U.S. Environmental
Protection Agency
Electric Power
Research Institute
Topics:
Particulates
Fabric filters
Electrostatic precipitators
Sulfur dioxide
Gaseous wastes
Environment
EPRI CS-4404
Volume 2
Project 1835-6
Proceedings
February 1986
Proceedings: Fifth Symposium
on the Transfer and Utilization of
Particulate Control Technology
Volume 2
Prepared by
Research Triangle Institute
Research Triangle Park, North Carolina

-------
REPORT SUMMARY
SUBJECTS Particulate control I Integrated environmental control / SOx control
TOPICS Particulates	Sulfur dioxide
Fabric filters	Gaseous wastes
Electrostatic precipitators	Environment
AUDIENCE Environmental engineers and operators
Proceedings: Fifth Symposium on the
Transfer and Utilization of Particulate
Control Technology
Volumes 1-4
From a speculative discussion of the future regulatory frame-
work in the opening sessions to the detailed treatments of par-
ticulate and fugitive emissions control in the following days, this
symposium updated the research community on the range of
promising technologies. The report includes the more than
100 papers presented.
In 1984, EPRI joined EPA as cosponsor of this symposium. The meeting-
sponsored in the past by EPA alonehas taken place at 18-month intervals.
	To promote the transfer of results from particulate control research to
potential users of those technologies.
	To provide an exchange of ideas among researchers active in the field.
The 430 professionals attending the symposium on August 27-30,1984, in
Kansas City, Missouri, represented utilities, manufacturers, state and federal
agencies, educational institutions, and research organizations. From more
than 100 presentations, they learned of developments in such areas as elec-
trostatic precipitators, fabric filters, fugitive emissions, and dry S02 control
processes. The discussions touched on many aspects of new and old
technologiesfrom economics to operation and maintenance to the devel-
opment and testing of advanced concepts.
KEY POINTS The proceedings, which include all formal presentations from the confer-
ence, report the research efforts of air pollution control equipment manufac-
turers, as well as EPA and EPRI. Of particularly broad interest are papers
addressing trends in particulate environmental regulations and their pos-
sible impacts on those manufacturers, utilities, and the iron and steel
Industry. Papers hawing a more detailed focus explore developments in elec-
trostatic precipitator controls and other performance-enhancing technolo-
gies, as well as materials and bag-cleaning methods for fabric fitters and
such new S02 control methods as dry sorbent furnace injection and spray
BACKGROUND
OBJECTIVES
APPROACH

-------
drying. In addition, several papers consider fugitive emissions control, a
growing environmental concern.
EPRI PERSPECTIVE These presentations stimulated new interest in promising technologies,
as evidenced by the many requests for further information both at the
conference and afterward. Subsequent developments in particulate con-
trol will be the focus of the sixth symposium, to be held in February 1986
in New Orleans.
PROJECT RP1835-6
EPRI Project Manager: Ralph F. Altman
Coal Combustion Systems Division
Contractor: Research Triangle Institute
For further information on EPRI research programs, call
EPRI Technical Information Specialists (415) 855-2411.

-------
Proceedings: Fifth Symposium on the
Transfer and Utilization of Particulate Control
Technology
Volume 2
CS-4404, Volume 2
Research Project 1835-6
Proceedings, February 1986
Kansas City, Missouri
August 27-30, 1984
Prepared by
RESEARCH TRIANGLE INSTITUTE
Cornwallis Road
Research Triangle Park, North Carolina 27709
Compiler
F. A. Ayer
Prepared for
U.S. Environmental Protection Agency
Office of Research and Development
401 M Street, SW
Washington, D.C. 20460
Air and Energy Engineering Research Laboratory
Research Triangle Park, North Carolina 27711
EPA Project Officer
D. L. Harmon
Electric Power Research Institute
3412 Hill view Avenue
Palo Alto, California 94304
EPRI Project Manager
R. F. Altman
Air Quality Control Program
Coal Combustion Systems Division

-------
ORDERING INFORMATION
Requests for copies of this report should be directed to Research Reports Center
(RRC), Box 50490, Palo Alto, CA 94303, (415) 965-4081. There is no charge for reports
requested by EPRI member utilities and affiliates, U.S. utility associations, U.S. government
agencies (federal, state, and local), media, and foreign organizations with which EPRI has an
information exchange agreement. On request, RRC will send a catalog of EPRI reports.
Copyright  1986 Electric Power Research institute, Inc. All rights reserved.
NOTICE
This report was prepared by the Electric Power Research Institute, Inc. (EPRI). Neither EPRI, members of EPRI,
nor any person acting on their behalf: (a) makes any warranty, express or implied, with respect to the use of any
information, apparatus, method, or process disclosed in this report or that such use may not infringe privately
owned rights: or (b) assumes any liabilities with respect to the use of, or for damages resulting from the use of,
any information, apparatus, method, or process disclosed in this report; or (c) is responsible for statements made or
opinions expressed by individual authors.

-------
ABSTRACT
These proceedings are of the Fifth Symposium on the Transfer and Utiliza-
tion of Particulate Control Technology, held August 27 to 30, 1984, in
Kansas City, Missouri. The symposium was sponsored by EPA's Air and
Energy Engineering Research Laboratory (formerly Industrial Environmental
Research Laboratory), located in Research Triangle Park, North Carolina,
and the EPRI Coal Combustion Systems Division, located in Palo Alto,
California.
The objective of the symposium was to provide for the exchange of knowl-
edge and to stimulate new ideas for particulate control with the goal of
extending the technology and aiding its diffusion among designers, users,
and educators. Fabric filters and electrostatic precipitators were the
major topics, but novel concepts and advanced technologies were also
explored. The organization of sessions was as follows:
Day 1
--Plenary session
--ESP: Performance Estimating (Modeling)
--FF: Practical Considerations
Economics
--Novel Concepts
Day 2
--ESP: Performance Enhancement I
--FF: Full-Scale Studies I (Coal-Fired Boilers)
--Fugitive Emissions I
--ESP: Performance Enhancement II
--FF: Full-Scale Studies II (Coal-Fired Boilers)
--Fugitive Emissions II
Day 3
--ESP: Advanced Technology I
--FF: Fundamentals/Measurement Techniques
iii

-------
--Dry SC>2 Removal I
--ESP: Advanced Technology II
--FF: Advanced Concepts
--Dry SO^ Removal II
Day 4
--ESP: Fundamentals I
--FF: Pilot-Scale Studies (Coal-Fired Boilers)
--Operation and Maintenance I
ESP: Fundamentals II
--Advanced Energy Applications
--Operations and Maintenance II
Volume 1 contains 19 papers presented at the Plenary, Advanced Energy
Applications, Economics and Novel Concepts Sessions.
Volume 2 contains 33 papers presented at the ESP: Performance Estimating
(Modeling), ESP: Performance Enhancement I and II, ESP: Advanced Tech-
nology I and IIj and ESP: Fundamentals I and II Sessions, plus one
unpresented paper.
Volume 3 contains 24 papers presented at the FF: Practical Considera-
tions, FF: Full-Scale Studies I and II (Coal-Fired Boilers), FF: Funda-
mentals/Measuring Techniques, FF: Advanced Concepts, and FF: Pilot-
Scale Studies (Coal-Fired Boilers) Sessions.
Volume 4 contains 29 papers presented at the Fugitive Emissions I and II,
Dry SO^ Removal I and II, and Operation and Maintenance I and II Sessions.
iv

-------
PREFACE
These proceedings for the Fifth Symposium on the Transfer and Utilization
of Particulate Control Technology constitute the final report submitted
to EPA's Air and Energy Engineering Research Laboratory (AEERL), Research
Triangle Park, North Carolina, and to the Coal Combustion Systems Divi-
sion, EPRI, Palo Alto, California. The symposium was conducted at the
Hyatt Regency Hotel at the Crown Center in Kansas City, Missouri, August
27-30, 1984.
This symposium (the first jointly sponsored by EPA and EPRI) was designed
to provide a forum for the exchange of knowledge and to stimulate new
ideas for particulate control with the goal of extending technology and
aiding its diffusion among designers, users, educators, and researchers.
In the opening session, an address was given on the regulatory framework
for future particulate technology needs followed by a series of addresses
on the impact of coming particulate requirements on the utility industry
and the iron and steel industry as well as the viewpoint of large and
small manufacturers. There were subsequent technical sessions on elec-
trostatic precipitator performance estimating (modeling), ESP performance
enhancement, ESP advanced technology, ESP fundamentals, practical con-
siderations for fabric filters, fabric filter full-scale studies (coal-
fired boilers), fabric filter fundamentals/measurement techniques, fabric
filter pilot-scale studies (coal-fired boilers), fugitive emissions, dry
S02 removal, operation and maintenance, and advanced energy applications.
Participants represented electric utilities, equipment and process sup-
pliers, state environmental agencies, coal and petroleum suppliers, EPA
and other Federal agencies, educational institutions, and research organ-
izations .
The following persons contributed their efforts to this symposium:
Dale L. Harmon, Chemical Engineer, Particulate Technology
Branch, Utilities and Industrial Power Division, U.S. EPA,
AEERL, Research Triangle Park, North Carolina, was a sym-
posium co-general chairman and EPA project officer.
Ralph F. Altman, Ph.D., Project Manager, Coal Combustion
Systems Division, EPRI, Chattanooga, Tennessee, was a co-
general chairman and EPRI project manager.
Franklin A. Ayer, Consultant, Research Triangle Institute,
Research Triangle Park, North Carolina, was the overall
symposium coordinator and compiler of the proceedings.
v

-------
TABLE OF CONTENTS
VOLUME 1, PLENARY, ADVANCED ENERGY APPLICATIONS, ECONOMICS, AND
NOVEL CONCEPTS
VOLUME 2, ELECTROSTATIC PRECIPITATION
VOLUME 3, FABRIC FILTRATION
VOLUME 4, FUGITIVE EMISSIONS, DRY S02, AND OPERATION AND MAINTENANCE
VOLUME 1
PLENARY, ADVANCED ENERGY APPLICATIONS, ECONOMICS,
AND NOVEL CONCEPTS
Section	Page
Session 1: PLENARY SESSION
Everett L. Plyler, Chairman
The Regulatory Framework for Future Particulate
Technology Needs 	 1-1
Sheldon Meyers
The Impact of Coming Particulate Control Requirements on the
Utility Industry 	 2-1
George T. Preston
The Impact of Coming Particulate Control Requirements on the
Iron and Steel Industry	3-1
Earle F. Young, Jr.
The Impact of Particulate Control Requirements: Large
Manufacturer's Viewpoint 	 4-1
Herbert H. Braden
Paper presented by Gary R. Gawreluk
Future Particulate Regulations: The View of the
Small Manufacturer	5-1
Sidney R. Orem
Session 2: ADVANCED ENERGY APPLICATIONS
George A. Rinard, Chairman
vii

-------
Section
Page
High-Temperature, High-Pressure Electrostatic
Precipitation, Current Status
P. L. Feldman* and K. S. Kumar
Test Results of a Precipitator Operating at High-Temperature
and High-Pressure Conditions 	 6-1
Donald E. Rugg", George Rinard, Michael Durham, and
James Armstrong
Evaluation and Development of Candidate High Temperature
Filter Devices for Pressurized Fluidized Bed Combustion. . . . 7-1
T. E. Lippert,v, D. F. Ciliberti, S. G. Drenker,
and 0. J. Tassicker
High Temperature Gas Filtration with Ceramic Filter
Media: Problems and Solutions 	 8-1
Ramsay Chang
The Development and High Temperature Application
of a Novel Method for Measuring Ash Deposits and
Cake Removal on Filter Bags	9-1
David F. Ciliberti", Thomas E. Lippert, Owen J. Tassicker,
and Steven Drenker
Session 3: ECONOMICS
John S. Lagarias, Chairman
Economics of Electrostatic Precipitators and
Fabric Filters	10-1
Victor H. Belba*, Fay A. Horney, Robert C. Carr,
and Walter Piulle
Estimating the Benefits of S03 Gas Conditioning on the
Performance of Utility Precipitators When Burning
U.S. Coals	11-1
Peter Gelfand
Microcomputer Models for Particulate Control	12-1
A. S. Viner", D. S. Ensor, and L. E. Sparks
The Impact of Proposed Acid Rain Legislation on Power
Plant Particulate Control Equipment	13-1
William H. Cole
Session 4: NOVEL CONCEPTS	14-1
Dale L. Harmon, Chairman
^Denotes speaker
viii

-------
Section	Page
Particle Charging with an Electron Beam Precharger 	 15-1
J. S. Clements", A. Mizuno, and R. H. Davis
Charging of Particulates by Evaporating Charged
Water Droplets	16-1
G.	S. P. Castle*, I. I. Inculet, and R. Littlewood
Role of Electrostatic Forces in High Velocity Particle
Collection Devices	17-1
H.	C. Wang, J. J. Stukel*, K. H. Leong, and P. K. Hopke
Hot-Gas Fabric Filtration 500 F - 1500 F, No Utopia but
Reality	18-1
Lutz Bergmann
The Prediction of Plume Opacity from Stationary Sources. . . .19-1
David S. Ensor*, Ashok S. Damle, Philip A. Lawless,
and Leslie E. Sparks
APPENDIX: Attendees 	 A-l
VOLUME 2
ELECTROSTATIC PRECIPITATION
Session 5: ESP: PERFORMANCE ESTIMATING (MODELING)
Leslie E. Sparks, Chairman
Microcomputer Programs for Precipitator Performance
Estimates	1-1
M. G. Faulkner*, J. L. DuBard, R. S. Dahlin,
and Leslie E. Sparks
Analysis of Error in Precipitator Performance Estimates. . . . 2-1
J. L. DuBard* and R. F. Altman
Use of a Mobile Electrostatic Precipitator for Pilot
Studies	3-1
Robert R. Crynack* and John D. Sherow
Prediction of Voltage-Current Curves for Novel
ElectrodesArbitrary Wire Electrodes on Axis	4-1
Phil A. Lawless* and L. E. Sparks
Numerical Computation of the Electrical Conditions in a
Wire-Plate Electrostatic Precipitator Using the Finite
Element Technique	5-1
Gregory A. Kallio* and David E. Stock
*Denotes speaker
ix

-------
Page
Section
Session 6: ESP: PERFORMANCE ENHANCEMENT I
Ralph F. Altman, Chairman
A Field Study of a Combined NH3-SO3 Conditioning System
on a Cold-Side Fly Ash Precipitator at a Coal-Fired
Power Plant			6-1
Robert S. Dahlin*, John P. Gooch, Guillaume H. Marchant, Jr.
Roy E. Bickelhaupt, D. Richard Sears, and Ralph F. Altman
Conditioning of Power Station Flue Gases to Improve
Electrostatic Precipitator Efficiency	7-1
Gemot Mayer-Schwinning" and J. D. Riley
Pilot-Scale Study of a New Method of Flue-Gas Conditioning
with Ammonium Sulfate	8-1
Edward B. Dismukes*, E. C. Landham, Jr., John P. Gooch,
and Ralph F. Altman
Power Plant Plume Opacity Control	9-1
J. Martin Hughes* and Kai-Tien Lee
Pulse Energization System of Electrostatic Precipitator
for Retrofitting Application 	 10-1
Senichi Masuda* and Shunsuke Hosokawa
Session 7: ESP: PERFORMANCE ENHANCEMENT II
B. G. McKinney, Chairman
Practical Implications of Pulse Energization of
Electrostatic Precipitators	11-1
H. Milde*, J. Ottesen, and C. Salisbury
Laboratory and Full-Scale Characteristics of Electrostatic
Precipitators with Rigid Mast Electrodes 	12-1
H. Krigmont", R. Allan, R. Triscori, and
H. W. Spencer, III
Full Scale Experience with Pulsed Energization of
Electrostatic Precipitators	13-1
K. Porle* and K. Bradburn
New Life for Old Weighted Wire Precipitators: Rebuilding
with Rigid Electrodes	14-1
Peter J. Aa* and Gary R. Gawreluk*
^Denotes speaker
x

-------
Section	Page
Pulsing on a Cold-Side Precipitator, Florida Power
Corporation, Crystal River, Unit 1	15-1
Joseph W. Niemeyer*, Robert A. Wright, and Wayne Love
Session 8: ESP: ADVANCED TECHNOLOGY I
Norman Plaks, Chairman
Field Study of Multi-Stage Electrostatic Precipitators . . . .16-1
Michael Durham, George Rinard, Donald Rugg,
Theodore Carney, James Armstrong", and
Leslie E. Sparks
Optimizing the Collector Sections of Multi-Stage
Electrostatic Precipitators	17-1
George Rinard*, Michael Durham, Donald Rugg,
and Leslie Sparks
Ceramic-Made Boxer-Charger for Precharging Applications. . . .18-1
Senichi Masuda*, Shunsuke Hosokawa, and Shuzo Kaneko
Precipitator Performance Enhancement with Pulsed
Energization	19-1
E. C. Landham, Jr.*, James L. DuBard, Walter R. Piulle,
and Leslie Sparks
Aerosol Particle Charging in a Pulsed Corona Discharge . . . -20-1
James L. DuBard* and Walter R. Piulle
Session 9: ESP: ADVANCED TECHNOLOGY II
Walter R. Piulle, Chairman
Performance of Large-Diameter Wires as Discharge
Electrodes in Electrostatic Precipitators	21-1
P. Vann Bush*, Duane H. Pontius, and Leslie E. Sparks
Technical Evaluation of Plate Spacing Effects on
Fly Ash Collection in Precipitators	22-1
Ralph F. Altman*, Gerald W. Driggers, Ronald W. Gray,
and James L. DuBard, and E. C. Landham, Jr.
Electrical Characteristics of Large-Diameter Discharge
Electrodes in Electrostatic Precipitators	23-1
Kenneth J. McLean* and Leslie E. Sparks
Laboratory Analysis of Corona Discharge Electrodes
and Back Corona Phenomena	24-1
P. Vann Bush* and Todd R. Snyder
*Denotes speaker
xi

-------
Section	Page
Session 10: ESP: FUNDAMENTALS I
Grady B. Nichols, Chairman
The Onset of Electrical Breakdown in Dust Layers	25-1
Ronald P. Young", James L. DuBard, and Leslie E. Sparks
Bipolar Current Probe for Diagnosing Full-Scale
Precipitators	26-1
Senichi Masuda*, Toshifumi Itagaki, Shigeyuki Nohso,
Osamu Tanaka, Katsuji Hironaga, and Nobuhiko Fukushima
A Method for Predicting the Effective Volume Resistivity
of a Sodium Depleted Fly Ash Layer	27-1
Roy E. Bickelhaupt* and Ralph F. Altman
Analysis of Air Heater-Fly Ash-Sulfuric Acid Vapor
Interactions	28-1
Norman W. Frisch
Session 11: ESP: FUNDAMENTALS II
Philip A. Lawless, Chairman
Experimental Studies of Space Charge Effects in an ESP .... 29-1
D. H. Pontius* and P. V. Bush
An Electrostatic Precipitator Facility for Turbulence
Research 	 30-1
J. H. Davidson* and E. J. Shaughnessy
On the Static Field Strength in Wire-Plate Electrostatic
Precipitators with Profiled Collecting Electrodes by an
Experimental Method	31-1
C. E. Akerlund
The Fluid Dynamics of Electrostatic Precipitators:
Effects of Electrode Geometry	32-1
E. J. Shaughnessy*, J. H. Davidson, and J. C. Hay
VOLUME 3
FABRIC FILTRATION
Session 12: FF: PRACTICAL CONSIDERATIONS
Wallace B. Smith, Chairman
Fabric Screening Studies for Utility Baghouse Applications . . 1-1
Larry G. Felix* and Randy L. Merritt
*Denotes speaker
xii

-------
Section	Page
Tensioning of Filter Bags in Reverse Air Fabric Filters. . . . 2-1
Robert W. Tisone* and Gregory L. Lear
Sound of Energy Savings	3-1
N. D. Phillips* and J. A. Barabas
Solving the Pressure Drop Problem in Fabric Filter
Bag Houses			4-1
Carl V. Leunig
Session 13: FF: FULL-SCALE STUDIES (COAL-FIRED BOILERS)
Robert P. Donovan, Chairman
Emission Reduction Performance and Operating
Characteristics of a Baghouse Installed on a
Coal-Fired Power Plant 	 5-1
David S. Beachler*, John W. Richardson,
John D. McKenna, John C. Mycock, and Dale Harmon
Evaluation of Sonic-Assisted, Reverse-Gas Cleaning
at Utility Baghouses 	 6_1
Kenneth M. Cushing*, Larry G. Felix,
Anthony M. LaChance, and Stephen J. Christian
Sonic Horn Application in a Dry FGD System Baghouse	7_i
Yang-Jen Chen*, Minh T. Quach, and H. W. Spencer III
Full Scale Operation and Performance of Two New
Baghouse Installations 	 8-1
C. B. Barranger
Session 14: FF: FULL-SCALE STUDIES II (COAL-FIRED BOILERS)
Robert C. Carr, Chairman
Performance of Baghouses in the Electric Generating
Industry	9-1
Wallace B. Smith* and Robert C. Carr
Flue Gas Filtration: Southwestern Public Service
Company's Experience in Design, Construction, and
Operation	10-1
John Perry
Start-Up and Operation of a Reverse-Air Fabric Filter
on a 550 MW Boiler	11-1
R. A. Winch and L. J. Pflug, Jr.*
*Denotes speaker
xiii

