation (in their
non-symmetrical geometry) was located above the corona wire. Note that the
relative importance of corona induced secondary flows decreases with gas
speed. It was noted, however, that the circulation increased with current
density. Under the illustrated conditions, the maximum current density was
near 3.5 mA/m^.
The calculations have shown that for gas speeds greater than 1 = 5 m/s,
secondary flow effects need not be considered as a contributor to
entrainment. However, at lower gas speeds, the secondary flow introduces a
limit on the minimum shear stress experienced by the surface particles. Such
a limit only exists in single-stage type electrostatic precipitators.
384
-------
( Flow Inlet )
plate
( Flow Outlet )
• •>-—- N ^ —^>i ^ — *•• * • • «-«• — —•»•«••• •
. . xy"* 7 y"^sA/i~***" ^ • « . v — •«.*. « . • •
» • i i • *f'I **O\ / —** v > * . • #^.— ^.^^%».
» . ^ 'Ik xi /^**» •»>''*.*<*.»..-
corona wire corona wire
U0« 0.5 ft/sec, V0= 15.0 kV, Sggj,- 10.57
Velocity Vectors of Flow Interaction with the Primary Flow for the
Iwo Wire Geometry
( Flow Inlet )
r-1 » 1 ^—
-orona wire
corona wire symmetry line i. Flow Out lei
+••*••»•
U0" 0.5 ft/sec, V0= 15.0 kV
Streamline of Flow Interaction wlrh r.he Primary Flow for the
TWo Wire Geometry
Figure 1.
385
-------
Outlet )
corona wire corona wire symmetry line
U0- 2.0 ft/sec, V0- 15.0 kV, NEHD- 2.65
( Flow Inlet )
1 - 1 - 1 - ( - 1
Velocity Vectors of Flow Interaction with the Primary Flow for the
Two Wire Geometry
corona wire corona wire symmetry line ( Flow Outlet )
1 - 1 - 1 I I • - 1- — I - 1 - 1 - 1 I • -J - 1 - 1 - ! - t- — ! - 1 - 1 - 1 - 1 - 1 - 1
plate
U0- 2.0 ft/sec, V0- 15.0 kV
Streamline of Flow Interaction with the Primary Flow for the
Two Wire Geometry
Figure 2.
386
-------
( Flow Inlet )
plate '
,(Flow pullet )
corona wire corona wire symmetry line
U0- 4.0 ft/sec, V0- 15.0 kV, NJJJJ,- 1.32
( Flow Inlet )
I 1 1 1 1 1 !-
Velocity Vectors of Flow Interaction with the Primary Flow for the
Two Wire Geometry
corona wire corona wire symmetry line ( Flow Outlet )
-) 1 ( I • ) 1 1 ! 1 1 • ! ! 1 ! 1 1 1 1 ! 1 ! 1 1 1-
plate
4.0 ft/sec, V0- 15.0 kV
Streamline of Flow Interaction with the Primary Flow for the
Two Wire Geometry
Figure 3.
387
-------
CONCLUSIONS
Dust collection in an electrostatic precipitator is not assured by
particle charging and subsequent migration. The particle must adhere or
"condense" onto the surface lattice.
In the present paper we have recognized that the assumption of dust
collection as a simple Markov process permits the isolation of surface
"condensation" from particle transport. This observation allowed a
factorization of the efficiency equation into a part dealing with particle
transport and a part representing surface phenomena.
The mathematical structure for describing the surface effects is greatly
simplified by separating the two-body particle interaction into stochastic
forces(molecular impacts, particle impacts, bulk lattice vibrations,
structural vibrations, rapping impulses, etc.) and quasi-steady forces
(adhesive, electrostatic, EHD shear stress, gravitational, etc.).
With these assumptions the surface of the layer can be described as a two
dimensional lattice fluid with collection occuring below an effective
temperature-determined by the ratio of the stochastic energy to the
quasi-steady binding energy. Thus, any effect such as back corona or
.secondary flow induced particle transport will increase the effective
temperature of the surface and inhibit condensation.
Condensation is closely related to entrainment. The electrostatic
entrainment of dust depends on the spectral response, range, and depth of the
adhesive potential. Both thermal sources and particle deposit collisions
contribute spectral components in the range appropriate to entrainment. Under
the assumptions of the present theory, the Kramers-Chandrasekhar model can be
adapted to describe entrainment. Then, if a "macroscopic" v.d. Waals adhesive
binding is assumed, the emission rate increases exponentially with the applied
field. In this case laboratory studies should provide microscopic information
on the adhesive vector (XA,U)A,A). Also, the analysis revealed the
importance of incident particle kinetic energy in determining the collection
of low resistiviy dusts.
The model was developed under the physical limitations of equal-sized,
isotropic spheres, near a 2-D square lattice. Of course, this is a gross
approximation. However, the simple model has revealed most of the qualitative
effects usually left to parameters or supplementary conditions. Refinements
of the model to include coupling effects and some experimental investigations
to confirm it are in progress.
388
-------
REFERENCES
1. Cooperman, P. Efficiency Theory and Practice in Electrostatic
Precipitation. Proc 4th Int Clean Air Cong (Tokyo) 46:835-8, May 1977.
2. Gothard, N. An Approach to Calculate the Collection Efficiencies of
Electrostatic Precipitators. Staub Reinhalt Luft 37:14-6, January 1977.
3. Robinson, M. A Modified Deutsch Efficiency Equation for Electrostatic
Precipitation. Atm Environ 1:193-204, May 1967.
4. Penney, G.W. and E.H. Klingler. Contact Potentials and the Adhesion of
Dust. Trans AIEE 81:200-4, July 1962.
5. Zimon, A.D. Adhesion of Dust and Powder. New York, Plenum Press, 1969.
pp 347-66.
6. Ruckenstein, E. Reversible Rate of Adsorption or Coagulation of Brownian
Particles - Effect of Shape of the Interaction Potential. J Coll Int Sci
66:531-43, October 1978.
7. Leutert, G. and B. Bohlen. The Spatial Trend of Electric Field Strength
and Space Charge Density in Plate-Type Electrostatic Precipitators. Staub
Reinhalt Luft 32:27-33, July 1972.
8. Kramers, H.A. Brownian Motion in a Fields of Force and the Diffusion
Model of Chemical Reactions. Physica 7:284-304, April 1940.
9. Chandrasekhar,' S. Stochastic Problems in Physics and Astronomy. Rev Mod
Phys 15:1-89, January 1943.
10. Zimon, A.D. op. cit. pp.22-34.
11. Cho, A.Y.H. Contact Charging in Intense Electric Fields. J Appl Phys
35:2561-4, September 1964.
12. Berger, S.K. Reduction of. Breakdown Voltage in Uniform and Coaxial Fields
in Atmospheric Air through Moving Conducting Spheres. In: IEE Conf Pub
118:380-4, 1974.
13« Goldman, M. Spin Temperature and Nuclear Magnetic Resonance in Solids.
Oxford, The Clarendon Press, 1970.
14. Lee, T.D. and C.N. Yang. Statistical Theory of Equations of State and
Phase Transitions. II. Lattice Gas and Ising Model. Phys Rev 87:410-9,
August 1952. (See Also, H.N.V. Temperley, The Mayer Theory of
Condensation Tested Against a Simple Model of the Imperfect Gas. Proc
Phys Soc (London) LXVII:233-8, March 1953-)
f
15. McCoy, B.M. and T.T. Wu. The Two-Dimensional Ising Model. Cambridge, The
Harvard University Press, 1973. pp 216-48.
389
-------
16. Nakamura, S. Computational Methods in Engineering and Science with
Applications to Fluid Dynamics and Nuclear Systems. New York,
Wiley-Interscience, 1977. pp.193-8, 418-23.
17. Yabe, A., Y. Mori, and K. Hijikata. Heat Transfer Augmentation around a
Downward-Facing Flat Plate by Non-Uniform Electric Fields. In: Heat
Transfer Annual Meeting (Toronto), pp.171-6, June 1978.
390
-------
ION CURRENT DENSITIES PRODUCED BY ENERGETIC
ELECTRONS IN ELECTROSTATIC PRECIPITATOR GEOMETRIES*
By:
W. C. Finney, L. C. Thanh, and R. H. Davis
Department of Physics, The Florida State University
Tallahassee, Florida 32306
ABSTRACT
A new laboratory test system for electron beam ionization in electro-
Static precipitator geometries has been constructed to measure ion current
densities as a function of voltage differences for clean (bare) plate condi-
tions. The new system incorporates improved electrodes which withstand a
driving voltage of ±55 kV, a factor of 5 increase over the previous test
system. A 3 MeV Van de Graaff accelerator produced ionizing electron beams
of 1.2 and 2 MeV and currents of 10.5 and 21 UA in place of corona wire ioni-
zation. Current densities of up to 130 mA/m were measured before breakdown
between the plates, and no current saturation was observed. A comparison of
I-V curves and sparkover voltages for various beam energies, currents, and
collimation are discussed and the need for measurements with good beam
geometry is reviewed.
391
-------
ION CURRENT DENSITIES PRODUCED BY ENERGETIC
ELECTRONS IN ELECTROSTATIC PRECIPITATOR GEOMETRIES*
INTRODUCTION
Experiments using electron beam ionization instead of corona wire ioni-
zation in electrostatic precipitators were introduced in an earlier paper
(Davis and Finney, 1978)l. In these earlier experiments, ion current densi-
ties were measured which were 65 times larger than those which could be ob-
tained in a duct-type electrostatic precipitator with corona ionization
(White, 1963)2. Since particulate charging and collection in electrostatic
precipitators depend on the ion current density within the interelectrode
space, electron beam ionization is being appraised as a replacement for
corona wire ionization.
In this paper, a new laboratory scale electrostatic precipitator test
system incorporating improved electrodes and higher operating voltages was
used to extend the earlier measurements which were limited by the available
power supplies. Current versus voltage curves for several electron beam
energies, beam currents, and beam profiles were measured, and increased ion
current densities were noted in all cases. No saturation of ion current
density was observed and the reasons for this will be discussed.
The earlier paper (Davis and Finney, 1978) presented data using a poor
geometry beam configuration in which the electron beam was not scanned or
trimmed by collimators to form a narrow corridor of ionization but instead
formed a plasma in and beyond the ionization zone. In a second set of
measurements, a collimator was used to confine the beam and restrict its
spreading. The present work incorporates uncollimated and collimated beam
profiles and these are discussed in conjunction with good geometry experi-
ments which are planned to further investigate electron beam precipitation.
EXPERIMENTAL APPARATUS
The objective of this work was to study the effects of beam energy,
current, and profile on the ion current characteristics of the laboratory
precipitator. A schematic diagram of the test system is shown in Figure 1.
The laboratory scale electrostatic precipitator constructed for these
experiments was composed of an upper and a lower set of 3 parallel plates
separated by a distance of 10 cm. The plates are h inch thick aluminum, and
the plate margins were rounded and polished to a smooth finish to reduce edge
effects on the field distribution of the system. The collecting area was
400 cm2 for plates b, b1, d, and d' and 200 cm2 for plates c and c1. The
outside plates served as guard electrodes while the current flow through the
inner two plates (c and c1) was measured. Two DC power supplies capable of
delivering 130 kV (optional polarity) and 28 mA were connected to the elec-
trodes so that the upper set was positive (anode) and the lower set was
negative (cathode), both with respect to ground.
The electron beam was delivered between the electrodes by the 3 MeV Van
de Graaff electron accelerator at the Florida State University (Davis, 1976)3,
392
-------
Accelerator
Tube ._ .
Foi
Window
s:
s'
t
t
b
t
Electron
Beam
1 <
-W4-
2L
•«
P.S.
Plates
Figure 1. Schematic diagram of the experimental apparatus. The electron
beam is delivered between the upper and lower sets of plates.
S = 5 cm, L = 10 cm.
which was operated to produce beam energies of 1 MeV and 2 MeV and initial
beam currents of 10.5 ]UA and 21 yA. The distance from the foil window to
the center of plates c, c' was 55 cm. To determine the effect of electron
beam collimation on the ion current characteristics of the system, a ±0.1
radian honeycomb collimator was mounted on the end of the accelerator tube
downstream of the foil window.
RESULTS
Current-to-voltage characteristics for two beam energies, two initial
beam currents, and with and without the collimator are shown in Figure 2.
Only the anode current (c) is plotted against voltage because the anode and
cathode currents were equal in magnitude. For all six experimental conditions
plotted, the I-V curves approximate straight lines, indicating an ohmic re-
lationship between current and voltage. No tapering off of current at the
highest voltage readings occurs, therefore no saturation in ion current den-
sity is observed for the experimental conditions reported here.
393
-------
3.00
2.50
2.00
1-50
Id
cr
cr
13
O
1.00
0.50
0
150
1.2 MeV, 10.5^ A,Uncoil.
l.2MeV, 10.5/zA.Coll.
Z.OMeV, 10.5/1 A,Uncoil.
2.0MeV, 10.5/xA.Coll.
1.2MeV, 2l/iA,Uncoil.
2.0MeV,2l/iA,Uncoil.
0
20 30
Voltage (kV)
40
50
0
60
Figure 2.
The effects of beam energy, beam current, and collimation on the
ion current characteristics. Plots are for the current read from
the anode (plate c). Uncoil. - Uncollimated beam; Coll. -
Collimated beam.
394
-------
Table 1. Sparkover Voltages and Maximum Ion Current Density for Various
Beam Conditions
Sparkover Voltage
(kV)
Maximum Ion Current Density
(mA/irr )
Energy
Uncoil.
10.5 yA
a
Coll.
21.0 yA
Uncoil.c
Uncoil.
10.5 yA
a
Coll.
21.0 ]JA
Uncoil/
No Beam 55
1.2 MeV 48
2.0 MeV 46
o
52 40 80 35 89
52 ca. 40 100 60 130
Unco11imated
Collimated
Large ion current densities were obtained in each case as shown in
Table 1. For 1.2 MeV beam energy, 10.5 yA beam current, and uncollimated
beam, an ion current density of 80 mA/m was obtained, six times the ion cur-
rent density of 13 mA/m2 found in earlier experiments under similar beam
conditions (Davis and Finney, 1978)1. The highest ion current density found
was 130 mA/m2 for 2.0 MeV energy, 21.0 yA current, and uncollimated beam.
Several general results are illustrated in Figure 2 for the values of
beam energy, beam current, and collimation conditions studied here. As beam
energy increased from 1.2 to 2.0 MeV, ion current density increased by ~25%
and 75% for uncollimated and collimated beam, respectively. A doubling in
beam current from 10.5 to 21 yA caused an ion current density increase of 33%
and 50% for 1.2 and 2.0 MeV energies, respectively. Finally, the uncollimated
beam produced ion current densities of 150% and 100% greater than the colli-
mated beam for 1.2 and 2.0 MeV beam energies, respectively.
As seen in Figure 2, the current versus voltage curves for each experi-
ment were extended until sparkover occurred. Sparking usually occurred be-
tween the lateral edges of plates c and c'. Further control of edge effects
may be important in preventing premature sparkover. A comparison of spark-
over values for various experimental conditions is summarized in Table 1.
For specific conditions, beam energy had little effect on sparkover voltage
(a 2 kV or less difference between 1.2 and 2.0 MeV). Doubling the beam cur-
rent from 10.5 to 21.0 yA causes sparkover voltage to decrease by 6-8 kV. A
slight (4 kV) increase in sparkover voltage is noted with collimated versus
uncollimated beam.
395
-------
DISCUSSION
Ion Current Characteristics
The relationship between the ion current and the applied voltage is
approximately ohmic for voltages up to sparkover value. This suggests that
most of the space between plates c and c' is occupied by a virtually grounded
plasma generated by the ionizing collision of the electron beam and air
molecules. The driving field is limited by the breakdown field strength,
which is, in turn, determined by the presence of a copious supply of ions and
the effective plasraa-to-plate distance. Therefore, saturated extraction of
ions is not expected, and, as shown in Figure 2, no current saturation occurs
in any of the I-V plates. Higher driving voltages not limited by sparkover
or lower ionization production rates are required to reach the saturation
point.
The linear relationship between current and voltage in the electron beam
test system is in contrast to the quadratic (lav2) (White, 1963)2 relation-
ship prevailing in corona wire ionization systems. This difference may repre-
sent an advantage of the electron beam precipitator over its corona wire
counterpart.
Ion current densities which are much higher than previously obtained
have been produced in these experiments. Using electron beam ionization with
the electrode system shown in Figure 1, the ion current density is 100 mA/m2
at 2.0 MeV energy, 10.5 yA current, uncollimated beam, and ±44 kV applied
voltage. For a similar plate geometry, conventional electrostatic precipita-
tors have a maximum ion current density of about 0.2 mA/m2 (White, 1963)2
which is space charge limited. Therefore, the laboratory electron beam pre-
cipitator yields an ion current density that is ~500 times greater than a
conventional corona precipitator, and to date it is limited only by sparkover,
which is in turn determined by beam geometry, plate spacing, and edge effects.
As shown previously, ion current density increased as beam energy and
beam current increased which is expected because with greater beam energy or
current more ionization takes place and a larger extraction of charge occurs.
The uncollimated beam produced larger ion current densities because the
collimator reduces the actual beam current delivered to the interelectrode
space while providing a better beam geometry with less beam spreading. Even
so, the ionization yield was not exhausted prior to sparkover.
In these poor geometry experiments, the spreading of the plasma column
generated by the electron beam reduces the effective distance between plates
c, c', and the grounded plasma. It has been estimated (Davis and Finney,
1978)1 that for a beam energy of 1 MeV and a beam current of 10.5 yA that the
penetration of the electric field adjacent to each electrode is 2.5 cm at a
current density of 13 mA/m2 and an electrode voltage +10 kV (anode) and -10
kV (cathode). The sparkover between parallel plates at this distance (5.0
cm) is ~137 kV (Meek and Craggs, 1978) **. A higher driving voltage would be
required to reach current saturation. The sparkover voltage is further re-
duced by field distortion caused by the edges of the electrodes. Collimation
396
-------
Limit of
.Electric
Field
V +
Plasma
lonization
Zone
Electron
Beam Electrode
V*
^* ••••••••t*»* *•»•»•/ — — .»— — • «
+ + + 4- + + +
+V
V"
a)
V
b)
V
c)
Figure 3.
Diagram of three different electron beam geometries. A poor
geometry beam configuration with a wide ionization zone is shown
in part a). The collimated geometry used in the present experi-
ments with a more restricted plasma is shown in part b). Good
geometry with a very narrow ionization zone is shown in part c).
of the beam confines the plasma column to a smaller cross-sectional area,
thus increasing the plasma-to-electrode distance and the effective inter-
electrode gap. Consequently, this results in a slight increase in sparkover
voltage as shown in Table 1.
Poor vs. Good Geometry Beam Configurations
To simulate the small radius ionization cylinder of a corona wire, an
electron beam should be prevented from spreading and instead retained within
a narrow region centered between the electrodes. As pointed out earlier, the
spreading of the plasma column is probably the major cause of early sparkover
because the effective gap distance from electrode to plasma is reduced. This
poor geometry beam configuration is diagrammed in Figure 3a, and shows the
limited reach of the electric field and the expanded ionization zone and re-
duced effective gap.
A honeycomb collimator which passes electrons within ±1/10 radian of the
beam direction was used to trim the edges of the beam and the result was less
397
-------
scattering, a wider electrode-to-plasma gap, and an increased sparkover vol-
tage (Figure 3b). The beam current is reduced because the surface area of
the collimator tubes presents an obstacle and eliminates the fraction of the
beam which is not projected straight through the honeycomb.
Although the honeycomb collimator does improve beam geometry, further
refinement is projected by reducing the surface area of the collimator in
the beam path while further reducing scatter. A good geometry beam configura-
tion, shown in Figure 3c, would restrict the ionization zone to a column with
small cross-sectional area, allowing more complete charge extraction and
raising sparkover voltage to a level at which ion current density saturation
may be found. Collimation of the beam just prior to its entry into the
electric field may serve to further define the ionization zone.
CONCLUSIONS
Earlier experiments using electron beam ionization in electrostatic pre-
cipitators were extended using a new laboratory scale test system and larger
power supplies. An approximately ohmic relationship between ion current and
plate voltage suggests the formation of a large grounded plasma region be-
tween the electrodes. Large current densities, as high as 500 times that of
conventional precipitators, have been observed with no saturation. Variations
in ion current densities with beam energy, current, and collimation are due to
the intensity of beam ionization in the interelectrode space. Sparkover vol-
tage is probably determined by beam geometry which controls the plasma-to-
electrode gap distance and the width of the ionization zone. Measurements
with good geometry are required to observe saturation and investigate the
interplay between ionization zone restriction and sparkover. Future work
(Thanh et al., in preparation) will include the use of a medium which
simulates fly ash and will address the phenomenon of back corona and its ef-
fect on sparkover.
*Supported in part by Department of Energy Contract No. Et-78-S-01-3199.
REFERENCES
1Davis, R. H. and W. C. Finney. Ionization by Electron Beams for Use in
Electrostatic Precipitators. Energy Research. 2:19-27, 1978.
2White, H. J. Industrial Electrostatic Precipitation. Reading, Mass.,
Addison-Wesley, 1963.
Davis, R. H. Ion Source and Accelerator Applications to Electrostatic
Precipitators. In: Conference Record 76CH1122-1-1A, IEEE Industrial Appli-
cations Society. Chicago, 1976. p. 328.
^Meek, J. M. and J. D. Craggs. Electrical Breakdown in Gases. John Wiley
and Sons, 1978. p. 608.
5Thanh, L. C., W. C. Finney, and R. H. Davis. The Effects of Back Corona
Discharge on Ion Current Density in Electron Beam Precipitators. submitted
for publication.
398
-------
EXPERIMENTAL STUDIES
IN THE ELECTROSTATIC PRECIPITATION
OF HIGH-RESISTIVITY PARTICULATE
BY:
John C. Modla
Robert H. Leiby
Thomas W. Lugar
Kris E. Wolpert
Buell Emission Control Division
Envirotech Corporation
Lebanon, PA 17042
ABSTRACT
A pulsed charging system is described that eliminates reverse ionization
in electrostatic precipitation. The motivating concept is a purely electrical
approach. The fundamental characteristic of this method is the relaxation
time of the precipitated particulate. Laboratory studies indicate that this
powering concept could increase precipitator performance on high-resistivity
particulate by more than two orders of magnitude. Pilot results are also
presented.