-------
Section	Page
Update on Australian Experience with Fabric Filters
on Power Boilers	12-1
F.	H. Walker
Session 15: FF: FUNDAMENTALS/MEASUREMENT TECHNIQUES
David S. Ensor, Chairman
Modeling Baghouse Performance	13-1
David S. Ensor", Douglas W. VanOsdell, Andrew S. Viner,
Robert P. Donovan, and Louis S. Hovis
Measurement of the Spatial Distribution of Mass on
a Filter	14-1
Andrew S. Viner*, R. P. Gardner, and L. S. Hovis
Laboratory Studies of the Effects of Sonic Energy on
Removal of a Dust Cake from Fabrics	15-1
B. E. Pyle", S. Berg, and D. H. Pontius
Cleaning Fabric Filters	16-1
G.	E. R. Lamb
Session 16: FF: ADVANCED CONCEPTS
John K. McKenna, Chairman
Modeling Studies of Pressure Drop Reduction in Electrically
Stimulated Fabric Filtration 	 17-1
Barry A. Morris*, George E. R. Lamb, and Dudley A. Saville
Flow Resistance Reduction Mechanisms for Electrostatically
Augmented Filtration 	 18-1
D. W. VanOsdell", R. P. Donovan, and Louis S. Hovis
Laboratory Studies of Electrically Enhanced Fabric
Filtration	19-1
Louis S. Hovis*, Bobby E. Daniel, Yang-Jen Chen,
and and R. P. Donovan
Pressure Drop for a Filter Bag Operating with a
Lightning-Rod Precharger	20-1
George E. R. Lamb" and Richard I. Jones
New High Performance Fabric for Hot Gas Filtration ..... .21-1
J. N. Shah
Paper presented by Peter E. Frankenburg
-''Denotes speaker
xiv

-------
Section	Page
Session 17: FF: PILOT-SCALE STUDIES (COAL-FIRED BOILERS)
Louis S. Hovis, Chairman
The Influence of Coal-Specific Fly Ash Properties Upon
Baghouse Performance: A Comparison of Two Extreme
Examples	22-1
Stanley J. Miller" and D. Richard Sears
Top Inlet Baghouse Evaluation at Pilot Scale 	 23-1
Gary P. Greiner* and Dale A. Furlong
Development of Woven Electrode Fabric and Preliminary
Economics for Full-Scale Operation of Electrostatic
Fabric Filtration. . 		24-1
James J. Spivey*, Richard L. Chambers, and Dale L. Harmon
ESFF Pilot Plant Operation at Harrington Station 	 25-1
Richard L. Chambers", James J. Spivey, and Dale L. Harmon
VOLUME 4
FUGITIVE EMISSIONS, DRY S02, AND
OPERATION AND MAINTENANCE
Session 18: FUGITIVE EMISSIONS I
Chatten Cowherd, Jr., Chairman
Technical Manual on Hood Capture Systems to Control
Process Fugitive Particulate Emissions 	 1-1
E. R. Kashdan", J. J. Spivey, D. W. Coy,
H. Goodfellow, T. Cesta, and D. L. Harmon
Pilot Demonstration of Air Curtain Control of
Buoyant Fugitive Emissions 	 2-1
Michael W. Duncan", Shui-Chow Yung, Ronald G. Patterson,
William B. Kuykendal, and Dale L. Harmon
Characterization of Fugitive Particulate Emissions from
Industrial Sites 	 3-1
K. S. Basden
Evaluation of an Air Curtain Secondary Hooding System	4-1
John 0. Burckle
Paper presented by William F. Kemner
"Denotes speaker
xv

-------
Section	Page
Session 19: FUGITIVE EMISSIONS II
Michael J. Miller, Chairman
Technical Manual on the Identification, Assessment,
and Control of Fugitive Emissions	5-1
Chatten Cowherd, Jr.*, John S. Kinsey, and
William B. Kuykendal
Quantification of Roadway Fugitive Dust at a Large
Midwestern Steel Mill	6-1
Keith D. Rosbury and William Kemner*
Evaluation of Street Sweeping as a Means of Controlling
Urban Particulate	7_1
T. R. Hewitt
Windbreak Effectiveness for the Control of Fugitive-Dust
Emissions from Storage Piles--A Wind Tunnel Study	8-1
Barbara J. Billman
Evaluation of Chemical Stabilizers and Windscreens for
Wind Erosion Control of Uranium Mill Tailings	9-1
Monte R. Elmore* and James N. Hartley
Session 20: DRY S02 REMOVAL I
Richard G. Rhudy, Chairman
Modeling of S02 Removal in Spray-Dryer Flue Gas
Desulfurization System 	 10-1
Ashok S. Damle* and Leslie E. Sparks
Fabric Filter Operation Downstream of a Spray
Dryer: Pilot and Full-Scale Results 	 11-1
Richard G. Rhudy and Gary M. Blythe*
Novel Design Concepts for an 860 MW Fabric Filter
Used with a Dry Flue Gas Desulfurization System	12-1
Michael F. Skinner, Steven H. Wolf,
John M. Gustke*, and Donald 0. Swenson
Start-Up and Operating Experience with a Reverse Air
Fabric Filter as Part of the University of Minnesota
Dry FGD System	13-1
J. C. Buschmann*, J. Mills, and W. Soderberg
^Denotes speaker
xv i

-------
Section	^
Spray Dryer/Baghouse Experiences on a 1000 ACFM Pilot
Plant	14-1
Wayne T. Davis*, Gregory D. Reed, and Tom Lillestolen
Session 21: DRY S02 REMOVAL II
Theodore G. Brna, Chairman
Design and Operation of the Baghouse at Holcomb
Station, Unit No. 1	15-1
B. R. McLaughlin* and R. D. Emerson
An Update of Dry-Sodium Injection in Fabric Filters	16-1
Richard G. Hooper*, Robert C. Carr, G. P. Green, V. Bland,
L. J. Muzio, and R. Keeth
Removal of Sulfur Dioxide and Particulate Using E-SOX	17-1
Leslie E. Sparks*, Geddes H. Ramsey, Richard E. Valentine,
and Cynthia Bullock
Comparison of Dry Injection Systems at Normal and
High Flue Gas Temperatures	18-1
Robert M. Jensen*, William Dunlop, George C. Y. Lee,
and Duane Folz
Acid Rain Control Options - Impact on Precipitator
Performance	19-1
Victor H. Belba*, Fay A. Horney, and Donald M. Shattuck
Session 22: OPERATIONS AND MAINTENANCE I
Richard D. McRanie, Chairman
Comparison of U.S. and Japanese Practices in the
Specification and Operation and Maintenance of
Electrostatic Precipitators	20-1
Michael F. Szabo*, Charles A. Altin, and
William B. Kuykendal
Operation and Maintenance Manuals for Electrostatic
Precipitators and Fabric Filters 	 21-1
Michael F. Szabo*, Ronald D. Hawks, Fred D. Hall,
and Gary L. Saunders
An Update of the Performance of the Cromby Station
Fabric Filter	22-1
M. Gervasi*, J. R. Darrow, and J. E. Manogue
*Denotes speaker
xvii

-------
Section
Critical Electrostatic Precipitator Purchasing Concepts. . . .23-1
Charles A. Altin* and Ralph F. Altman
Reducing Electrostatic Precipitator Power Consumption	24-1
Joseph P. Landwehr* and George Burnett
Session 23: OPERATIONS AND MAINTENANCE II
Peter R. Goldbrunner, Chairman
Design Considerations to Avoid Common Fly Ash
Conveying Problems	25-1
Gus Monahu,v and Walter Piulle
Feasibility of Using Parameter Monitoring as an Aid
in Determining Continuing Compliance of Particulate
Control Devices	26-1
Joseph Carvitti", Michael F. Szabo, and William Kemner
Air Pollution Control: Maintenance Cost Savings
from the Washing, Patching and Reuse of Bags Used
in Fabric Filters	27-1
Frank L. Cross, Jr.
Paper presented by Lutz Bergmann
Optimizing the Performance of a Modern Electrostatic
Precipitator by Design Refinements	28-1
Donald H. Rullman* and Franz Neulinger
Weighted Discharge Electrodes - A Solution to
Mechanical Fatigue Problems	29-1
John A. Knapik
PAPER PRESENTED AT THE FOURTH SYMPOSIUM ON THE TRANSFER
AND UTILIZATION OF PARTICULATE CONTROL TECHNOLOGY BUT NOT
PUBLISHED IN PROCEEDINGS
Measurement of the Electrokinetic Transport Properties
of Particles in an Electrostatic Precipitator	30-1
Wallace T. Clark III*, Robert L. Bond, and
Malay K. Mazumder
UNPRESENTED PAPER
Electrostatic Precipitator Bus Section Failure:
Operation and Maintenance	31-1
Louis Theodore, Joseph Reynolds, Francis Taylor,
Alan Filippi, and Steve Errico
^Denotes speaker
xviii

-------
Session 5: ESP: PERFORMANCE ESTIMATING (MODELING)
Leslie E. Sparks, Chairman
U.S. Environmental Protection Agency
Air and Energy Engineering Research Laboratory
Research Triangle Park, NC

-------
MICROCOMPUTER PROGRAMS FOR PRECIPITATOR
PERFORMANCE ESTIMATES
M. G. Faulkner
J. L. DuBard
R. S. Dahlin
Southern Research Institute
Birmingham, Alabama 35255
L. E. Sparks
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
ABSTRACT
The EPA/SRI mathematical model of electrostatic precipitation has been
coded in FORTRAN for the Radio Shack TRS-80 Model III or 4 microcomputer. The
TRS-80 model retains all essential features of the main-frame computer model,
but omits several user options. The precipitator simulation is divided into
four separate programs: a simplified data entry menu, a choice of a rigorous
computation of ideal collection efficiency or a rapid estimation procedure, and
approximate corrections for non-ideal gas flow and rapping reentrainment. A
supplementary program can be used to estimate the electrical operating points
that are required for the precipitator simulation. Given values of fly ash
resistivity and precipitator plate spacing, this program estimates the elec-
trical operating points by using average data from cold-side utility fly ash
precipitators. Another supplementary program can be used to compute the plume
opacity, given an optical path length and the particulate emissions computed by
the precipitator simulation. All six microcomputer programs are included in
Revision III of the main-frame computer model.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
1-1

-------
INTRODUCTION
The Electrostatic Precipitator Performance Model (ESP Model) developed by
Southern Research Institute for the Environmental Protection Agency was first
released in 1975 (1). Since that time, the model and its revisions (2,3) have
been widely used to study proposed and existing precipitators. Revision 3 of
this model (4) features a faster algorithm and many new options, including
calculation of plume opacity, V-I curve generation, flexible rapping schedule,
metric input data, and a routine to check input data for errors. The FORTRAN
code for this model requires 221,000 bytes of memory. This is well within the
memory limits of a modern main-frame computer.
In order to make this program more widely available, a microcomputer
version of it has been written. The microcomputer version uses the same algo-
rithms as the main-frame version to compute collection efficiency and plume
opacity, and to estimate operating voltages and currents if these quantities
are not known. The system for which this program was developed is the Radio
Shack Model III (or Model 4) computer with a minimum of 48 kilobytes of memory,
two disk drives, and a line printer. Although some of the newest and largest
microcomputers have enough memory for the entire main-frame program, at the
time that this project was undertaken, 48 kilobytes was about the maximum
available.
CALCULATION OF COLLECTION EFFICIENCY
Like the main-frame model, the microcomputer model performs a detailed
mathematical simulation of the precipitation process along a single gas
passage. For calculation purposes, the precipitator is divided into sections,
permitting different electrical and gas conditions in each section, and then
further divided into incremental lengths on the order of the precipitator wire-
to-wire spacing. The efficiency calculation is performed sequentially on each
incremental length to determine the total collection efficiency. The use of
small length increments allows the calculation of the changing effects of space
charge, electric fields, and particle concentrations throughout the precipi-
tator. To compensate for the different charging and collecting rates of large
and small particles, the input dust load is divided into particle size bands,
each of which is handled separately in the efficiency calculation.
A flow chart for the ESP Model is shown in Figure 1. In the ESP Model the
efficiency calculation is performed using the Deutsch-Anderson equation for
each particle size band in each incremental length. The Deutsch-Anderson
equation is given by
1 = 1 - exp (-Ato/Q)	(1)
where
n = the fractional collection efficiency,
A  the collection plate area,
w = the migration velocity of the charged particles, and
Q  the gas volume flow rate.
1-2

-------
MIGRATION VELOCITY
ENTER DATA
E FIELD
AT PLATE
SPACE CHARGE EFFECTS
PARTICLE
CHARGE
EFFICIENCY
NEW ESTIMATED
EFFICIENCY
		}	
NO
YES
NON-IDEAL EFFECTS
PRINT OUTPUT
Figure 1. Flow chart for the ESP Model.
JEST
EFFICIENCY
CONVERGENCE
1-3

-------
The migration velocity is calculated from
w = qEpC/6irau	(2)
where
q = the charge of the particle,
Ep = the electric field at the plate,
C = the Cunningham slip correction factor,
a = the radius of the particle, and
u = the viscosity of the gas.
Thus, the calculation of collection efficiency requires the calculation of the
charge for each particle size and of the electric field at the plate, in each
incremental length. These quantities can be obtained quickly by estimation
formulas or more rigorously by using a unipolar ionic charging theory (5) and a
detailed analysis of the electric field in the interelectrode space of the
precipitator.
In the particle charging calculation, the entire operating current density
specified by the user is assumed to be useful for unipolar ionic charging.
No allowance is made for back corona. In the rigorous charging calculation,
the algorithm sums the time rates of charging for field charging (until the
saturation limit is reached for a given particle size), charging by thermal
diffusion of ions, and field-enhanced thermal diffusion charging. The algo-
rithm calculates a fourth-order Runge-Kutta numerical solution to the dif-
ferential equation for the charging rate. The charge acquired by an aerosol
particle depends on the particle size, the density of ions, the exposure time,
and the average electric field in the interelectrode space.
In the rigorous computation of the electric field at the plate, the ESP
Model does not assume any explicit spatial distribution of suspended charged
particles. Instead, the ESP Model performs a self-consistent iterative compu-
tation of the interelectrode space charge distribution in each computational
length increment, including both ionic and particulate space charge. The
algorithm simultaneously solves Poisson's equation and the current continuity
equation in the interelectrode space. Starting with an estimate of space
charge density at the corona electrode and the electrostatic potential distri-
bution for wire-plate geometry, local values of electric potential and space
charge density are computed alternately, one from the other, on a rectangular
coordinate grid until convergence is obtained. The results at the boundary
representing the collecting plate are then tested against the current density
specified by the user. If the test fails, the estimate of space charge density
at the corona electrode is adjusted and the whole procedure is repeated until
the boundary condition is satisfied. This procedure requires the simplifying
assumption that the average combined motions of ionic and particulate charge
carriers can be described by a single effective mobility, reduced from the
ionic mobility. The ESP Model is not sensitive to the geometrical details of
different types of corona electrodes or different collecting plate baffles.
After the ideal collection efficiency is calculated, the ESP Model cor-
rects the efficiency for various non-ideal effects. Corrections for gas
1-4

-------
sneakage and non-uniform gas velocity are applied by adjusting the migration
velocity for each particle size band by a factor based on theoretical and
experimental studies of a pilot-scale precipitator (6). Following this, the
losses due to rapping reentrainment are calculated. This correction is based
on the amount of mass collected in the last section of the precipitator. A
fraction of the collected mass is fit to a log-normal size distribution and
added by size bands to the dust penetration predicted by the ideal efficiency
calculation. The collection efficiency is then recomputed based on the
combined effluent. The log-normal size distribution used to describe the
reentrained particles has default parameters of a 6 ym mass median diameter
and a standard deviation of 2.5. Other size distribution data may be used in
addition to these values if desired. However, the total mass of the reen-
trained particles is fixed by a relationship based on data derived from a study
of rapping reentrainment in six utility fly ash precipitators (7).
In order to fit these calculations into a 48 kilobyte TRS-80, it was
necessary to reduce the program in scope. A comparison of the features of the
TRS-80 Model and Revision 3 of the main-frame model is presented in Table 1.
Those functions which are included in the TRS-80 Model are divided into six
programs. In addition, in order to fit the rigorous efficiency calculation
into the TRS-80, it was necessary to reduce the sizes of some of the arrays,
as shown in Table 2.
Four of the six programs apply directly to the collection efficiency
calculation. These are shown in Figure 2. Three of these four programs must
be used in order to calculate the efficiency. The programs are connected by
data files stored on diskette, so that all of the programs need not be used in
one session.
The first task in the use of the TRS-80 ESP Model is to enter data into
the computer. The ESP Model requires a rather complete specification of the
precipitator operating parameters. The necessary parameters are summarized in
Table 3. These data are entered into the TRS-80 using the first program in the
sequence, named DATA. This is an interactive data entry routine which prompts
the user for data, displays the data on the screen, and allows the data to be
easily changed. Figure 3 shows a sample data entry screen. This program also
reads old data files so that the data can be updated. The final data from this
routine is saved in a file for later use in the efficiency calculation.
As Figure 2 implies, the user has a choice of two routines for calculating
the ideal collection efficiency. Program MAIN uses the rigorous calculations
of electric field and particle charge. Program ESTIMATE calculates efficiency
using much simpler estimations of field and charge. The difference in these
routines is that the rigorous calculation is more accurate and takes much more
time to run. Table 4 shows a comparison of calculation accuracy for one
utility precipitator. The calculated results from the DEC 2020 computer are
taken as the correct values. Note that in this case, the rigorous and
estimated microcomputer calculations fall about equally on either side of the
main-frame value. In other examples, the differences between the rigorous and
estimated values have been much greater.
Table 5 shows the times required to run the data set used for Table 4.
The rigorous run time is a factor of 16 greater than the estimation time for
1-5

-------
TABLE 1. OPERATIONS PERFORMED BY TRS-80 ESP MODEL
Main-Frame Calculations
Availability in TRS-80
1.	Inlet mass distribution
2.	Charge on particles (estimation or rigorous)
3.	Field at plate (estimation or rigorous)
4.	Sneakage and non-uniform gas flow
5.	Rapping reentrainment (empirical or dynamic)
6.	Plume opacity
7.	Input data error check
8.	Unknown electrical operating conditions
9.	Metric data
Yes
Two separate programs
Two separate programs
Yes
Empirical only
Separate program
No
Separate program
No
TABLE 2. ARRAY SIZES
Variable
TRS-80
Maximum
Main-Frame
Maximum
Main-Frame
Typical
Particle size bands
ESP electrical sections
ESP incremental lengths
Non-ideal conditions
Rapping data
Characters in title
Electric field calculation grid
12
5
30
5
2
45
8 x
20
10
99
15
10
80
15 x
15
16
4
40
3
1
40
10 x
10
1-6

-------
ADJUST
MAIN
DATA
ESTIMATE
Figure 2. TRS-80 ESP Mode! program sequence.
TABLE 3. ESP MODEL INPUT PARAMETERS
1)	Precipitator geometry in each gas passage of each electrical field.
2)	Operating voltage and current in each electrical field.
3)	Inlet aerosol parameters.
Gas temperature, pressure, viscosity
Gas velocity
Particle mass loading
Particle size distribution
4)	Non-ideal correction parameters.
Number of baffled sections
Fractional gas sneakage past each baffled section
Normalized standard deviation of the gas velocity distribution
1-7

-------
###################################################
DATA FOR SECTION 1
PAGE 2 OF 3
############################
H#####################
VOLTAGE
-> 0.000E+00 V
CURRENT
O.OOOE+OO A
PLATE AREA
O.OOOE+OO FT2
CORONA WIRE RADIUS
0.000E+00 IN.
WIRE-TO-PLATE SPACING
0.00 IN.
WIRE-TO-WIRE SPACING
0.00 IN.
PRESS > FOR NEXT PAGE, < FOR PREVIOUS PAGE, ? = HELP, @ = RET
Figure 3. Sample data screen for the TRS-80 ESP Model.
TABLE 4. CALCULATED RESULTS
DEC 2020	TRS-80	TRS-80
Rigorous	Rigorous Calculation Estimation
Quantity	Calculation	Value	Error	Value Error
No rap penetration, %
No rap efficiency, %
Penetration with
rapping, %
Efficiency with
rapping, %
0.19	0.17
99.81	99.83
0.32	0.30
99.68	99.70
10	0.21 10
99.79
6	0.35 9
99.65
TABLE 5. CALCULATION TIME COMPARISON
DEC 2020
TRS-80
Rigorous calculation
Estimation
3 min
11 sec
3 hr
2 min
1-8

-------
the DEC computer. But for the TRS-80, this difference is a factor of 90. In
fact, the TRS-80 took about the same amount of time to run the estimation as
the DEC did for the rigorous calculation. However, other factors inhibiting
the use of a main-frame computer should be considered  factors such as
terminal availability, computer Load, job priority, and job cost. Thus, in
spite of the greater computer time, a microcomputer may be the instrument of
choice. The choice of running a rigorous or estimated efficiency calculation
must be determined on a case-by-case basis. Although the rigorous calculation
is more accurate, there are many cases in which the accuracy difference is
negligible. In such cases, the faster run time makes the estimation routine
the procedure of choice.
The third routine in the program sequence shown in Figure 2 is ADJUST.
This routine reads the data files produced by DATA and MAIN (or DATA and
ESTIMATE). The data are then corrected for sneakage, non-uniform gas velocity,
and rapping reentrainment. Finally the data are printed. ADJUST prints the
full input data set, followed by particle charge data, precipitator outlet
statistics by particle size band, and overall outlet statistics with and
without rapping.
CALCULATION OF PLUME OPACITY
An additional program may be used to calculate plume opacity. This pro-
gram, named OPAC, reads the data files created by DATA and MAIN (or DATA and
ESTIMATE), makes the specified non-ideal corrections, and calculates opacity
for each non-ideal data set. The only additional data needed are optical path
length (stack diameter) and complex index of refraction (if different from the
default value of 1.5 - O.Oi).
The mathematical relationships required to predict plume opacity were
derived from the theory of light scattering and absorption by Gustov Mie in
1908 (8,9). These relationships, which are used in all opacity calculations,
have been discussed in detail by Ensor (10). The equations have been tested
against experimental data obtained under controlled laboratory conditions and
found to be accurate to within a typical error of 5 percent in the total
extinction coefficient. The major source of error in opacity calculations
appears to be the accuracy of the particle size distribution used in the
opacity calculation. Opacity is a sensitive function of particle size and
concentration. Opacity calculations have been based on log-normal particle
size distributions with varying degrees of accuracy. The present trend is to
use a particle size histogram instead of a log-normal distribution, as precip-
itator effluents tend to deviate significantly from the log-normal curve. In
this program, the plume opacity is calculated from the particle size histogram
generated in the collection efficiency calculation.
Figure 4 compares calculated opacities with in-stack transmissometer
readings made during field tests. These data are from tests of 11 pulverized-
coal-fired power plants. To compare accuracy, opacities were calculated using
both particle size histograms and log-normal curves. The particle size histo-
grams were measured by cascade impactors during the field tests. Log-normal
curves were then derived from the histograms. The opacity values calculated
1-9

-------
iiii i i i r
O MEASURED SIZE DISTRIBUTION
 LOG-NORMAL SIZE DISTRIBUTION  <. /'
&
<0/
<&/
J?/
A/
rS>S
/
 
-------
from the particle size histograms, shown by open circles in Figure 4, have
an average error of 26 percent. The opacity values calculated from the log-
normal curves, shown by closed circles in Figure 4, have an average error of
59 percent. The data comparison shown in Figure 4 clearly demonstrates the
superiority of basing opacity calculations on particle size histograms.
ESTIMATION OF ELECTRICAL OPERATING CONDITIONS
Whenever possible, the electrical operating conditions used to model ESP
efficiency should be based on field test data. However, for cases where this
is not feasible, another program may be used to estimate the electrical opera-
ting points. The empirical approach used in this program is the result of a
critical review and compilation of comprehensive field test data available to
SRI, sponsored by the Electric Power Research Institute (11). The data on
electrical operating conditions were restricted to cold-side fly ash pre-
cipitators. There were 17 complete data sets, obtained on 13 different
precipitators at 12 different plant sites. The principal data consisted of
V-I characteristic curves for each electrical section, the V-I operating point
in each electrical section, and simultaneous in situ fly ash resistivity
measurements. Fly ash resistivity was determined by the spark method, using
the SRI point-plane resistivity probe. It should be emphasized that these data
represent operating precipitators having a variety of electrode geometries,
dirty electrodes, and particle-laden gases.
The relationship between the operating voltage and current density and
the fly ash resistivity was developed by (1) obtaining a correlation between
measured values of fly ash resistivity and estimated useful current density,
and (2) obtaining a correlation between useful current density and operating
voltage. The prediction of operating current density is based on a least-
squares fit to the precipitator data for each electrical field. Figure 5 shows
the data for the first electrical field. The solid line represents the least-
squares fit. The dashed lines flanking the solid line represent the 90 percent
confidence interval.
Empirical relations for useful current density, averaged over all elec-
trical fields of a precipitator, have been given by H. J. Hall (12) and have
been used to predict electrical operating points. The most conservative
relation given by Hall corresponds to a restriction of the average electric
field inside a fly ash layer on the collecting plate (E = PJ) to less than
1 kV/cm. This line is superimposed on Figure 5. The more recent data show
that this line is overly conservative for high-resistivity fly ash.
In the program named VI, the operating current densities are obtained from
the specified ash resistivity and the least-squares fitted line for each field
of the precipitator. Then another set of correlations is used to obtain the
average interelectrode electric field in each precipitator field. The opera-
ting voltages are obtained by multiplying the average electric fields by the
specified wire-to-plate spacing.
1-11

-------
100
SECTION 1
log (J) = 6.455 - 0.5013 log (p>
R2 = 0.79
10"
ASH RESISTIVITY, ohm-cm
Figure 5. Current density vs. fly ash resistivity for section 1.
1-12

-------
SUMMARY
The microcomputer programs described in this paper are written in FORTRAN
specifically for the Radio Shack TRS-80 Model III or 4 computer. With the
exception of program DATA, these programs may be used with any microcomputer
which has a FORTRAN compiler. Program DATA uses routines which are specific
to the TRS-80 computer. Transcribing this program to another computer would
require extensive rewriting of the interactive screen portion of the program.
An alternative approach would be to replace the interactive screen portion with
statements vrtiich display the data in list form and ask for the number of the
data item to be changed. This would be followed by an appropriate GO TO state-
ment and prompt statements pertaining to that specific data item.
The functions of the six microcomputer programs described in this paper
are included in the main-frame ESP Model. These functions are:
1.	calculation of collection efficiency,
2.	calculation of plume opacity, and
3.	estimation of operating voltages and currents.
Table 6 summarizes the advantages and disadvantages of each of these calcula-
tions, as performed by the microcomputer. A major advantage is that these
calculations have been successfully tested against field data. The accuracy
of the collection efficiency calculation compared to measured efficiencies is
discussed in the following paper (13). The accuracy of the opacity calculation
is discussed in this paper and summarized in Figure 4. The V-I curve generator
is an empirical relationship derived from fly ash precipitators operating under
typical conditions with dirty electrodes and particle-laden gases. These
programs allow a microcomputer user to simulate an existing or planned electro-
static precipitator under realistic operating conditions.
REFERENCES
1.	Gooch, J. P., J. R. McDonald, and S. Oglesby, Jr. A Mathematical Model of
Electrostatic Precipitation. EPA-650/2-75-037, U.S. EPA, IERL, Research
Triangle Park, NC, 1975.
2.	McDonald, J. R. A Mathematical Model of Electrostatic Precipitation
(Revision 1): Volumes I and II. EPA-600/7-78-llla,b, U.S. EPA, IERL,
Research Triangle Park, North Carolina, 1978.
3.	Mosley, R. B., M. H. Anderson, and J. R. McDonald. A Mathematical Model
of Electrostatic Precipitation (Revision 2). EPA-600/7-80-034, U.S. EPA,
IERL, Research Triangle Park, NC, 1980.
4.	Faulkner, M. G., and J. L. DuBard. A Mathematical Model of Electrostatic
Precipitation (Revision 3): Volumes I, II, and Source Code Tape. EPA-
600/7-84-069a,b,c. (NTIS PB84-212-679, PB84-212-687, PB84-232-990). U.S.
EPA, IERL, Research Triangle Park, NC, 1982.
1-13

-------
TABLE 6. PROGRAM ADVANTAGES AND DISADVANTAGES
EFFICIENCY MODEL
Advantages
	Menu driven data entry
	Allows different electrical conditions in each ESP field
	Broken down into logical program blocks
	Accuracy tested against field test data
Disadvantages
	Longer run time than main-frame version
	Limited memory size
OPACITY MODEL
Advantages
	Very little input data (most data is generated by efficiency model)
	Accuracy tested against field test data
Disadvantages
	Requires ESP efficiency model
V-I CURVE GENERATOR
Advantages
	Derived from field test data
	Estimates electrical operating points in each ESP field
Disadvantages
	Applies only to cold-side fly ash precipitators
1-14

-------
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.	Preszler, L., and T. Lajos. Uniformity of the Velocity Distribution of
a Flowing Gas Upon Entry into an Electrostatic Precipitator. Staub
Reinhalt. Luft (in English), 32(11):1-7, 1972.
7.	Gooch, J. P., and G. H. Marchant, Jr. Electrostatic Precipitator Rapping
Reentrainment and Computer Model Studies. EPRI Report FP-792, Electric
Power Research Institute, Palo Alto, CA, August 1978.
8.	Mie, G. Beitrage zur Optik truber Medien, Speziell Kolloidaler
Metallosunge. Ann. Fhysik. 25:377-455, 1908.
9.	van der Hulst, H. C. Light Scattering by Small Particles. John Wiley
and Sons, New York, 1957.
10.	Ensor, D. S. Smoke Plume Opacity Related to the Properties of Air
Pollutant Aerosols. Ph.D. Dissertation, University of Washington, 1972,
11.	DuBard, J, L., and R. F. Altman. Prediction of Electrical Operating
Points for Use in a Precipitator Sizing Procedure. EPRI Report CS-2908,
Electric Power Research Institute, Palo Alto, CA, April 1983.
12.	Hall, H. J. Trends in Electrical Energization of Electrostatic Precipi-
tators. Electrostatic Precipitator Symposium, Birmingham, AL, February
23-25, 1971.
13.	DuBard, J. L., and R. F. Altman. Analysis of Error in Precipitator Per-
formance Estimates. The Fifth Symposium on the Transfer and Utilization
of Particulate Control Technology, Kansas City, MO, August 27-30, 1984.
1-15

-------
ANALYSIS OF ERROR IN PRECIPITATOR PERFORMANCE ESTIMATES
J. L. DuBard
Southern Research Institute
Birmingham, Alabama 35255
R. F. Altman
Electric Power Research Institute
Chattanooga, Tennessee 37411
ABSTRACT
The EPA/SRI mathematical model of electrostatic precipitation has been
analyzed for the propagation of error in estimated input data and for the
agreement between measured and computed collection efficiency, using both
measured and estimated input data. The estimated aerosol data and electrical
data are based on measurements performed at 33 utility boilers and 13 utility
fly ash precipitators. The numerical parameters used to make approximate
corrections for non-ideal gas flow are based on a comparison of 18 sets of
measured and computed efficiency, obtained for precipitators of SCA from 135 to
745 ft2/kACFM. Numerical confidence intervals in the efficiency computed using
estimated input data are established for four test cases with fly ash resis-
tivity of 5E10 and 1E12 ohm-cm and SCA of 250 and 500 ft2/kACFM. In each case,
the lower 90% confidence margin in computed collection efficiency requires an
increase in SCA of approximately 125 ft2/kACFM to attain the nominal collection
efficiency.
2-1

-------
INTRODUCTION
The electrostatic precipitator performance modeL (ESP MODEL) is described
in documentation available from the National Technical Information Service (1).
This paper describes an error analysis of the computation of fly ash collection
efficiency. The numerical specification of error makes the ESP MODEL more
reliable for users. The objectives of this work were to develop
1)	a numerical procedure for estimating input data for the computation:
	fly ash particle size distribution and inlet mass loading,
	operating voltage and current in each electrical field,
	non-ideal gas flow correction parameters, and
2)	a numerical specification of the effect of error in estimated input
data on the error in computed collection efficiency.
The first part of this paper describes tests of the intrinsic accuracy
of the numerical simulation of electrostatic precipitation. The measured
collection efficiencies from 18 field tests of utility fly ash precipitators
are compared with ESP MODEL computations, using measured input data for fly ash
aerosol parameters and electrical operating points. This comparison yields an
objective determination of the parameters used by subroutine ADJUST to make
approximate corrections for non-ideal gas flow.
The second part of this paper describes an analysis of the propagation
of error from the input data to the computed collection efficiency. This
procedure requires a means of estimating input data for fly ash aerosol
parameters and electrical operating points, and a numerical determination of
the uncertainty in those estimates. The error analysis is based on compil-
ations and correlations of the precipitator data base available to Southern
Research Institute. Finally, the input data are sorted into five groups that
are approximately uncorrelated mathematically, in order to obtain a numerical
estimate of the uncertainty in computed collection efficiency.
ESP MODEL COMPUTATIONS WITH MEASURED INPUT DATA
The precipitator performance data base is summarized in Table 1, in order
of increasing specific collection area (SCA) of the precipitators. Table 1
presents 18 different tests of 14 different cold-side utility fly ash precipi-
tators. Three of the precipitators were tested with and without flue gas
conditioning. One precipitator was tested at both normal and reduced boiler
load. Using measured input data for fly ash aerosol parameters and electrical
operating points, and two different sets of correction parameters for non-ideal
gas flow, the ESP MODEL was executed for each precipitator.
2-2

-------
TABLE 1. DATA BASE FOR PRECIPITATOR COLLECTION EFFICIENCY
Test Coal Gas T Resistivity	Duct	SCA	Efficiency
Number Type	F	ohm-cm	Inches ft2/kACFM percent
3
SB
260
5E11
9
135
77.
4
SB
260
1E11
9
135
97.5
22
B
260
2E9
9
150
95.5
21
B
260
2E9
9
155
99.7
27
L
370
1E10
9
165
98.5
7
SB
250
7E12
9
175
91.3
25
B
285
2E9
9
175
88.
26
B
260
2E9
9
220
99.3
2
B
315
2E10
11
245
99.6
10
B
315
2E10
10
255
99.8
11
SB
320
1E12
9
310
88.
12
SB
320
6E10
9
310
99.1
5
B
350
4E11
9
360
99.90
6
B
350
3E10
9
360
99.96
23
L
330
2E10
12
390
99.97
13
SB
300
2El 1
12
570
99.92
9
SB
220
5E11
9.75
585
99.8
8
SB
300
6E12
12
745
99.7
A user of the ESP MODEL must specify the fractional gas sneakage (S) in
each electrical field and the normalized standard deviation (ogas = SD/Vavg^ of
the inlet gas velocity distribution. Conservative values of those parameters
have been estimated as S = 0.10 and  0 = 0.25. Figure 1 shows the comparison
gdS
of measured and computed collection efficiency using those values. The hori-
zontal error bars show the range of scatter, day-to-day, in the measurements of
collection efficiency.
The performance of high-efficiency precipitators is consistently under-
estimated in Figure 1. The plants with measured collection efficiency less
than about 97% are outliers, as well as Plant 9. Plants 3,7, 9, and 11
operated in severe back corona. Reduced values of useful voltage and current
density (estimated from the measured V-I curves) were used in the mathematical
modeling. However, the computed collection efficiency remained higher than the
measured collection efficiency. Plant 4 is the same as Plant 3, and Plant 12
is the same as Plant 11, but tested with flue gas conditioning to eliminate the
back corona problem. Plants 22 and 25 suffered from excessive reentrainment of
collected fly ash. Plant 26 is the same as Plant 22, but tested under low-load
2-3

-------
conditions to eliminate the reentrainment problem. Figure 1 shows that when
the operating difficulties with back corona and reentrainment were removed,
the comparison between measured and computed collection efficiency became con-
sistent with the general pattern of data.
Modern, high-efficiency precipitators may be designed to reduce the gas
sneakage to as little as 5%, and a more typical measure of the inlet gas
velocity distribution is about 15% normalized standard deviation. Figure 2
shows the comparison of measured and computed collection efficiency, using
S = 0.10 and crgas = 0.15. (The vertical error bars in Figure 2 are the subject
of the second part of this paper.) In this figure, the data points above 97%
are scattered closely about the perfect agreement line. Below 97%, however,
the precipitator performance is degraded by non-ideal operating conditions that
are not well simulated in a routine execution of the ESP MODEL. This analysis
of precipitator performance data leads to two conclusions.
1)	The comparisons in Figures 1 and 2 give an indication of the intrinsic
accuracy of the numerical simulation of electrostatic precipitation. With
reliable input data, the ESP MODEL reliably computes the precipitator collec-
tion efficiency.
2)	The numerical approximations in subroutine ADJUST give a reliable simu-
lation of the degrading effects of non-ideal gas flow. The comparisons in
Figures 1 and 2 give an objective estimate of appropriate numerical values of
S and ogas that are consistent with typical gas flow in modern precipitators.
ESP MODEL COMPUTATIONS WITH ESTIMATED INPUT DATA
COMPILATION OF PRECIPITATOR DATA
The operating data that are most critical for a numerical simulation of
electrostatic precipitation are the fly ash aerosol properties and the elec-
trical operating points. The operating data available to Southern Research
Institute have been described in detail in published EPRI conference proceed-
ings (2,3). Measurements of fly ash particle size distribution and total mass
loading were obtained at 33 pulverized-coal-fired utility boilers. The
measurements were correlated with boiler operating parameters, ultimate coal
analyses, and chemical analyses of the flue gas and fly ash. The T-R secondary
operating voltage and current in each electrical field, and the V-I charac-
teristic curves, were obtained in 17 tests of 13 cold-side utility fly ash
precipitators. The electrical data were compared with simultaneous in situ
measurements of fly ash resistivity, using a point-plane probe and the spark
method.
The data for inlet mass loading are plotted in Figure 3 versus the percent
ash in the coal. The data for inlet particle size distribution were reduced to
the form of cumulative percent mass contained in particles of diameter less
than 1, 2, 5, and 10 um. The average values of cumulative percent mass are
shown in Figure 4 for the two separate groups of fly ashes from bituminous and
2-4

-------
100
99 99
Measured Penetration, %
10	1	01
0 01
|1111 I I 1 I	|llll I 1 lI7|0.01
6^
99.9 
c
o
u
<1)
o
o
o


-------
o 2 4 6 8 10 12 14 16 18 20 22 24 26
Percent Ash in the Coal
Figure 3. Fly ash mass loading versus percent ash in the coal
60
50
40
30
20
10
I I I
5 
2
1
05
0 2
0.1
J	I	L
1
10<-
t	r
ii i i |
bituminous coals
 SUBBITUMINOUS COALS
 i I t i
1
10'
Figure 4.
Particle Diameter, um
Fly ash particle size distributions for bituminous
and subbituminous coals.
2-6

-------
subbituminous coals. For each group, the average values of cumulative percent
mass lie almost exactly along a log-normal straight line. However, there is a
lot of scatter in the fly ash data from subbituminous coals, as shown by the
wide error bars. These data are presented in more detail in Figure 7.
The values of useful current density in the inlet electrical field are
shown in Figure 5 versus in situ fly ash resistivity. A least-squares-fitted
straight line and its 90% confidence interval (given by 1.645 times the
standard deviation of the fitted ordinate intercept) are superimposed on the
data. The operating voltage in each electrical field was divided by the wire-
plate spacing to obtain an average interelectrode electric field for the
electrical data correlation. The data for the inlet electrical field are shown
in Figure 6. Given a specification of precipitator wire-plate spacing, the
fitted straight line in Figure 6 becomes in effect an average V-I character-
istic curve for the inlet electrical field. With a further specification of
fly ash resistivity, the fitted straight line in Figure 5 determines the
operating point on that V-I curve. Data of the sort shown in Figures 5 and 6
form the basis of the estimation procedure for electrical operating points.
The field test data for fly ash aerosol properties and electrical oper-
ating points, and the correlations of data, have been described in detail
(2,3). Correlations of the data provide a means of estimating input data for
the ESP MODEL, and a means of estimating uncertainty in the input data.
However, the estimation procedure is influenced by the mathematical method of
the error analysis.
METHOD OF THE ERROR ANALYSIS
The penetration (P) of a fly ash precipitator, computed by the ESP MODEL,
is a dependent function of the several input data,
Pi = f(xi, y^, zi> ---),
where the subscript denotes the ith execution of the ESP MODEL. The variables
{x, y, z, 	} represent input data such as particle mass loading, electrical
operating points, and gas temperature. A sample of N independent deter-
minations of penetration is obtained (conceptually) by allowing random
fluctuations about the point {x, y, z, 	} in the multi-dimensional coordinate
space of the several input data. The probability of obtaining any particular
result P^ = f(xj_, y^, z^,	) from a particular execution of the ESP MODEL is
assumed to be governed by the "normal" or gaussian probability distribution.
The same probability distribution is assumed to govern the random fluctuations
in each of the several input data. The sample mean value of penetration is
given approximately by P = f(x, y, z, 	). The sample variance in penetration
is given by
ff 2 - _L_ i (p - P)2
P N-l i-1 1
The fluctuations in P can be calculated approximately by writing a Taylor
series expansion of P about the point {x, y, z 	}.
2-7