399
-------
EXPERIMENTAL STUDIES
IN THE ELECTROSTATIC PRECIPITATION
OF HIGH-RESISTIVITY PARTICULATE
INTRODUCTION
Electrostatic precipitation is the removal of suspended particulate
from a gas stream by the application of an electric field. This method of
pollution control appears simple on a microscopic level, but in reality the
phenomenon of electrostatic precipitation on a microscopic level is a very
complex process. The fundamental processes of electrostatic precipitation
consist of corona generation, particulate charging, and particulate collec-
tion. Corona generation is the formation of a self-sustained electrical
discharge between a high-voltage electrode (usually at a negative potential)
and a grounded or passive electrode. Negative ions are formed by the attach-
ment of electrons to electronegative gases and are accelerated by means of
an electric field. The ions become attached to the suspended particulate
within the gas stream by a combination of direct collisional and diffusional
processes. The unipolar charged particulate is attracted to the passive
electrode for subsequent removal. The role of the corona generation is
twofold. The first is to charge the particulate. Secondly, together with
the static electric field, the unipolar space charge density also aids in
the process of particulate collection.
PKECIPITATOR LIMITATIONS
The surface and bulk electrical resistivity of the particulate is a
crucial variable that limits the performance of an electrostatic precipitator.
It is a well-established fact that for materials with a resistivity in the
range of lO^O-lO^ ohm cm the phenomenon of reverse or back ionization occurs,
thus reducing the effectiveness of the precipitator. Reverse ionization,
as associated with electrostatic precipitation, is categorized as an abnormal
electrical discharge emanating from the dust-electrode interface. The light
signal from the electrical discharge consists of two light waves: a primary
light wave that rises very rapidly in time and a secondary light wave that
rises more slowly. The former proceeds into space while the latter proceeds
along the precipitated dust layer surface. When a sufficiently high voltage
pulse is applied at low pressure, reverse ionization is triggered by free
electrons. As the pressure is increased, the triggering mechanism is due
to negative ions.
400
-------
CRITERIA FOR REVERSE IONIZATION
Reverse ionization is a phenomenon whereby the gas molecules among the
porous precipitated dust layer are electrically overstressed to such an extent
that the gas molecules become ionized. Positive ions are formed which, in
some cases, produce a local surface disruption within the dust layer. The
positive ions are accelerated toward the emitting electrode, thus neutralizing
the negative space charge density. A simple analysis of the occurrence of
reverse ionization is given by the following equation:
E = Jp1
where
E: electric field strength; volt/m
J: current density, amp/m2
p1: resistivity, ohm m
As can be seen from the above equation, a large electric field strength
can be obtained for modest current densities through the precipitated dust
layer, provided the resistivity of the particulate is large.
The occurrence of reverse ionization generally causes a reduction in the
breakdown voltage of the gas thus leading to a spark, a substantial increase
in the precipitator current, and unsteadiness in the current-voltage charac-
teristic curve (hysteresis). The immediate effects of reverse ionization as
far as pollution control is concerned are a reduction in the charging and
precipitating electric fields, a substantial increase in particulate
reentrainment from the collecting plates, and an increase in emitting wire
vibrations.
METHOD OF CONTROL
A common method of minimizing the process of reverse ionization is to
condition either the gas-entrained particulate, the flue gas, or both. The
conditioning process is achieved by the addition of chemical additives to the
main gas stream. There are two schools2 of thought which attribute the re-
duction in reverse ionization to the increase in surface and bulk conductivity
of the particulate or to a modification in the electrical characteristics of
the flue gas itself (i.e., the formation of ion clusters, which leads to
lower ion mobilities).
Another method^ of minimizing reverse ionization is to revert to the
concept of "wide space" precipitators. The concept of wide spacing refers
to the separation between emitting electrodes and collecting electrodes.
The likelihood of the occurrence of a disruptive spark due to reverse
ionization is reduced with larger spacings.
Very recent studies4 on the minimization of reverse ionization have
concentrated on the control of the current density through the precipitated
401
-------
dust layer. This method of control can be accomplished by electrical means.
The idea behind this control strategy is based on the electrical properties
of the flue gas and particulate. The criteria5 for the initiation of reverse
ionization are based on steady state conditions and the continuity of the
electric field and current density across a boundary. The crucial parameters
that determine the occurrence of reverse ionization are the relaxation time
of the flue gas compared to the relaxation time of the precipitated particu-
late dust layer. The relaxation time is defined as the product of the
permittivity of the media in question and the resistivity Of the same media.
The initiation of ionization will occur when the relaxation time of the dust
layer exceeds the relaxation time of the flue gas. These criteria imply
that reverse ionization will occur because the conductivity of the flue gas
is larger than the conductivity of the dust layer. This observation suggests
a way to minimize reverse ionization by correlating the electric field due to
the ionic space charge density to the relaxation time of the precipitated
dust layer.
RELAXATION TIME APPROACH
The Tri-Electrode Precipitator6 is a device which minimizes the occurrence
of reverse ionization. The concept is based on the relaxation time of the
precipitated dust layer. A derivation of this relaxation time approach is
given below:
Consider the continuity equation:
y-j = - _E; p: space charge density
3t
But J = ^r-
P
Assume that the resistivity of the dust layer is uniform
(no spatial variation)
V(E/p') = V(l/p')-E+p/ep' = sp/gt
p(t) = Poe't/1; T = p'e
where PQ is the space charge density at time t = 0,
and e is the permittivity of the particulate.
The above equation describes how the space charge density decays with time.
This equation is analogous to the decay of a typical capacitor-^resistor cir-
cuit. To minimize the occurrence of reverse ionization, the space charge
density must be allowed to relax to a level below the breakdown strength of
the flue gas.
402
-------
ARCHITECTURE OF THE TRI-ELECTRODE PRECIPITATOR
The basic architecture of the Tri-Electrode Precipitator, as compared
to a conventional precipitator, is the additon of a third non-ionizing
electrode. The purpose of the non-ionizing electrode is to further enhance
the performance of the precipitator by producing a strong long-range static
electric field. The device employs two separate pulsating power supplies
for the ionizing and non-ionizing electrodes. The two sets of electrodes
are operated with variable DC reference voltages with superimposed high
voltage pulses. The pulse forming network is a resonant charging scheme
which is discussed by Glaso and Lebacqz.? By connecting many thyristors
in series with the proper voltage grading circuits, the desired voltage
rating can be achieved. This leads to a highly reliable solid-state
switch8, as compared to using relays or vacuum tubes. The power supply has
the capability of varying the pulse amplitude, pulse duration, pulse frequency,
and rise and fall times of the pulses.
MODE OF OPERATION
The mode of operation is to pulse the corona emitting electrode at a
rate which correlates with the relaxation time of the precipitated dust layer.
The duration of the pulse (pulse width) and the amplitude of the pulsed
voltage (more specifically, the peak electric field) determines the quantity
and rate of negative ion production (i.e., the ionic space charge density).
This, in turn, determines the degree of charge acquired by the particulate.
The pulsed voltage that creates the corona current can have a pulse width
of one to four milliseconds. Since the pulse duration is short, the applied
voltage to the emitting electrodes can be larger than the charging voltage
of a conventional precipitator. These short corona pulses have been found
to be effective in charging the particulate.
The frequency of the pulsed corona voltage depends on the resistivity of
the particulate. As an example, for particulate with a resistivity of Id11
ohm cm and a relative dielectric constant of 2, the relaxation time is 180
milliseconds. This means that the space charge density requires 180 milli*-
seconds to decay to 37% or e'1 of its initial value. This implies that if
the electric field associated with the initial space charge density is near
the threshold value for reverse ionization, the corona emitting electrode
should be pulsed at a rate not exceeding 30 Hertz. Too long a period between
pulses results in a low electric field through the dust layer (poor collection),
while too short a period between pulses permits so little decay in the ionic
space charge density that the dust layer cannot accept the full pulse current
(reverse ionization).
With the emitting electrode at a low potential (i.e., below the corona
emitting potential), the unipolar charged particulate migrates towards the
collecting electrode. This process of particulate collection is categorized
as space charge precipitation. This technique of pollution control also has
a characteristic relaxation time which is quite long. Thus, to aid in the
particle collection, a non^corona emitting electrode is pulsed with a high
voltage whenever the emitting electrode is at a voltage below corona generation.
403
-------
In essence, the Tri-Electrode Precipitator is basically a two-stage
precipitator itself. The pulsed corona emitting electrode charges the
particulate, while the non-corona emitting electrode aids in the collection
process. The pulsing scheme controls the flow of current through the
precipitated dust layer, so the occurrence of reverse ionization is
minimized.
LABORATORY TEST
Laboratory tests were conducted under the supervision of Professor
GaylordPenney at Carnegie Mellon University for the Buell Division of
Envirotech Corporation. A sketch of the test apparatus is shown in Figure 1.
The passive electrodes are three-foot-square aluminum sheets mounted four
inches apart. The spacing from wire to ground is two inches, with three
inches between wires. The third electrodes (non-emitting) are tubes that
are mounted midway between the wires. The system is operated in a closed-
loop fashion. A calibrated orifice is used to meter the air flow. Com-
pressed air is supplied to a f luidized-bed dust feeder which, in turn,
supplies the test dust. Leaving the feeder, the entrained particulate
passes through a small-diameter, high-efficiency cyclone to remove agglom-
erates. The temperature of the air is adjusted to give the desired resistivity
as measured by a resistivity cup located in the outlet chamber of the
precipitator. The system is seasoned before actual efficiency measurements
are taken.
A schematic of the pulsing scheme for resistivity in the range of
ohm cm is shown in Figure 2. The pulse width of the corona emitting electrode
is two milliseconds, with a millisecond on the rise and fall time.
A comparison of the performance of the Tri-Electrode Precipitator and
a typical Research Cottrell precipitator over the resistivity range of lO^
to lO-^ ohm cm is shown in Figure 3. To normalize comparison, the efficiencies
for the two-electrode precipitator were taken with the non-emitting electrodes
present and removed. The measurements indicate that the pulsed system of
operation is far superior.
Scale-up studies with the Tri-Electrode Precipitator, based on scale
factors obtained from the Deutsch-Anderson equation, are shown in Figures 4
and 5. The scale factor is two on the wire-to-plate spacing with additional
plate area. Also shown in the same figures are results obtained for a con-
ventional precipitator by the removal of the third electrode and the pulsed
power supply. Operating under identical conditions, the Tri-Electrode
Precipitator is roughly two orders of magnitude more efficient on the high
resistivity materials.
PILOT STUDIES
A pilot unit has been constructed in a 45-foot-long, 13. 5- foot-high, drop
frame - tandem axle trailer. The primary elements of the pilot unit are
listed in Table I. The geometrical arrangement of the three-electrode system
is shown in Figure 6.
404
-------
An initial series of tests has been completed at a utility located on
the east coast. A chemical analysis of the fly ash is given in Table II.
Laboratory measurements of the fly ash resistivity using planar electrodes
varies from 2 x lO^ohm cm to 5 x 10i:Lohm cm within a temperature range of
232°F to 268°F. Figure 7 shows the experimental results of the efficiency
as a function of the gas velocity for both the Tri-Electrode and conven-
tional precipitator design. Note that the conventional design is powered
with a DC voltage. The migration velocity, (0, can be obtained from the
efficiency measurements via the Deutsch-Anderson equation. Figure 8
displays the relationship between 0) and gas velocity- Initial conjecture
as to the reason for this phenomenon may be related to the role of ionic
wind, i.e. electrical relaxation time, in precipitation.
Further tests are planned to quantify the effect of the rise time of
the pulsed voltage. The effect of a positive pulsed scheme is also under
investigation.
REFERENCES
1. Masuda, S., "Research on Electrostatic Precipitation - 1975",
Dept. of Electrical Engineering, University of Toyko.
2. Loeb, L.B., "Fundamental Processes of Electrical Discharge in
Gases", 1939, John Wiley & Sons, Inc.
3. Private communication with personnel from Sumitomo Metal Mining
Co., Ltd.
4. Penney, G.W., Gelfand, P.C., "The Tri-Electrode Precipitator
for Collecting High Resistivity Dust", APCA Journal, Vol. 28,
No. 1, Jan. 1978.
5. Cooperman, P., "Back Corona and Relaxation Time", IEEE Transactions
on Industry and Applications, Vol. 1A-12, No. 1, Jan. 1976.
6. U.S. Patent 3,915,672.
7. Glaso, G.N., Lebacqz, J.V., "Pulse Generators", McGraw Hill,
Vol. 5, MIT Radiation Laboratory Series, 1948.
8. Shoup, J.F., Mason, C.A., "A High Voltage Thyristor Valve for
Precipitator Applications", future paper in Electronics Journal.
405
-------
TABLE I
TRI-ELECTRODE PILOT PRECIPITATOR
DESIGN PARAMETERS
Number of Fields
Depth of Field
Plate Height
Total Plate Area
Gas Passages
Gas Volume
SCA
Plate Spacing
Emitting Electrode
Total Emitter Length
Tube Electrode
Dust Removal
Hopper
1
9 ft.
6 ft.
600 ft.
6
7100 ACFM
84.5
9"
Rigid FrameRods
430 ft.
Standard Pipe
Vibrators
Continuous Discharge System,
Drag and Screw Conveyer
406
-------
TABLE II
CHEMICAL ANALYSIS OF FLY ASH
Parameter
Si 02
Fe2 03
A12 03
CaO
MgO
K20
Na20
P2°5
TiO2
Li20
Percent
55.91
7.80
25.71
3.80
1.60
3.13
0.43
0.31
1.19
0.04
407
-------
1. Corona Wires
2. Tubular Third Electrode
3, Grounded Dust Collecting Electrode
tj. Compressed Air Inlet
5, Fluidized Bed Oust Feeder
6, Cyclone Oust Collector
7. Fan
8. Motor
9. Dust Delivery Pipe
10, Distribution
\\, Inlet Sampling Point
12. Orifice
13, Outlet Sampling Point
1^, Enclosure
15, Heater And Fan
16, Heater And Fan
17, Resistivity Test Cup
Fig. 1. Tri-Electrode Precipitator test model
408
-------
TIME
0
50 msec
-10
O
UD
r
U4
O
«=C
~1
-20
H h
2 msec
4 msec
Fig. 2. Typical static and dynamic potential
for resistivity of 5 x lO^-1 ohm-cm
-------
80
*>
.. 60
:*
u
2
U
H
U
H
EM
£ & 40
20
00 ' ' •
oo » • • • *
... • • • . •
-on
o o o
O O x%
0 °
"" o o —
Q 0
o o
o o o
>— • Tri-Electrode O- ~
O
O O
o Conventional
0 o o
- o -
o
( 1
10
10
10
11
RESISTIVITY, OHM CM
Fig. 3A.
Efficiency comparison of laboratory
Tri-Electrode Precipitator with
conventional precipitator
10
13
-------
o
1—4
LU-
LL.
LU
100
80
60 -
40
20
O
• Tri-Electrode
Conventional
1
50
70
90 TOO
AIR SPEED, FT/MIN
130
Fig, 3B. Precipitator performance as
a function of air speed
150
-------
80
• •
60
>-
CJ
•ZL
LU
i—i
O
i—i
*t
-&
0
o
0
0
0
o
0 0
0
Tri-Electrode
20
O Conventional
1012
1013
RESISTIVITY, OHM CM
Fig. 4. Laboratory full scale Tri-Electrode-conventional precipitator tests
-------
80
60
O
Z
UJ
LU
40
20
o
o
TRI-ELECTRODE
0 CONVENTIONAL
o
o
o
Q
o o
o
o
1012
RESISTIVITY, OHM CM
1013
Fig. 5. Efficiency comparison with tubes present
-------
O \ o 0 1 o 0 \ O
O O 0
IN-—i-——OO 1 O D
Fig. 6. Pilot Tri^Electrode
Precipitator gas passage
414
-------
80
60
O
H
u
Tri^-Electrode
en
° Conventional
20
GAS VELOCITY, FT/SEC
Fig. 7. Pilot study of Tri-Electrode Precipitator
-------
8 6
H
I
o
H
2
•
• o
-° o
o
o
TRI -ELECTRODE
0 CONVENTIONAL
2
GAS VELOCITY, FT/SEC
Fig. 8. Calculated migration velocity
from pilot tests
-------
COLLECTION OF FLY ASH WITH HIGH ELECTRICAL RESISTIVITY
ELECTROSTATIC PRECIPITATOR PRECEDED BY
THE EPA-SoRI PRECHARGER
L. E. Sparks, G. H. Ramsey and B. E. Daniel
U. S, Environmental Protection Agency
Industrial Environmental Research Laboratory
Particulate Technology Branch
Research Triangle Park, N. C. 27711
417
-------
INTRODUCTION
It is well known that dust with high electrical resistivity is
difficult to collect in electrostatic precipitators (ESP). The difficulty
is primarily due to poor particle charging. This poor particle charging
is not because particles having high electrical resistivity are intrinsically
difficult to charge, but because back corona (which results from deposition
of material on the collection electrode) produces a bipolar ion field.
When ions of both positive and negative polarity are present in the
charging region, the competing effects of the two plus low values of
electric field produce low electrical charge on particles.
EPA has contracted with Southern Research Institute (SoRI) to
•develop a novel ESP consisting of a precharger and a downstream collector.
The precharger has been built. The downstream collector is still being
designed. The experiments described in this paper were designed to
demonstrate the effectiveness of the precharger and to provide information
needed to design the downstream collector. The data presented here are
the first published data demonstrating the effect of the EPA/SoRI precharger
on particle collection in an ESP.
CONCLUSIONS
Although the ESP was not the optimum downstream collector, particle
collection by the precharger/ESP combination was excellent. With a
specific collector area (SCA) of 29.2 m2/m3/s (128 ft2/103 acfm) the
particulate emissions from the ESP without precharger were 2.5 times the
particulate emissions from the precharger/ESP combination wtien collecting
12
a dust with electrical resistivity greater than 10 ohm-cm. The particulate
emissions from the ESP alone were 15 times the particulate emissions
from the combination when collecting a dust with electrical resistivity
of about 10 ohm-cm. The graded penetration (penetration versus particle
diameter) curves show results consistent with those of the mass emission
tests.
It appears that the precharger more than doubles the effective SCA
of the ESP. Initial capital cost estimates indicate that the precharger
-------
will cost about a third to half the cost of one conventional electrical
section. A conventional electrical section might increase the SCA of a
small ESP, such as used for high sulfur coal, by as much as 33 percent
and the SCA of a large ESP, such as used for low sulfur coal, by no more
than 17 percent. Thus, the cost effectiveness of the precharger appears
to be excellent.
The data were used to estimate the SCA needed to meet an emission
standard of 13 ng/J (0.03 lb/106 Btu) when collecting fly ash with
12
electrical resistivity of about 10 ohm-cm. The estimated SCA is 70
23 23
m /m /s (355 ft /10 acfm) which compares with an estimated SCA of at
least 180 m2/m3/s (930 ft2/103 acfm) for a conventional ESP.
3 3
Pilot scale {% 0.5 m /s) and prototype scale U 14 m /s) field
tests plus more pilot plant tests are planned to confirm these initial
results.
THE EPA/SoRI PRECHARGER
The EPA/SoRI precharger has been fully described by Pontius and
Sparks (1978) and Pontius et al. (197.8). The basic concept is presented
in the following paragraphs quoted from Pontius et al. (1978).
"In research sponsored by the Particulate Technology Branch of
the Industrial Environmental Research Laboratory, Environmental Protection
Agency, Research Triangle Park, North Carolina, Southern Research Institute
has devised and investigated the performance of a three-electrode system
for controlling the effects of back corona in a particle precharger. In
! this device two of the three electrodes are the conventional corona
discharge and passive electrodes. The third is a screen electrode placed
near the passive electrode. Separate power supplies are provided for
the corona discharge and screen electrodes. The passive electrode is set
at ground potential.
"The general principle of operation is based on the use of the
screen electrode as a sink for ions generated at the passive electrode
as a result of back corona effects. Consider, for example, a two electrode
419
-------
system where the corona discharge electrode is at a high negative potential
with respect to the grounded passive electrode. Now, locate an equipotential
surface near the passive electrode and insert a conducting screen coincident
with that equipotential surface. If the screen voltage is set equal to
the original potential on the surface the electric field will be practically
undisturbed in comparison with the original field. Only the non-zero
thickness of the wires in the screen will cause very localized modifications
to the field. A corona current originating at the discharge electrode
will be distributed such that a fraction of the total equal to the ratio
of open area to total surface of the screen will reach the passive
electrode. The remainder of the current will be intercepted by the
screen.
"Now, if the potential on the screen electrode is made more negative
the field near the screen will become distorted in such a way that
negative ions from the discharge electrode will be repelled from the
screen wires and forced toward the open area, through which they can
proceed to the plate. If we introduce high resistivity particulate
material into the system, deposition will occur on both the plate and
the screen electrodes. Since negative ions from the discharge electrode
are being repelled by the screen, it must have a lower current density
than the plate, and hence corona from the screen electrode would probably
not occur. If back corona occurs, the positive ions from the passive
electrode would be attracted to the screen electrode, where many would
be captured and removed from the system. If most of the positive ions
resulting from back corona can be captured by the screen electrode, the
ion field between the screen and the discharge electrode would be
essentially unipolar."
Pontius and Sparks (1978) and Pontius et al. {1978} presented data
showing the effectiveness of the precharger in charging particles—but
did not present data on particle collection in a device preceded by the
precharger.
EXPERIMENTAL APPARATUS
The experiments were conducted in a one-lane pilot plant ESP located
in EPA's Industrial Environmental Research Laboratory, Research Triangle Park, N.C.
-------
The ESP was operated with plate-to-plate spacing of 22.9 cm (9 in.).
The discharge electrodes were selected to maximize the electric field
for a given current density. Several different discharge electrode
geometries were tried including: screen electrodes, 0.318 cm diameter
wire, 2.54 cm apart;, 0.635 diameter wires, 5.08 cm apart; and 0.635 cm
diameter wires, 2.54 cm apart. Once the discharge electrode wires
became dirty they all gave essentially the same voltage current (V-I)
curves. A set of typical dirty wire/dirty plate V-I curves is shown in
Figure 1.
The precharger was installed in the inlet duct to the ESP. The
precharger was a two-lane precharger with plate-to-plate spacing of
about 13 cm (5 in.). Each lane had a single discharge electrode.
Sampling was conducted using sampling ports located after the
precharger and before the ESP, and at the ESP outlet. Mass samples were
collected using an out-of-stack filter. Impactor sampling was conducted
using Meteorology Research, Inc. cascade impactors. The cascade impactors
were calibrated by personnel of EPA/Industrial Environmental Research
Laboratory's Particulate Technology Branch (PATB), using the procedurr
described by Calvert et al. (1976).
The dust used in these experiments was a fly ash which was injected
into the ESP through two sand blast guns. The particle size distribution
of the resuspended fly ash was approximately log-normal with a geometric
mass mean physical diameter of about 8.5ym and a geometric standard
12
deviation of about 3. The resistivity of this fly ash is about 10
ohm-cm at 150°C when no steam is injected and 10 ohm-cm at 20°C.
Steam injection at 150°C reduced the resistivity to about 5 x 10 ohm-
All data were reduced using TI-59 programmable calculator programs
reported by Sparks (1978).
RESULTS
The average graded penetration curve for the high resistivity fly
ash (MO ohm-cm) for both precharger-off and precharger-on is shown
421
-------
in Figure 2. These graded penetration curves were used to calculate
effective migration velocities, wfi(di), for various particle diameters.
we(d.) = In Pt(d.)/(A/V) (1)
where Pt(d^ is the penetration of particles with diameter d^ and
A/V is the SCA.
The effective migration velocities are plotted in Figure 3.
The graded penetration curves indicate that rapping reentrainment
may be a problem. The penetration of large particles is higher than
expected and is probably due to reentrainment. An optimized downstream
collector should minimize the rapping reentrainment problem.
The graded penetration curve for the low resistivity case is shown
in Figure 4.
DESIGN TO MEET NSPS
The effective migration velocity curves can be used to estimate the
SCA required give a given penetration or a given emission such as required
to meet New Source Performance Standards (NSPS). The overall penetration
can be estimated from
exp [f. (A/V) we (d.)] (2)
1=
where f.. is the fraction of the total mass at the inlet with diameter d.,
w (d.j) is the effective migration velocity of particles with
diameter d..,
and A/V is the SCA.
Equation (2) was solved for a typical fly ash size distribution,
various values of SCA, and w (d^) taken from Figure 3. The particle size
distribution was assumed to be log-normal with d = 20 ym and a = 4.5.
g q
The results of these calculations are shown in Table 1. Based on these
calculations it appears that an SCA of 70 m2/m3/s (355 ft2/103 acfm)
may be adequate to meet the proposed NSPS of 13 ng/J (0.03 lb/106 Btu).
This is less than half the SCA required for a conventional ESP.
422
-------
REFERENCES
Calvert, S., C. Lake, and R. Parker (1976), "Cascade Impactor
Calibration Guidelines," EPA-600/2-76-118 (NTIS PB 252-656/AS),
April 1976.
Pontius, D. H. and L. E. Sparks (1978), "A Novel Device for Charging
High Resistivity Dust,", J. Air Poll. Control Assoc. 28. 698.
Pontius, D. H., P. V. Bush, and L. E. Sparks (1978), "A New
Precharger for Two Stage Electrostatic Precipitation of High Resistivity
Dust." In Symposium on the Transfer and Utilization of Particulate Control
Technology: Volume I. Electrostatic Precipitators, EPA-600/7-79-044a,
February 1979.
Sparks, L. E. (1978), "Cascade Impactor Data Reduction with SR-52 and
TI-59 Programmable Calculators," EPA-600/7-78-226 (NTIS PB 290 710),
November 1978.
423
-------
Table 1. ESTIMATED OVERALL PENETRATION FOR VARIOUS SCA IN NOVEL ESP
COLLECTING FLY ASH WITH HIGH RESISTIVITY
SCA
m2/m3/s (ft2/103 acfm)
CURRENT, roA
0 -ft M b*
25 (127)
40 (203)
50 (254)
60 (304)
70 (355)
75 (381)
80 (406)
I I I
1
I I IV
) 10 20 30
Pt0 B
0.087 91.30
0.031 96.9
0.016 98.4
0.0083 99.17
0.0044 99.56
0.0032 99.68
0.0023 99.77
i 1
40 50 E
VOLTABE.kV
Figure 1. Typical dirty-wire/dirty-plate V-I curves for high resistivity
case.
424
-------
cc
t-
uu
UJ
O.
1.0 rr
0.5
0.4
0.3
0.2
T~T~"I I I I I Ml II I I I I I hH
xPRECHARGEROFF
/
PRECHARGERON
V
SCA*25m2/m3/s
p«2X 10*2 ohm-cm
0.1
0.1
$2 0.3 0.40.5 1.0 2.0 3.0 4.05.0
PARTICLE DIAMETER, ;zm
10
Figure 2. Graded penetration curves for high resistivity runs.
•5 10
o
2 5
UJ
3
e
S 2
o
Ul ,
u. 1
ii i i N i MI it i mi
P.RECHAR6ER ON
\^ PRECHARGEROFF
ohm-cm
1 I I II MJU 11 I II I i II
0.1 0.2 0.3 0.40.5 1.0 2.0 3.0 4.05.0 10
PARTICLE DIAMETER, jum
Figure 3. Effective migration velocity versus particle diameter for
high resistivity runs.
425
-------
F= T\T
: \
50
- 10
s
^ 5
LU
1.0
D.5
0.10
0.1
i i M i
i i
*
PRECHARGEROFF
0.2 0.3 0.5
PRECHARGERON
1
I
t
i
i
I
\
\\\\\ \\ i i urn
Figure 4.
2.0 3.0 5.0 20 30 50
PARTICLE DIAMETER, Mm
Graded penetration curves for low resistivity runs.
100
426
-------
PILOT PLANT/FULL SCALE EP SYSTEM DESIGN
AND PERFORMANCE ON A BOF APPLICATION
By:
Douglas Ruth, P.E.
United McGill Corporation
Columbus, Ohio 43216
And:
David Shilton
CF & I Steel Corporation
Pueblo, Colorado 81002
ABSTRACT
An ESP pilot plant study was done on emissions from a BOF
process which is cyclic with very high and low gas volumes,
temperature and grain loadings. Data collected were EP performance
vs. gas velocity, and collection area. Also measured was particle
size distribution, dust resistivity, and grain loading vs. opacity
at the EP outlet. From this data, a full size EP system was
designed and installed. Recent performance tests indicate the data
agrees well with the initial pilot plant study. Emissions have been
reduced from 11 grs/scf to less than .033 grs/scf and opacity to 20%
or less.
427
-------
INTRODUCTION
The EOF shop where these tests were done, CF & I, Pueblo
Colorado, is operated with two (2) 90 metric ton vessels, with one
vessel active at any given time; the other vessel concurrently
undergoing a reline. A precipitator system installed would
accordingly be designed to handle the emission from one active
vessel. Specifically, the volume to the precipitator would be
300,000 acfm at 350°F, with moistures between 10 to 30%. This
shop utilizes an open hood concept with a spark box cooling chamber
immediately after the vessel. This system removes the bulk of the
heavy solids and is directed with a separate clarifier - lagoon
water treatment system.
The emission from a EOF furnace is cyclic, with high grain
loading for a period of twenty minutes, then a quiescent period
lasting 20 minutes. After the first 20 minute period, oxygen lance
reblows may be necessary to satisfy metalurgical requirements on the
steel being melted. The highest grain loadings are usually
associated with reblows. Scrap iron is poured into the converter,
which is filled with molten iron, then oxygen lanced. This process
takes about 20 minutes, and it is during this time that the outlet
flue volume, gas temperature and particulate emissions increase
markedly. During this blow period, emissions go from essentially
zero to around 11 gr/scf, the temperature goes from 175°F to near
400°F, and the gas volume increases to it's maximum value of about
300,000 acfm. The particulate emission is primarily Fe2 0%, 90%
of which is below 10 micron in size, with 10% of these below 1
micron. The dust resistivity at flue gas conditions is about 1010
ohm-cm. It was the desire of CF & I to reduce these emissions to
.033 grs/scf and not to exceed a visual stack opacity of 20%. For
these requirements. United McGill designed and installed an
electrostatic precipitator system.
PILOT PLANT STUDY
One of the best possible ways to specify the design and size of
a full size electrostatic precipitator system is to do a pilot plant
study on the site where it is to be used. This is so the actual
stack emissions to be collected are specifically characterized and
actual operating data are obtained with a unit exactly like the full
scale system, only reduced in size. This was done at CF & I during
April 1976 by United McGill Corporation, using one of their Mobile
EPs and testing laboratories. This Mobile is a completely
self-contained EP system, which has three (3) fields, each of which
has its own TR set and voltage control, rapper system, hopper, and
rotary valve. Also included are the fan, duct work, outlet stack,
evaporative cooler, control panel and inlet/outlet testing ports
including opacity transmissometers. Each field uses the McGill
needle/pi ate design, with needles on the leading and trailing edges
of the discharge plates, which are held at a positive potential of
about 25 KV. The Mobile EP cross-sectional area is 25 ft2, and
428
-------
the fan is capable of pulling 8000 cfm at temperatures up to
800°F. The entire EP system is mounted on a trailer for ease of
transportation. It can usually be set up and be operating in eight
(8) hours. The accompanying laboratory trailer is fully equipped
for performing the standard EPA stack tests, gas analysis by infared
spectrophotometry, and some chemical analysis.
At CF & I, the Mobile EP was connected to a breech in the
existing downcomer from the process by a 85 ft long by 20 inch
diameter insulated duct. Figure 1 shows the Mobile EP at this
installation. Figure 2 is another view of the installation.
Emission collection performance of the pilot EP was done only
during the blow or reblow part of the EOF cycle, since this is the
period of maximum emission. A crew from CF & I tested at a point
before the EP inlet using the W.P.-50 apparatus, while a
simultaneous EP outlet test was done by a United McGill crew, using
EPA Method 5. An observer was located in the EOF shop who gave
signals to the test crews at the beginning of a cycle. Within one
minute, the equipment would be readied and a portion of a test
completed during the 20 minutes blow cycle. The EP inlet test could
only last 8 minutes due to the high grain loading. The outlet tests
were done for 2 blow/reblow cycles, but with testing halted during
the quiet cycle. Figure 3 shows the EP outlet during a blow cycle
with all fields de-energized. Figure 4 shows the outlet with one
field on, Figure 5 shows 2 fields on and Figure 6 with all three
fields energized. The white plume at this point is merely water
condensation. Figure 7 and 8 show that indeed the EP collects the
dust.
The primary data to be obtained were that of optimum EP
flow-through velocity, specific collection area, and rapping
timing. Operation temperature is more or less fixed by the process,
which was 360°F during the periods of heavy emission, so no
attempt was made to change this, except to insulate the duct. Also
of interest, was the chemical make up of the dust, dust resistivity,
particle size, and opacity measurements.
The EP velocity was varied by changing the volume flow with a
damper control on the fan section and was varied between 1.5 to 3.0
ft/sec. The specific collection area was also varied in this
manner, and also by de-energizing one or two fields from 400 to 800
ft2 per 1000 acfm. From this data, graphs were generated showing
the relation between outlet grain loading EP flow-through velocity,
and specific collection area. From this, one could then determine
the full size EP design.
Figure 9 shows the result of a dust resistivity measurement vs.
temperature and moisture. Figure 10 shows a typical particle size
distribution, a correlation between grain loading and opacity was
done, with results shown in Figure 12, for a 48 inch diameter stack.
429
-------
EP DESIGN
From a graph relating EP gas velocity and outlet grain loading,
obtained from the pilot testing, the cross-sectional area of the
full size unit was chosen. Also taken into account was the particle
size, especially in regards to re-entrainment. This influenced the
choice in the number of fields required, as well as the velocity.
As for plate area or total collection area, the following procedure
was used. A curve of the form
outlet = e-(a x SCA + b)
was fit to the data by means of regression analysis, ie., least
square fit.
This technique finds values for the fitting parameters a, b.
Once these are determined, one can calculate the required SCA needed
to achieve the desired outlet grain loading, providing the operating
conditions for the EP, such as inlet loading, temperature and
moisture, are in the range that they were during the pilot testing.
It was assumed that this would be so since the process would not be
expected to change. The equation is only a fitting technique and no
physical meaning is given to it, it just happens to fit this type of
data. It might be argued that it is a modified Deutsch - Anderson
equation, however, one could find more argument against this than
for it.
As mentioned, the basic EP size was determined by the EP
velocity/gas volume relation, which in this case, dictated a 1800
ft2 chamber. The total collection surface area required
determined by the above technique, was found to be about 238,000
ft2 to treat 300,00 acfm of gas. This would yield a predicted
outlet grain loading of .033 grs/scf. For these requirements, it
was decided that a McGill Model 5-600x4 EP would be the best
choice. This consists of 4 chambers, each 600 ft2 in
cross-sectional area, each with five fields, and each with about
79,000 ft2 of collection area. Four, rather than the necessary
three units was chosen so that there would be 3 active units and one
standby. Using three units for 300,000 cfm of gas at 360°F and 11
grs/scf inlet loading gives an average EP flow through velocity of
2.7 fps, an SCA of 790 for an expected outlet of .033 grs/scf and
20% or less visual opacity. The EP's are paired, and each pair
share the same TR set on the first field. Fields 2, 3 and 4, 5 on
each EP have their own TR set. These were designed to operate at 25
KV at 500 ma each.
Since each of the four precipitators is the same, flow modeling
was done on only one unit. A 1/16 scale model was constructed of
plexiglas. The velocity pressure was set the same in the model as
would be found in the full size system (Euler Modeling) and a
430
-------
velocity profile measured across the face of the model. The
standard technique of fitting turning vanes and perforated plates to
provide an even flow distribution was used. The final configuration
installed was diagonal grid of channel pieces with a 6" opening,
followed by a perforated plate and then a set of splitter vanes.
Another design consideration was dust handling. Each row of EP
hoppers are connected by a screw conveyor. These in turn empty into
a pnuematic conveyor to a dust silo. Dust from the silo is
pelletized and carried away by rail car. This hopper dust has a
bulk density of around 60 lbs/ft3. One third of the dust from the
process is quite heavy, near 140 Ibs/ft3. For this, a single
dropout chamber was built before the EP system inlet. This dust is
also conveyed to the silo for pelletizing and subsequent re-use.
One final concern was that of possible explosion due to the
high CO levels possible. In light of this, a CO monitoring system
was designed to de-energize the fields. This reduces the
possibility of a spark induced explosion in the EP.
FABRICATION AND INSTALLATION
The unique McGill design allows for a modular construction;
that is, each EP is made up of one or more of five standard
modules. These individual modules are manufactured complete before
shipping, including plates, side skins, and rapper boxes where
required. The top structures, hoppers and transitions are also shop
produced and shipped ready for installation. In this particular
case, the 600 model is made of 4-150 ft2 modules, stacked 2x2,
each field. Therefore, each of the four EP's is a 5 x 4 array of
modules for a total of 80 modules to make up the system. Field
erection then consists only of pouring the foundation, setting up
the support structure and hoppers and then lifting the modules in
place. Top structures and transitions are then installed.
Finally the wiring, piping, insulation and siding is
installed. Using this method, the quality of the final product is
higher, because most of the EP is shop produced and assembled, under
controlled supervision. This means that the quality of the
installation is not totally in the hands of a contractor who may not
have the necessary experience in areas where critical assembly is
needed. This is especially true in the case of plate alignment.
FULL SCALE SYSTEM PERFORMANCE
Colorado regulations require that the full system emission be
less than 40 Ibs/hr, and less than 20% opacity. The three unit
system operation easily meets the process weight requirement.
Figure 12 shows a comparison between the pilot plant data and the
full system performance data. As can be seen, the agreement between
the two sets of data is good.
431
-------
With three of the four units in operation, there are occasional
excursions above 20% opacity. Most typically there readings over
20% opacity are brief and associated with commencement of the blow
or reblows. These puffs are indigenous to the EOF process and well
controlled with all four units operating.
OPERATION AND MAINTENANCE
All four EP's were installed and operating within two years
after the pilot plant study. The biggest operating problem with the
system has been associated with the high density dust transporting
system; evidenced by plugging in the line or defective gate valves.
Periodic cleaning of all the transporters is needed to maintain
continuity of dust removal.
pelleti
However
It was also found necessary to attend to the "automatic"
;tizer full time, due to the required water feed adjustments.
However, one eight hour shift can pelletize dust from a 24 hour
collection period. It is also necessary to blow off the ledges
between modules with compressed air once every shift, due to the
high buildup in this area.
The biggest problem area with EP operation seems to be failure
of the insulators in the rapper boxes, especially in colder
weather. A re-piping of the purge air system has reduced this
problem somewhat.
SUMMARY
The pilot plant testing and subsequent data has provided for a
successful full size EP design and installation for this BOF
application. The problems unique to this process were solved at the
early stage of the pilot plant program when design changes were
easily made. As a result, the required emission level and opacity
is within that allowed by state regulation with no major problems in
day to day operation.
432
-------
Figure 1 - Pilot Plant installation at BOF plant
Figure 2 - Another view of Mobile EP at BOF site
433
-------
.pa.
00
Mobile EP outlet with all
fields off during blow cycle
Figure 4 - Mobile EP outlet with one field on
-------
-p"
CO
en
Figure 5 - Mobile EP outlet with 2 fields
Figure 6 - Mobile EP outlet with 3 fields
-------
Figure 7 - Mobile EP hopper outlet
F,igure 8 - Mobile EP plates with EOF
436
-------
FIGURE 9, BOF DUST RESISTIVITY
10
ii
P 2
to
12.o9
10
8
10% H20
20% H20
150 200 250 300 350
TEMPERATURE, »F
400
437
-------
FIGURE 10, PARTICLE SIZE DISTRIBUTION
.p.
CO
CO
100
80
60
40
20
Blow
4 8 12 16
AERODYNAMIC EQUIVALENT PARTICLE SIZE, AJL
-------
FIGURE II, OPACITY vs ESP OUTLET
I.Or
.5
o
o
u.
<£ .2
.1
.05
UJ
-J
H-
O
.02
.01
§ •_
5 10 20 40 60 100
OPACITY, %
439
-------
.30
.25
o
o.20
o
o •> *
UJ
.IO
.05
FIGURE 12
Performance of Pilot Plant (•)
vs
Full Siie System (x)
2345678
COLLECTION AREA, NORMALIZED UNITS
10
-------
THE SELECTION AND OPERATION OF A NEW PRECIPITATOR SYSTEM
ON AN EXISTING BASIC OXYGEN FURNACE
By:
Douglas Ruth, P.E.
United Me Gill Corporation
2400 Fairwood Avenue
Columbus, Ohio 43216
And:
David Shilton
CF&I Steel Corporation
P. 0. Box 316
Pueblo, Colorado 81002
ABSTRACT
The selection of a new precipitator installation on an existing
basic oxygen furnace facility is discussed. The selection process
began with a search of technical alternatives and ultimately was
based on a mobile unit test unit program. The completed precipitator
system was then tested after initial operation. Performance data
and operating practices are then summarized.
441
-------
THE SELECTION AND OPERATION OF A NEW PRECIPITATQR SYSTEM
ON AN EXISTING BASIC OXYGEN FURNACE
SEARCH AND SELECTION
The original precipitator was installed as part of the Basic Oxygen
Furnace installation completed in the summer of 1961. At that time, perform-
ance on precipitators, or for that matter, any other air pollution control
device, was not predicated on opacities, but on efficiencies or outlet grain
loadings. A satisfactory level of partlculate emissions on a pounds-per-hour
basis was usually established between the purchaser and the supplier of the air
pollution control device. The precipitator supplied with the original Basic
Oxygen Steel Shop in 1961 was commensurate with the air pollution control
regulations then in force or anticipated.
CF&I operates a two 90 metric ton vessel EOF shop with one vessel active
at any given time; the other vessel concurrently undergoing a reline. The
precipitator system installed was accordingly designed to handle the emissions
from one active vessel. Specifically, the volume to the precipitator was rated
at 300,000 ACFM at 350° F. with moisture levels varying between 10 to 30%.
CF&I's EOF shop utilizes an open hood concept with a "spark box" cooling
chamber immediately after the vessel. This system removes much of the heavy
solids which are then directed to a clarifier-lagoon water treatment system.
It should be noted that a spark box or a somewhat similar pre-treatment device
would be included in a majority of air pollution control systems installed on
the existing EOF. The original precipitator system was designed with an
efficiency of 97.5%. This efficiency was established between the incoming
emissions to the precipitator, but after the spark box, and the emissions
emanating from the stack.
This two-chamber three field precipitator was first tested in October,
1962. Results of that test indicated an operating efficiency of 97.26% with
volumes near 300,000 ACFM at 350° F. Although not at the design criteria, it
was acceptable at that time, due to the difficulty in verifying all of any of
the idiosyncrasies which may have been evident during the testing. It is worth
noting that opacity considerations were not of primary importance during this
era of air pollution control.
During the first ten years of operation of the original precipitators,
trends became apparent. The principal problem was severe corrosion of the
collecting plates primarily near their lower connection point. Complete re-
placement of most of the internal components was necessary approximately every
three years. An early problem with severe corrosion and subsequent breakage of
the emitter electrode wires was substantially solved by the use of Carpenter 20
wiring to replace the original 316 stainless; however, during the periodic
repair periods, plates, wires, and frames were all replaced. The danger of
442
-------
explosions with high levels of CO was not only realized, but actual explosions
did occur. Most of the explosions were quite minor and only on rare occasions
was more than a few minutes of production time lost. Another major drawback of
the original precipitators was the two-chambered limitation. Shutting down one
half of the precipitator to work on the remaining chamber resulted in excessive
stack emissions. Additionally, with only an insulated dividing wall, the shut
down chamber took a long time to cool down sufficiently for workmen to enter.