-------
100
1 1 Ir-r-TTT]	1	1 I I hi;
CM
E
o
c
 r-<
CO
c
Q>
Q
C
CD
U
3
O
SECTION 1
lo| IJI  6 4BB - 0.5013 log (pi
Fly Ash Resistivity, ohm-cm
Figure 5. Current density versus the fly ash resistivity
in the first electrical field.
100
T	1I V I
I I	II I I I IT	I > II I I I !
10
CM

G
<
G
CO
c
OJ
n
C
u
u
3
CJ
,7*.J5
6'
13 X
1?
u
*1
\
#11
SfCTION 1
I09IJI  5 104 5  1 1354 log (.'JI
R2' 0.99
OH-
IO1
I I	I I I I 11
J	' 	I I I I I 11
7*v
10'
n13
E/J, ohm-cm
Figure 6. Current density versus the effective interelectrode
ohm-cm of the first electrical field.
2-8

-------
50
40
1 I 'Ml
LINEAR LEAST-SQUARE FIT /
WITH 90% CONFIDENCE BAND /
- R*  0.72	/
log(mmd)  1.324 i 0.393 O
mmd  21.1 fjm	^
/
/
/
/
O /
/
&
1	' l 11 11
20
Figure 7.
2	4 6 8 10
Particle Diameter, um
ParticLe size distributions for fly ashes
from 16 subbituminous coals.
2-9

-------
p. - P = (x- - x) -. + (y- - y) + (z- - z) +	,
L	1	3x	1	9y	1	8z
where the partial derivatives are evaluated at {x, y, z, 	}. This expression
must be squared and summed in order to calculate the sample variance in
penetration. The sample variance is assumed to be approximately the same as
the square of the standard deviation appearing in the normal probability
distribution of all determinations of penetration.
In the process of squaring and summing, terms involving (x^ - x)2 are
related to the sample variance in each ESP MODEL input parameter (each
independent variable) as
N
2 _ 1 V I - 'N
a = ___ l (x,- x) .
x N-l i=l 1
If the independent variables are not correlated in their fluctuations, all
cross-product terms vanish in the calculation of <*p2j and
n2 = x2 <>2 + v2 <)2 + Z2 )2 + 
P x 3x	y 9y	z 9z
Some input data to the ESP MODEL are obviously correlated  for example, flue
gas temperature and fly ash resistivity (which influences the electrical
operating points.) Other input data, such as inlet particle mass loading and
negative ion mobility, are obviously uncorrelated. The error analysis of the
ESP MODEL proceeds by defining five groups of input data, each of which is
approximately independent of the other groups in its effect on the computed
penetration.
1)	Inlet mass loading.
2)	Negative ion mobility in the flue gas.
3)	Inlet particle size distribution (mmd and a of a log-normal distribution).
4)	Flue gas temperature (with correlated fluctuations in fly ash resistivity,
inlet mass loading, flue gas pressure and viscosity, negative ion mobility,
and flue gas velocity).
5)	Fly ash resistivity (with correlated fluctuations in operating voltage and
current density in each electrical field).
With this approximation, the calculation of variance in the penetration is a
simple sum of five numerical terms, each of the form 0x2(3P/3x)2. Each
partial derivative is evaluated at the nominal values of ESP MODEL input data
(x, y, z, 	) for a specified test-case precipitator.
Four test-case precipitators were chosen to represent approximately most
existing precipitator applications to pulverized-coal-fired utility boilers in
the United States. The four cases are listed in Table 2. The four cases are
listed in order of increasing value of the nominal collection efficiency
(100 - P) computed using nominal values (x, y, z, 	) of the ESP MODEL input
data. The nominal input data are listed in Table 3. All four test cases
represent cold-side precipitators with 9-inch plate spacing.
2-10

-------
TABLE 2. TEST CASES FOR THE SENSITIVITY ANALYSIS OF THE ESP MODEL
Case
SCA
(ft 2/kACFM)
Coal
Resistivity
(ohm-cm)
Efficiency
(%)
1
250
Subb ituminous
1 x 1012
95.6
2
250
Bituminous
5 x 1010
99. 1
3
500
Subb ituminous
1 x 1012
99.7
4
500
Bituminous
5 x 1010
99.97
TABLE 3. NOMINAL INPUT DATA FOR THE SENSITIVITY ANALYSIS OF THE ESP MODEL
1. Mechanical Parameters
Number of Electrical Sections
Electrical Section Length
Collection Area per Section
Wire-Wire Spacing
Plate-Plate Spacing
Electrical Parameters
Ash Resistivity, ohm-cm
Current Density, yA/ft2
Section 1
Section 2
Section 3
Operating Voltage, kV
Section 1
Section 2
Section 3
3(SCA 250 ft2/kACFM)
6(SCA 500 ft2/kACFM)
9 ft
33,333 ft2
6 in.
9 in.
5 x 1010
12
17
19
41
40
38
1 x 1012
2.5
3.5
6.0
34
33
33
3.	Inlet Aerosol Parameters
Gas Temperature
Gas Velocity
Gas Flow
Inlet Mass Loading
Particle Size Distribution
Ash from Bituminous Coals
Ash from Subbituminous Coals
4.	Non-Ideal Correction Parameters
Gas Sneakage per Electrical Section
Gas Velocity Distribution
300 F
5 ft/s
400,000 ACFM
3.0 gr/SCF (2.1 gr/ACF)
MMD = 16.3 ym, a - 3.4
MMD = 21.1 ym, a - 4.8
gas
0.10
= 0.25
2-11

-------
Of the five terms in each of four calculations of variance in the
penetration, an example is given here of the one term representing the inlet
particle size distribution, for Case 3. This test case is a moderately large
precipitator (500 ft2/kACFM) collecting high-resistivity fly ash (1E12 ohm-cm)
from a western subbituminous coal. The particle size distribution data from
Figure 4 are shown in individual detail in Figure 7, along with the linear
least-squares-fitted mean and 90% confidence interval.
The sensitivity of the ESP MODEL computation to the inlet mass loading of
fine particles is illustrated in Figure 8. The large difference in the two
curves in Figure 8 does not result entirely from the difference in particle
size distribution, but rather mostly from the different nominal values of
electrical operating point. That is, bituminous coal was associated with
moderate-resistivity fly ash, 5E10 ohm-cm, and subbituminous coal with high-
resistivity fly ash, 1E12 ohm-cm. Figure 8 shows the computed penetration
versus the cumulative percent mass contained in particles of diameter less
than 2 ym, only to illustrate the importance of the inlet particle size distri-
bution. This set of curves was not actually used in the error analysis.
For the error analysis, the logarithm of the mass median diameter was
treated as the independent mathematical variable. That is, the standard
deviation of the log-normal particle size distribution was fixed at the fitted
value of 4.8; while the log-normal distribution was allowed to fluctuate
within the 90% confidence interval in log(MMD). This method fixed the slope of
the solid line in Figure 7; while shifting the line sideways in the range
between the dashed lines. From Figure 7, the 90% confidence interval in
log(MMD) is  0.393, where the horizontal decade from 1 ym to 10 ym represents
one unit in log(MMD). Thus, an estimate of the standard deviation in log(MMD)
is 0.393/1.645 = 0.239. The variance is the square of this number.
A curve of computed penetration versus mass median diameter is shown in
Figure 9. The penetration for fly ash from subbituminous coals is given along
the right-hand vertical axis. The nominal value is about 0.003, corresponding
to the collection efficiency of 99.7% listed for Case 3 in Table 2. The
partial derivative required for this numerical example is given by the slope
of a straight line drawn tangent to the curve at the nominal operating point
marked by a triangle in Figure 9. The value of that slope is 8P/91og(MMD) =
-0.00895. Thus, the numerical value of this one term in the calculation of the
variance in penetration is (0.239)^(-0.00895)2 = 4.58E-06. This number appears
in the summary of results for Case 3 in Table 4.
RESULTS OF THE ERROR ANALYSIS
The numerical components of overall error in the ESP MODEL computation are
listed in Table 4. Errors arising from uncertainty in the estimated data for
inlet mass loading and negative ion mobility were computed for only two cases.
Table 4 shows that these two components contribute little to the overall error.
Far more significant are the uncertainties in estimates of the inlet particle
size distribution, the fly ash resistivity (as it affects the electrical
operating points) and the flue gas temperature (as it affects the fly ash
resistivity and electrical operating points.) These calculations demonstrate
2-12

-------
a
o
u
4-1
 " 11012 ohm cm
Figure 8.
6	10	20
Cumulative Percent Mass < 2 pm
Fractional penetration in a test-case precipitator
(500 ft2/k.ACFM) versus cumulative percent mass < 2
Vim.
jC 0.0006
to
rt
to
3 0 0005
0.0004
0.0003
0 0002
0.0001
0
10

1
1 1 1
SIX ELECTRICAL SECTIONS


SCA 500 t2/kACFM AT 300 F


O ASH FROM BITUMINOUS COALS


O" 3 4 


p* 5 k 10" ohm-cm


A ASH FROM SUBBITUMINOUS COALS


0' 4 8
 \

P" 1  10^2 ohm-cm _


A _

1
1 1 1
Figure 9.
0 006
0 005
0 004
0.003
0.002
0 001
30
40
SO
15	20
Fly Ash Particle mmd, pm
Fractional penetration in a test-case precipitator
(500 ft2/kACFM) versus mass median diameter.
2-13

-------
TABLE 4. RESULTS OF THE ERROR ANALYSIS OF THE ESP MODEL
Case
SCA
Resistivity
Efficiency
Penetration
0.6745 0
1.6449
1
250
1E12
2
250
5E10
3
500
1E12
ax2 (3P/9x)2
95.6%
0.0441
0.0235
0.0573
99.1%
0.0087
0.0053
0.0129
99.7%
0.00309
0.00213
0.00521
4
500
5E10
Inlet Mass Loading


1.03
E-7


1.80
E-10
Negative Ion Mobility


3.32
E-7


3. 72
E-9
Inlet Particle Size
6.30
E-4
5.88
E-6
4.58
E-6
7.03
E-9
Gas Temperature
1.31
E-5
3.73
E-6
1.08
E-7
6.46
E-9
Ash Resistivity
5.69
E-4
5.18
E-5
5.33
E-6
1.40
E-7
op2 = ox2 ( 3P/ 9x)2
1.21
E-3
6.19
E-5
1.00
E-5
1.57
E-7
99.97%
0.00029
0.00027
0.00065
that the ESP MODEL computation of collection efficiency is not very sensitive
to the value of inlet mass loading, within the range of fly ash data shown in
Figure 3. That is, for fixed values of other data, the outlet mass loading
will vary approximately in proportion to the inlet mass loading. However, much
higher values of inlet mass loading (as, for example, from a cement kiln or
Kraft recovery boiler) can have a very strong adverse effect on the electrical
operating points.
The calculations of error for bituminous coal and moderate-resistivity fly
ash (Cases 2 and 4) show that the variance in penetration is dominated by the
uncertainty in the estimate of fly ash resistivity. The uncertainty in fly ash
resistivity is due in part to an uncertainty in the effect of naturally
occurring SO^ vapor in the flue gas. This source of error does not enter into
the calculations for Cases 1 and 3. The equilibrium concentration of S03 vapor
is assumed to be zero in the flue gas from combustion of western subbituminous
coals. In Cases 1 and 3, the three components of variance in the penetration
are comparable within a factor of two.
Reading left-to-right across the calculated values of a 2 in Table 4, the
absolute magnitude of the variance decreases by four orders of magnitude. As
the penetration decreases, the rate of change in penetration with changes in
the ESP MODEL input data also decreases. In other words, these calculations
demonstrate that more efficient precipitators have more certainty in a
computation of their efficiency, using estimated precipitator operating data.
2-14

-------
CONCLUSIONS
One conclusion from the error analysis can be illustrated with the
vertical error bars superimposed on the data points in Figure 2. From a
comparison of Tables 1 and 2, the measured data from precipitator Test #7 are
fairly similar to the hypothetical data for Case 1 of the error analysis.
Precipitator Tests #2 and #12 can be compared with Case 2, and precipitator
Test #6 with Case 4. The variance in penetration of the test-case precipita-
tors, computed from estimated operating parameters, can be applied to the
penetration values in Figure 2 computed from measured operating parameters.
This procedure yields the vertical error bars which show the probable error, or
50% confidence interval, in computed penetration. A comparison of vertical and
horizontal error bars leads to the conclusion that the probable error in
computing the precipitator collection efficiency, using estimated input data
for the ESP MODEL, is comparable to the day-to-day scatter in measurements.
The results of the error analysis were used to estimate the additional
plate area necessary to compensate for the uncertainty in the collection
efficiency computed by the ESP MODEL. A utility fly ash precipitator was
assumed to have been designed and built using estimated operating parameters.
The measured emissions were assumed to exceed the designed emissions at the 90%
confidence limit in computed penetration. This scenario was simulated with the
ESP MODEL, for each of the four test-case precipitators, by using input data to
represent higher-than-expected values of fly ash resistivity and fine particle
concentration. Then, the SCA of each precipitator was increased until the
designed performance level was recovered. For the two test cases with SCA 250
ft2/kACFM constructed, a 45% increase in SCA was necessary to recover the
designed performance level. For SCA 500 ft2/kACFM constructed, 25% to 30% more
SCA was necessary., Thus, in all four test cases, the estimated add-on SCA was
about 125 ft2/kACFM. For SCA 500 ft2/kACFM, the computations are plotted in
Figure 10. Figure 10 shows also how the Matts-Ohnfeldt approximation
(using exponent k = 0.5) underestimates the increase in collection efficiency
with large increases in SCA. The conclusion is that the lower 90% confidence
limit in collection efficiency, computed from estimated precipitator operating
parameters, corresponds to a design margin of approximately 125 ft2/kACFM.
REFERENCES
1.	Faulkner, M. G., and J. L. DuBard. A Mathematical Model of Electrostatic
Precipitation (Revision 3): Volumes I, II, and Source Code Tape. NTIS
PB84-212-679, PB84-212-687, and PB84-232-990. US EPA, IERL, Research
Triangle Park, NC, 1982.
2.	DuBard, J. L., and R. F. Altman. Prediction of Electrical Operating
Points for Use in a Precipitator Sizing Procedure. EPRI Report CS-2908,
Electric Power Research Institute, Palo Alto, CA, April 1983.
3.	Dahlin, R. S., and R. F. Altman. Prediction of Mass Loading and Particle
Size Distribution for Use in a Precipitator Sizing Procedure. Ibid.
2-15

-------
0.02
99.98
CASE 4
BITUMINOUS COAL
0.029# DESIGNED
0.04
99.96
OBTAINED
0.094
99.90
0.10
99 80
020
CASE 3
SUBBI TUMI NOUS COAL
0.309 # DESIGNED
99.60
0 40
0.830 # OBTAINED
99.00
1.00
i
750
600
700
550
650
450
500
SCA, ft2/kACFM
Figure 10. SCA added to correct measured emissions exceeding
designed emissions at the 90% confidence Limit.
2-16

-------
USE OF A MBILE ELECTROSTATIC PRECIPITATOR
FOR PILOT STUDIES
Robert R. Crynack, John D. Sherow
Wheelabrator Air Pollution Control
One of the Signal Companies
5100 Casteel Drive
Coraopolis, PA 15108
ABSTRACT
This paper describes the use of a mobile electrostatic precipitator for
pilot studies. The advantages and disadvantages of using and operating a
pilot ESP are reviewed along with cost estimates. The paper includes a
description and the features of the mobile ESP, which include variable
plate spacing, three field operation, tapped transformers and reactors,
automatic controls, and the ability to quickly change electrodes. Four
case studies using the mobile ESP are discussed. The first study was to
determine the scale factor between the pilot unit and a cold side, full
scale unit and to compare the performance of various electrode
configurations. The second study was to obtain performance data on a new
fuel, solvent refined coal. The third study was to evaluate ESP
performance where a conditioning agent was injected on a hot side, high
resistivity application. The fourth study was the application of an ESP
following a spray dryer operating under various conditions.
3-1

-------
MOBILE ESP EQUIPMENT DESCRIPTION
The Wheelabrator Air Pollution Control (WAPC) mobile electrostatic
precipitator (MESP) is a three field unit complete with automatic voltage
controls, transformer/rectifier sets, rappers, hoppers and an induced
draft fan. The MESP is mounted on a trailer 8 ft. wide by 54.5 ft. long
by 13.5 ft. high, as shewn in Figure 1. Each of the three fields consists
of sets of discharge electrodes and collecting plates. Although almost
any kind of discharge electrode or collecting plate may be installed in
the MESP, the basic design is set up for normal WAPC discharge and
collecting electrodes as shewn in Figure 2 and Figure 3, respectively.
The collecting plate is 6.5 ft. high and consists of four
interlocking strips, providing a length of 6.25 ft. for each field. The
plate spacing can be varied to produce gas passage widths from 10 inches
to 20 inches in 2 inch increments. The variable plate spacing results in
varying numbers of gas passages and collecting areas. A summary of this
information is provided in Table 1.
TABIE 1. NUMBER OF GAS PASSAGES AND COLLECTING AREA FOR
VARIOUS PLATE SPACINGS
PLATE SPACING S PASSAGES	COLLECTING AREA
(INCHES)	(#)	(SQ.FT.)
10	5	1218
12	4	975
14	3	731
16	3	731
18	2	488
20	2	488
The standard discharge electrodes used in the MESP consist of pipe
frames that are 5 ft. high, 6.25 ft. long, and have 12 wires per frame.
Two standard wires are normally available in the pipe frame, a star-shaped
and a spiked wire.
There are three large top access doors, which allow for removal and
installation of internal components. Four smaller oval shaped man doors
are provided on the side for inspection and maintenance.
3-2

-------
FIGURE 1. MOBILE ELECTROSTATIC PRECIPITATOR
3-3

-------


V

	
0
SPIKED
J	STAR
FIGURE 2. STANDARD DISCHARGE ELECTRODES
FIGURE 3. STANDARD COLLECTING ELECTRODE
3-4

-------
Each collecting curtain is individually rapped at the bottom with a
rotating, falling hammer. All the discharge electrodes and pipe frames in
each field are cleaned with a side mounted vibrator. Both the plate
rappers and the discharge electrode vibrators are manually controlled for
each field.
The ash or dust removal system consists of two trough hoppers per
field. Each trough hopper has a screw conveyor which empties into a
pyramidal hopper at the end of the screw conveyor. The ash is then
emptied into 55 gallon drums using an industrial vacuum system.
The induced draft fan is capable of drawing 9,000 acfm at 700F.
The volumetric flew rate is controlled by an opposed louver damper. The
specific collecting area, SCA, (square feet per thousand acfm) for varying
gas volumes and varying plate spacings are given in Table 2. Gas
velocity as a function of gas flow and plate spacing is given in Table 3.
TAB IE 2. SPECIFIC COLLECTING AREA* AS A FUNCTION OF
GAS FLOW AND PLATE SPACING
FLOW RATE
(ACFM)
10
12
PLATE SPACING (INCHES)
14 16 18
20
3000
406
325
244
244
163
163
5000
244
195
146
146
98
98
7000
174
139
104
104
70
70
9000
135
108
81
81
54
54
~SPECIFIC
COLLECTING AREA
IN FT2/1000 ACFM


TABLE 3. GAS VELOCITY*AS A FUNCTION OF GAS FLOW AND PLATE SPACING
FLOW RATE
(ACFM)
10
12
PLATE !
14
SPACING
16
(INCHES)
18
20
3000
1.9
1.9
2.2
1.9
2.6
2.3
5000
3.1
3.2
3.7
3.2
4.3
3.9
7000
4.3
4.5
5.1
4.5
6.0
5.4
9000
5.5
5.8
6.6
5.8
7.7
6.9
*GAS VELOCITY IN FT/SEC
3-5