A gradual deterioration of the precipitator was evident after the first ten
years of operation.
The advent of the Colorado State Regulation I, with a stringent 20%
opacity limitation on any given stack reading, made it necessary to re-evaluate
the performance of the EOF precipitator. Opacity observations on the EOF
precipitator stack were initiated on a routine basis in 1974. Observations
indicated compliance with the 20% limit a great deal of the time in cool or
cold damp weather, but readings above 20% opacity in hot, dry conditions. The
affect of ambient conditions on the precipitator was immense - a phenomenon
which we have never fully understood.
In order to appraise the operating conditions of the precipitator, an
efficiency test was performed in August 1974. Testing was performed at the in-
let to the precipitators and the existing stack. Dust sampling commenced at
the start of the oxygen blow and included any reblows which occur. Usually,
the oxygen blow would last 19 to 21 minutes with reblows 30 seconds to 2
minutes. Test data was fractionated to various parts of the cycle and averages
also developed for the whole process. These tests indicated an average efficiency
of 97.5%, actually higher than the original tests in 1962; a maximum inlet
grain loading of 11 GR/SCFD, an average inlet grain loading of 7 GR/SCFD, and
an average outlet loading of .1261 GR/SCFD. The maximum inlet grain loadings
were usually associated with reblows. Stack opacities observed during these
tests averaged 68%. The measured volume of gas entering the precipitator thru
the ductwork had decreased to 212,000 ACFM at 350° F., indicative of leaks
which had accumulated in the precipitator system during 13 years of operation.
The calculated emission of 55 Ibs./heat with over 10,000 heats per year es-
tablished an operating level of 300 tons per year of emissions. Assuming 1.3
heats per hour, this translated to approximately 72 Ib./hr. of emissions. The
Colorado Process Weight Regulation allowed for 40 Ib./hr. Subsequent tests in
1975, even after one-half of the precipitator was rebuilt, indicated emission
levels near the 300 ton-per-year level. The actual search and selection of a
system of the EOF primary emission source now had its design criteria estab-
lished; performance at less than 20% opacity at any time and a process weight
level below 40 Ib./hr. of emissions at the inlet conditions established.
It should be noted that the 20% opacity requirement is stated "at any
time". Colorado opacity regulations do not allow averaging of opacity readings;
any instantaneous reading over 20% opacity is technically a violation. Hence,
our design requirements could actually be described as a clear stack during
normal operation with a 20% opacity limit for non-optimum performance during a
reblow.
CF&I virtually accepted the fact that upgrading the existing precipitators
electrically would not meet the more stringent requirements. The limitation of
443
-------
collection surface area would remain with the two chambered, three-field pre-
cipitator. A closed-hood, non-combustion system had been discussed several
years earlier, but was prohibitively difficult to install in an existing EOF
shop, and the merits of its installation from an air pollution control stand-
point were rather dubious. The scope of inquiries for new or additional equip-
ment was narrowed down to baghouses, scrubbers, and precipitators.
Baghouses were discussed but never seriously considered due to the burn up
potential caused by the wide variation of temperatures inherent with the EOF.
Any baghouse would need a complex conditioning tower ahead of it to relieve
moisture problems and to protect it from high temperature excursions.
Two options remained; scrubbers and precipitators. The trend in 1975 and
still today was to install scrubbers on EOF shops. Considerably more than half
of the EOF shops in North America use scrubbers for primary control. Pre-
liminary proposals on a scrubber installation at the CF&I shop were extremely
expensive; in retrospect, approximately double that of the ultimate precipi-
tator installation. Outside of the initial capital cost, scrubbers were ob-
viously considerably more energy intensive than precipitators, due to the high
horsepower requirements of higher operating pressure differentials. The Pueblo
Plant's water system and future discharge permit conditions also could have
been affected with the addition of a water-based, solid-waste removal system
associated with a scrubber. The most attractive features of a scrubber system
were the non-explosive nature of its operation with carbon monoxide and the
ability of a scrubber to operate under the many wide process changes which
accompany a basic oxygen furnace operation. CF&I then sought precipitator
equipment which could reduce the disadvantages of scrubbers and mitigate pre-
cipitator maladies.
Several precipitator alternatives were sought by CF&I. Consideration was
given to adding one or two more precipitators to the existing unit so as to
permit maintenance on one unit and comply with the control regulations with the
residual operating units. All designs of precipitators were initially con-
sidered: rigid frame (European design), wire-weight, stiff electrode, and
various other electrode configurations. Consideration was given to rebuild the
existing precipitators with new internals of a different manufacturer. Many of
the precipitator vendors did not wish to guarantee a system with their internal
components in another manufacturer's 15-year-old unit. With all the alter-
natives being evaluated, the underlying principle of the modifications to be
made was to meet the opacity and process weight requirements with one unit
(precipitator) off line.
Through the course of evaluating precipitator proposals, the ultimate
efficiency and control needed to satisfy the opacity requirement became ques-
tionable. Installations on BQE- shops in Europe and Great Britain were visited,
as well as other industrial operations in the United States. It was observed
that the German EOF installation did not need to satisfy any opacity require-
ments, only a concentration allowance of .07 GR/SCFD. A British Steel Oper-
ation needed to satisfy an outlet emission of .05 GR/SCFD. Opacity vs. Grain
Loading tests done on the existing CF&I EOF precipitator in December 1974 did
not conclusively provide a means to verify the actual grain loading corres-
ponding to 20% opacity. Because the levels of performance on the BOF's oper-
444
-------
ating overseas did not show any visible emissions, it was thought that a grain
loading below those reported GR/SCFD would conservatively satisfy a 20% opacity
requirement. Additionally, manufacturers who were members of the International
Gas Cleaning Institute (IGCI) were reluctant to guarantee any precipitator
system to an opacity proviso. Consequently, the number of proposals received
which did accept this opacity guarantee was limited and those which did de-
creased the outlet option to .033 GR/SCFD in order to comply with the opacity
limit.
In order to verify the actual size of the precipitator system which was
required to meet the grain loading and opacity limits, the United McGill
mobile EP unit was operated on a slip stream at the CF&I BQF plant in April
1976. Results from the mobile test unit agreed quite favorably with the
previously discussed test data. This data provided the basis of the design of
the precipitator units with sufficient capacity to meet the desired perform-
ance and maintenance levels. It also provided data to alleviate some of the
problems with precipitators on BOF's, such as carbon monoxide levels and the
unit size needed to handle process variations.
In the final analysis, it was the United McGill Corporation, with the
supportive data from their mobile EP unit, who provided CF&I with an economically
attractive proposal [four (4) completely separate five (5) field precipitator
units as ultimately constructed], which would guarantee opacity limits below
20% and satisfy process weight limitations. These conditions were to be main-
tained with three of the four units operating, thus allowing one unit down for
maintenance while the system was on line.
445
-------
FULL SCALE TESTING
Units 1 and 2 were initially placed on line in late July 1977. After a
short period of trial operation, the old precipitator inlet duct system was
removed in August 1977. Stack appearance of the new two unit system was much
improved over the old system.
It was decided to stack test the two units to gauge the performance with
only two units operating against both the past performance of the original
precipitator and future performance of the complete new precipitator. The
units were tested in October 1977. The volume and temperature of the gas were
found to be very close to design conditions, 300,000 ACFM at 350° F. Outlet
particulate loadings were tested at .073 GR/DSCF, approximately double the
expected design value of .033 GR/DSCF for three unit operation.
The 300 ton per year level of emissions which was previously discussed in
the interim had become a state approved variance level of operating perform-
ance. Since the construction format of the United McGill precipitator system
called for the demolition of the existing precipitators, it became obvious that
half of the new McGill precipitator system would meet the variance level. In
fact, emissions for the first two units out of the four total were calculated
at 171 tons per year, slightly over half as much as the 300 ton per year vari-
ance level.
The second two EP units were placed in service in early March 1978. After
final checkout and trouble shooting, the complete precipitator system was
tested during the week of April 10, 1978. Since the system contained four
units and any three units were to operate and perform to design specifications,
all combinations of three EP operation plus four EP operation were tested.
Inlet and outlet points were tested simultaneously. The inlet ports were
located in the 8 ft. diameter downcomer duct after the "Y" junction to accomo-
date either vessel operating. The outlet ports were on the main stack at the
90 foot level on the 110 foot high, 10 foot diameter stack.
The test results of April 1978 indicated that all combinations of three
unit operation operated well below the process weight allowance of 40 Ibs./hr.
Maximum grain loadings to the precipitator system were previously estimated at
11 GR/DSCF. An average value of 7 GR/DSCF had also been determined during
mobile EP testing and prior experience with the old precipitator. The initial
performance tests bore out both of these concentration levels.
Performance guarantees on the United McGill precipitators specified that
with three unit operation outlet grain loadings would be .033 GR/DSCF; with
446
-------
four unit operation .022 GR/DSCF. All four combinations of three unit oper-
ation readily met the .033 GR/DSCF criteria. The four unit operation was less
than one half the guaranteed outlet of .022 GR/DSCF.
Using the design maximum inlet grain loading of 11 GR/DSCF and design
outlet grain loading of .033 GR/DSCF, the efficiency of the system would then
be 99.7%; 99.8% with .022 GR/DSCF outlet grain loading. The performance tests
showed that these efficiencies indeed were obtained with three unit operation
(99.7%) and four unit operation (99.8%). Although the inlet concentrations
were not as great as expected, the outlet concentrations were also appro-
priately lower to achieve the design efficiencies.
It should be noted that average moisture levels for these tests were
above the 15% minimum desired. With the temperatures at the inlet to the
precipitator also staying somewhat above 350° F, the resistivity condition
during the test was such that near optimum precipitation was achieved.
A follow up test run about a year later, May 24, 1979, with four (4)
units operating, showed the same high level of performance with an efficiency
of 99.81% based on actual test data, flow and grain loadings, and an outlet
grain loading of .019 GR/DSCF. This test covered the full blowing time of
four heats including six reblows. Opacity observations made during each of
the heats averaged less than one (1) and no readings were over 20%.
For the sake of clarity, efficiency measurements were based on inlet
concentrations before the drop out chamber and outlet concentrations measured
in the stack. It is impossible to measure dust concentrations after the drop
out chamber just ahead of the precipitators due to duct restrictions. There-
fore, efficiencies should be referred to as system efficiencies and not pre-
cipitator efficiencies.
Opacity levels as read by both an in-stack transmissometer and by certi-
fied smoke readers (including State and local air pollution control personnel)
during the performance tests in April 1978 indicated that they were maintained
below 20% with four unit operation. A few short duration excursions slightly
above 20% opacity were observed with three unit operation, generally associ-
ated with initial blowing of oxygen and reblows of oxygen.
Several inspections by State and local air pollution control personnel
have been made since April 1978 and compliance with the Colorado opacity code
has been maintained.
447
-------
OPERATION AMD MAINTENANCE
As stated previously, four unit operation of the precipitator system
began shortly before March 1, 1978. Units 1 and 2 had been operating since
July 1977. No major operating problems were encountered with these two units
from July 1977 to March 1978. The first two precipitator units had been
subject to overloading in order to handle total gas and dust volume during
this period, thus the lack of operating problems was very gratifying.
Upon initiation of the four unit operation, the A.V.C.'s (Automatic
Voltage Controllers) inappropriate installation resulted in overheating and
some of the wiring was destroyed. A decrease in efficiency was noted by way
of the precipitator stack appearance. After the A.V.C. wiring was corrected,
stack appearance improved dramatically with four unit operation of the system.
One A.V.C. unit was later replaced in November 1978; however, overall ex-
perience with the A.V.C. operation since the initial problem period has been
quite satisfactory.
The most persistent operating problem with the precipitator system since
April 1978 has been associated with the high density pneumatic dust transpor-
ting system. Almost weekly problems on any one of the transporter lines has
been evidenced and usually associated with plugging in the line or a defective
gate valve. Gate valves have been replaced in some of the transporters.
Periodic cleaning of all the transporters is needed to maintain continuity of
dust removal. Another problem which was encountered and essentially solved
was periodic malfunctions of the compressed air dryer supplying the trans-
porter system. In general, preventive maintenance based on experience has
alleviated major problems with the transporter system and its overall oper-
ation has been satisfactory.
A pelletizer was installed underneath the dust storage silo in mid April
1978 to further assure total dust control in the system including fugitive
emissions associated with dust handling. Initial operation of the pelletizer
involved adjustment to the water feed rates which determine the size of the
balls formed by the pelletizer. Although the pelletizer system was designed
for automatic operation to start when a high level was reached in the dust
silo and to stop at a low dust level, it has not been possible to leave the
pelletizer unattended. The dust level probes have not worked reliably and
frequent adjustments are required on the spray water. Usually eight (8) hours
of semi-attended operation on one shift is sufficient to pelletize all of the
dust collected in 24 hours. The 6000 ACFM baghouse system associated with the
pelletizer and silo has required some bag replacement. It presently appears
that annual bag replacement will be necessary. As with the transporters,
448
-------
periodic plugging of the silo has been noted but not serious enough to jeopar-
dize the operation of the pelletizer during attended operation.
The EP units proper have not experienced any problems with intrinsic
hardware. Most notably, the emitting electrode plates and collecting plates
have not become distorted to cause chronic shorts or arcing. Accordingly, no
emitter spikes or plates have needed replacement. It has, however, been
necessary to use compressed air through a permanent piping system to blow dust
off ledges between modules, blowing each field at each level once every shift
to avoid dust buildup which can cause a ground.
The complete four unit system operated during the winter season of 1978-
1979- January 1979 was the coldest January in Pueblo on record, and also for
Pueblo, one of the wettest. A marked increase in insulator failures was noted
in the rapper boxes during this past winter, especially around January. A
"Catch 22" appears to be evident in this particular case in that increased
ambient air into the rapper boxes tends to induce moisture on the insulators
but maintains air flow through the boxes to keep internal components clean; if
the boxes are completely sealed, moisture and the effects of ambient air are
kept out but dust accumulation in the boxes tends to increase drastically re-
sulting in excessive insulator failure and limit switch failure due to high
temperatures. The changing ambient conditions due to the seasons dictates
that some equilibrium condition must be established to prevent the chronic
rapper insulator failures which have proliferated this past winter. Piping to
the rapper boxes has been revised to more efficiently clean them and purge air
holes have been enlarged to accomodate more air. The insulators located in
the High Voltage energizing system have had no failures. These insulators are
enclosed in heated compartments in the penthouse section of the EPs and operate
without fan generated purge air.
Air leaks have been noted through the seals in the pneumatic cylinders.
All four units have had the air seals replaced in these cylinders. The cylin-
der air leakage has not been evident since these replacements.
One of the major drawbacks of the installation of a precipitator system
on a Basic Oxygen Furnace is the potential of explosions inherent with the
combination of high voltage and moderate amounts of carbon monoxide in the
presence of oxygen. An explosion occurred in May 1978 in the EP system.
However, because explosion doors were designed into the system, the doors
opened and no structural damage was suffered by the precipitators. CF&l's BOF
primary gas cleaning system has two CO monitors; one (the original) near the
operating pulpit and the new monitor on the downcomer just upstream of the
inlet sampling location. Both monitors are arranged to stop the production
process and shut off the precipitator upon sensing high CO. The pulpit moni-
tor analyzes a sample of gas withdrawn from the duct, but the downcomer monitor
senses directly through the 8 ft. diameter duct, reading the CO content in a
fraction of a second. If the CO should exceed 14 percent at either analyzer,
the power to the precipitator fields shuts off to prevent ignition of an
explosion. Simultaneously, at the BOF vessel, the oxygen blow stops and the
lance raises. The explosion in May 1978 reached reactive levels below the set
points of the monitors which are set at 14%. Improved sampling and calibration
techniques are being investigated by CF&I on the CO monitor located on the
downcomer.
449
-------
Closely associated with the CO hazard is the operation of the main I.D.
fans. The fan inlet dampers are operated hydraulically by a current control
system to maintain a pre-set amperage to the fan motor. When the duct system
is cold after a EOF down day, the normal current control can result in in-
sufficient draft. It has been found desirable to hold a higher amperage
during the first three heats after startup. Tests were conducted in February
1979 to determine what amperage levels and fan louvre settings were needed
after start up to maintain sufficient draft to nullify CO buildup. It was
determined that the fan motors do indeed have sufficient power to maintain the
4" t^O draft necessary during start up and that the amperage levels may need
changing during various climatic changes. As with the rapper insulators, the
CO and air flow relationship is subject to cold weather idiosyncrasies and
will have to be monitored.
450
-------
PERFORMANCE
As mentioned before in the discussion involving initial performance
tests, the four unit EP system has satisfied the design specifications.
System performance accomodates the "instantaneous" 20% opacity limits with
four units operating. Since it is the intent of CF&I to operate with four
units whenever possible, the Colorado Opacity Regulation of 20% is satisfied
virtually continuously. Three unit system operation easily meets process
weight requirements (again, refer to initial performance tests). However, as
the initial performance tests indicated, occasional excursions above 20%
opacity may occur with three unit operation. Most typically, these readings
over 20% opacity are very brief and associated with commencement of the blow
or reblows. These puffs are indigenous of the EOF process and well controlled
with four units operating. Three unit operation is limited to the extent
possible; however, on occasion one unit has been shut down. Reasons for such
shut downs have been plugged hoppers, plugged screws, or broken insulators.
The durations of the shut downs usually have been less than one day. A pre-
ventive maintenance program was instituted in March 1979 which involves the
isolation of one precipitator from the active gas system for one 8-hour period
per week. As initially discussed, this was one of the maintenance features
that was designed into the precipitator system.
The most repetitive performance malady with the precipitator system has
been with high CO tripping off power to the units. In the past, there has
been one or two instances of these cutoffs most every month. As previously
discussed, knowledge obtained from fan amp changes and improved calibration
and maintenance of the CO monitor system should alleviate such power outages
on the precipitators.
Daily operation of the system with four unit operation indicates no
visible emissions during any stage of the process. Performance of the four
unit system appears to be at least equivalent to that of a high energy scrubber.
451
-------
CONTROL OF FINE PARTICLE EMISSIONS
WITH WET ELECTROSTATIC PRECIPITATION
By:
Steven A. Jaasund
Flu id-Ionic Systems
a division of Envirotech Corporation
2525 East Magnolia
Phoenix, Arizona 85034
This paper presents the results, both pilot and full-scale, of experience
with the application of wet electrostatic precipitation technology for the
control of fine particle emissions from industrial processes. Performance
data involving the collection of such difficult-to-clean emissions as sulfuric
acid mist and recovery boiler salt fume are presented. The measured wet pre-
cipitator performance parameters (e.g., SCA, effective migration velocity) are
compared to those reported in literature for dry precipitators operating on
equivalent processes. Particular emphasis is directed toward examining wet
electrostatic precipitator performance levels in controlling very high concen-
trations of submicron fume with the accompanying problem of space charge corona
quenching. Relationships between these performance data and specific wet pre-
cipitator operating parameters such as operating voltage and current density
are also examined.
452
-------
CONTROL OF FINE PARTICLE EMISSIONS
WITH WET ELECTROSTATIC PRECIPITATION
INTRODUCTION
Considerable attention has been focused recently on the harmful
effects of airborne particles less than two microns in diameter. Particles
in this size range have been shown to have a high potential to cause lung
damage, both directly and with adsorbed gases. Also, these particles are
known to be a chief cause of visibility degradation in polluted areas. As a
consequence, the emphasis in applied and basic research concerning particu-
late emission control has been directed toward the development of more effi-
cient means for collecting these fine particles.
Electrostatic precipitation is known to be among the most promising means
of achieving high fine particle collection efficiencies. To date, the great
majority of the research on the collection of fine particles by electrostatic
precipitation has been concerned with the collection of fly ash from the com-
bustion of coal for power generation. The exclusive attention given to fly
ash collection with dry electrostatic precipitators (ESPs) is unfortunate
because:
' Dry ESP technology is often limited in applicability when
collection of particulate matter other than fly ash is re-
quired; e.g., condensible materials cannot be collected in
a dry ESP.
' Conventional dry ESP technology is limited by inherent flaws
and nonidealities in design; e.g., rapping/reentrainment
losses and high resistivity dusts often make the attainment
of very low outlet loadings impossible.
The wet electrostatic precipitator (WESP) can be an effective means of
overcoming most of the design deficiencies of dry ESPs so that the maximum
potential for fine particulate collection may be achieved. It is the object
of this paper to present selected WESP operating and performance data in order
to illustrate the potential of the wet electrostatic precipitation process to
reduce fine particle emissions to very low levels. To this end, the following
text wi11:
' report actual operating and performance data collected at
both full and pilot-scale WESP installations operating on
fine particle emission sources,
453
-------
' compare the performance of the individual WESP facilities
with the performance of other conventional technologies
reported for similar applications, and
' explore the operating factors which affect WESP performance.
Specifically, performance data gathered for WESP applications on the
following fine particle emission sources and categories will be discussed:
' Mineral wool cupola
' Sintering windboxes
' S03 hydration (sulfuric acid mist)
• Sulfite recovery boilers
' Phosphorus furnace
' Chemical incinerators
DESCRIPTION OF WET PRECIPITATION PROCESS
The wet precipitation process is fundamentally the same as the dry pre-
cipitation process. The distinguishing feature of the wet precipitation
process is that the particles are collected on a liquid medium which then acts
as both a collecting surface and a removal medium. In dry precipitators, the
collection function is performed by a grounded metal surface and removal of
collected material is achieved by mechanical impact. The simple change from
dry collection with mechanical rapping to collection and removal in a liquid
medium frees the wet process from some of the problems associated with the
dry process. That is, with the wet process:
' resistivity of the collected material is not a performance-
influencing factor,
' rapping and reentrainment losses are avoided, and
' condensible emissions are readily collected.
Of course, the use of a flushing medium, usually water, can introduce
new considerations and problems to the application of the electrostatic pre-
cipitation process. The principal problems introduced are:
' increased corrosion potential
' need for water treatment.
The embodiment of the wet precipitation process has taken many forms since
the first WESP was conceived. WESP designs employing continuous and intermit-
tent flushing, either with sprays or overflow weirs, have been tried on a wide
454
-------
variety of applications with varying degrees of success. The specific perfor-
mance and operating data presented in this paper were all gathered at operating
installations of the HYDRO-PRECIPITROL wet electrostatic precipitator.
The HYDRO-PRECIPITROL is a continuously-irrigated, concentric-cylinder
wet electrostatic precipitator with integral prescrubber and gas distribution
system. As shown in Figure 1, hot dirty gas enters tangentially at the bottom
where a bank of hydraulic spray nozzles cool the entering gases and provide
some degree of large particulate removal. The tangentially spinning gas next
passes upward through gas straightening vanes which remove the tangential com-
ponent of the gas velocity and render the flow profile essentially uniform
across the entire cross section. The cooled, partially-cleaned gases next
enter a series of annular electrostatic zones; these are made up of seven con-
centric collection cylinders, with six concentric high-voltage, expanded-metal
discharge electrode cages suspended from above and located equidistant between
the cylinders. The discharge electrode cage assembly is supported by a cen-
trally located insulator as shown. As the gas passes up into the zone between
the discharge electrode and the wall of a collecting cylinder, the particles
are charged by a high-voltage corona discharge and driven to the collecting
cylinder by the high-voltage field. Each collection cylinder is continuously
irrigated from above by circular liquor distribution pipes mounted on top of
each cylinder. The distribution pipes employ a unique overflow design of 1/2-
inch vertical nipples mounted four inches apart around the circumference of
each distributor. This arrangement provides a smooth, contiguous film of
liquor flowing over all collecting surfaces. This water distribution system
is essentially nonplugging due to the large discharge orifice size employed;
it also is insensitive to the leveling difficulties encountered with weir-
type water distributors. With the exception of the discharge electrodes, the
entire device is constructed of corrosion-resistant, fiberglass-reinforced
plastic (FRP). The discharge electrodes are made of metals suitable for the
corrosion potential of the applications; e.g., high-grade alloys, such as
Hastelloy C-276, may be specified for very corrosive gas streams.
The HYDRO-PRECIPITROL design described above offers the following
operating features which help maximize the collection of fine particles:
' excellent gas flow distribution (RMS deviations of less than
10% from the mean velocity are achievable),
' smooth irrigating film to minimize electric field distur-
bances and increase operating voltage, and
' no sneakage due to the vertical/concentric design.
In fact, a careful scrutiny of the design features show that this design
eliminates all of the nonidealities occurring in other precipitator designs,
both wet and dry. The following data show that the HYDRO-PRECIPITROL is
able to achieve performance levels on fine particulate applications signifi-
cantly higher than reported for any othar available precipitation technology.
455
-------
CLEAN GAS A DISCHARGE
HOOD
ACCESS
MANWAY
PRECIPITATOR
HIGH VOLTAGE
CABLE
PRECIPITATOR
BASE
ACCESS
MANWAY
PRECONDITIONER
SPRAYS
GAS
INLET
PRECONDITIONER
WATER
DISTRIBUTOR
WATER INLET
CYLINDER
ELECTRODE
CAGE
VENTURI/
DRAIN GUTTER
STRAIGHTENING
VANES
PRECIPITATOR
DRAIN
PRECONDITIONER
DRAIN
Figure 1 "HYDRO-PRECIPITROL" wet electrostatic precipitator.
456
-------
FINE PARTICLE COLLECTION RESULTS
Fluid-Ionic Systems, a division of Ervirotech Corporation, has been manufac-
turing and marketing the HYDRO-PRECIPITROL wet electrostatic precipitator
since 1974. Since then, WESP operating and performance data has been acquired
on 62 different pilot and full-scale applications. Some of the test programs
at these applications have been extensive, lasting over a,year and involving
up to 100 individual stack tests. Also, many of these applications have in-
volved the collection of high concentrations of particles less than two
microns in diameter. The following data are selected from the application
experience with the HYDRO-PRECIPITROL specifically to illustrate the capa-
bility of the WESP to collect fine particles. As a point of reference, per-
formance parameters will be given in terms of the Deutch equation,
EFFICIENCY = 1 - EXP (-^-)
where
EFFICIENCY = Inlet - Outlet
LH-IL1LNLY
A = Collecting surface
(adjacent to discharge electrodes)
Q = Area volumetric gas flow rate
(at saturated conditions)
W = Effective migration velocity
Performance parameters were calculated from test data obtained with vari-
ous sampling techniques; i.e., in-stack filter, EPA Method No. 5, and EPA
Method No. 8. Other than acid mists, condensible concentrations measured were
not included in the calculation of performance parameters.
Mineral Wool Cupola
A pilot test program was conducted at a Bethlehem Steel Corporation
Mineral Wool Plant to demonstrate the capability of the WESP to effectively
clean the top gas emissions from two hot-blast, slag-melting cupola furnaces.
The particulate emissions from the cupola consisted chiefly of silica, alkali,
and elemental sulfur fume. Particle size analyses of the aerosol showed that
80% was less than 1.0 micron in diameter. Prior pilot testing with a high
457
-------
energy scrubber showed that the emission was exceptionally hard to clean; i.e.,
over 80 inches W.G. pressure drop was reouired to reduce the concentration to
0.04 grains/SCFD, the state outlet criterion.
The results of the pilot testing demonstrated the ability of the WESP
to achieve extremely low outlet loadings. During the final phase of testing,
a series of eight inlet/outlet efficiency tests wece conducted. The inlet and
outlet particulate loadings averaged 1.08 grams/NM (0.470 grains/SCFD) and
0.005 grams/NM (0.002 grains/SCFD) respectively. The specific collection
area and the migration velocity averaged 30 M/M /second (153 ft /1000 ACFM)
and 18 cm/second respectively. Field strengths of 5 kv/cm and current densi-
ties of 5 na/cm were measured for these tests.
Sintering Plants
Bethlehem Steel also tested the WESP at two sintering plants for the
control of windbox emissions. A pilot test program was conducted at the
Lackawanna Sintering Plant, and a subsequent full-scale demonstration test was
conducted at the Johnstown Sintering Plant. The,results of these test pro-'
grams have been reported by Mazer et. al. (1976) .
No specific size distributions are available for either WESP application.
However, it is known that sintering windbox emissions usually present an ex-
tremely difficult gas cleaning task due to both high concentrations of fine
particles and the high dust resistivity often encountered. In a recently
published report, Szabo^and Gerstle (1978)? state that specific collection
areas in excess of 80 M/M /second (400 ft /1000 ACFM) may be required to
achieve 99% efficiency on windbox emissions with dry ESPs.
The data obtained at Johnstown and Lackawanna may be compared to perfor-
mance curves also reported by Szabo and Gerstle for dry ESPs operating on
windbox emissions. This comparison, shown on Figure 2, illustrates the dra-
matic increase in precipitator performance that can be achieved by switching
from conventional dry technology to a well-designed wet system. It is note-
worthy that the data reported in Figure 2 for a dry precipitator are for
operation on the raw windbox emissions, while the WESP data reported are for
emissions that had been treated by cyclones (Lackawanna) or an existing pre-
cipitator (Johnstown). Hence, the dry precipitator performance curves are for
an inherently easier gas cleaning situation. It should also be noted that the
WESP efficiency includes the scrubbing effect of the sprays at the gas inlet.
These sprays undoubtedly perform a significant scrubbing function on any coarse
particulate entering the WESP. Unfortunately, because of the WESP geometry,
it is nearly impossible to quantify this effect by field testing.
Sulfuric Acid Mist
The release of anhydrous SOn into a humid environment is known to form
an extremely fine sulfuric acid mist aerosol. La Mer et. al. (1950) showed
that sulfuric acid mist aerosols formed in this manner had mass mean diameters
which ranged from 0.1 to 0.4 microns and occurred in very narrow size ranges.
458
-------
cn
<£>
O -JOHNSTOWN FULL SCALE TESTS
' LACKAWANNA PILOT SCALE TESTS
10
50 100 200 300 400500
SCA (FT2/ 1000 ACFM )
1000
Figure 2 Dry and wet precipitator performance on sinter windbox emissions.
-------
During a 1978 EPA sponsored pilot-scale evaluation of the HYDRO-
PRECIPITROL as an S02 removal device, tests were conducted on the WESP to
quantify its ability to capture sulfuric acid mist. The test aerosol was
formed as described above—anhydrous SOo was continuously bled into a warm
(49°C, 120°F) saturated stream of simulated boiler flue gas to form approxi-
mately 0.39 grams/NIT (0.17 grains/SCFD) of sulfuric acid mist. The gas
stream, laden with acid mist, was then passed through the WESP pilot plant.
Inlet/outlet testing was conducted according to the specifications of EPA
Method No. 8. To quantify the effect of voltage on precipitator performance,
the operating voltage was deliberately varied from test to test. The results
of the efficiency tests showing the effect of voltage variations are plotted
in Figure 3.
Sulfite Recovery Boiler
In the sulfite pulping process, the recovery boiler flue gas is scrubbed
in a series of spray and packed absorption towers for the recovery of SO?
necessary to the pulping process. At facilities where saltwater- borne fogs
are processed, the gases exiting the absorption towers are often contaminated
with heavy concentrations of submicron salt fume and require further particu-
late removal steps. Figure 4 illustrates the extremely fine nature of the
aerosol encountered at two such operations.
To identify the technology necessary to reduce the fine particle emission
problem at these two pulp mills, pilot scale WESP performance evaluations were
conducted. At one of these mills a full-scale WESP was subsequently installed
and evaluated. The results of all three test programs, shown below in Table 1,
clearly demonstrated the capability of the WESP to effectively abate this fine
particulate emission.
TABLE 1
WESP PERFORMANCE PARAMETERS ON
SULFITE RECOVERY BOILER EMISSIONS
Plant 1 Pilot
Plant 1 Full
Scale
Plant 2 Pilot
COLLECTION
EFFICIENCY
(%)
96.1
94.3
97.2
MIGRATION
VELOCITY
(cm/ sec)
21.3
22.0
25.0
SPECIFIC
COLLECTION
2 A!FA
M /M /sec
,-and
(FTV1000CFM)
15.3 & (78)
13.4 & (68)
14.6 & (74)
FIELD
STRENGTH
(kv/cm)
7.3
4.3
5.8
CURRENT
DENSITY
(na/cm )
—
3.3
—
460
-------
100
2468
AVERAGE FIELD STRENGTH ( KV/CM )
Figure 3 WESP performance on sulfuric acid mist.
-------
v/i
Z
o
a*.
u
4.0
3.0
2.0
< 1.0
5 °-9
0.8
S 0.7
0.5
Q.
u 0.4
a
O
0.3
0.2
I I I I
I I I I I 1 I
99.8 99.5 99 98 95 95 80 70 60 50 40 30 20
PER CENT LESS THAN STATED DIAMETER
10
Figure 4 Sulfite recovery boiler salt fume size distributions.
-------
It is noteworthy that the full-scale WESP performs effectively despite an
extremely low current density.
Phosphorus Furnace
The formation of ?2^^ and related compounds from the rapid oxidation of
elemental phosphorus is known to result in an extremely fine aerosol. An
observable measure of this fineness is the dense white smoke which results from
the combustion of phosphorus. Pilot tests on such emissions from the tapping
hole of a phosphorus furnace again demonstrated the capability of the WESP to
achieve high efficiencies on submicron particles. A summary of the results of
the 22 pilot tests conducted is shown below in Table 2.
TABLE 2
WESP PERFORMANCE ON P205 AEROSOL
Average effective migration velocity - 22.3 cm/second
Average specific collection area
Average P-Or removal efficiency
Particle size distribution
- 8.1 M2/M,/second
- (41.2 ftVI000 ACFM)
- 84%
- Greater than 90% by weight
below 1.0 micron in diameter
Chemical Incinerators
The incineration of liquid and solid waste chemicals results in the
generation of varying concentrations of submicron particulate matter. The
concentration of submicron material is often extremely high when waste com-
pounds containing certain noncombustible constituents (e.g., sodium, calcium,
or potassium) are being burned.
The WESP was pilot tested on two chemical incinerators and later in-
stalled for full-scale duty on one. Both of these incinerators burn waste
chemicals and generate high concentrations of fine particles due to the
presence of noncombustible constituents in the wastes. A summary of the
results of the pilot tests at both facilities, plus corresponding aerosol size
distribution information, is given in Table 3.
The high inlet loads encountered at chemical incinerators are of parti-
cular concern in the application of single field precipitators, such as the
HYDRO-PRECIPITROL, because of the problem of space-charge corona quenching.
The phenomenon of severe corona quenching, illustrated in Figure 5, results
463
-------
when very high concentrations of fine aerosol enter a precipitator and become
charged to the polarity of the discharge electrode. The resultant "space
charge" of the aerosol has the effect of reducing the electric field at the
discharge electrode which, in turn, reduces the corona current density. By
contrast, the space charge increases the field at the collection surface.
Since the theoretical migration velocity is a function of both the particle
charging conditions at the electrode and the precipitating field, it is not
apparent whether space charge actually increases or decreases precipitator
migration velocities.
TABLE 3
PILOT WESP PERFORMANCE
AT CHEMICAL INCINERATORS
Plant 1
Plant 2
COLLECTION
EFFICIENCY
(%)
99.5
93.2
EFFECTIVE
MIGRATION
(cm/sec)
39.4
13.2
SPECIFIC
COLLECTION
~ AREA
MVMVsec
9 and
(FT/1000CFM)
13.4 & (68.3)
20.3 & (103.3)
FIELD
STRENGTH
(kv/cm)
--
5.6
CURRENT
DENSITY
(na/cm )
—
59.1
SIZE DISTRIBUTION
Plant 1
Plant 2
% LESS THAN 1.0 MICRON
90
92
MASS MEAN DIAMETER (MICRONS)
--
0.67
The full-scale WESP performance data gathered at the chemical incinerator
installation, shown in Figure 6, points to the possibility that the increase in
the precipitating field more than compensates for the loss of particle-charging
field and current density. However, as there are no data or calculations to
further support such a conclusion, we must reserve judgment as to the actual
cause of this interesting relationship. It should be noted, however, that the
increase in migration velocity should not be attributed to an increase in
average particle size with increasing inlet load. The full-scale WESP was in-
stalled downstream of a medium-energy venturi scrubber which, by nature, tends
to yield a relatively uniform outlet particle size distribution regardless of
large changes in the inlet distribution.
464
-------
CT)
en
23
20
CM
2E
15
O
_ 10
D
U
O 5
OS
O
u
^^
234
INLET PARTICULATE LOADING (GRAMS/NM3)
Figure 5 Effect of particulate loading on corona current - chemical incinerator,
-------
cri
11
10
ui
V)
u
>
>
z
o
1—
o
UJ
>
}-
u
uy
a
w
©
— ©
1
1234
INLET PARTICULATE LOADING (GRAMS/NM3)
Figure 6 Effect of particulate loading on WESP effective migration velocity.
-------
CONCLUSIONS
The data presented in this paper clearly demonstrate the ability of
the wet electrostatic precipitation process to attain very high performance
levels when operating on high concentrations of fine particles. The data
also demonstrate the dependence of WESP performance on operating voltage,
but refute the notion that very high inlet concentrations necessarily have
a limiting effect on performance due to space charge corona quenching.
467
-------
REFERENCES
1. Mazer, M. R., S. T. Herman, and S. A. Jaasund. Adaptation of Wet
Electrostatic Precipitators for Control of Sinter Plant Windbox
Emissions (presented at the annual meeting of the Air Pollution
Control Association, Toronto, Ontario, June 1976, paper #77-6.1).
2, Szabo, M. S. and R. W. Gerstle. Operation and Maintenance of
Particulate Control Devices on Selected Steel and Ferroalloy
Processes, PEDCO ENVIRONMENTAL CORPORATION, Cincinnati, Ohio,
March 1978, PD 282256.
3. La Mer, V., E. Inn, and I. Wilson. Journal of Colloid Science.
5;471, 1950.
468
-------
TUBULAR ELECTROSTATIC PRECIPITATORS
OF TWO-STAGE DESIGN
By:
Harish S. Surati
Michael R. Beltran
and
Isaac Raigorodsky
Beltran Associates, Inc.
Syosset, New York 11791
ABSTRACT
Although the "two stage" electrostatic precipitator concept was first de-
veloped in 1910, until recently most of their use was confined to in-plant air
cleaning. In the last decade, plate type designs have been modified to make
them suitable for industrial applications involving organic emissions, most not-
ably in Asphalt saturating, Plastic curing, Food processing, Printing, Textile
finishing and Heat treating industries. However, plate type designs are inade-
quate in applications where very high loading and/or high particulate content
are involved.
Tubular design with wider spacing and higher voltages incorporates best
features of both the single stage and two stage type precipitators. These
units have been used in Molybdenum roasting, Zirconium Calcining, Ammonia
scrubbing of oxides of sulfur, Meat broiling, Foundry exhaust, etc.
The paper discusses design parameters, field test data, and operating
data on these units. Comparisons with single-stage type precipitators are made
wherever applicable. Advantages and disadvantages of two stage precipitators
are also discussed.
469
-------
TUBULAR ELECTROSTATIC PRECIPITATORS
OF TWO-STAGE DESIGN
INTRODUCTION
The two-stage electrostatic precipltator concept was first conceived and
patented by Schmidt in 1910. Unlike the single-stage or Cottrel type precipi-
tator, the two-stage precipitator has separate ionizing and collecting sections.
Although the first commercial use of the two-stage precipitator design was re-
ported in 1937, applications were limited to air cleaning in offices, bars
and hospitals. In the past fifteen years design modifications have enabled
the plate type two-stage precipitator to be used in various industrial ap-
plications. The use of two-stage precipitators for hydrocarbon emission con-
trol has gained wide popularity. The two-stage units have virtually replaced
thermal incinerators for emission control in textile, rubber, vinyl, asphalt,
carpet, printing, grain drying and food industries. A detailed discussion of
the use of two-stage precipitators in various applications can be found in
other publications of the authors. >>
Theory and experiment both indicate that particle charging takes place very
rapidly, as compared to their separation from the gas stream. Moreover, the
charging process requires a nonuniform field, whereas a uniformly high field
is required for the most efficient separation of the particles from the air
stream. Both these facts indicate the need for separate ionizing and collect-
ing sections. The treatment time for the industrial single-stage type unit
varies from 1 to 10 seconds. Thus, long residence times are needed for the
separation and subsequent collection of the particles from the gas stream.
Saturation charging can be achieved in 0.01 seconds or less for most industrial
applications encountered. Consequently, in single-stage precipitators the
corona power is wasted over the major portion of the ionizing electrode.
Figure 1 shows a typical arrangement of tubular precipitators, both single-
stage and two-stage design. Table 1 gives some typical comparison figures
for single-stage precipitators and tubular precipitators of two-stage design.
The tubular single-stage units have a large grounded cylinder known as
the collecting electrode and, coaxial with it, a high potential wire called the
discharge electrode. In the two-stage design the precipitator consists of a
short ionizing section followed by a comparatively longer collecting section.
The discharge electrode is in the form of a rod or tube with a sharp needle at
the end and is centered in the tube. Various tube geometries have been utilized
over the years, the most common being the round and hexagonal. The hexagonal
shape is more space efficient than the round shape. The square configuration
shown in Figure 1 is a slight variation of the hexagonal shape and was chosen
because of manufacturing ease. The corona is generated on the needle when
high voltage is applied to the discharge electrode. The whole length of the
rod then acts as a nondischarging electrode still providing the electric field.
This arrangement provides a nonuniform electric field in the ionizing section
and a uniform electric field in the collecting; section.
470
-------
THEORETICAL ASPECTS
The electrostatic precipitation phenomenon can be divided into three main
categories: (1) lonization or particle charging
(2) Collection of the charged particles
(3) Removal of the collected particles
lonization Mechanism:
When electrical potential is applied to the discharge electrode (the rod
and the needle assembly) located in the center of the grounded tube (collect-
ing electrode), an electric field is created that varies from a high strength
near the sharp needle point to a low strength near the grounded electrode.
When the potential difference between the discharge electrode and the collect-
ing electrode reaches the "critical corona voltage", local ionization of the
air stream surrounding the needle occurs and an onset of blue corona can be
observed on the needle point tip. As the potential is increased, the corona
glow increases in strength and volume. In some cases when sufficiently high
voltages are applied to the thin needle, a corona sheath can be observed on
the whole shank of the needle. Such corona was observed with the present de-
sign at AOkv using 1/16" diameter needle. The corona was observed to be of
the inverted umbrella shape, with the needle as the shaft of the umbrella.
When the applied potential reaches sparkover voltage, complete ionization of
the air stream occurs, resulting in a short circuit.
With the discharge electrode at the negative potential, which is the
usual arrangement in most industrial applications, positive ions generated are
attracted to the needle and negative ions to the grounded tube. Although an
equal number of positive and negative ions are generated in the corona region,
99 percent of the gas space between the needle and the tube is filled with only
negative ions. This forms a space charge of high opposing voltage which limits
the corona current.
Particle charging occurs by way of two known mechanisms: (a) field
charging, wherein particles are charged in the corona region by the bombard-
ment of negative ions moving toward the grounded tube in the applied field, and
(b) diffusion charging which occurs because of the random motion of the particles
themselves. The diffusion charging mechanism is relatively insignificant for
particles above 1 micron in diameter. Sub-micron particles have significant
random motion and, hence, the diffusion charging mechanism is important to
achieve the saturation charge. Typical charge values are 300 electron charges
for one micron and 30,000 electron charges for 10 micron particles. This satu-
ration charge is achieved in less than .01 seconds. In tubular two-stage design,
an ionization residence time of between .02 to .05 seconds is desirable. In con-
ventional single-stage design, a charging time of between 5 to 10 seconds is pro-
vided. This results in wasted corona power. In practice, power requirements
for single-stage precipitators are much greater than indicated by theoretical
calculations because only a small fraction of the power is used to transport
the particles. The theoretical energy requirements may be approached in prac-
tice for two-stage designs. The two-stage design requires less than one-third
of the power of the single-stage design for equal efficiency.
471
-------
Various tube and discharge electrode geometries have been tried in the past.