-------
The MESP is equipped with automatic voltage controls and high voltage
transformer/rectifier sets. Variable tap reactors are provided in the
primary control circuit to produce an impedance of 30%, 50%, or 70%.
Variable taps are also provided on the transformer/rectifier sets to
provide a rated output of 45 KV, 60 KV, and 75 KV for the same primary
voltage. A polarity switch provides for selection of positive or negative
corona. The secondary current rating of the transformer/rectifier sets
is 50 ma. With the maximum collecting area for five 10 inch gas
passages, a current density of 0.12 ma/ft2 can be obtained.
INSTALLATION CF MESP
Even though the MESP is a self-contained unit, installation and setup
at the test site generally takes about one week. High voltage feed-
through insulators, insulator housings, hoppers, rappers, and vibrators
are shipped separately and must be installed en the unit before
operation. To provide access under the mobile ESP for installation of
the pyramidal hoppers and to facilitate ash removal, the mobile ESP is
normally raised off the ground an additional one to two feet. Inlet and
outlet ductwork must be connected to the MESP along with 480 volt, three
phase paver, if the gases are to discharge into the atmosphere rather
than being returned to the process ductwork, a stack must be erected. A
support frame surrounding the fan is capable of sustaining the weight of
five 10 ft. long sections of stack. Guy wires are used for stabilizing 30
to 50 ft. high stacks.
OPERATION AND TESTNG
Test coordination and operation of the MESP is handled by one
engineer or technician. Normal operation of the MESP includes such tasks
as (1) adjusting gas flew, (2) obtaining voltage-current curves, (3)
adjusting and monitoring operating voltages and currents, (4) operating
the manual rapping system, and (5) removing and disposing of ash.
The size of the test crew depends upon the types of testing to be
conducted. For conducting particle mass concentration tests, standard EPA
Method 5 or 17 is used. Only one person is required to set up and conduct
these tests at each test location. If inlet and outlet tests are
conducted, then a two man test crew is necessary. However, for expendency
a third test crew member is valuable. If laboratory facilities are
available on site, the third test crew member can process the samples and
filters and provide results on the following day. Availability of
immediate results of tests are important in determining test validity and
in making decisions on hew to proceed with the test program. This is an
effective cost and manpower procedure that is beneficial for a successful
field test program. Additionally, the third test crew member can assist
the other two test crew members in their routine, time consuming
activities.
3-6

-------
The observed scatter of data on pilot test programs suggests that a
minimum of six tests be conducted at any given condition. Normally a set
of three emission tests are adequate for showing guarantee or compliance.
However, for research purposes variations in performance that may be
acceptable for guarantee or compliance purposes, are not acceptable in
evaluating new process technologies or in making comparative studies of
electrode geometries.
In pilot scale testing many factors must be considered to minimize
their effect on the test results. Although temperature loss, sneakage,
and in-leakage are of concern in a full-scale precipitator, they can be
devastating in pilot ESP testing. Consider first the temperature loss due
to the ductwork from the take-off point of the slip stream to the pilot
ESP. Temperature loss in ductwork is primarily due to two affects,
radiation and in-leakage. A 2 to 3 inch layer of thermal insulation
covered by a waterproof covering such as sheet metal can provide adequate
insulation to minimize heat loss due to radiation. Bv far the more
difficult loss in gas temperature is due to in-leakage. This is
particularly true of pilot studies where long runs of ductwork are
required. Generally, small sections of ductwork are assembled to form the
long duct runs. This produces many opportunities for in-leakage due to
poor sealing at the joints of the sections. It is recommended that a
heavy flange with a substantial number of bolts and a good gasket be used
at each joint. Bolts must be torqued adequately to assure adequate
compression of the gasket to minimize the in-leakage. All joints should
be tested for in-leakage before insulation of the joints is made.
Temperature loss and in-leakage can also occur in the pilot unit
itself. The three large doors, four man doors, and six pyramidal hoppers
of the MESP provide ample opportunity for a significant amount of air in-
leakage. Care was taken in the design of the door flanges and gasket
seats. Gasket material must be resilient and acid resistant to minimize
air in-leakage over an extended period of time.
Large industrial ESPs normally utilize purge air systems to prevent
contamination of insulators. The amount of purge air relative to the
large gas volume is small and does not have a significant effect on gas
temperature or ESP performance. However, the percentage of purge air can
be substantial with the smaller gas volumes in pilot units. For
relatively high resistivity flyash, purge air of the insulators in a pilot
unit can be minimized or even eliminated. However, low resistivity ashes,
particularly those with a high carbon content, require a substantial
amount of purge air. A substantial drop in temperature and the effect on
pilot ESP performance can be observed if extensive purge air is required.
In one of the pilot tests discussed later, in-leakage primarily due to
purge air, was 15 to 20 percent.
The pilot ESP can be a significant source of heat radiation and heat
loss due to a relatively large surface area. As with the ductwork, 3
inches of good thermal insulation is adequate to minimize heat loss when
3-7

-------
there is a reasonable gas flew through the mobile ESP to act as a
continuous source of heat. In the case of the MESP this is around 4,000
acfm or more. In two uses of the mobile ESP only one or two gas passages
were used. A single gas passage was used in one application where the gas
flew was less than 1,000 acfm and two gas passages were used in another
test program where the gas flow was approximately 2500 acfm. With the low
gas flew the heat loss can be substantial and results in a substantial
temperature drop across the pilot precipitator. To offset this heat loss,
supplemental heaters were required.
Just as in a full-scale precipitator, the gas sneakage can be
detrimental to the performance of a pilot ESP. With proper baffling at
the top, bottom, and sides gas sneakage can be minimized. In the MESP the
side baffling is made adjustable because of the potential change in the
number of gas passages and the width of those gas passages.
In monitoring the operation of any pilot ESP, it is important to
obtain voltage-current curves for each electrical section. For the mobile
ESP voltage-curent curves are taken on a daily basis during normal
operation. Additionally, airload voltage-current curves are obtained to
verify clean insulators and good electrical clearances. In addition to
voltage-current curves, power reaadings on all three fields are taken in
20 to 30 minute intervals. This verifies that the operation of the pilot
ESP is steady. Any variation during a test of more than 2 KV raises
concern about the operation of the precipitator or changes in the process.
For high efficiency full-scale precipitators, it has been shown that
rapper reentrainment can account for a significant amount of particulate
emissions. Because rapper acceleration levels and even the type of
rapping may be different in a pilot unit as compared to a full-scale unit,
it is difficult to account for the rapping reentrainment of full-scale
unit frcm that determined in a pilot unit. During normal operation of the
mobile ESP, testing is conducted without rapping. The rappers are
manually operated before and after each test, but not during the test.
Therefore, in analyzing the results and in predicting full-scale
efficiency, seme consideration has to be given to rapping reentrainment
losses.
CASE STUDIES
The MESP has been used in four field test programs. The first study
was to determine the scale factor between the pilot unit and a full-scale
unit with a coal fired boiler and to compare the performance of various
electrode configurations. The second study was to obtain ESP performance
data on a new fuel, solvent refined coal. The third study was to evaluate
ESP performance where a conditioning agent was injected on a hot-side,
high resistivity application following a mini-cambustor. The fourth study
was the application of an ESP following a spray dryer operating under
various conditions. Each of these four fields studies will be discussed
briefly.
3-8

-------
TEST PROGRAM #1
The purpose of this test program was (1) to determine a scale factor
between the mobile ESP and a full-scale WAPC precipitator ana (2) to
evaluate the relative performance of several electrode configurations.
Tests on these configurations were conducted at various gas volumes and
corona power levels. Tests were conducted April through June of 1982.
Table 4 shews the performance based on migration velocity of four
different electrode geometries, named A, Bf C, and D. These tests were
conducted at a gas volume of 6,000 acfim and specific corona powers of 200
and 350 watts per thousand acfm. For the higher specific corona power
each of the four electrode configurations showed a higher migration
velocity wk (centimeters/second). In comparing the performance of the
four electrode configurations, configurations A and B were the first and
second ranked configurations for both high and low corona power levels.
However, the ranking of C and D reverse for the two corona power levels.
TABLE 4. PERFORMANCE OF FOUR ELECTRODE GEOMETRIES
AS A FUNCTION OF GftS FLOW AND CORONA POWER
GAS FLOW CDRONA POWER ELECTRODE MIGRATION VELOCITY,Wk
(acfim) (watts/1000 acfm) GEOMETRY	RELATIVE
6000	200	A	1.00
B	0.94
C	0.93
D	0.93
6000	350	A	1.00
B	0.92
C	0.78
D	0.84
Table 5 shows the results of varying gas volume for one electrode
configuration. For gas volume flow rates of 4,000, 6,000, and 8,000 acfm
the specific corona pewer was kept at 200 watts per thousand acfim. As the
gas flew increased, the efficiency decreased while the residual and
migration velocity increased.
3-9

-------
TABLE 5. MESP PERFORMANCE FOR VARIOUS GAS FLCWS
GAS FLOW	EFFICIENCY	RESIDUAL	MIGRATION VELOCITY,Wk
(acfim)	(%)	(gr/acfm)	RELATIVE	
4000	99.69 .0020	1.0
6000	99.40 .0035	1.1
8000	99.34 .0042	1.4
Tests were conducted to show the effect of corona power on the
performance of an electrostatic precipitator. Using one configuration and
gas volume flow of 6,000 acfim, the corona power was reduced frcm 311 to
103 watts per ttousand acfm. Table 6 shows the efficiency, residual and
migration velocity for the four values of specific corona pcwer. The
table shows good consistency of results, with both migration velocity and
efficiency decreasing with decreasing corona power.
TABLE 6. ESP PERFORMANCE AS A FUNCTION OF SPECIFIC CORONA. PCWER
CORONA POWER
(watts/1000 acfim)
EFFICIENCY
(%)
RESIDUAL
{gr/acfm)
MIGRATION VELOCITY, Wk
RELATIVE
311
99.6
.0024
1.00
192
99.4
.0035
0.86
131
99.0
.0048
0.77
103
98.8
.0054
0.65
In order to use the mobile ESP effectively in pilot studies, it is
important to establish a scale factor between the pilot ESP and the
full-scale precipitator. This site has a full-scale WAPC electrostatic
precipitator and provided an excellent opportunity to make such a
comparison. Table 7 shows the comparison of parameters for the
operatic*! of the full-scale and the mobile ESP. It is impossible to
operate a pilot ESP with all the same parameters as the full-scale ESP.
In this study the gas velocity, current density ancl power density of the
full-scale and mobile ESP were approximately the same. The results shew
that even though the efficiency of the full-scale precipitator was much
higher and the residual much lower, the full-scale migration velocity was
lower. In comparing the calculated migration velocities, the ratio of the
mobile to full-scale values is 1.85. This would be defined as the scaling
factor for these tests. The closer to unity the scaling factor is, the
more confident one might be in sizing a full-scale ESP from pilot ESP test
results.
3-10

-------
TABIE 7. COMPARISON OF FULL-SCALE AND M3BILE ESP OPERATING PARAMETERS
PARAMETER	FULL-SCAIE MOBILE
PIATE SPACING (in)
12
12
NO. FIELDS
4
3
C&S FLOW (acfm)
436,500
6000
SCA (ft2/1000 acfm)
513
163
GAS VELOCITY (ft/s)
3.7
3.8
CURRENT DENSIGY (ma/ft2)
0.025
0.027
POWER DENSITY (watts/ft2)
1.47
1.33
CORONA POWER (watts/1000 acfm)
756
200
INLET CONCENTRATION (gr/acf)
0.525
0.637
EFFICIENCY (%)
99.87
99.40
RESIDUAL (gr/acf)
0.00055
0.0035
MIGRATION VELOCITY,Wk (cnv/s)
46.7
86.4
During this program the MESP was turned over to the host utility for
their own evaluation of the program. The purpose, description, and
results of this test program have been reported and will not be discussed
in this paper. (1)
TEST PROGRAM #2
The International Coal Refining Company was under contract frcm the
Department of Energy to study the technical feasibility and marketing
potential of solvent refined coal (SRC) fuels for oil fired boiler
applications. The technical aspects of the boiler fuel system conversion,
combustion performance, and emission control system performance were
studied through an SRC combustion test program. This program was
conducted at the Pittsburgh Energy Technology Center (PETC). Testing was
conducted on a 700 hp water tube industrial boiler, designed for oil
firing. Baseline data was conducted by firing #6 fuel oil. Feasibility
data were obtained by firing three types of SRC fuels: (1) pulverized SRC
solids, (2) SRC residual oil, and (3) SRC - water slurry. (2)
WftPC was contracted to supply and operate the MESP and to perform
mass emission tests. Tests were conducted between August 1982 and
February 1983. Most tests were conducted with a gas volume of 5,000
acfm. Results for similar precipitator operating conditions for the four
3-11

-------
fuels that were burned are summarized in Table 8. The migration
velocity varies slightly for the three SRC fuels, but all three are far
above that of the #6 fuel oil. Part of the reason for the low migration
velocity for the #6 fuel oil may be the low inlet mass concentration which
results in low efficiency.
TABLE 8.
SUMYIARY OF TEST RESULTS FOR SOLVENT
REFINED COAL
FUEL
INLETT ESP
(lb/lOTBtu)
OUTLET ESP
(lb/10T3tu)
EFFICIENCY
(%)
RELATIVE
MIGRATION
VELOCITY, Wl,
NO. 6 FUEL OIL
0.050
0.004
92.35
1.0
SRC FUEL #1
0.959
0.005
99.46
4.3
SRC FUEL #2
0.208
0.002
99.08
3.3
SRC FUEL #3
0.960
0.006
99.34
4.2
TEST PROGRAM #3
This test program was conducted to evaluate the effect of a
conditioning agent for a full-scale, hot side electrostatic precipitator.
This hot side precipitator exhibits deterioration of performance similar
to that reported many times in previous literature. The cause for the
deterioration was assumed to be the sodium depletion as described by
Bickelhaupt. (3)
The mobile ESP was used on a small oombustor burning the sane coal as
in the full-scale unit at a rate of 125 lb. per hour. Just as with pilot
precipitator tests, care must be taken with small scale oombustor tests.
The first phase of the test program was to modify the existing ccmbustor
to produce the same flyash as that of the full-scale unit. Only when this
was completed could the evaluation of the performance of the mobile ESP be
made. With a gas volume of less than 1,000 acfm, only one gas passage was
used for this evaluation. As discussed previously, with small gas volumes
the heat loss can be substantial.
Table 9 shows a summary of the operating voltages and currents of
the precipitator with and without the conditioning agent. Due to a
transformer problem that developed due to shipping of the mobile ESP,
fields #2 and #3 were tied together and powered frcm one
transformer/rectifier. It can be seen that the inlet field power changed
very little, but the outlet two fields deteriorated significantly and then
improved substantially with the addition of the conditioning agent.
3-12

-------
T&BLE 9, MESP POWER LEVELS FOR ESP CONDITIONING STUDY
CONDITION
w1

^2,3
m2,3
INITIAL
32.0
1.5
35.0
4.5
AFTER nEJIERIQRATIQN
28.0
0.5
31.0
3.5
AFTER CONDITIONING
29.0
1.0
38.0
7.0
TEST PROGRAM #4
WAPC and the U.S. Department of Energy jointly sponsored a program to
evaluate the potential of retrofit applications of spray dryer flue gas
desulfurization (FGD) systems to existing coal-fired furnaces using
electrostatic precipitators for particulate emission control. The
objective of this project was to characterize the emissions (SO^, N0X,
and particulates) frcm a spray dryer FGD system while burning different
types of ooal, and to study the effectiveness of an electrostatic
precipitator in reducing emissions. Three different coals (lew, medium,
and high sulfur) were burned in a 500 lb. per hour combustion facility.
The flue gas was then treated by a rotary atomizer spray dryer and the
MESP. For each type of fuel a three part test program was conducted,
consisting of: (1) baseline ESP performance tests with no spray dryer
operation to determine precipitator requirements for flyash collection
only, (2) parametric tests defining SC>2 removal efficiency versus lime
stoichiometric ratios for both the spray dryer and ESP at various spray
dryer operating conditions, and (3) recycled tests in which particulate
collected by the MESP was recycled into the spray dryer system. A summary
of the data is presented in Table 10 for one ooal.
TABIE 10. ESP PERFORMANCE FOR A SPRAY DRYER APPLICATION
CONDITION
S00 REMOVAL
(%)
EFFICIENCY
(%)
RESIDUAL
(gr/acfm)
BASELINE
0
99.4
0.003
SPRAY DRYER
70
99.3
0.005
SPRAY DRYER
80
99.1
0.008
SPRAY DRYER
90
99.3
0.010
PILOT STUDY COST ESTIMATES
The cost of any pilot ESP test program is site specific and depends
upon many factors. Because of the wide range of potential costs depending
3-13

-------
upon the length of the program, scope of the program, and location of the
program, it is difficult to precisely estimate the cost of a pilot test
program. Table 11 is an approximate cost estimate for a pilot ESP test
program that would last four weeks. Most of the costs are fixed, and
therefore, extending the program beyond the four weeks would primarily
increase the testing and operating costs.
TABIE 11. COST ESTIMATE FOR A PILOT ESP TEST PROGRAM
TRANSPORTATION & INSTALLATION
5,000
FABRICATION OF DUCTWORK
10,000
INSTALLATION OF DUCTWORK
10,000
MISC. SUPPLIES
3,000
TESTING
40,000
PILOT OPERATOR/COORDINATOR
10,000

$78,000
Transportation and installation costs range frcm $2,000 to $20,000
depending upon site location. The cost of hauling the MESP from
Pittsburgh to California was $10,000 one way. However, a reasonable cost
would be $5,000. Ductwork fabrication installation costs are also site
specific and depend upon length and diameter. Approximate costs for 100
ft. of insulated ductwork, 16 in. in diameter and fabricated in 10 ft.
sections with flanges on each end, would be approximately $10,000.
Installation costs of the ductwork vrould probably be about the same.
There are always a myriad of miscellaneous consumable supplies that are
required on a pilot test program. These costs are estimated at $3,000.
The cost of testing depends upon the scope of the testing that is to be
conducted. A typical cost for a three man test crew that would conduct
two to three efficiency tests per day would be approximately $10,000 per
week. This would include testing, analysis, and reports. The cost of an
operator and test coordinator would be an additional $10,000 for the four
week program.
The estimated oost for this four week program would be $78,000. This
cost estimate does not include the cost of man power associated with
planning the test program, technical and administrative management of the
program, data analysis, or report writing. It is again emphasized that
the cost figures are rough estimates only, and are so site specific that
it is difficult to produce an estimate with reasonable accuracy.
3-14

-------
SUNMARY
The Wheelabrator Air Pollution Control mobile electrostatic
precipitator is a fully self-contained unit, complete with fan, high
voltage transformer/rectifier sets and controls. This three field unit
was designed to readily change electrodes or gas passage spacing. Care
must be taken in using a pilot ESP to minimize the effects of thermal
loss, inleakage, sneakage, and purge air. Data from four case studies
shew a wide variety of test programs are possible using a pilot ESP.
Although a scale factor between pilot and full scale ESP performance can
be determined, care must be taken in using the data. The mobile ESP
provides an excellent tool to obtain relative performance data for a
variety of operating conditions.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore, the contents do not
necessarily reflect the views of the Agency and no official endorsement
should be inferred.
REFERENCES
1.	Kisha, M., Goerss, E., Schwinn, M. Investigation of acidic matter
emissions at the Niles Generating Station. Paper presented at the
1983 Annual Meeting of the Air Pollution Control Association, Atlanta,
Georgia. June 19-24, 1983.
2.	Harrison, W., Nicholson, J., Boyd, K., Lennon, D. SRC burn test.
Southern Company Services. January, 1984.
3.	Bickelhaupt, R. An interpretation of the deteriorative performance
of hot-side precipitators. Journal of the Air Pollution Control
Association . August, 1980. 882 pp.
3-15

-------
PREDICTION OF VOLTAGE-CURRENT CURVES FOR NOVEL ELECTRODES
ARBITRARY WIRE ELECTRODES ON AXIS
Phil A. Lawless
Research Triangle Institute
P. 0. Box 12194
Research Triangle Park, North Carolina 27709
L. E. Sparks
Industrial Environmental Research Laboratory
U. S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
ABSTRACT
Theoretical prediction of voltage-current characteristics for ESPs has
been limited to single wires or infinite arrays of wires in the wire-plate
geometry. As part of a program investigating the prediction for arbitrarily
arranged electrodes, a model for wire electrodes has been developed that
accurately predicts starting voltages and V-I curves for finite numbers of
wires of mixed diameters and spacings. This model is implemented on a
microcomputer, providing an interactive approach for investigating electrode
arrangements. The accuracy of the model is demonstrated, and some of the
implications for operating ESPs are explored.
This paper has been reviewed in accordance with the IT. S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
INTRODUCTION
The prediction of voltage-current (V-I) curves for operating
electrostatic precipitators (ESPs) shows a large gap between the theory for
smooth clean wires and the observations for dirty complex electrodes. This is
because of the variety of unmodeled effects in the theory and the uncontrolled
conditions under which operating ESPs can be observed.
The calculation of corona onset voltages and V-I curves in wire-duct
precipitators, for an infinite series of wires, has been addressed theoreti-
cally by Cooperman.(1) In reality, the number of corona wires in any one
electrical section is finite and may be quite small. The corona onset
condition was experimentally determined by Peek(2) for smooth, clean wiresa
condition that leads to a different type of corona from that encountered in
ESPs. Particle space charge in operating units also shifts the characteris-
tics of the corona in a way that is difficult to model, because it involves
highly interacting effects of charging and collection as well. Back corona is
another electrical effect that complicates the modeling.
The work to be described here is aimed toward the whole problem of
predicting the operating point of a given ESP. It involves, at this stage,
the prediction of corona onset voltages and V-I relations for round wires; the
wires may be arranged conventionally or in unusual configurations. The
4-1