The relation between the collection efficiency and the corona current is very
unclear. It does not conform to any set trends. lonization produced by a sharp
needle point was found to have the most efficient utilization of the corona cur-
rent. These findings are presented in graphical form in Figure 2. An exten-
sive research program is underway to determine the effect of tube geometry and
discharge electrode spacings. It is believed that more than one discharge
point in the horizontal plane produces overlapping corona fields. The over-
lapped region would have markedly decreased concentration of negative ions, re-
sulting in decreased overall efficiency. Addition of the electrical discharge
points in the direction of the airflow was found to marginally increase the ef-
ficiency.
Collection Mechanism:
The particles charged in the corona region surrounding the needle enter
the collecting section. The collecting section is composed of a large positive
(grounded) tube and a nondischarging rod at a high negative potential. The
particles are repelled by the rod and attracted by the tube wall. The parti-
cles are accelerated toward the collecting electrode by the Coloumb force and
this motion is resisted by inertial and viscous forces. Since, for fine par-
ticles inertial forces are insignificant, the particle attains a velocity deter-
mined by the equilibrium between the Coloumb and Stokes forces. This velocity
is known as the migration or drift velocity. The typical drift velocity in the
tubular two-stage type precipitator ranges between 0.3 to 0.7 ft/sec for sub-
micron particles. It is a function of collecting field strength, gas medium,
gas temperature and pressure, and the nature and size of the particles.
In tubular precipitators of both single-stage and two-stage design, the
most critical cross-section is that at which the discharge electrode enters the
collecting electrode. In this plane the electric field is highly concentrated
and is distorted by edge effects. In order to avoid sparkover in this region,
various methods have been tried, the most notable being the "necking down"
of the electrode at that point. Structural and mechanical considerations limit
the amount of necking that can be employed.
Extensive work done by Dr. Masuda in Japan has shown that particle migra-
tion velocity in the tubular precipitator increases with increase in inter-
electrode spacing. A variety of explanations have been offered, most of them
involving a redistribution of the electric field. In this explanation, if the
current is held constant as the interelectrode space increases, the total space
charge in the gap also increases. The space charge, being of the same polarity
as the discharge electrode, more effectively shields the discharge electrode
from the tube. Thus, the field is reduced near the discharge electrode and, in
compensation, is raised near the grounded electrode. The intensified field near
the wall is more effective in charging particles.
Removal Mechanism:
Tubular precipitators are mainly used as mist precipitators and, hence,
do not have reentrainment problems if the precipitator throughput velocity
is kept below 7-8 ft/sec. If the collected material is not very viscous, it
472
-------
drains by gravity. Most of the industrial applications where two-stage tubular
precipitators are used have saturated exhaust streams, and the high moisture
content provides continuous flushing of the collecting surface.
Precipitator Efficiency:
The exponential Deutsch-Anderson equation is extensively used in practice
for the design of the single-stage precipitators. Field experience has shown
that, with little modification, the same equation can also be applied to two-
stage precipitator design. The more refined expression developed by Penny
(1937) for the two-stage precipitator is not used in industry.
In its simplest form, the Deutsch-Anderson equation is given as:
Eff = 1- exp (-A/Vw) (1)
The collecting electrode area, A, and volumetric flow rate, V, are calculated
from the known geometry. The drift velocity, w, is generally determined from
data collected on similar applications.
From Figure 3 for duct precipitators,
A = 2Lh = L_
V 2Shv Sv (2)
So,
EffDp = 1- exp (-Lw/Sv) (3)
From Figure 3, for tubular precipitators, (single-stage design)
A = 2RL = 2L
V ^ R2V Rv
Also from Figure 3, for tubular precipitators, (two-stage design)
A = 4(2RL) = 2L
V 4RZ v Rv (5)
Therefore,
EffTp = 1- exp (-2Iw/R ) (6)
Comparison of Equations (3) and (6) indicates:
(i). For a given efficiency a tubular precipitator may be operated at
twice the gas velocity of a duct precipitator of equal electrode length and
interelectrode spacing.
(ii) . For a given gas flow (volumetric flow rate) efficiency of the duct
precipitator is independent of duct spacing.
473
-------
(ill). For a given volumetric flow rate, efficiency of a tubular precipi-
tator increases with increase in tube diameter.
DESIGN ASPECTS
Polarity:
Electrical energization is the single most important design consideration
in the two-stage precipitator design. The two-stage precipitator developed by
Penny used positive polarity to reduce the amount of ozone produced. This
practice is still continued in air cleaning units. However, units designed
for industrial duty use negative ionization to achieve higher sparkover volt-
ages. The fouling of the discharge electrode is less, in the case of negative
ionization, compared to positive ionization.
Voltage Waveform:
Unfiltered voltage waveforms are generally used in the single-stage de-
sign, both plate and tube type. Unfiltered waveforms allow for higher peak
operating voltages and reduced intensity of sparks. The plate type two-stage
units generally use filtered or low ripple waveforms. Although the particle
approaches a higher final value for the unfiltered waveforms, the effective
charging time constantly increases. This is because the particle accepts a
charge only during the time interval in which the applied electric field, in the
vicinity of the particle, exceeds the self field from the charged particle.
Two-stage tubular precipitators are operated at comparatively high volt-
ages compared to its plate type counterpart and, hence, it is not appropriate
to lable them as "low voltage" precipitators, the name commonly used for the
plate type units of two-stage design. The high potential and a short ioniza-
tion zone make it virtually mandatory to use sophisticated silicone controlled
rectifier (SCR) automatic controllers to keep the operating voltage at the
maximum. The capacitance of the tubes themselves is used to smooth the wave-
form. The use of a three-phase power supply with Saturable Core Reactor (SR)
is also under investigation. The typical power supply for the tubular units
have ratings of about 10KVA and, hence, the use of external capacitors to re-
duce the ripple is not advisable. The intensity of a spark, in such a case,
would be deleterious to the tube material.
The shift in phase angle, or more appropriately the conduction angle, be-
cause of the space charge effect, can also alter the shape of the waveform
significantly. Space charge occurs with a medium concentration of fine fumes
or with a heavy concentration of large particles. The particles being charged
to the same polarity tend to reduce the electric field at the discharge elec-
trode and, in some cases involving very high concentration, may reduce the
potential to the incipient corona starting voltage. The corona current is then
reduced by a factor of 100 or more. Incorporation of a variable inductance or
a resistor in the primary circuit is necessary to reduce the effect of the
phase angle shift.
474
-------
Sectionalization:
Sectionalization of electrical energy is a very important considera-
tion in the design of a precipitator. The expression for sparkover volt-
age for negative discharge electrodes energized by a single power supply is
given by:
Vs,n = Vs,l -C In n (7)
The reduction in voltage, Cln n, is because: (i) The instantaneous poten-
tial of all parallel-connected discharge electrodes is set by the reduced
potential of whichever electrode in the system is experiencing sparkover at
that moment and, (ii) The greater the total discharge electrode length in the
section, the higher is the probability of the occurrence of conditions which
tend to lower the sparkover potential. This problem can be circumvented,
partially, by introducing adequate resistance in each leg of the parallel-
connected discharge electrodes. The other obvious solution is to reduce the
number of discharge electrodes energized by a single power supply. The economic
considerations preclude a high degree of Sectionalization in spite of its ob-
vious advantage. Tubular units generally use one power supply per 20,000 CFM
per pass. Several sections in series are necessary where high current suppres-
sion conditions due to space charge is observed.
Velocity Distribution:
Velocity through the precipitator is the only flow parameter involved in
the Deutsch equation. The Deutsch equation assumes plug flow velocity distri-
bution. In practice, however, this condition is very difficult to achieve.
Nonuniform gas flow can affect the precipitator performance by scouring the
plates in the localized regions of high gas velocity and by reducing perform-
ance due to the exponential relation between efficiency and the gas flow.
Poor gas flow can decrease the precipitator efficiency as much as 20 to 30%.
For single-stage units, 1/10 or 1/16 scale model studies are usual. Such ex-
pensive studies are rarely undertaken for the two-stage units. IGCI gas flow
quality criteria or their equivalent are generally required in the single-stage
units. Although such requirements are not included in the specifications for
two-stage units, well engineered two-stage system designs generally adhere to
them. Two-stage units are smaller in size and do not have much internal
structural framework to disturb the air flow quality. Various air distribu-
tion devices are used in the two-stage units. A well engineered equipment lay-
out can also materially contribute to the efficient air flow distribution in
the precipitator.
INDUSTRIAL APPLICATIONS
Two-stage design tubular electrostatic precipitators are ideally suited for
the control of the sub-micron organic mist and inorganic acid mist emissions.
High collection efficiency in sub-micron particulate size range and the light-
weight compact design make the two-stage tubular precipitator ideal for such
applications.
475
-------
Organic emission occurs when high boiling point organic compounds volatilize
and evaporate in any part of an industrial process. Organic impurities are
generally found with most of the ores. More organic matter is introduced during
floatation operation. Lubrication oils, cutting oils, mineral oils, etc., are
also sources of organic matters. When materials containing organic compounds
are roasted, calcined, sintered, or reduced at high temperatures, organic emis-
sions occur. When the fumes come in contact with cold outside air, the hydro-
carbons condense to form fine aerosol. Condensation phenomena produces sub-
micron size particles. Very fine mist is also produced during gas phase
chemical reaction where the product of reaction has very low vapor pressure at
the reaction temperature and, hence, condenses from the gas stream. The reac-
tion of sulfur trioxide and water vapor to form sub-micron sulfuric acid mist
is probably the best example of fine mist formation during gas phase reaction.
The sub-micron particle size range is optimum for scattering the light,
hence, a very small weight fraction produces highly visible plume. Moreover,
these fine particles can also bypass the body's respiratory filters and pene-
trate deeply into the lungs. Because of their high surface area, some fine
particles have been identified as transport vehicles for gaseous pollutants,
both absorbed and reacted and, hence, can produce synergistic effects dele-
terious to human health.
Typical Applications:
Tubular two-stage type units have been tried in several industrial appli-
cations involving very tough corrosion problems. A special conductive fiber-
glass resin was developed to combat the problem. This eliminates the need for,
and also the problems associated with, maintaining water film on the collecting
electrode. The tubular design lends itself easily to various types of materials
of construction.
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 two-stage tubular
unit treats the exhaust stream before venting it to the atmosphere. The ex-
haust stream entering the precipitator contains organic compounds and sulphuric
acid mist with some oxides of molybdenum, sellenium and mercury. Chlorides and
trace amounts of fluorides are also present in the air stream. Tantalum is the
only metal that could be used as material of construction. Fiberglass rein-
forced plastic (FRP) with synthetic fiber veil on the inside surface proved to
have excellent corrosion resistant characteristics in this service also. Col-
lection efficiency in excess of 99 percent was obtained.
Pulp and Paper Industry:
Recovery of spent sulfite liquor is widespread in the pulp and paper in-
dustry. The spent sulfite liquors are burnt to recover the sulfur values.
The scrubbing of the oxides of sulfur is accomplished by the use of ammonium,
sodium, calcium or magnesium base liquors. A significant amount of fine partic-
ulate matter is generated during this absorbing operation. The exhaust plume
476
-------
from the scrubber is thus highly visible. The particle size is extremely fine,
between 0.1 to 0.5 microns in diameter. The particles are partially insoluble
and, hence, plug the fiber bed type devices used to control these emissions.
Moreover, considerable energy is spent to overcome the gradually increasing
pressure drop through the fiber bed. Unavailability of suitable and economical
materials of construction hindered the use of electrostatics in these applica-
tions. The conductive FRP construction was tested for corrosion resistance in
this service. No measurable signs of corrosion were observed after a few months
of tests. Collection efficiencies in excess of 98 percent were obtained under
the entire range of process conditions. Throughput velocities of 6 to 7 fps
were found to be optimum. The collection efficiency was measured by EPA method
5 and also by a forward light scattering photometer. Excellent agreement was
found between these two measurements. Although the inlet particulate concen-
tration of 0.6 grain/CF is not heavy enough to cause current suppression, the
secondary current was reduced by a factor of three when exhaust gases were passed
through the precipitator. This is believed to have been caused by extremely fine
particle size (.1-.5 micron) and enormous moisture loading (saturated stream at
160 F). Additional fogging was not required to keep the unit clean.
Zirconium Calcining:
Zirconium and Hafnium are used in the nuclear industry for the full rod
casings. These metals are immune to corrosion attack from most of the chemicals
and can withstand very high temperature. Neutrons pass through Zirconium,
whereas they are absorbed in Hafnium. Thus the nuclear reaction can be con-
trolled by use of Zirconium and Hafnium tubes. Zirconium and Hafnium are mined
as Zircon sand. The ore is chlorinated, selectively precipitated, passed
through separation 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 gases are highly corrosive to most of the metals. Zirconium or FRP
are acceptable materials of construction. The exhaust is first treated in a
soda ash packed bed scrubber to reduce oxides of sulfur. The tubular two-
stage precipitator is used to remove fine particulates and acid mist. Severe
current suppression is observed because of high moisture loading. The collected
liquor washes off the tubes as it drains making the unit self-cleaning.
!>
Metal Industry:
Primary and secondary metal production usually involves smelting of the
ore in a reduction furnace. Sulfur oxides generated during this operation have
to be treated before they can be vented to the atmosphere. Sulfur dioxide con-
centration in most of the smelter gases varies from 6 to 14 percent. The con-
tact process for the manufacturing of sulfuric acid requires 7 to 14 percent
concentration of sulfur dioxide for the most efficient conversion to sulfur
trioxide. The smelter gases are thus ideal for a sulfuric acid production.
However, metal oxide impurities and acid mist from the exhaust stream has to
be removed to prevent black acid formation. Tubular precipitators are ideal
for this application. Wet electrostatic tubular lead precipitators have been
used most extensively for this application. FRP tubulars are less costly,
477
-------
easier to construct and maintain, and are equally corrosion resistant. A very
high collection efficiency can be obtained with a two-stage tubular design.
The tubular two-stage unit can also be used to control the emission from the
interpass tower or the final absorber tower in the contact acid plant train
using spent acid as the source of sulfur. FRP construction is not good if
more than 50 percent concentrated acid is being handled. ASTM 316 steel or
carpenter-20 alloy might be required for such applications.
The tubular two-stage units can also be used to control emissions from
sinter plants, forging, casting, foundry exhaust, meat broiling exhaust,
scarfing operations, etc.
CONCLUSION
Tubular precipitators of two-stage design have a very high collection ef-
ficiency in a sub-micron region. The compactness and ease of manufacture of
these units make them ideally suitable for corrosive acid mist applications
involving particulate emissions. New and different applications for the two-
stage tubular units are numerous and are vigorously being explored. Increased
use of these units in applications having high resistivity problems is also
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 M. R. Beltran. Heat Recovery On Organic Electro-
static 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. Penny, G. W. A New Electrostatic Precipitator. Elect. Eng., January
1937. pp 159-163
5. Harrington, R. E. JAPCA, vol. 24 pp 927-931
6. White, H. J. Industrial Electrostatic Precipitations. Addison-Wesley
Publishing Company, Inc., 1963. pp 180
7. Hall, H. J. Technical Report HAR79-218.
478
-------
TABLE-1 TYPICAL OPERATING DATA7
Single-Stage Two-Stage
Ionizing field, KV/in. 13.3 16.67
Collecting field, KV/in. 13.3 25
Tube dimension 9 in 0 3x3 in. square
Current consumption MA/1000 CFM 15-30 5-10
Power consumption, watts/1000 CFM 400 125
SCA, ft2/1000 CFM gross 400 100
Treatment time, sec. 4-10 1-2
479
-------
A-A'
r
r
E>
COLLECTION
SECTION
N
tOMZATlOM ]
SECTION
J
xl
ft
(
f
V
u.
— — * *
-^-^
T
••»%. ,
h—s
I
1
~l
B
SINGLE STAGE
TWO STAGE
FIG. I
480
-------
f UEEDLE
DISC
8NEEDLES
SZ
10 IS tO
MA/ )000 CFM
. 2
481
-------
k
A
I
(a)
I I
u
(c)
(a) Single-Stage Duct Precipitator
(b) Single-Stage Tubular Precipitator
(c) Two-Stage Tubular Precipitator
FIG.3
482
-------
PRESENT STATUS OF WIDE-SPACING TYPE PRECIPITATOR IN JAPAN
By:
Senichi Masuda
Department of Electrical Engineering
Faculty of Engineering, University of Tokyo
7-3-1, Kongo, Bunkyo-ku, Tokyo, Japan 113
ABSTRACT
The wide-spacing type electrostatic Precipitator (WESP) has
aquired an established position in Japan in the recent decade,
and 160 WESPs had been installed until the end of 1978 by 10
manufacturers in almost all the application fields. The
collection performance remains unchanged with increasing duct
spacing up to ca. 400 - 600 mm or more at a constant precipitator
volume, so far as the applied voltage be raised correspondingly.
As a result, WESP results in a substantial reduction of
installation cost in case of a large gas volume, amounting to 15
- 20 % under the most favourable conditions. Moreover, the
stability in operation is largely increased, so that the initial
high collection performance can be maintained. The maintenance
work becomes also much easier. WESP-design can be applicable
also in the wet ESPs and the novel type ESPs recently developed.
483
-------
PRESENT STATUS OF WIDE-SPACING TYPE PRECIPITATOR IN JAPAN
1. INTRODUCTION
The wide-spacing type electrostatic precipitator (WESP)
using a duct spacing of 400 - 600 mm or more, substantially
larger than the conventional spacing of ca. 250 mm, represents
one of the largest technical impact occured in the recent decade
in the field of ESP. In Japan 160 WESPs had been installed
until the end of 1978, and now WESP has aquired an established
position. In the case of high performance precipitators, common
under the present severe emission standard, the collection
efficiency remains almost unchanged even if the duct spacing, D,
is increased from its conventional value of D = 250 - 300 mm up
to a large value of D = 400 - 600 mm or more, so far as the
applied voltage be raised correspondingly. This effect, hard to
understand from the classical Deutschian concept, should be
refered to as the "WESP-ef f ect". At any rate WESP produces a
large economical advantage over the conventional spacing ESP in
the case of medium or large gas volume. Moreover, the stability
in operation is largely increased, and the the maintenance work
becomes much easier. WESP-design can also be successfully
applied in the wet ESPs and the novel type ESPs recently
developed.
WESP in Japan has originated from two different sources: one
being Nippon Kai Heavy Industries Co., Ltd, and another being
Ninon Kogei Co., Ltd. Nippon Kai started the commercialization
of WESP with the license of R.F. and D.O- Heinrich. Its first
WESP was installed for a cement clinker cooler in October 1971,
and showed a spectacular performance with no visible stack
emission (Figure 1). Ninon Kogei started from a completely
different design, using a large horizontal grounded duct with
more than 2 m in diameter as a collecting electrode and a very
fine horizontal wire with less than 0.5 mm in diameter spanned
along the duct axis as a discharge electrode. The voltage to be
applied was 200 kV.
The advantages of WESP, both economical and technical, have
been widely recognized by the process operators and manufacturers,
and 160 WESPs had been installed until the end of 1978 by 10
manufacturers in Japan, as shown in Table 1. In this paper is
described the present status of WESPs in Japan.
2. THEORETICAL EXPLANATIONS AND EXPERIMENTAL RESULTS
The WESP-effect caused a big controversy on the Deutschian
concept of precipitation itself. The dust migration velocity, w,
in the Deutsch equation has long been considered as a process
parameter specific to each gas and dust condition and independent
484
-------
of D. However, the existence of WESP-effect requires the
migration velocity, w , to rise proportionally with D. Many
theories have been presented to explain this effect, whereas a
number of experiments performed to examine this effect in more
details.
Heinrich proposed recently a theoretical explanation based
on the classical Deutsch equation and the approximate expressions
of corona field intensity (Heinrich (1978) ). According to this
theory, w could rise proportionally with D, provided the total
corona current be kept unchanged.
T. Misaka et. al. made thorough experimental investigations
on the test WESPs in air at NTP using fly ash samples and
changing D from 250 to 750 mm ?with the inter-wire spacing kept
constant (Misaka et. al. (1978) ). The applied voltage, U, was
changed proportionally with D so that the average inter-electrode
field intensity, Ea = U/D, could be kept constant at Ea = 4.0 and
3.2 kV/cm. In this case, the average current density per unit
collection area, i, remained almost constant independent of D.
The solid curves in Figure 2 represent the magnitudes of w
derived from the collection efficiencies at these two Ea-values
and plotted against D. w shows a rise slightly less than
proportional to D. They further measured the space distribution
of field intensity, E, inside the corona space using the dynamic
probe method. The magnitude of E at the surface of collecting
electrode, Es, increased with D even though Ea was kept
unchanged, evidently due to the space-charge effect. Misaka et.
al. explained the WESP-effect from this phenomenon, considering
the collection process of very fine particles to be govererned by
their Coulombic transport in the boundary layer region. If the
values of Es measured at different Ds are used for E in the
expression of the migration velocity, w = k E , the dotted
curves in Figure 2 are obtained, where the constant c was
derived from the data at D = 250 mm. The ageement between the
solid and dotted curves is quite satisfactory, providing some
support to their explanation.
S. Matts presented a different explanation, consider-
ing the fact that the WESP-effect appears only in the case of
very high collection efficiency, say 99 %, where the reentrainment
loss plays a major role in collection performance (Matts
(1978) ). This loss will be lowered in WESPs because of the
decreased number of collection plates.
P. Cooperman presented a non-Deutschian expression for the
collection efficiency assuming the existence of reverse transport
of dust particles due to turbulent diffusion (Cooperman (1976) ).
He pointed out that the WESP-effect could be derived from this
new expression.
0. Gtipner, H. Lau, R. Aureille et. al., and St. Nibeleanu
also presented the experimental results supporting the existence
485
-------
of the WESP-effect under suitable conditions (Gu^pner (1976) ),
(Lau ,£1969) ), (Aureille and Blanchot (1971) ), (Nibeleanu
(1979)°).
The author and his co-workers carried out the field tests of
wet WESP at a dry-process cement rotary kiln (Ago et. al.
(1975) ). The applicable voltage at D = 250 mm was considerably
low at U = 40 kV/cm (Ea = 3.2 kV) owing to a large sparking
tendency. This was due to the Taylor cones of water appearing
at the lower edges of the collection plates. By doubling the
duct spacing to D = 500 mm, the operation became completely
stable, and U = 100 kV could be applied without any sparking,
resulting in the increase in Ea up to 4 kV/cm. The magnitude of
w could be more than doubled (2.16 times) by doubling D at Ea
unchanged (3.2 kV/cm) . By increasing Ea up to 4.0 kV/cm, the
further rise in w (2.51 times) could be obtained. In addition,
the savings of irrigation water and its treatment system were
quite substantial.
Noso et. al. carried out.recently the thorough field tests
on WESPs (Noso et. al. (1978) ). Two large series of tests
were performed: one using a large test precipitator located at
the research station, and the other using a pilot plant located
at a cement mill. In the first series of test, where the sample
dusts were injected into the oil-burned gas, the electrical
characteristics and collection performance were measured at D =
250 - 800 mm under different corona current density, i, and inlet
dust loading, Ci. The sparking voltage, V showed a linear
rise with increasing D in air load condition. Whereas, in dust
load condition, V at a very large duct spacing D = 800 mm
showed a substantial, drop at C. beyond 7g/Nm owing to excessive
sparking caused by the space-charge effect. Figure 3 indicates
the effect of i on w, where w is normalized by its value at i =
0.4 mA/m , w._n ,, and reduced to D = 250 mm as
1~ \J • ^f
w* = (w/wi=0>4)/(D/250). (1)
3
The value i = 0.4 mA/Nm is considered as the current threshold
beyond which a satisfactory collection performance can be
expected under the normal operating conditions. The curves of
w coincide well each other, and rise with i up to a certain
value. The more than proportional rise of w with D can be seen
from this figure, which tendency is enhanced with the increase in
i, but becomes saturated at i = 0.6 mA/m owing to the increased
sparking,, The optimum current density lies in the gange of 0.4 -
Oft6 mA/m independent of D. Figure 4 shows the effect of C. on
w , where the magnitude of w at C . = 5 g/Nm , wr-_Q 5>* is used
instead of w-_0 A for normalization in Equation (i). w at D =
250 mm continues to increase with C., whereas that -for WESP shows
the saturation in its increase^ beyond C. = 5 g/m . Figure 5
indicates the effect of D on w at different C.-values, where w
3_ j
at D = 250 mm, wn = 250' ^s usec* f°r normalization. Again, the
more than proportional WESP-effect is shown to exist in the low
486
-------
3
C. range below 5 g/Nm , which, however, becomes saturated at D =
600 mm. Noso et. al. came to the conclusion that the optimum
value of D in view of both stability in electrical operation and
collection performance seemed to lie in the range of D = 400 -
600 mm, and that the minimum overall cost calculated from the
economical assessment also coincides with the same duct spacing
range. In the second series of test the duct spacing was set at
both D = 250 mm^and D = 400 mm, and the effects of gas velocity,
v, and C. on w were investigated. Figures 6 and 7 show the
results obtained. The effect of v is the same for both D = 250
mm and D = 400 mm. The effect of C. is #lso the same except for
the very high range of C... beyond 601g/Nm . From these and many
other results, Noso et. al. concluded that the design data
obtained at the conventional spacing ESP could be well transfered
to WESP through the concept of normalized migration velocity.
Finally, R. Ito and K. Takimoto presented recently a method
of estimating the collection efficiency of WESP from the data of
the conventiona^ spacing ESP and dust to be collected (Ito and
Takimoto (1977) ). They expressed the fractional loss out of
WESP as
Ft = 1 - y = (h(d±) - h(di+1)) . exp(-w(D,di).f) (2)
where y = fractional collection efficiency, d. = i-th diameter of
particle, h(d.) = dust cummulative mass fraction above d.,
w(D,d.) = dust migration velocity for d. at D, and f = specific
collection area. If the magnitude of w"\D,d.) is proportional to
D, it can be expressed, using its value at the conventional
spacing ESP with D = 250 mm, wD=25o^d^' as folJLows:
w(D,di) = (D/250) . wD=25Q(di) (3)
Hence, if w ,. (d.) is known from the data of the conventional
spacing ESP and h(cl.) measured for the dust to be collected, the
overall collection performance can be immediately calculated from
Equation (2). Ito and Takimoto confirmed that this procedure
could provide a satisfactory estimation for performance of WESP.
3- CHARACTERSITIC FEATURES OF WIDE-SPACING PRECIPITATOR
From the field experiences thus far obtained, the characteri-
stic features of WESP can be summarized as follows:
(1) The technically optimum value of duct spacing lies usually in
the range of D = 400 - 600 mm. The corresponding voltage is 80
- 150 kV.
(2) The increase in D at a constant precipitator volume produce
the cost reductions in the discharge and collecting ellectrodes,
rapping elements, construction members and ground, whereas it
causes, through the rise of voltage needed, the cost increases in
487
-------
the power pack, insulators, bushings etc. It also causes the
increase in dead space owing to the large gas gap needed for
insulation and, therefore, the quantity of heated cleaning air to
be purged into the insulator boxes. The magnitude of D
providing the maximum economical advantage, therefore, depends
largely on the precipitator size. In case of small sized ESPs,
the weight of cost increase due to increased voltage becomes
dominant, so that the economical advantage of WESP appears only
in the medium or large sized ESPs with a gas volume beyond, say,
1,000 m /tnin. The cost reduction by 15 - 20 % can be obtained
in large precipitators under most favorable conditions. The
economically optimum value of duct spacing lies in the range of D
= 400 - 600 mm, which coincides with the technically optimum
value cited above.
(3) WESP has a large stability in operation owing to much less
sparking tendency. This is because the relative variation of
the inter-electrode gap due to deformatios of electrode and
construction members as well as irregular dust deposit becomes
largely diminished, A high voltage and, therefore, a high
collection performance at the start of operation can be
maintained during its entire period.
(4) Ease in the maintenance work enabled by the increased gap is
its another large technical merit (see Figure 8; D = 640 mm)).
(5) The davantages of WESP are preserved also in the case of wet
precipitators. In this case, the stability in operation becomes
much more pronounced. In addition, the quantity of irrigation
water needed and the capacity of the water treatment system are
reduced. The lowering of gas temperature is decreased owing to
reduced contact area to water film..
(6) WESP-design is especially suited to the Roof-Monted—
Type precipitator, because of its reduced total weight.
(7) WESP is more sensitive to the dust space-charge effect
causing the reduction or quenching of corona current and
excessive sparking especially in the inlet field. This is
because the space-charge effect is a size effect to be enhanced
by the increase in D. Among two factors enhancing this effect,
the high inlet dust loading and the fine dust particle size, the
latter proved to have much more detriorating effect, because the
fine particles cannot be quickly collected and remain suspended
for much longer time. Therefore, when the very fine dust or mist
with a considerable mass loading is encountered, the decrease of
D or use of the conventional duct spacing value becomes
imperative. On the contrary, no reduction of D is needed in the
case of a large mass loading of coase dust.
(8) The effect of back discharge on the performance of WESP is
one of the most controversial problems. Back discharge is
observed to occur also in WESPs collecting high resistivity
488
-------
dusts, and the size of WESP has to be more or less increased
accordingly. ^Jn case of the sinter main dust of steel plants
with ca. 10 ohm-cm resistivity, an excellent collection
performance could be obtained by the WESPs having a considerably
large size, whereas, in case of very high resistive low sulphur
fly ash, somewhat pessimistic prospect is suggested by the tests
so far made. Anyway, the economical advantage of WESP seems to
become more or less diminished in the case of high resistivity
applications.
5. DESIGN OF WIDE-SPACING TYPE PRECIPITATORS
WESP is mostly constructed in the vertical plate design
identical to the conventional ESP (Figure 1). However, the
horizontal and vertical cylinder geometries, both single and
multiple, are also used in a few special cases. Figure 9
illustrates the multiple horizontal duct construction named "ESCS
(House Type)" (Electrostatic Space Cleaner - Super) which has a
house-like cross-sect ion and a Yfry large duct spacing at D =
1,200 - 1,800 mm (Masuda (1978) ). It has been developed by
Nippon Steel Corporation with the license of Nihon Kogei, and
preserve well its original design. The House-Type ESCS,
although showed an excellent collection performance for very high
resistive dusts from the ore sinter machine, proved to be fairly
expensive. As a result, it has been modified now to the
ordinary vertical plate construction named "ESCS (Straight-Type)".
Figure 10. shows the photograph of a huge Straight-Type ESCS
(30,000 m /min) installed at a sinter machine of Nippon Steel
(Wakamatsu Works; sinter main gas). The typical data of ESCS
are given in Table 2. The vertical multiple duct construction,
as shown in Figure 11 is also used in the case of medium gas
volume.
The high voltage power pack is mostly mounted on the top of
WESP, and the conduit type connection is used for feeding of
current at beyond 150 kV. In case of lower voltage, the high
voltage cable is also used for feeding. The voltage in the range
of 80 - 150 kV is preferably used in most of the applications at
D = 400 - 600 mm because of the reasons previously described.
The use of voltage as high as 200 kV is now restricted to a few
special applications with D beyond this range.
The magnitude of D used in WESP differs considerably from
one manufacturer to another. Some manufacturer uses D = 500 mm
as a standard design in all of the fields, and uses D = 400 mm
only in the case when the space-charge effect is substantial.
Another manufacturer uses a narrower duct spacing at the inlet
field, considering the space-charge effect, and D = 400 - 800 mm
in the outlet field: D = 400 - 500 mm for smaller gas volume to
lower the cost of power pack, D = 600 - 700 mm for larger gas
volume to lower the electrode and construction costs, and D = 800
mm in the high resistivity applications.
489
-------
The estimation of the necessary specific collection area, f,
can be made from the data in the conventional spacing ESPs using
the methods already described.
6. APPLICATION FIELDS OF WIDE-SPACING TYPE PRECIPITATORS
The application fields of WESP cover almost all the areas
where the conventional spacing ESPs, both dry and wet, have been
used (see Table 1), such as:
(1) Cement industry: rotary kiln, clinker cooler, clinker mill
and dryer.
(2) Steel industry: sinter machine, sinter cooler, electric
furnace, basic oxygen furnace and open hearth furnace.
(3) Metal and mining industry: melting furnace, smelter furnace,
holding furnace, sinter-machine, sinter-cooler, electric furnace,
casting furnace, rotary kiln and sulphuric acid plant.
(4) Chemical plant: sulfuric acid plant, spray dryer and
inc inerator.
(5) Glass industry: glass melting furnace.
(6) Oil refinery: FCC plant.
(7) Electric power industry: oil-fired boiler, coal-fired boiler
and desulfurization plant.
(8) City gas supplier: coke oven and coke mixing room.
(9) Firebrick industry: baking furnace.
(10) Municipal incinerator plant.
(13) Automobile tunnel.
WESP-design have also been successfully applied in the novel
types of precipitators recently developed, including:
(a) Hybrid-Type ESP ( Onoda Cement Mfg. Co., Ltd.): A small wet
field is located after one or two dry fields in a-common casing
to prevent the rapping loss (Masuda et. al. (1976) ). The wide
duct spacing are used in both the wet and dry fields.
(b) Bias-Controlled Pulse Charging ESP (IHI Heavy Industries Co.,
Ltd.): A very large duct spacing at D = 600 mm is used in the
tri-electrode fields (Masuda (1978) ).
(c) CEEP-Type ESP (Nippon Kai Heavy Industries, Co., Ltd.): The
collecting electrodes consist of the water-cooled square pipes
490
-------
and comprize the mechanical scrapers moving vertically to remove
dust deposit (Figures 12 and 13). The very high resitive dust
from FCC process of an oil refinery plant is collected effectively,
and the heat recovered from the cooling water is used for
preheating of boiler feed water. This economical gain amounts to
more than 32,000 $/year. The second CEEP has started its
operation recently at a cement rotary kiln.
(d) Roof-Mounted-Type ESP (Sumitomo Heavy Industries Co., Ltd.;
Sumitomo Metal Mining Co., Ltd.; Hitachi Plant Enug-ineering &
Construction Co., Ltd.): See (Nomura and Sakai (1977) ) and (Ito
and Takimoto (1978) ) .
(e) RCCES-Type ESP (Rotary Cylindrical Collecting Electrode with
Scraper) (Koyo Iron Works & Construction Co., Ltd.): A large
rotary cylinder with 1.5 - 2.0 m in diameter and 10 - 20 m in
length serves as a collecting electrode, and a number of long
needles (40 cm in length) planted radially on , the center rod
serve as the discharge electrodes (Isahaya (1978) ) (Figure 14).
A chain scraper is attached on the entire length of the bottom of
the cylinder (Figure 15). The rotation of the cylinder brings
the deposit on its inner wall to its bottom position where the
deposit is mechanically removed. The applied voltage is 200 kV.
The very viscous or sticky tar fume (50 - 300 Englar degree) can
be effectively collected.
7. CONCLUSION
WESP has aquired an established position now in Japan, and
is expected to be more widely accepted in future. From the
experiences of WESPs, amounting to 160 in number at the end of
1978, the following conclusions are derived:
(1) WESP can be used in almost all the applications of the
conventional spacing dry and wet ESPs.
(2) The optimum duct spacing lies in the range of 400 - 600 mm
depending upon the gas volume, emission level required, and
concentration and resistivity of dust to be collected.
(3) The voltage of 80 - 150 kV is preferably used in most of the
applications. The use of a higher voltage as 200 kV is
restricted in a few application where a very large duct spacing
is needed.
(4) Its economical advantage appears only in the medium or large
sized precipitators with the gas volume beyond, say, 1,000
m /min, and it becomes more pronounced with increasing size.
The cost reduction to be obtained amounts to 15 - 20 % under the
most favorable conditions.
(5) WESP has an excellent stability in operation, so that its
491
-------
high initial performance can be maintained throughout the whole
operation period. WESP also has an advantage of ease in
maintenance work.
(6) WESP is affected by the dust space-charge much more
sensitively than the conventional spacing ESP. The very fine
particles with a considerable mass loading causes the most
detrimental space-charge effect, so that the use of the smaller
duct spacing is imperative in this case.
(7) The economical advantage of WESP is likely to be diminished
in the case of high resistivity applications.
(8) The design of WESP can be made from the data obtained in the
conventional spacing ESP.
(9) WESP-design can also be successfully applied in the novel
type ESPs.
REFERENCES
1. Heinrich, D.O. Der grosse Gassenabstand im Elektrofilterbau.
Staub-Reinhaltung der Luft, 38 (11): 446-451 (November,
1978) .
2. Misaka, T., K. Sugimoto and H. Yamada. Electric Field
Strength and Collection Efficiency of Electrostatic
Precipitators Having Wide Collection Pitches. Proc. CSIRO
Conf. on Electrostatic Precipitation (Leura, N.S.W.,
Australia), Paper No. 11 (August, 1978).
3. Matts, S. Some Experiments with Increased Electrode
Spacing. Proc. CSIRO Conf. on Electrostatic Precipitation
(Leura, N.S.W., Australia), Paper No. 13 (August, 1978).
4. Cooperman, P. Non-Deutschian Phenomenon in Electrostatic
Precipitation. APCA 69th Annual Meeting (Portland, Oregon),
Paper No. 76-42.2 (June/July, 1976).
5. GUpner, 0. Einflusse auf die Wanderungsgeschwindigkeit,
nachgewiesen mit einem Versuchselektrofilter. Dissertation,
Univ. of Essen (1976).
6. Lau, H. Mit Wechselspannung betriebene Elektrofilter.
Staub-Reinhaltung der Luft, 29 (8): 311-314 (August, 1969).
7. Aureille, R. and P. Blanchot. Experimentelle Untersuchungen
des Einflusses verschiedener Parameter auf den Wirkungsgrad
492
-------
eines Elektrofliters. Staub-Reinhaltung der Luft, 31
(9):371-375 (September, 1971).
8. Nibeleanu, St. Der Einfluss des Gassenabstandes auf den
Entstaubungsgrad von Elektrofiltern bei StMuben mit hohem
Widerstand. Staub-Reinhaltung der Luft, 39 (2): 41-44
(February, 1979).
9. Ago, S., T. Itoh, H. Saito, N. Furuya and S. Masuda. Use of
Large Electrode Spacing in A Wet-Type Electrostatic
Precipitator. Proc. 1975 Annual Conf. of Inst. Elect.
Engrs. Japan, Paper No. 921 (1975).
10. Noso, S., Y. Nishimura, S. Yokawa, S. Koosaki, T. Tomita, K.
Hayashi, A. Tanaka, T. Noda, M. Mieno, S. Kuroki and M.
Uragami. Research on Dry-Type Electrostatic Precipitator.
Technical Journal of Sumitomo Heavy Industries Co., Ltd., 26
(28): 65-74 (December, 1978) (Japanese).
Ito, R. and K. Takimoto. Wide Spacing EP Is Available in
Cleaning Exhaust Gases from Industrial Sources. Proc. 1978
EPA-Sumposium on Transfer and Utilization of Particulate
Control Technology (Denver, Colorado), 1: 297-305 (July,
1978) .
11.
12.
13.
14.
15.
16.
Masuda, S.
Industries.
Utilization
Colorado), 2:
Electrostatic Precipitation in Japanese Steel
Proc. 1978 EPA-Symposium on Transfer and
of Particulate Control Technology (Denver,
309-318 (July, 1978).
Masuda, S., S. Ago, T. Itoh, H. Saito and N. Furuya.
Hybrid-Type Electrostatic Precipitator. APCA 69th Annual
Meeting (Portland, Oregon), Paper No. 76-42.1 (June/July,
1976).
Masuda, S. Novel Electrode Construction for Pulse Charging.
Proc. 1978 EPA-Symposium on Transfer and Utilization of
Particulate Control Technology (Denver, Colorado), 1: 241-251
(Junly, 1978).
Nomura, T. and M. Sakai. Roof-Mounted Electrostatic
Precipitator for Blast Furnace. Proceedings of Inst.
Electrostatics Japan, 1 (2): 82-87 (July, 1977).
Isahaya, F. and K. Ishida. Development of New Electrostatic
Precipitator Using Rotary Cylinder Collecting Electrode,
Attaching Rotary Chain Scraper. Proc. CSIRO Conf. on
Electrostatic Precipitation (Leura, N.S.W., Australia), Paper
No. 20 (August, 1978) .
493
-------
TABLE 1. STATUS OF WIDE-SPACING TYPE ESP IN JAPAN
(l66 units in total at the end of Dec. 1979 ; D> 300 mm )
Manufactur er
Number
Application Field
1. Nippon Kai Heavy Heavy 65
Industries Co., Ltd.
2. Sumitomo Metal & Mining 6k
Co., Ltd.
3. Nippon Steel Corporation 11
. Onoda Engineering Co., Ltd. 6
5. Mitsubishi Heavy Industries 2
6. Hitachi Plant Engineering & 2
Construction Co., Ltd.
T. Koyo Iron Works & Const- 2
ruction Co., Ltd.
8. Sumitomo Heavy Industries 6
Co., Ltd.
9. Chiyoda Plant Engineering & 1
Construction Corporation
10. Ninon Cement Mfg. Co., 1
Ltd.
cement rotary kiln, cement clinker
cooler, glass melt, furnace, FCC
process in oil refinery plant,
incinerator plant, desulfurization
plant for oil burning boiler, and
sinter main gas in steel plant.
sinter cooler, electric furnace,
sulfuric acid plant, roof-mounted type
ESP for electric furnace, cement plant,
incinerator plant.
sinter main gas in steel plant (large
gas volume).
cement rotary kiln, cement clinker
cooler, Hybrid-type ESP for cement
Lepol-kiln.
FCC process in oil refinery plant,
Hot-ESP for low-sulfur-coal burning
boiler (power plant).
incinerator plant, roof-mounted type
ESP for electric furnace.
tar fume from baking furnace (rotary
drum type ESP).
cement rotary kiln, sinter cooler in
steel plant
roof-mounted type ESP for converter.
cement clinker cooler.
TABLE 2. RECORDS OF ESCS-TYPE ESP INSTALLED IN NIPPON STEEL
CORPORATION (at Sinter Main Gas)
Specification
Gas Flow Rate (m^/min)
Gas Temperature (°C)
Gas Pressure (mm HvjO)
Ci (g/Nm^)
Co (g/Nm3)
House Type
19,200
Normal 130
±100
0.3
< 0.05
Straight Type
30,000
Normal 150, Max. 200
-2,200
0.5
< 0.05
Straight Type
26,000
130-150, Max. 250
±100
0.35
< 0.05
494
-------
-P
•H
O
O
O
-H
Cn
-H
g
Q)
>
•H
4J
rfl
rH
cu
—O--I
Figure 2
•Calculated from Collection
Efficiency with Pilot
Precipitator
•Caluculated from Measured
Field Strength near
Collecting Plate
= 4.0
Average Field Strength
(kV/cm)
I
I
200 400 600
Duct Spacing (mm)
800
Relative dust migration velocity vs
duct spacing (Misaka et.al.)
Figure 1 Wide-Spacing precipitator in
operation (cement clinker cooler;
2,300 mVmin, 150 °C)
-------
150
125
*, 100
75
100
° D =6 250 mm
• D = 400 mm
50
150
10
50 100 500
Ci (g/Nm3)
0
500 1000
D (mm)
Figure 7 Effect of inlet dust Figure 5 Effect of duct spacing
loading Ci on w* D on w*
(Noso et. al. test series II) (Noso et. al.; test series I)
125r
100
75
50
o D = 250 mm
• D = 400 mm
100
50
o D = 250 mm
• D = 400 mm
o> D = 800 mm
0.5 1.0
i (mA/m2)
Figure 3 Effect of current
density i on w*
(Noso et. al.; test series I)
150
100
50
D = 250 mm
D = 400 mm
D = 800 mm
5 10 50
C± (g/Nm3)
v (m/s)
Figure 6 Effect of gas velocity Figure 4 Effect of inlet dust
v on w* loading Ci on w*
(Noso et. al.; test series II) (Noso et. al.; test series I)
-------
Figure 8 Collection field of WESP
( D = 640 mm ; by the courtesy of
Nippon Kai Heavy Ind.,Co.,Ltd.)