-------
effects of rusty/dirty wires will be included through a combination of
theoretical and experimental relations for negative corona but will be
reported elsewhere.
The eventual scope of the work will include the prediction of character-
istics for rather arbitrarily shaped electrodes, the full space charge inter-
actions, and the integration of the electrical model into a full ESP perfor-
mance model. It will be aimed at evaluating the various aspects of design and
testing that go Into a proposed ESP installation.
The current model is built around the capabilities of an advanced per-
sonal computer, but the basic aspects of the model can be run on any of
several popular microcomputers.
CALCULATION OF THE STARTING VOLTAGES
Starting Voltage Relation for Coaxial Geometry
In a coaxial ESP, the voltage on a round-wire electrode can be determined
from the electric field at the wire surface, provided no space charge is
present:
where V is the voltage, E is the electric field, r is the wire radius, and
R is the radius of the outer electrode. It is generally assumed that the
electric field at corona onset is not significantly different from the
electrostatic field just below corona onset.
The corona onset field was first determined by Peek(2) in the coaxial
geometry, among others, for a wide range of wire and tube radii and environ-
mental conditions. The electric field so determined will be called the Peek
field, E ; when it is present at the surface of a clean smooth wire, corona
will ensue.
Single Wire Between Planes
For a single wire located midway between two planes, a conformal mapping
of cylindrical geometry onto the planes results in the following relation for
the starting voltage:
V = E x r x ln(R/r ),
www	w
(1)
V = E x r x ln(d/r ),
c c w	w
(2)
where d is the equivalent cylinder radius, given by:
d = 4 x b/ir,
(3)
where b is the wire-plate separation.(1)
4-2

-------
This relation does assume that the electric field near the wire is
unaffected by the relatively distant plates; but for wire diameters much
smaller than the plate-plate spacing, this is not a major restriction.
This expression for the single wire lends itself to expression in an
infinite series of wires, where the effects of all the other wires can be
calculated.(1) When the wire-wire spacing in such a series becomes much
larger than the wire-plate spacing, the terms in the series converge so
rapidly that Equation 2 is accurate enough for most uses.
Effect of One Wire on the Starting Voltage of Another
A wire operating at a given voltage produces an electric field at all
points in space at some distance from itself. If that electric field is
integrated from the plate to the location of another wire, a potential differ-
ence arises. The resulting potential difference is a bias that must be added
to the second wire's single-wire starting potential before the Peek field can
be reached at the surface of the second wire. The integrated potential at the
second wire, because of a potential on the first wire, is given by:
V12 = Vx * F12i	(4)
where F^2 is a purely geometrical factor, given by:
cosh(irci2/2b) - cos (-rr^/^b)
T cosh(TTCi2/2b) - cos (Trr2/2b) "I
ln l cosh(itci2/2b) + cos(irr2/2b) J
[" 1 - cos(irri/2b) "j
(_1 + cos(irri/2b) J
12	r 1 - cosfirri /2b) "1	^
where c^2 is the distance between wires 1 and 2, r^ and r2 are the radii of
wires 1 and 2, respectively, and b is the wire-plate spacing, as before.
Operation of All Wires at a Common Voltage
For a multi-wire ESP with all wires operating at a common voltage, a
given wire will be at its corona threshold when the following relation is
satisfied:
vi - vd+  Fu - v
where V^ is the corona threshold of the i-th wire in the presence of all the
others; V is the single-wire corona threshold voltage of the i-th wire,
given by Equations 2 and 3; and the geometry factors, , are given by
Equation 5, with i and j substituted for 1 and 2. V. appears on both sides of
Equation 6 because it assumes that all wires are at the same voltage.
4-3

-------
In general, Equation 6 is most easily evaluated iteratively since the
geometrical factors are usually much less than 1, giving a rapidly converging
process:
First approximation:
v.. = v + T x V	(7a)
il ci f-f. ij ci
Z* i
second approximation:
n-th approximation:
Vi2 " Vcl *  Fij x Vil'	<7b>
Vi * Vci +  V) " Vin-1"	<7c>
When the wires are large in diameter or close together, the convergence
of the approximations is slower and should be monitored by means of the
changes between successive values.
Consequences and Limitations of the Interaction of Wires
One major point becomes apparent when calculating the starting voltages
of an array of equally spaced wires of the same diameter: the first and last
wires in the array have noticeably lower starting voltages than the central
wires. This is the result of having fewer strong interactions with neighbor-
ing wires and becomes most apparent when the wire-wire spacing is less than
the wire-plate spacing.
This difference in starting voltages means that attempts to raise the
overall corona onset point by grouping wires close together will fail because
the outermost wires of the grouping will have much lower starting voltages
than all the others.
On the other hand, placing a large diameter wire near a small diameter
wire will raise the threshold of the small wire considerably. This effect may
be so large that wires near large diameter mast supports do not ever reach
their corona threshold in normal operation.
The major limitation for applying the algorithm of Equations 6 and 7 is
that it does not include the electric field effects due to the presence of the
ionic space charge once a wire has begun corona generation. The ionic elec-
tric field would have the effect of raising the threshold of the others wires
above that calculated electrostatically. The difference would be most appar-
ent for wires of equal diameters, whose corona thresholds would otherwise be
quite similar. The amount of the shift due to the ion-produced electric field
depends on the current from each wire at the corona threshold being calcu-
lated, a consideration that requires knowledge of the V-I curve of each wire.
4-4

-------
CALCULATION OF WIRE V-I RELATIONS
Uniform Electrical Conditions at the Wire
The V-I relation for a single wire between plates is based upon the
solution for the similar case in the coaxial geometry. In the coaxial
geometry, Poisson's equation for the ionic space charge can be solved if
certain assumptions are made about the nature of the corona discharge.(3,4)
These amount to knowing the electric field and ion density at the boundary of
the corona. For both the negative glow and positive Hermstein glow(5), the
corona region appears to be cylindrically uniform around the corona wire. If
the diameter of the bipolar generation region is assumed to be approximately
equal to the wire radius, then the governing differential equation can be
solved analytically.
In the case of the wire-plane geometry, the electric field around the
wire can be shown to be nearly uniform for small wire diameters; therefore,
the ion generation should also be uniform. From Cooperman's solution for the
potential around the wire, the ratio of the field directed parallel to the
plate to the field directed toward the plate, at the wire's surface, is given
by:
E.I /E , = sin(ira /2b)/sinh (ira /2b),	(8)
II 1	c	c
where a is the wire radius and all other terms are as before. In the limit
as a /bCtends to zero, this ratio is unity; but for large diameter wires, the
electric field from wire to plate is stronger than that directed parallel to
the plates, as would be expected.
Provided a /b is kept to less than 0.1, the variation of E around the
wire will be less than about 0.003. At this level, the corona generation
should remain nearly uniform as well, although It is very sensitive to the
electric field variation.
V-I Curve in Cylindrical Geometry
In the cylindrically symmetric coaxial ESP, the voltage	can be computed
for a given current/length of wire analytically.(3,A) Under	the simplifying
assumption that the wire diameter is much less than the tube	diameter, an
expression for the operating voltage is:
V - V + Ec * a (1 + B)1/2 - 1 - In ) [1 + (1 + B)1/2]/2 }J ,	(9)
where V is the corona starting voltage, B is a reduced current/length of
wire, and the other terms are as before. The reduced current/meter is given
by:
B - I/fc x b2/[2ire0u(Ec x ac)2],	(10)
where I is the current, I is the length of wire, eQ is the free-space permit-
tivity, p is the mobility, and the other terms are as before. The term [b/(Ec
* a )] represents the electrostatic part of the electric field at the outer
4-5

-------
boundary. The current density at any point on the outer cylinder is related
to the current/meter by:
Of course, below the corona onset voltage, no current flows and Equa-
tion 9 is inapplicable.
Current Density in Wire-Plate Geometry for a Single Wire
The current density on the plate directly opposite the wire has almost
the same value as in the cylindrical geometry for the same wire-plate spacing.
This is a consequence of assuming a uniform generation of ions around the
circumference of the corona wire, a uniform electric field around the circum-
ference, and the fact that the ions must travel the same distance in the two
cases. In those instances where the uniformity provisions no longer hold,
then the current density will depart from the coaxial geometry value.
The implication is that the V-I density relation for points directly
under the wire is a universal one, given by the relations in Equations 9
through 11. That is, if B is written as:
then Equation 9 gives the voltage corresponding to that local current density.
To obtain the full current flowing from the wire, an auxiliary condition
is needed, relating the current density under the wire to the current density
at other positions. Such a relationship is being developed experimentally and
theoretically(6), but most ESP applications can be given with sufficient
accuracy by:
where j(6) is the current density at a point on the plate at an angle 6,
measured between a normal to the plate which passes through the wire and a
plane passing through the wire and point of interest.
In more accurate terms, the power of the cosine is a function of current
density, ranging from somewhat more than 2.5 at zero to about 4 at extremely
high current densities. The choice of 3 as an average value gives good
results over the conventional range.
The current/meter is obtained by integrating over the plate at angles for
which current may flow from the wire:
1Q = (I/)/(2irb).
(11)
(12)
j(9)/j0 = cos3(0)
(13)
(14)
"01
4-6

-------
where 0^ and 02 are the limiting angles. For the single wire case, both will
be equal to ir/2. For integral powers of the cosine function, the integral may
be written explicitly, and for the cubic power is:
l/l = 2j^b fsin(02) + sin(8i)].	(15)
Thus, the V-I relation for the single wire is obtained on the basis of a
1 m length.
Current Density in Wire-Plate Geometry for Multiple Wires
For the multiple wire ESP, similar relations hold. The starting voltages
for each wire have already been described. The normalized current densities,
B, for each wire can be calculated according to Equation 12; no compensation
for the increased electric field at the plate is needed, because of a cancel-
lation in terms involving E^.
The cosine power law has been found to remain a good representation of
the distribution of current density, even though the currents flowing from
adjacent wires do not overlap.(6) If the angles 0i and 02 are computed from
the points on the plate midway between adjacent wires, then the total current
from a given wire is adequately represented by Equation 15.
As was the case with the starting voltage, these relations do not take
into account the effects of the ionic space charge on the currents flowing
under adjacent wires. Nonetheless, experimental evidence does support the use
of these relationships, even for tightly spaced wires.
COMPARISON OF STARTING VOLTAGES WITH THEORY
In addition to the electrode effects which are at the heart of this work,
there are polarity effects for the corona starting voltage. Since the direc-
tion of movement of the ion-generating free electrons is different for posi-
tive and negative polarity, it is to be expected that there would be differ-
ences in the critical field for self-sustained ionization. Based on V-I
curves measured in a small coaxial ESP, the following expressions for the
critical fields have been derived. These are similar to the one determined by
Peek but clearly distinguish between the positive and negative coronas.
Positive:
E	- 3.126E06[d + 0.0266(d /a )1/2].	(16a)
c	r	r c
Negative:
E	= 3.126E06[d + 0.0301(d /a )1/2J.	(16b)
c	r	r c
Here d is the gas density relative to its value at normal temperature and
pressure. The constants were chosen by least squares fits, requiring the
leading constant, the critical field at infinite radius, to be the same for
both polarities.
4-7

-------
With these relations for the critical field, experimental values were
compared with the predicted ones, and the results are in Table I. The
measurements were made at the EPA's optical ESP at Research Triangle Park,
North Carolina. It comprises a single-lane pilot-scale ESP, made of Plexiglas
and operated at ambient temperature. The apparatus is constructed so that it
can be easily disassembled in the quickest manner for adjustment or
modification.
The plate spacing in the ESP is adjustable up to 30.5 cm (12 in.), and
discharge wires can be inserted with any desired spacing. Air is circulated
through the ESP in a closed loop at an adjustable velocity, but a portion of
the flow is exhausted from the loop to prevent buildup of gaseous discharge
products. For this work, only enough air flow was used to keep the gas in the
discharge region fresh.
Each plate is 1.2 m (4 ft) square and is covered with aluminum strips
isolated from one another so that the current to each can be measured. On one
plate, the center strips are 1.27 cm (1/2-in.) wide for a distance in the
direction of flow of 25.4 cm (10 in.); the remaining strips on that plate are
5.08 cm (2 in.) wide. On the opposite plate, all strips are 5.08 cm wide.
The strips are used to measure the current from the corona wire (or wires)
independently of one another.
Several different voltmeters were used during the measurements, because
of limitations on the maximum voltage which could be applied to each. Pending
a satisfactory calibration of all the meters used, it is assumed that the
measured values are accurate to no better than 5 percent.
The predicted clean wire voltages in Table 1 are in good agreement with
the measured values, mostly within the 5 percent limit. There does appear to
be a systematic underprediction of the corona starting voltages, which may be
due to a calibration problem. Nonetheless, both measured and predicted values
show the same trends, with positive corona starting at lower voltages than
negative corona, and center wires of an array having higher starting voltages
than those at the ends of the array.
When it is possible to compare the starting voltage at the center and end
of an array, the prediction is usually better for the end wire. This is a
consequence of not including the ionic space charge effects of a given wire on
another: the center wire should have a higher starting voltage due to the
other wires' ion currents than the (electrostatic) calculation predicts.
It is also apparent that the discrepancies between theory and experiment
increase as the wires become more closely spaced. This is partly due to the
experimental difficulties in ensuring a smooth clean wire, and partly due to
inadequacies in the theory, particularly with regard to the applicability of
the Peek field criterion.
COMPARISON OF V-I CURVES WITH THEORY
For comparing the full V-I curves, polarity effects again intrude on the
values of ionic mobility chosen. In air at room temperature, both positive
4-8

-------
TABLE 1. Comparison of Calculated and Experimental Starting Voltages
Setup
<->
b
(cm)
c
(cm)
(+kV)
V
c
(-kV)
V
(+kV)
exp
(-kV)
A
0.159
15.2

+39.7
-41.7
+41.7
-43.8
A
0.159
11.A

+37. 3
-39.2
+ 38.9
-40.0
A
0.308
11.4

+55.7
-58.1
+58.0
-59.2
A
0.159
7.62

+ 33.9
-35.7
+ 35.0
-37.0
B (center wire)
0.159
11.4
*10.6
+ 38.8
-40.8
+41.7
-42.2
B (outer wire)
0.159
11.4
11.4
+ 38.0
-40.1
+40.5
-40.8
C (center wire)
0.159
11.4
8.89
+40.7
-42.8
+42.5
-43.0
C (outer wire)
0. 159
11.4
8.89
+ 38.9
-41.0
+40.8
-41.0
D (center wire)
0.159
11.4
6.35
+45.9
-48.3
+46.6
-47.8
D (outer wire)
0.159
11.4
6.35
+41.2
-43.3
+42.6
-43.0
D (center wire)
0.159
11.4
5.08

-49.5

-53.0
D (outer wire)
0.159
11.4
5.08

-46.0

-46.0
D (outer wire)
0.159
11.4
3.81
+49.4
-52.0
+49.0
-50.0
E (center wire)
0.159
11.4
3.81
+70.6
-74.3
+60.0
-64.0
E (center wire)
0.159
11.4
5.08
+52.6

+53.0

E (center wire)
0.159
11.4
6.35

-48.3

-49.6
F (center wire)
0.159
11.4
*10.6
+38.9

+40.6

F (center wire)
0.159
11.4
8.89
+40.7

+41.0

G (outer wire)
0.159
11.4
6.35
+41.2
-43.3
+42.7
-44.0
H (outer wire)
0.159
15.2

+46.6

+47.0

H (inner wire)
0.159
15.2

+65.4

+66.0

H (outer wire)
0.159
11.4

+40.6

+41.0

H (inner wire)
0. 159
11.4

+51.3

+56.0

J (outer wire)
0.159
11.4
--
+42.0

+42.5

J (inner wire)
0.159
11.4

+51.6

+53.5

Single wire between plates.
Five equal diameter wires spaced 20.3 - 22.9 - 22.9 - 20.3 cm apart
Five equal diameter wires spaced equally.
Seven equal diameter wires spaced equally.
Seven wires-center wire 0.308 cm diameter, all others 0.616 cm diameter, spaced equally.
Five wires-center wire 0.308 cm diameter, all others 0.616 cm diameter,
spaced 20.3 - 22.9 - 22.9 - 20.3 cm.
H-	W^r"-n out 0-308 cm diameter, all others 0.616 cm diameter, spaced equally.
 Mast, 3.17 cm diameter; supporting 0.308-cm wires at 7.62 cm and 22.9 cm from center
or mast.
J: Same mast aB H, plus 0.616-cm wires at 41.9 cm from center of oast.
4-9

-------
and negative ions have mobilities of about 1.8 cm2/Vs, but they are dependent
on the level of moisture. In flue gas atmospheres, the presence of
electronegative gases (acid gases) can modify the mobility as much as
temperature does. In the computer model used to obtain the present theoreti-
cal values, the relations of Lawless and Sparks(7) are used, but should be
extrapolated to other conditions with suitable caution.
Three basic cases were used to test the quality of the prediction of the
V-I algorithm: a small three-wire ESP(8), a pilot-scale ESP with five
wires(9), and a high-voltage, high-current laboratory ESP. (10) In addition,
V-I curves for a near-full-scale laboratory ESP section were available to
check the general performance of the model, although the conditions of the
wires did not correspond to the clean conditions required for the theory.(11)
The results of the comparisons of these experimental values with the
model predictions are shown in Figures 1 through 3. The first V-I relation,
from the three-wire ESP, was done with positive corona. The wires were all
the same diameter; the central wire was made of brass and the outer wires of
steel. This may explain the large difference in starting voltage observed
experimentally. Figure 1 has two theoretical curves: one with corona starting
voltages calculated as outlined above, and the other with the starting voltage
of the center wire artificially raised to about 38 kV. This is a value
corresponding to an inflection in the experimental curve, where the corona
start was observed visually on the center wire.
Although the model does not accurately predict the starting voltages of
all the wires, the V-I curve has the correct slope when corona is flowing from
all the wires. The alteration of the starting voltage for the center wire
brings the model into excellent agreement with the measurements.
Figure 2 shows data taken on the EPA pilot scale ESP, with all three
sections configured identically: wire diameter 0.308 cm, plate-plate spacing
22.8 cm, and five wires spaced 22.8 cm apart. Special care was taken to
smooth and clean the wires, resulting in nearly identical V-I curves. The
theoretical model also is in excellent agreement from the starting voltage up
to about 70 nA/cm2, which covers the usual operating conditions of industrial
ESPs. The growing discrepancy above that current density is due mainly to the
approximation in Equation 12: it has been observed that the high current
limit of the exponent of the cosine dependence is 4, which will be approached
at voltages well above the starting voltage.
Figure 3 shows two experimental situations, a normal wire size and plate
spacing, and a small wire size at larger-than-normal plate spacings. The
exact arrangements were not clear from Reference 10, but the assumption was
made that the data correspond to the central wire of an array. Even so, the
starting voltages are not in agreement with the model.
For Figure 3a, the experimental starting voltage, 37 kV, corresponds to a
single wire, while for five wires, the starting voltage for the center wire is
45 kV. Except for the starting voltage, the remainder of the curve is in
excellent agreement with the data. For Figure 3b, the experimental starting
voltage, 25 kV, cannot be obtained with a wire of diameter 1.78 x 10~2 cm, as
4-10

-------
80
Figure 1. Laboratory ESP with three wires (diameter = 0.308 cm, plate-plate
spacing = 15.2 cm, wire-wire spacing = 15.2 cm).
	 Complete model prediction.
Model with adjusted starting voltage for center wire.
140
120
-




 Section 1
o /
~/
100

a Section 2
o/


~ Section 3
o/
8/
80

 Model
o/
60



40



20



0

 *
 '
Voltage (kV)
Figure 2. Pilot-scale section with five wires (diameter = 0.308 cm, plate-
plate spacing = 22.8 cm, wire-wire spacing = 22.8 cm).
4-11

-------
SO	100 150
Vottag* (kV)
200
Figure 3a. Laboratory ESP (wire diameter = 0.277, plate-plate spacing =
30.5 cm, wire-wire spacing = 15.2 cm). Five wires assumed with
model predicting current density of center wire.
B0	100
Woltags (kV)
200
Figure 3b. Laboratory ESP [wire diameter = 1.78 x 10 2 cm, plate-plate
spacing  45.7 cm, wire-wire spacing = 22.9 cmj. Five wires
assumed; model predicts current density of center wire; starting
voltage of center wire adjusted to 25 kV.
4-12

-------
given in Reference 10; the highest reasonable voltage is about 10 kV. The
model was artificially adjusted to give the experimental starting voltage,
producing a curve in good agreement with the data.
In spite of the difficulties with Reference 10, it is useful because of
the extreme current densities and voltages found in the data. The model shows
good agreement up 1000 nA/cm2, far above the normal ESP range. The normally
excellent predictions of starting voltage are so far in error that either the
data or the description of equipment is suspect.
The model predictions and the near-full-scale V-I curves(11) were in good
agreement only at voltages well above the smooth-wire corona-start voltage.
The wire-wire spacing was large enough that the interactions between wires was
barely noticeable, but the rough condition of the wires led to corona onset
(both positive and negative) at voltages well below the smooth-wire value. It
is useful to note that the smooth-wire V-I curve, as predicted by the model,
is rapidly approached by the real curve above the smooth-wire starting
voltage.
CONCLUSIONS
The results of the initial phases of this work have demonstrated that the
theory for corona starting can be extended very well to wires of varying
spacings and diameters, using Peek's criterion for the electric field at the
surface of the wire. Such discrepancies as there are seem to be due to
imperfections in the model relating to the influence of ions from wires where
corona has started.
An important result of the starting voltage predictions is that arrays of
wires of the same size may have very different starting voltages, depending on
the wire's location within the array. This means that unusual electrode
arrangements can have important consequences for the operation of the electri-
cal section, in that some wires may fail to even reach the corona threshold
before other electrical effects limit the applied voltage. Mast electrode
supports were shown to have a very significant effect on the starting voltages
of adjacent corona wires.
In addition to the starting voltage, the V-I curve of a wire follows a
universal law in which the current density directly under the wire is deter-
mined by the equivalent current density in the coaxial geometry. For areas
not under the wire, the current density is given by a simple cosine relation
to the current density under the wire. This holds over a range of current
densities normally encountered in ESPs and has been experimentally verified to
much higher levels.
Although the model is strictly limited to uniform modes of corona, it has
applicability wherever the smooth-wire corona threshold is exceeded. Further
work on non-uniform corona is being addressed as part of this effort.
4-13

-------
REFERENCES
1.	Cooperman, P. A theory for space-charge-limited currents with applica-
tion to electrical precipitation. AIEE Transactions. 79: 47-50, 1960.
2.	Peek, F. W. Dielectric phenomena in high-voltage engineering.
McGraw-Hill, New York, 1929.
3.	Oglesby, S., Jr. and Nichols, G. B. Electrostatic precipitation. Marcel
Dekker, New York, 1978.
4.	Waters, R. T. and Stark, W. B. Characteristics of the stabilized glow
discharge in air. Journal of Physics D: Applied Physics. 8: 416-426,
1975.
5.	Loeb, Leonard B. Electrical coronas  their basic physical mechanisms.
University of California Press, Berkeley and Los Angeles, California,
1965.
6.	Lawless, P. A., McLean, K. J., Sparks, L. E., and Ramsey, G. H. Work in
progress.
7.	Lawless, P. A. and Sparks, L. E. Measurement of ion mobilities in air
and sulfur dioxide-air mixtures as a function of temperature. Atmos-
pheric Environment. 14: 481-483, 1980.
8.	Lawless, P. A., Damle, A. S., Viner, A. S., Shaughnessy, E. J., and
Sparks, L. E. Laser doppler anemometer measurements of particle velocity
in a laboratory precipitator. In: Third Symposium on the Transfer and
Utilization of Particulate Control Technology, Vol II. Electrostatic
Precipitators. EPA-600/9-82-005b (NTIS PB83-149591). U. S. Environ-
mental Protection Agency, Research Triangle Park, North Carolina, 1982.
p 25-34.
9.	Lawless, P. A. and Sparks, L. E. A mathematical model for calculating
effects of back corona in wire-duct electrostatic precipitators. J.
Applied Physics. 51: 242-256, 1980.
10.	Cooperman, G. A new current-voltage relation for duct precipitators
valid for low and high current densities. IEEE Transactions on Industry
Applications. IA-17: 236-239, 1981.
11.	Davidson, J. H. Ph.D. Thesis, Duke University, 1984.
4-14

-------
NUMERICAL COMPUTATION OF THE ELECTRICAL CONDITIONS IN A WIRE-PLATE
ELECTROSTATIC PRECIPITATOR USING THE FINITE ELEMENT TECHNIQUE
Gregory A. Kallio and David E. Stock
Mechanical Engineering Department
Washington State University
Pullman, WA 99164-2920
ABSTRACT
The calculation of electrical conditions inside electrostatic precipi-
tators is integral in the development of performance models and design of
hardware. Specifically, accurate knowledge of the electric field strength
and space charge density is needed to evaluate such processes as particle
charging, coulombic drift, electric wind, and back corona. In the present
paper, electrical conditions in a wire-plate ESP are computed by utilizing
the finite element method to solve the governing electrostatic equations.
This method offers distinct advantages in this geometry over conventional
finite difference numerical techniques: i) nodal spacing can be selectively
adjusted oyer the solution domain to optimize computational accuracy and
economy; ii) curved boundaries are easily and accurately accommodated; iii)
internodal solutions can be readily evaluated. The solution yields voltage-
current characteristics in addition to space charge density, electric poten-
j.,' anc^ electric field distributions. The results are compared to finite
difference solutions and existing experimental data.
5-1

-------
INTRODUCTION
Accurate knowledge of the electric field and space charge density dis-
tributions inside electrostatic precipitators is vital to the modeling and
analysis of device performance. These quantities play an important role in
characterizing particle charging, coulombic particle drift, and the effects
of electric wind. Past studies have determined electrical conditions inside
wire-plate precipitators by numerically solving the governing partial dif-
ferential equations (PDE's), namely, Poisson's and current continuity equa-
tions. Yabe, et al. (1), Yamamoto and Velkoff (2), and Ushimaru and Butler
(3) have used finite difference techniques to solve the governing PDE's, but
their models required experimentally measured current versus voltage data to
establish the charge density boundary condition at the wire electrode.
Leutert and Bohlen (4), McDonald, et al. (5), and Lawless and Sparks (6)
employed finite difference solution methods without the reliance of empiri-
cally derived boundary conditions. The electrical model of McDonald, et al.
has subsequently become an integral part of the well-known Southern Research
Institute (SRI) electrostatic precipitator model which has been widely used
for performance predictions (7). Felder and Arce-Medina (8) have recently
suggested improvements to the SRI electrical model through the use of con-
vergence acceleration methods to minimize computation time.
The finite difference method has an inherent weakness when applied to
the solution of Poisson's equation over the domain of the wireplate geomet-
ry due to the strong gradients in potential and electric field near the wire
electrode. In order to obtain accurate solutions near the wire, grid size
must be significantly smaller than the wire radius. Typical wire-plate geo-
metries would then require grids of at least 200 x 200 points, and solving
such finite difference grids would consume unreasonable amounts of computa-
tion time. Alternatively, a variable size grid system can be employed to
reduce the number of computation points, but the difference equations become
awkward to solve. The scheme of Lawless and Sparks (6) numerically solves
the integral forms of the governing electrostatic equations to yield elec-
tric field and charge density solutions. The method appears to predict the
gradients near the wire with a coarse grid, however, an ad hoc power law
assumption with an arbitrary exponent is required to compute values within
the wire grid cell. The SRI model overcomes the aforementioned problems by
requiring solution accuracy only near the precipitator plate. A 10x10 grid
is normally utilized which yields satisfactory electric field values near
the plate and current-voltage characteristics. However, values of poten-
tial, electric field, and space charge density in the midsection of the
precipitator generally are not accurate.
The finite element method is well-suited to the solution of elliptic
PDE s such as Poisson's equation. The method is especially advantageous in
the present application since grid (mesh) size can be continuously varied
throughout the computational domain to accommodate large gradients in the
solution. The potential and electric field solutions to the wire-plate
precipitator are further complemented by the following properties of the
finite element method: i) curved boundaries, e.g., at the wire electrode,
are accurately accounted for; ii) the functional form of the solution allows
direct evaluation of internodal values; iii) solution derivatives, e.g., the
5-2

-------
electric field, are easily evaluated by differentiation of the elemental
solution functions.
The present scheme utilizes the finite element method to solve
Poisson's equation in order to obtain more accurate potential and electric
field values over the entire wire-plate domain. Triangular elements are
employed with quadratic polynomials as basis functions. The nonlinear con-
tinuity equation is not amenable to standard finite element techniques, so
it is solved using a backward difference method to yield the space charge
density distribution. Since the charge density solution does not typically
possess large gradients, a rather coarse finite difference grid is adequate.
Aside from the finite element solution technique, the solution generally
follows the procedure of the SRI computer model (subroutine EFLD2). The
space charge is assumed to consist solely of corona-generated unipolar ions
of constant mobility. The present computational technique is used to deter-
mine potential, electric field, and charge density distributions in addition
to current-voltage characteristics. Comparisons are made with finite dif-
ference solutions in terms of accuracy and computation time.
THEORETICAL DEVELOPMENT
The electrical conditions inside an electrostatic precipitator are
determined by the strongly-coupled applied electric field and corona-
generated space charge (in the absence of particles). The corona, or glow,
discharge is characterized as a stable process of ion emission from a highly-
stressed electrode in a nonuniform electric field. This stable region typi-
cally occurs over a wide range of applied voltage, from a minimum threshold
called "corona onset" to an eventual spark breakdown. The currents within
this range are relatively small (a few milli-amperes at most) and they in-
crease nearly proportional to voltage. This is referred to as the Ohm's law
regime. The process is thus described by the steady-state Maxwell equations
and a constitutive law relating current and voltage. With the assumption
that the magnetic field due to the corona current is negligibly small, the
governing electrostatic equations are
VE = p/e
V-j - 0
J = pbE
(1)
(2)
(3)
Equation (1) is the point form of Gauss's law, Eq. (2) represents the contin-
uity of steady currents, and Eq. (3) is the point form of Ohm's law, assum-
ing isotropic conduction. In the present case, the charge carriers are
singly-charged unipolar ions. Recent experimental measurements by Lawless
and Spares (9) show that negative ions have a mobility (b) of approximately
1.82x10 m2/V/s in simulated flue-gas conditions. Since the electric field
is irrotational (conservative) in the region of interest, it is derivable
from the gradient of a potential (V):
E = -VV	(4)
5-3

-------
Equations (1) to (A) can be rearranged to yield governing forms for the two
primary dependent variables, charge density and electric potential
V2V = -p/e
o
p2 - e VV*Vp = 0
o
(5)
(6)
Equation (5) is the well-known Poisson's equation and Eq. (6) is a special
form of the continuity equation. These two equations compose a coupled pair
of PDEs, one of which is nonlinear. With appropriate boundary conditions
and solution method, the scalar functions V and p can be numerically com-
puted along with the vector fields E and J from Eqs. (A) and (3).
Figure 1. Numerical solution domain for the wire-plate electrostatic
precipitator.
The solution domain in the wire-plate precipitator geometry is shown in
Figure 1. Due to the symmetry of the multi-wire precipitator, the domain
consists solely of the rectangle ABCD. It is assumed that the solution will
hold for all similar rectangles within the precipitator. The nomenclature
and coordinate system are similar to those used by McDonald et al. The fol-
lowing boundary conditions are applied to the specified domain:
i) E =0 along line AB;
ii) EX = 0 along lines BC, CD, and AD;
iii) V^= 0 along line CD;
iv) V = Vq at point A.
Formulation of the boundary conditions near the discharge wire poses
some difficulties. The corona discharge establishes an ionization region
around the wire which is characterized by a radius (r^). The ionization
radius is typically the same order of magnitude as the wire radius (a).
This radius separates the unipolar space charge (outer) region from the very
complicated bipolar ion processes contained in the ionization region. This
GAS
FLOW
CORONA
WIRE
COLLECTING
PLATE
X = Sx
5-4

-------
narrow ionized sheath around the wire is generally not modeled due to its
complexity, so inner boundary conditions are placed at r=r rather than r=a.
Establishing values or expressions for voltage (V) and charge density (p) at
r=r^ is the crux of formulating an accurate electrostatic model. Past stu-
dies have often determined these boundary conditions from experimentally
measured current versus voltage corona characteristics. This reliance on
empirical data obviously weakens a theoretical model. The present study
adopts the theoretical charge density condition of McDonald et al., who for-
mulated the following expression from the observed constancy of the ion
amplification factor
Pi = (0.002 SyJp)/irba{31f [6+0.0308 (6/a)^] }	(7)
where f is the wire roughness factor (f=l for smooth wires), 6 is the gas
density normalized to STP, and J is the averagp current density at the
collecting plate. The quantity %lf[6+0.308(6/a)^] is the semi-empirical
relation of Peek (10) for the corona onset electric field of a coaxial wire
in a cylinder. All quantities are measured in mks units except wire radius
(a), which is in cgs units. The onset ionization radius estimated by Cobine
(11) is employed, which is thought to be nearly constant in the Ohm's law
regime.
r_^ s a + 0.3/a~	(8)
where the wire radius (a) is again given in cgs units. McDonald et al.
chose to use the applied wire voltage (V ) as the appropriate value at r=r ,
however, it appears that a significant0voltage drop can exist across tiie
ionization region. A more accurate expression for V may be obtained by
using the coaxial cylinder expression for potential near the wire (assuming
that the ionization region contains no net space charge).
vi = v0 t [Jln(ri/a)]/[n(rc/a)] }	(9)
where is the effective radius of the collecting plate, which is very
nearly equal to the wire-plate spacing (Sx).
METHOD OF SOLUTION
The governing equations (5) and (6) with boundary conditions are nu-
merically solved in two dimensions using a combination of finite element
(FE) and finite difference (FD) methods. The solution procedure is an iter-
ative scheme similar to that of the SRI model, where an average current den-
sity is specified_and the applied voltage adjusted until current density
computed from J=pbE at the plate agrees with the desired value. The compu-
ter code produces values of p, V, E, and J at any point in the computational
domain in addition to the current-voltage characteristics.
POISSON'S EQUATION
Poisson's equation (5) is solved using finite elements with the
Galerkin method of weighted residuals (12). The desired electric potential
solution, V(x,y), is replaced by a finite series approximation V(x,y) over
5-5

-------
each element,
m
V(x,y) = I V Ux,y)	(10)
j = l
where . are basis functions defined for each node of the element and V, are
undetermined coefficients. The basis functions are chosen to be quadratic
polynomials,
<)>. = a. + b x + c y + d.x2 + e.xy + f.y2,
J j j j J	J	J
which must allow V to meet the imposed boundary conditions and possess con-
tinuity across element boundaries. The Galerkin method requires that the
integral of each node's residual, when weighted with the basis function, be
equal to zero. In this way, the undetermined coefficients are selected to
minimize the residual, or error. Imposing this requirement to Poisson's
equation and applying Green's theorem yields a linear matrix equation of the
form Kc=b where
m 9$. 3(|> 9
K- '	i"1'2 "
b = 1/e / f p (j) dxdy, i=l,2 ... m
o	i i '
c Vj $ j1,2, ... m.
Calculation of the above integrals is simplified by mapping each element
onto a master element by an isoparametric transformation. This allows the
use of elements with curved boundaries. The matrix equation is solved for
c, which contains the approximate potential value at each node. Once solved,
internodal potentials can be directly evaluated from Eq. (10). Furthermore,
the electric field is determined by simply differentiating Eq. (10):
(11)
(12)
The present finite element scheme uses 52 triangular elements with a total
of 111 nodes over the solution domain as shown in Figure 2. Each element
has six nodes to accommodate the quadratic basis functions. The computer
code automatically generates a mesh whose nodal arrangement depends upon a,
S , and S . Near the wire, the mesh has circular symmetry where the radial
positions^progress as
r = (a/S )1-11/10 n=i, 2, ... 8
Ii	X
This provides a high nodal density where the solution gradients are high.
The elements have curved boundaries which match the wire radius exactly and
also allow the nodal positions to diverge in a uniform fashion. Away from
the wire, the nodal arrangement transforms into a rectangular pattern with
seven equi-spaced nodes along lines BC and CD. The residual intergrals in
the matrix equation are evaluated over the master element using a seven-
point Gaussian quadrature scheme. Assembly of the integral terms yields a

m
9
Ex(x,y) =
- E V
j = l 2
j
3x

m
9<|>
Ey(x,y) =
- 1 V.
3 = 1 3
1
9y
5-6

-------
CORONA WIRE
^-COLLECTING PLATE
Figure 2. Finite element mesh arrangement.
5-7

-------
banded, positive definite matrix equation which is solved by a LINPACK sub-
routine. UNPACK is a package of machine independent Fortran subroutines
which analyze and solve various systems of simultaneous linear algebraic
equations (13).
CONTINUITY EQUATION
The special form of the continuity equation (6) is approximated by a
finite difference formulation, based upon a constant-size rectangular grid.
Using backward differences on Vp and the notation of the SRI model, Eq. (6)
becomes
p = - a + (a2+g)2	(13)
o
where a = he (a E +a E )/a a
o y xo x yo x y
3 = e (a E p+a E pQ)/a a
o y xo 2 x yo 3 x y
The subscript "o" refers to the computational point of interest, while sub-
scripts "2" and "3" refer to the "upstream" points in the x and y direction
respectively. Values of p are successively computed at each grid point by
progressing outward from the known condition (Eq. 7) at the wire electrode.
COMPUTATIONAL PROCEDURE
The solution procedure begins by specifying an average current density
(J ) at the collecting plate and a corresponding estimate (guess) of the
app?ied wire voltage (V ). The electric field at the FD grid points is
initialized by using Cooperman's space charge-free analytic solution (14).
Charge density is computed at all points from Eqs. (7) and (13). The values
of p at the FE nodes are then interpolated from the FD point values using a
bilinear approximation over each grid rectangle. This allows the FE compu-
tation of potential at all nodes from Poisson's equation. Next, charge
density is recomputed from the FD equations using the new values of electric
field found from Eqs. (11) and (12). Likewise, potential is recomputed us-
ing FE with new values of interpolated charge density. This computation
cycle is repeated until changes in potential values become less than 0.1
percent. Average current density is then determined at the plate using the
computed charge density and electric field values.
1 N
J = 77 E p.bE .	(14)
P Nj=l J x3
where N is the number of equi-spaced grid points along the plate. The value
of J calculated from Eq. (14) is compared to the specified value. If they
do n8t agree to within 1 percent, the value of V is adjusted and the entire
procedure repeated until J converges. The applied voltage is adjusted
according to a power law relationship in order to speed convergence (8),

-------
The exponent y is determined from the preceding two steps as
y = n(V(n_1)/V(n"2))/to(J(n_1)/J(n~2))
o	o	P	P
When the voltage is first adjusted, an arbitrary value of y=0.9 is used to
initialize the sequence. This algorithm produces a significant decrease in
convergence time over the SRI model, typically requiring only three itera-
tions to satisfy the current density convergence criterion.
RESULTS
The finite element (FE) and finite difference (FD) solutions are com-
pared in Figures 3 through 7 in terms of electric potential, electric field,
space charge density, and current versus voltage characteristics. The FD
model was run with a 10x10 grid (121 points) for all square domains, which
is comparable to the 111 nodes used in the FE model. The computations repre-
sent standard gas (air) conditions (6=1) and discharge wires that are clean
and polished (f=l). Figures 3(a) and 3(b) display the potential and elec-
tric field profiles, respectively, for zero space charge (J =0) along line
AD between the wire and plate of a typical geometry. Accompanying the nume-
rical solutions is Cooperman's analytic solution. As expected, the coarse-
grid FD solution poorly approximates the exact solution due to the high gra-
dients in potential near the wire. The FE solution closely follows the
exact solution due to the high nodal density near the wire. Note that the
FD electric field is discrepant near the plate, indicating that inaccuracies
initiated near the wire tend to propagate throughout the computational do-
main. The discrepancy between the FD and FE (or exact) solutions lessens as
the wire size is increased, however, even for a=1.27 mm, the error was found
to be significant; FD potential solutions were approximately 15 percent
higher over the entire profile while electric field values ranged from 15 to
75 percent high.
Figures 4(a) and 4(b) compare FD and FE^ solution profiles for a moder-
ate and high average current density, J =10 and J =10 A/m2. The presence
of space charge tends to decrease the gradients inpelectric potential, how-
ever, the FD model still has difficulty predicting the potential and elec-
tric field profiles when J =10 A/m2. Again, note that the electric field
values at the collecting pUate are highly discrepant. At J =10 A/m2, the
space charge is sufficient to reduce the gradients in potential over the
entire wire-plate gap, except very near the wire. Consequently, the FD and
FE solutions agree closely over most of the region.
Figure 5 displays space charge density profiles along line AD for both
low and high current densities, J =10~ and J =10~ A/m2. Note that both
solution techniques show close agreement, which^is expected since each model
employs FD to solve the continuity (space charge) equation. Also note that
low current density gives rise to a nearly uniform charge density distribu-
tion while high current density promotes a decreasing profile toward the
plate, representing a space charge-limited current.
5-9