497
-------
High-voltage DC power source
Figure 9
House-Type
ESCS
(Nippon Steel
Corp.)
PusagtofduM
LT~
Dust hopper
Dust conveyor
, • ' V ,'. . «
021
Figure 10. Straight-Type ESCS
(Nippon Steel Corp., Wakamatsu Works;
Sinter main gas; 30,000 m3/min; 150°C;
Ci - 500 mg/Nm3; co <50 mg/Nm3 )
498
-------
Figure 11 Vertical-Duct Type WESP
(Sumitomo Metal Mining Co., Ltd.)
499
-------
Discharge Electrode
Collecting electrode
(water cooled)
Figure 12 CEEP - Type WESP with water-cooled collecting electrodes
and mechanical scrapers
(Nipponkai Heavy Ind., Co., Ltd.)
Figure 13
CEEP - Type WESP at FCC-Plant
(Idemitsu oil Co., Ltd., Chiba Refinery; 3,570 m3/min;
gas temp. = 180 - 185 °C; outlet gas temp. = 160 - 165 °C;
Ci = 90 mg /Nm3, Co = 43 mg/Nm3)
500
-------
Roto ring cyindricol
Collecting electrode
Special radial needle type
discharge electrode
Brush for cleaning the roller chain
and the scraper
Sprocket wheel
Gear box
Figure 14 RCCES - Type WESP with a large rotating cylinder and
chain scraper
(Koyo Iron Works & Construction Co., Ltd.)
Figure 15 RCCES - Type WESP at firebrick baking furnace
(Shinagawa White Firebrick Co., Ltd: 200 m3/min; 20 - 30 °C;
tar fume; C^_ = 30 - 50 mg/Nm3; Co = 2.4 - 4.0 mg/Nm3)
501
-------
LOW FREQUENCY SONIC CLEANING APPLIED
TO ELECTROSTATIC PRECIPITATORS
By:
Stewart B. Smith
Inspiration Consolidated Copper, Inc.
Inspiration, Arizona
Joseph A. Schwartz
KVB, Inc.
17332 Irvine Blvd.
Tustin, California 92680
ABSTRACT
Inspiration Consolidated Copper was the first company in the United
States to apply low frequency horns to an electrostatic precipitator. These
low frequency acoustic devices (horns) have successfully demonstrated the
ability to solve many of the dust collection problems associated with the
operation of electrostatic precipitators. The type of problem addressed
includes: inlet distribution plate clogging, hopper clogging and bridging,
bridging between collector plates and structure, and build-up on wires.
While these horns have been widely used in Europe since 1969, they were
only recently introduced in the United States by KVB.
Dust particles clinging to surfaces are dislodged by sound wave vibra-
tional energy (145 dB, 250 Hz). The wave pressure fluidizes particles by
breaking their bond with other particles and the surfaces to which they cling.
Once "fluidized," particles will flow from surfaces by gravity or gas stream
pressure.
Noise levels outside the precipitator are well below OSHA regulations.
Prior to use of the horns the electric furnace precipitator Number 4 was
shut down for cleaning every four to seven days. With three KVB horns
installed, the precipitator operated for six weeks before a mechanical problem
caused a shut-down. Inspiration has since installed horns on two precipitators,
a cyclone, and a converter gas cooler.
This paper describes the installation, operation, and economics of
the Inspiration Consolidated Copper application.
502
-------
LOW FREQUENCY SONIC CLEANING APPLIED
TO ELECTROSTATIC PRECIPITATORS
INTRODUCTION
Inspiration Consolidated Copper Company (ICC) was the first company in
the United States to solve some very serious precipitator dust collection
problems by installing special purpose, low frequency, high energy acoustic
horns. Faced with such severe dust problems that a precipitator had to be
shut down and cleaned after less than a week of operation, ICC decided to
proceed with an evaluation program in the summer of 1978. The results, as
described below, have been extremely satisfactory, with benefits extending
beyond the precipitator to include improvements in emission controls, acid
plant operations, and mineral recovery. Economic savings in precipitator
cleaning alone provided a payback period of about two months.
THE PROBLEM
The copper smelting process employed at ICC, as shown in Figure 1,
produces approximately 80 tons of dust per day from the electric smeltzng fur-
nace. Flue gas from the electric furnace passes first through a cyclone, where
about 65 tons per day are collected, then through an electrostatic precipitator,
and finally goes to the sulfuric acid plant. In less than a week's operation
the precipitator would experience one or more of these problems.
Inlet distribution plate clogging of 60% or more
Dust build-up of two to three feet in the cat-walks between fields
Dust build-up on wire electrodes and collector plates of one to
two inches
Bridging between collector plates and structure
Bridging between hoppers and structure
Not only did this require frequent precipitator shut-downs for cleaning,
but the reduced efficiency of the precipitator—a drop from 95% to less than
50% in less than a week—adversely affected the sulphuric acid plant. Dust
overloads tended to clog the acid plant converter catalyst beds, necessitating
shut-downs for cleaning every four to six months. Also adversely affected by
the reduced precipitator efficiency was the recovery of copper and other
minerals from the dust lost downstream.
503
-------
To Acid Plant
Hot Gas Fan
o
Emergency
Stack
en
O
Converter
Pr e ci pi t ato rs
Furnace
Precipitator
Cyclones
Furnace
Gas Cooler
Electric Furnace
Figure 1. Inspiration Consolidated Copper Process.
-------
On the average, it took 15 work shifts—about one week—to cool the hot
precipitator down, get inside and clean out the dust, and put the unit back
into service. In addition to being time-consuming and costly, the nature of
the cleaning operation was particularly distasteful.
PRECIPITATOR DESIGN CHARACTERISTICS
The Number 4 precipitator at ICC consists of three nine-foot fields
separated by one foot wide cat-walks. Each field had 30 collector plates with
nine inch spacing between plates. Each plate is 20 feet high. The inlet
distribution plate measures about 22 feet by 20 feet.
The gas flow is about 24,000 SCFM at a temperature of 750 to 800°F
(399 to 427°C) and contains 4% S02 with a dew point of 430 °F (221°C).
KVB HORN DESCRIPTION
Figure 2 is a picture of the KVB horn. They are low frequency, high
energy acoustic devices—250 Hz and 145 dB. The bell is made of cast stain-
less steel and is good for temperatures of 1900°F (1040°C). There is only
one moving part—a titanium diaphragm.
The horns operate on compressed gas—at ICC compressed air—in the 75 to
100 psi range. Plant air is satisfactory and instrument air is not required.
Air consumption during insonations is 40 to 80 SCFM. They are turned on and
off by actuating a 110 volt solenoid valve. In operation this valve is con-
trolled by an automatic electric timer. The solenoids may be located remote
from the horn.
Each horn weighs 55 pounds and measures about two feet in length and
one foot in diameter at the mount of the bell. The bell is bolted to the
housing to facilitate installation. Diaphragms—which should last six months
to one year—are replaced by simply unbolting the back of the diaphragm
housing. In operation the diaphragm is cooled by a continuous air flow of
three to five SCFM.
Soot and other particles clinging to surfaces are dislodged by sound wave
vibrational energy. The wave pressure fluidizes particles by breaking their
bond with other particles and the surfaces to which they cling. Once
"fluidized," particles will flow from surfaces by gravity or gas stream
pressure.
The "fluidizing" effect is compounded in an enclosed space where large,
solid surfaces reflect sound waves, creating a homogeneous vibrational
energy field.
505
-------
Figure 2. KVB horn.
506
-------
HORN INSTALLATION
In July 1978, three KVB horns were installed in precipitator Number 4
as shown in Figure 3—one in the side of the inlet plenum, and one in each
cat-walk on opposite sides of the unit. Total installation time, including
electrical and pneumatic supply lines, is estimated at 40 hours.
All three horns are insonated simultaneously for 25 seconds every 25
minutes. Outside noise levels are well below OSHA regulations. During insona-
tions, the outside noise level is affected by sound coming through the wall
of the precipitator, by air coming out the solenoid valve exhaust, and by
sound coming through the back of the horn. A sheet metal box filled with
about three inches of rockwool insulation was built around each horn housing.
The solenoid valve exhausts to the atmosphere without any muffler. Noise
levels five feet from the precipitator measured about 90 dBA.
OSHA permissible noise exposures are shown below. Note that the limit
for one half hour per day is 110 dBA—25 seconds every 25 minutes results in
a daily insonation time of less than 30 minutes.
Duration Per Day Sound Level dBA
Hours Slow Response
8 90
4 95
2 100
1 105
1/2 110
RESULTS
The overall results of utilizing the KVB low frequency acoustic horns
were very satisfactory and produced several significant benefits to Inspira-
tion Consolidated Copper including:
An increase in precipitator running time
A reduction in the time required to clean the precipitator
. An increase in the ratio of running time to cleaning time from
one to one to six to one
An increase in operating time of the acid plant prior to screening
the converter catalyst bed
. A significant labor cost saving
With three KVB horns installed, precipitator Number 4 operated for six
weeks before a mechanical breakdown caused a shut-down. In fact, since
installation of the horns, factors other than dust build-up, i.e. mechanical
and electrical problems, now dictate the length of service of the precipitator,
507
-------
en
O
CO
Original
Horn Installation
Horn added after
initial evaluation
Figure 3. Horn installation precipitator No. 4
-------
Before installation of the horns, the operating time for the precipitator
equaled the cleaning time, on the average, about 15 shifts. After the horns,
continuous precipitator operating time (not counting problems unrelated to
the horns) jumped to 30 shifts, while clean-up time dropped to five shifts.
This, alone, is a labor cost saving of $4,260.
The acid plant converter catalyst bed can now be operated up to eight
months prior to shut-down for screening, compared with four to six months
before the horns were being used.
Another interesting result was obtained. During a one month period, the
precipitator was operated without any power on one of the fields. After
shut-down visual inspection of this section showed it to be as clean as the
other fields.
A fourth horn was subsequently added to deal with the dust build-up
occurring on the field-wire weight bottle guideways. This build-up was
allowing the bottles to hang up, thereby creating slack in the wire elec-
trodes, a condition leading to electrical shorts. The fourth horn was located
at the level of the bottles and was successful in eliminating the dust build-
up (Figure 3).
Cyclone Application
ICC was also experiencing severe clogging problems with the cyclone
handling flue gas between the electric furnace and precipitator. The dust
is fairly sticky and would build up on the cyclone walls and then slough off
due to iitc own weight. This would plug the one foot diameter dust collection
duct on the order of once or twice per week. Three mechanical vibrators had
been installed on the cyclone without producing a material improvement.
In December 1978 ICC installed a horn in the top of the cyclone as shown
.in Figure 4. The operating cycle was ten seconds on every minute. The
horn successfully solved this problem and to date they have not experienced
a plugging situation.
The result is a more efficient cyclone and a significant reduction
in the manpower associated with dust removal and handling.
CONCLUSIONS
The use of low frequency, high energy acoustic horns significantly
reduced dust collection problems at Inspiration Consolidated Copper. Three
horns installed on the electric furnace precipitator extended the operating
period between shut-downs for cleaning from less than one week to about six
weeks. In fact, since installation of the horns, factors other than dust
build-up, i.e. mechanical and electrical problems, now dictate the length of
service of the precipitator. A horn was also successfully used in a cyclone
to prevent frequent clogging.
509
-------
KVB Horn
Figure 4. Cyclone installation.
510
-------
ADDENDUM
ANALYSIS OF ACOUSTIC WAVE EFFECTS ON FINE P ART I GUI "^E MATERIALS
To put the nature of the acoustic field into perspective we should note
that at a point where the intensity is 145 decibels, the local pressure
fluctuation is of the order of one lb/ft2. At the conditions of a precipi-
tator this corresponds to a local air velocity fluctuation of about 0.5
ft/sec.., and when this velocity acts upon a particle of approximately one
micron radius, the particle experiences an acceleration on the order of
2500 g. These forces are usually intense enough to fluidize a lightly packed
parti culate volume or remove particles that are lightly bound to surfaces.
In this section we shall carry out an approximate calculation that will disclose
some of the mechanisms of the acoustic-particle interactions and illustrate
how we may expect the results to scale when particle size is varied.
The first problem relates to the penetration of the acoustic cone into
a lightly packed volume of particles such as might be encountered at various
locations of the precipitator. Only these particles will be initially
fluidized if the local acoustic velocities are sufficient to accelerate the
particles tp perhaps 50 g. Thus the rate of attenuation of the acoustic
velocity below the surface of the particulate volume is the quantity which
we shall require to estimate the depth of f luidization.
To analyze this, consider a particulate volume
Particle
Volume
that may be considered one-dimensional with x measuring the distance below the
surface,. A beam of plane waves moves with acoustic velocity c toward the
surface, and this beam may be represented analytically as
P1 u „ i(wt - kx)
Pe
YPo c
511
-------
where P1 and u are the acoustic pressure and velocity fluctuations, Po and c
are the reference (atmospheric) pressure and acoustic velocity respectively,
P is the magnitude of the wave, w is the angular frequency (= 2r\ x 250 for
present application) and the wave number K = w/c for the wave before it enters
the particle pile. Within the particle volume, the air motions are acted upon
by the resistance of the particles. In the present instance, this force may
be represented roughly by the Stokes interaction between moving gas and
stationary particles. It may be shown that this force is
Fp = (n) (6n ay) (u)
where n is the number of particles per unit volume having a radius a, y is
the gas viscosity, and u is the local gas velocity. This results in an
acoustic equation that can be represented as
3u 23n Kp u _ „
i r\ ^ L/
2 w a 2 T T 3t
Ot dx
where Kp is the ratio of particle mass to air mass in the volume and
m
T =
nay
is the "relaxation time" of the particles where m is the mass of a single
particle. For usual solid particles of one micron radius, T ^ 10~4 seconds.
The relaxation time scales as the square of the particle radius. The particle
leading Kp is of the order 102 and hence the parameter (Kp/ut), which enters
the solution of the acoustic problem within the particle pack, is quite large.
The approximate solution, valid for our range of interest, is then
. . . w /Kp w / Kp
U ^ ^ iwt -i — /r^T x - — / —7- x
— = Pe . e c / 2(jJt . e c v cot
c
which matches appropriately with the impinging wave at x = 0. The attenuation
within the pack is given by the last factor, which may also be written
2fT / Kp
— — / —— x
e X / GOT
o<.
where > is the acoustic wave length in free air. The value of x, the depth at
which the acoustic velocity has dropped to e of its initial value, is
x _ 1_ /oJt
X ~ 2TT /
-------
To obtain a better estimate of this fluidization depth we may note that
the acceleration experienced by any one particle is
6npy _ u
" ~~
when us is the local gas velocity. Therefore u/gT is the acceleration in
gravitational units which, if we set to 50 as a requirement for fluidization,
gives
S - »
as the minimum effective velocity. Therefore,
so <>
c IDT
gives the actual depth of fluidization x* as
which differs from the previous result by the logarithmic factor. For our
previous example, this factor is 8.97 so that the actual fluidization depth
is nearly five inches. The scaling of this value with particle size, gas
conditions and particle material is now more complex but is evident from the
above relation.
513
-------
AUTHOR INDEX
AUTHOR NAME PAGE
Ariman, T. 111-222
Bacchetti, J. A. 1-529
Bernstein, S. 11-125
Bibbo, P. P. 11-219
Bickelhaupt, R. E. 1-154
Blackwood, T. R. IV-312
Bloomfield, D. P. III-145
Brackbill, E. A. III-472
Brines, H. G. 1-351
Brookman, E. T. IV-274
Brown, J. T. (Jr.) III-439
Buchanan, W. J. 11-168
Burckle, J. 0. III-484
Bush, J. R. IV-154
Carlsson, B. III-260
Carr, R. C. 1-35, III-270
Chang, C. M. 11-314
Chapman, R. A. 1-1
Chmielewski, R. III-l
Cooper, D. W. III-127
Cowen, S. J. IV-424
Cowherd, C. (Jr.) IV-240
Czuchra, P. A. III-104
Darby, K. 1-15
Daugherty, D. P. IV'182
514
-------
AUTHOR NAME PAGE
Dennis, R. 1-494
Dietz, P. W. III-429
Donovan, R. P. 1-476
Drehmel, D. C. IV-170
Durham, M. D. IV-368
Dybdahl, A. W. IV-443
Ellenbecker, M. J. III-171, III-190
Engelbrecht, H. L. 11-279
Ensor, D.S. 111-39
Ernst, M. IV-30, IV-42
Eschbach, E. J. 11-114
Evans, J. S. IV-252
Fasiska, E. J. IV-486
Faulkner, M. G. IV-508
Fedarko, W. IV-64
Ferrigan, J. J. 1-170
Finney, W. C. 11-391
Furlong, D. A. !"4^5
Garrett, N. E. IV-524
Gastler, J. H. IV-291
Gavin, J. H. ni"81
Giles, W. B. IV"387
Gooch, J. P. I"132
Gooding, C. H.
Grace, D. S.
Guiffre, J. T.
515
-------
AUTHOR NAME PAGE
Hall, F. D. IH-25
Hardison, L. C. IH-382
Hoenig, S. A. IV-201
Hudson, J. A. 1-263
linoya, K. II1-237
Isoda, T. 111-16
Jaasund, S. A. 11-452
Kalinowski, T. W. III-363
Kallio, G. A. III-344
Kearns, M. T. 111-61
Kelly, D. S. 1-100
Kinsey, J. S. 111-95
Kolber, A. R. 1-224
Ladd, K. L. 1-317
Lamb, G. E.R. III-209
Lane, W. R. 1-410
Langan, W. T. 1-117, 11-256
Larson, R. C. II1-448
Leonard, G. 11-146
Lipscomb, W. 0. 1-453
Malani, S. 1-570
Marcotte, W. R. 1-372
Martin, J. R. 1-591
Masuda, S. 11-65, 11-334, 11-483
McCain, J. D. IV-496
McDonald, J. R. 11-93
516
-------
AUTHOR NAME PAGE
Mitchell, D. A. III-162
Modla, J. C. 11-399
Mosley, R. B. 11-45
Mycock, J. C. 1-432
Neundorfer, M. 11-189
Nixon, D, 1-513
Noll, C, G. 11-374
Nunn, M. 11-369
Ondov, J. M. IV-454
Ostop, R. L. 1-342
Parker, R. IV-1
Patch, R. W. IV-136
Patterson, R. G. IV-84
Pearson, G. L. 1-359
Pedersen, G. C. III-416
Petersen, H. H. 11-352
Pilat, M. J. 1-561
Potter, E. C. 1-184
Ranade, M. B. 1-538
Raymond, R. K. 11-173
Rinard, G. H-31, IV-127
Roehr, J. D. H"208
Rolschau, D. W. III-251
Ruth, D. H-427, 11-441
Samuel, E. A. H-l
Schliesser, S. P. 1-56
517
-------
AUTHOR NAME PAGE
Self, S. A. III-309
Severance, R. L. IV-321
Shale, C. C. 1-390
Smit, W. 1-297
Smith, S. B. n-502
Spafford, R. B. 1-202
Sparks, L. E. II-417, IV-411
Stenby, E. W. 1-243
Stock, W. E. IV-333
Surati, H. 11-469
Szabo, M. F. III-508
Tendulkar, S. P. IV-338
Tennyson, R. P. III-117
Tsao, K. C. IV-14
Umberger, J. H. 11-296
VanOsdell, D. W. II-74
VanValkenburg, E. S. IV-351
Wang, J. C.F. IV-396
Weber, E. IV-98
Wybenga, F. A. 11-242
Yung, S. IV-217
518
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
REPORT NO.
EPA-600/9-80-039b
2.
IERL-RTF-1062
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
Second Symposium on the Transfer and Utilization of
Particulate Control Technology (Denver, July 1979)
Vol. II. Electrostatic Precipitators
5. REPORT DATE
Sept. 1980 Issuing Date.
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
8. PERFORMING ORGANIZATION REPORT NO.
P.P. Venditti, J.A. Armstrong, and Michael Durham
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Denver Research Institute
P.O. Box 10127
Denver, Colorado 80210
10. PROGRAM ELEMENT NO.
EHE624
11. CONTRACT/GRANT NO.
R805725
12. SPONSORING AGENCY NAME AND ADDRESS
Industrial Environmental Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Proceedings; 6/79-6/80
14. SPONSORING AGENCY CODE
EPA/600/13
15. SUPPLEMENTARY NOTES
IERL-RTP project officer is Dennis C. Drehmel, MD-61, 919/541-2925.
044a thru -044d are proceedings of the 1978 symposium.
EPA-600/7-79-
16. ABSTRACT
The proceedings document the approximately 120 presentations at the EPA/IERL-
RTP-sponsored symposium, attended by nearly 800 representatives of a wide variety of
companies (including 17 utilities). The keynote speech for the 4-day meeting was
by EPA's Frank Princiotta. The meeting included a plenary session on enforcement.
Attendees were polled to determine interest areas: most (488) were interested in'
operation and maintenance, but electrostatic precipitators (ESPs) and fabric filters
were a close second (422 and 418, respectively). Particulate scrubber interest
appears to be waning (288). Major activities of attendees were: users, 158;
manufacturers, 184; and R and D, 182. Technical presentations drawing great interest
were the application of ESPs and baghouses to power plants and the development of
novel ESPs. As important alternatives to ESPs, baghouses were shown to have had
general success in controlling coal-fired power plant emissions. When operating
properly, baghouses can limit ^missions to<^5 mg/cu run at pressure drops of<^2 kPa.
Not all baghouse installation^ have been completely successful. Both high pressure
drop and bag loss have occur/ed ( at the Harrington Station), but these problems
appear to be solved. '
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Pollution Scrubbers
Dust Flue Gases
Aerosols
Electrostatic Precipitators
Filters
Fabrics
Pollution Control
Stationary Sources
Particulate
Baghouses
13B
11G
07D
131
14G
HE
07A
21B
is. DISTRIBUTION STATEMENT
Release to Public
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
535
20. SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA Fofm 2220-1 (Rev. 4-77) PREVIOUS EDITION is OBSOLETE
U.S. GOVERNMENT PRINTING OFFICE: 1980--657-165/0158
519
------- |