-------
Jp  0
a = 0.127 mm
Sx = 0.1143 m
Sy= 0.1143 m
	COOPERMAN'S
SOLUTION
FD SOLUTION
FE SOLUTION
ANALYTIC
0	0.2 0.4 0.6 0.8 1.0
NORMALIZED DISTANCE FROM WIRE(x/Sx)
>
N
K
CO
K
UJ
Q
_l
Ul
1.2
^ 1.0
o
CE
I
o
Ul
_l
Ul
o
Ui
N
_J
<
2
ex.
o
z
0.8
0.6
0.4
0.2
Jp=0
a = 0.127mm
S* =0.1143 m
Sy = 0.ll43m
	COOPERMAN'S
ANALYTIC SOLUTION"
FD SOLUTION
  FE SOLUTION
Figure 3.
*0 0.2 0.4 0.6 0.8 1.0
NORMALIZED DISTANCE FROM WIRE(x/Sx)
Electric potential (a) and electric field (b) distributions along
line AD for zero space charge (Jp = 0).
5-10

-------
"I I	1	1	1	1
a - 0.127 mm
b = 1.82 x I0"4m2/Vs
S*= 0.1143 m
Sy = 0.1143 m
-O-O- FD SOLUTION
FE SOLUTION
Jp = IO"3A/m2
Jp = IO~4A/m2
NORMALIZED DISTANCE FROM WIRE(x/Sx)
_ 1.4
3?
N
K
CO
x
Ld
1.2
UJ
111
o
cr
h-
O
UJ
_l
llJ
o
UJ
N
rr 1.0
0.8
0.6
cc
o
z
0.4
0.2.
a = QI27 mm
b=l.82xlO"4m2/V.s
Sx = 0.1143 m
Sy =0.1143 m
O-O-FD SOLUTION
FE SOLUTION
Jp = IO"4A/m2
0 0.2 0.4 0.6 0.8 1.0
NORMALIZED DISTANCE FROM WIREU/S*)
Figure 4. Electric potential (a) and electric field (b) distributions along
line AD for
Jp = 10'
"4 10~3 A/m2.
5-11

-------
Figure 5.
1.0
<*.
0.8 _
to
Z
liJ
 0.6
UJ
o
cc
<
o 0.4
O
UJ
N
-J 0.2 _
2
oc
o
a = 1.27 mm A _
1.82 x 10" m /V*s
Sx = 0.1143 m
Sy = 0.1143 m
FD SOLUTION
FE SOLUTION
X
X
X
X
X
X
0	0.2 0.4 0.6 Ofl 1.0
NORMALIZED DISTANCE FROM WIRE
(X/Sx)
Space charge density distribution along line AD for -Jp=10
10"3 A/m2.
-5
Figure 6 presents computed current versus voltage characteristics for
two wire sizes, a=0.127 mm and a=1.27 mm. The FD results display a gener-
ally higher current density for a given applied voltage due to the higher
electric field predicted at the plate. For a=0.127 mm, no|e that the FD and
FE curves merge together at high current density (J =10 A/m2) since the
electric field solutions of the two methods become comparable for such con-
ditions. Likewise, for a=1.2^ mm, the current-voltage curves merge at
approximately 115 kV for Jp=10 A/m2 (not shown on plot).
Figure 7 shows the electric potential profiles along BC, i.e., from the
midpoint between wires to the plate. Experimental data from Penney and
Matick (15) compared to the numerical solution predictions. These potential
measurements were taken with a shielded fine-wire probe in which voltage was
applied to the shield to overcome the excess potential of the wire probe due
to the ionic charge transfer. The plots show good agreement over the entire
region for both FD and FE predictions. Unfortunately, accurate experimental
measurements of potential or electric field are not available in regions and
geometries where the two numerical techniques give discrepant solutions.
5-12

-------
i	1	1	r
b = I.82xl0*4m2/Vs
Sx = 0.1143 m
Sy = 0.1143 m
oFO SOLUTION
 FE SOLUTION
a = 0.127
Figure 6.
** 10 20 30 40 50
APPLIED VOLTAGE (V.)t kV
Current density versus applied voltage for a=0.127, 1.27 mm.
CONCLUSIONS AND DISCUSSION
The prediction comparisons of the finite difference and finite element
electrical models can be summarized as follows:
i)	For low to moderate current densities, the FD potential and elec-
tric field solutions possess significant inaccuracies over the
entire wire-plate region due to the high gradients in potential
near the wire. For such conditions, the FE solution is suggested.
ii)	For high current densities and large wire sizes, the discrepancies
between the FD and FE predictions are minimal due to the lower
potential gradients near the wire. Either solution method is ade-
quate for these cases.
iii) The space charge predictions of each numerical model agree closely
due to the similar treatment of the space charge PDE.
Of further interest is the accuracy and required computation time for
each numerical model. Accuracy of FD methods is easily defined in terms of
Taylor series expansions. In the SRI FD model, electric potential is
second-order accurate due to center differencing, while the electric field
5-13

-------
a = 1.016 mm
0.5
Sx = 0.1143 m
Sy= 0.0735 m
J0= 4.84 x 10 A/m
0.4
2 0.3
E 0.2
FD SOLUTION
FE SOLUTION
EXPERIMENTAL
MEASUREMENTS
by Penney and Matick
0 0.2 0.4 0.6 0.8 1.0
NORMALIZED DISTANCE FROM
POINT MIDWAY BETWEEN WIRES
(X/Sx)
Figure 7. Comparison of electric potential predictions along line BC with
experimental measurements of Penney and Matick.
and space charge density solutions are first-order accurate from backward
differencing. The accuracy of finite element techniques are somewhat dif-
ficult to characterize and compare due to their variable nodal spacing.
Since the basis functions are quadratic, the accuracy of the potential solu-
tion is approximately second-order, while the electric field is approximate-
ly first-order accurate due to differentiation. However, the nodal spacing
in the FE formulation is much smaller than the FD grid size in those regions
where higher-order accuracy is needed (near the wire). Therefore, the rela-
tive accuracy of the two methods is best described by the ratio of FE nodal
spacing to the FD grid spacing in the region of interest.
The improved accuracy of the FE technique is slightly offset by an in-
crease in computation time. Even though the number of iterations between
Poisson's equation and the continuity equation are drastically reduced by
using the FE method, the time required to invert the K-matrix and solve the
matrix equation is lengthy. In addition, the alternating conversion between
FD and FE grid systems consumes a significant amount of time. As an example,
computation of the potential and electric field profiles for a=0.127 mm
(Figure 4) on a Prime 400 minicomputer yields the following:
total iterations	CPU time
FD method	686	72 sec
FE method	13	206 sec
5-14

-------
The accuracy and computational efficiency of the present finite element
model could be further improved through several modifications. First, the
use of Hermitian (cubic polynomial) basis functions in the FE approximation
would provide third-order accurate potentials, but more importantly, it
would also provide continuity in the first derivative (electric field)
across element boundaries. The present quadratic approximation does not
insure first derivative continuity, so electric field values at element boun-
daries are somewhat ambiguous. These small discontinuities are noticed in
the slightly ragged appearance of the presented electric field profiles.
Second, the required computation time of the model may be reduced by solving
the continuity (space charge) PDE with the FE method rather than the FD
method. This would eliminate the time consuming conversion between FE and
FD grids and would provide second-order accurate space charge density solu-
tions. However, the application of the Galerkin method to the hyperbolic,
nonlinear continuity equation does not appear straightforward (12).
The work described in this paper was not funded by the U.S. Environ-
mental Protection Agency and therefore the contents do not necessarily
reflect the views of the Agency and no official endorsement should be
inferred.
ACKNOWLEDGEMENTS
The authors wish to thank Associate Professor David S. Watkins for his
helpful advice on utilizing the finite element technique, Jack R. McDonald
for information about the SRI electrical model, and James R. Ferguson for
providing general discussion of the numerical model.
In addition, the support from the U.S. Department of Energy in the form
of a research fellowship is gratefully acknowledged.
5-15

-------
NOMENCLATURE
a
radius of discharge wire

a
x
length of FD grid cell in x-direction

a
V
length of FD grid cell in y-direction

j
b
mobility of ionic charge carriers

E
electric field strength

E
x
electric field component in x-direction

E
V
electric field component in y-direction

J
f
wire roughness factor

J
current density

J
p
average current density at collecting plate

r
r
radial distance from wire

ri
radius of ionized sheath

r
c
effective radius of collecting plate in a wire-cylinder geometry
S
X
wire-to-plate spacing

S
v
one-half wire-to-wire spacing

j
V
electric potential, or voltage

V
FE approximation to electric potential

V
o
applied voltage at wire

Vi
voltage at ionization radius (r^)

X
coordinate position toward collecting plate from
center of wire
y
coordinate position parallel to collecting plate
from center of wire
a
6
parameter used in Eq. (13)
parameter used in Eq. (13)

6
relative density of gas
-12
permittivity of free space (8.854x10 F/m)


O

Y
exponent in Eq. (15) x

~j
P
FE basis function associated with node j

space charge density

Pi
space charge density at ionization radius (r^

5-16

-------
REFERENCES
1.	Yabe, A., Mori, Y. and Hijikata, K. (1978), EHD study of the corona wind
between wire and plate electrodes, AIAA Journal 16_, 340-345.
2.	Yamamoto, T. and Velkoff, H.R. (1981), Electrohydrodynamics in an elec-
trostatic precipitator, J. Fluid Mech. 108, 1-18.
3.	Ushimaru, K. and Butler, G.W. (1984), Numerical simulation of gas-solid
flow in an electrostatic precipitator, Proceedings of the Symposium on
Gas Solid Flows, 7th Annual Energy-Sources Technology Conference & Exhi-
bition, February 12-16, 1984, New Orleans.
4.	Leutert, G. and Bohlen, B. (1974), The spatial trend of electric field
strength and space charge density in plate-type electrostatic precipi-
tators, Staub _32. 27-33.
5.	McDonald, J.R., Smith, W.B. and Spencer, H.W. (1977), A mathematical
model for calculating electrical conditions in wire-duct electrostatic
precipitation devices, J. Appl. Phys. 4jB, 2231-2243.
6.	Lawless, P.A. and Sparks, L.E. (1980), A mathematical model for calcula-
ting effects of back corona in wire-plate electrostatic precipitators,
J. Appl. Phys. 51, 242-256.
7.	McDonald, J.R. (1978), A mathematical model of electrostatic precipita-
tion, Vol. I, U.S. Environmental Protection Agency Report No. EPA-600/
7-78-111A.
8.	Felder, R.M. and Arce-Medina, E. (1984), Calculation of voltage and
space charge distributions in a wire-plate electrostatic precipitator,
J. Electrostatics 1_5, 3-13.
9.	Lawless, P.A. and Sparks, L.E. (1980), Measurement of ion mobilities in
air and sulfur dioxide-air mixtures as a function of temperature,
Atmos. Environ. ^4, 481-483.
10.	Peek, F.W. Jr. (1924), Dielectric Phenomena in High Voltage Engineering,
3rd edition, McGraw-Hill, Inc., New York, 66.
5-17

-------
Session 6: ESP: PERFORMANCE ENHANCEMENT I
Ralph F. Altman, Chairman
Electric Power Research Institute
Chattanooga, TN

-------
A FIELD STUDY OF A COMBINED NH3-SO3 CONDITIONING SYSTEM ON A COLD-SIDE
FLY ASH PRECIPITATOR AT A COAL-FIRED POWER PLANT
Robert S. Dahlin
John P. Gooch
Guillaume H. Marchant, Jr.
Southern Research Institute
Birmingham, Alabama 35255
Roy E. Bickelhaupt
Bickelhaupt Associates, Inc.
Carbondale, Illinois 62901
D. Richard Sears
University of North Dakota
Energy Research Center
Grand Forks, North Dakota 58202
Ralph F. Altman
Electric Power Research Institute
Chattanooga, Tennessee 37411
ABSTRACT
This paper presents the results of a field study of a combined NH3-SO3
conditioning system used to improve the performance of a cold-side electro-
static precipitator at a coal-fired power plant. The NH3 was injected between
the air heater and the precipitator at a rate of 5 ppm, and the S03 was
injected ahead of the air heater at a rate of 9 ppm (in addition to 5 ppm
of SO3 produced naturally in the boiler). The measured concentrations of NH3
and SO3 just ahead of the precipitator were below 0.5 ppm, suggesting that the
two agents were adsorbed on the fly ash and/or reacted together to form a
sulfate particulate. Laboratory studies showed that the adsorption of these
quantities of W^and S03 on the ash produced a surface film that reduced the
electrical resistivity of the ash by about one order of magnitude. The
reduced resistivity was directly responsible for the excellent electrical
operating conditions of the precipitator, which made it possible to achieve
a collection efficiency of 98.5% with an SCA of only 166 ft2/kacfm.
6-1

-------
INTRODUCTION
This paper describes the results of a field study of a combined NH3-SO3
conditioning system used to improve the performance of a cold-side electro-
static precipitator at a large coal-fired power plant. The combined NH3-SO3
conditioning system was placed into continuous service after numerous tests by
plant personnel showed that the combined NH3-SO3 system was more effective
than the use of either NH3 or SO3 alone. In the tests by plant personnel, the
effectiveness of various injection locations and injection rates of one or
both agents was judged on the ability to maintain an acceptable stack opacity.
The evaluations performed by plant personnel also included resistivity mea-
surements, particle sizing, and emissions testing. Based on the results of
these tests, the plant management selected a combined conditioning system in
which NH3 is injected between the air heater and the precipitator at a rate of
5 ppm, and S03 is injected ahead of the air heater at a rate of 9 ppm. This
was in addition to 5 ppm of SO3 that is produced naturally in the pulverized
coal-fired boiler. In most applications of flue gas conditioning, the S03 is
injected downstream of the air heater; however, it is believed that injection
ahead of the air heater, as was done at this plant, may result in a more uni-
form distribution of the SO3 in the flue gas and more effective utilization of
the S03.
Although the tests by plant personnel were successful in identifying a
conditioning system that would improve precipitator performance and reduce
stack opacity to an acceptable level, the tests provided no information con-
cerning the mechanisms involved, or why the combined NH3-SO3 system was more
effective than either NH3 or S03 alone. Thus, the present study was under-
taken to obtain a more fundamental understanding of the combined NH3-SO3 con-
ditioning process, and to hopefully learn what makes the combined process more
effective than NH3 or S03 alone. The present study provides some new under-
standing of the mechanisms, but it stops short of a definitive analysis. This
is largely because field studies could not be done with one or both of the
conditioning agents turned off. Nevertheless, laboratory studies have made it
possible to infer significant information concerning the role of NH3 and S03
in the combined conditioning process.
OBJECTIVES AND SCOPE
The work described in this paper was jointly sponsored by the Electric
Power Research Institute (EPRI) and the Grand Forks Projects Office of the
U. S. Department of Energy (DOE). The EPRI-sponsored portion of the work was
primarily concerned with laboratory studies aimed at developing a better
understanding of the effects of NH3 and S03 on the electrical resistivity of
the ash. The DOE-sponsored portion of the work was primarily concerned with
field studies aimed at: evaluating the performance of the conditioned
precipitator; obtaining data on the concentration, composition, and size
distribution of the conditioned fly ash; defining the injected and residual
levels of NH3 and SO3 in the flue gas and the levels of NHit+ and SO^-2 on the
fly ash; quantifying the overall and fractional collection efficiency of the
precipitator; and collecting coal and sizefractionated ash samples for
6-2

-------
further characterization. Along with similar data from other field tests, the
data listed above are being used by the Grand Forks Projects Office in the
assessment of particulate control technologies for low-rank Western coals.
The scope of the test program encompassed the following elements:
(1)	Determination of inlet and outlet mass loadings of fly ash and
overall fly ash collection efficiency by EPA Method 17 (1),
(2)	Measurement of inlet and outlet particle size distributions of
fly ash and fractional collection efficiency using calibrated
cascade impactors (2),
(3)	Documentation of boiler and precipitator operating conditions
during testing,
(4)	Sampling and analysis of coal by ASTM methods (3),
(5)	Sampling of fly ash in glass fiber thimbles and analysis of
major elements by atomic absorption spectrometry (4),
(6)	Analysis of the fly ash for NHI++ content by specific ion elec-
trode (5) and for SO4""2 content by ion chromatography (6),
(7)	Measurement of injected and residual levels of NH3 in the flue
gas by collection in a 0.02 N H2SO4 solution,
(8)	Measurement of injected and residual levels of S03 in the flue
gas by the Cheney-Homolya method (7), and
(9)	Laboratory resistivity studies with the conditioned ash and with
ash in vdiich the conditioning effect had been removed by thermal
treatment.
Space limitations preclude a detailed presentation of all of the above; the
interested reader is referred to the DOE report on this work (8).
SITE DESCRIPTION
PRECIPITATOR
The precipitator layout is illustrated in Figure 1. The flue gas stream
leaving the boiler and economizer is split and passes through two air heaters
and then through four separate precipitators in a double-chevron arrangement
as shown in the layout. The precipitator designated ESP 2-1 was the subject
of this testing. It is identical to the other precipitators: divided into
two chambers, each with two electrical fields in the direction of gas flow.
All fields are equipped with standard 0.109 in. weighted discharge wires and
9 ft by 30 ft collection plates spaced 9 in. apart. The precipitator is
small by modern standards, with a specific collection area of only 166 ft2/
1000 acfm. This is the actual operating specific collection area which is
slightly greater than the design value of 156 ft2/kacfm. It is operated at a
6-3

-------
^ !
	
\ /
X
o
-<
A
/ \
FAN
/ \
( \
OOC)

ooo



rjL
Hi
8 ^
0 '
o ^
ESP 2-4
Coo
ooooooooo
o
^>8
r O
\ /

V i
X
o
-n
X
/ \
z
f \
0 oo
)

ooo
o
8 "N
o 
e "N
7,:
f/.:
IIS
/JDOO

[OOO '
ESP 2-3	ESP 2-2
PLAN VIEW

6 0 0 0 0
ooo
X )t A

ESP 2-1
AIR PREHEATER
AIR PREHEATER
Figure 1. ESP layout.
6-4

-------
fairly high temperature (~370F) compared to most cold-side precipitators that
normally operate at about 300F. Both the small size and the high temperature
are factors that contributed to poor precipitator performance before this was
corrected by the use of the combined NH3-SO3 conditioning process.
BOILER
The steam generator that supplies the flue gas to ESP 2-1 is one of two
identical units. Each unit is a tangentially-fired, dry-bottom boiler. Both
units are normally base-loaded with a nominal generating capacity of 575 MW.
During the testing, the unit of interest, designated Unit 2, was operated at a
fairly constant load in the range of 566 to 580 MW with excess air in the
range of 13 to 19%.
FUEL
Both generating units burn a Wilcox-group Texas lignite that is mined on
site at the plant. Lignite samples taken from one of the Unit 2 feeders dur-
ing the testing showed an average heating value of 7000 Btu/lb and an average
ash content of 17% on an as-received basis. Daily samples showed considerable
variability in the lignite composition. The results of several proximate and
ultimate analyses are given in Table 1.
TEST PROCEDURES AND RESULTS
MASS LOADINGS AND OVERALL EFFICIENCY
Mass loadings were measured at the inlet and outlet of ESP 2-1 by EPA
Method 17 (1). Each sampling location was traversed with mass trains equipped
with in-stack filters. Isokinetic sampling was approximated as closely as
possible at each sampling point. The measured mass loadings are given in
Table 2. During this testing the boiler load was maintained fairly constant
in the range of 566 to 580 MW, with excess air in the range of 13 to 19%. The
stack opacity recorded by the plant transmissometer was between 26 and 32%.
During the test period the average operating SCA was 166 ft2/1000 acfm,
slightly higher than the design value. The average inlet and outlet mass
loadings were 6.64 and 0.103 gr/dscf, yielding an average efficiency of
98.45%. Previous measurements made by plant personnel yielded efficiencies in
excess of 99%, somewhat higher than the present data. The difference may be
related to differences in coal composition, boiler operating conditions, and
ESP operating conditions.
PARTICLE SIZE DISTRIBUTIONS AND FRACTIONAL EFFICIENCY
The inlet and outlet particle size distributions were determined from in-
stack cascade impactor sampling (2) using modified Brink impactors at the
inlet and University of Washington Mark III impactors at the outlet. Seven
real impactor runs and one blank run were performed in separate sampling ports
at the inlet, while six complete traverses of duct area were performed at the
outlet with two blanks. Glass fiber substrates were used in the impactors.
The blank runs did not reveal any appreciable substrate interference problems.
6-5

-------
TABLE 1. PROXIMATE AND ULTIMATE COAL ANALYSES
(AS-RECEIVED BASIS)3
Proximate
% Moisture
% Ash
% Volatile
% Fixed carbon
Btu/lb
% Sulfur
Sample 1
21.07
14.41
42.21
22.31
8,671
0.91
Sample 2
29.37
23.05
22.42
25.16
5,978
0.85
Sample 3
31.51
12.91
50.45
5.13
6,452
0.73
Ultimate
% Moisture
% Carbon
% Hydrogen
% Nitrogen
% Chlorine
% Sulfur
% Ash
% Oxygen'3
21.07
51.49
2.81
1.12
0.06
0.91
14.41
8.13
29.37
35.22
2.39
0.71
0.05
0.85
23.05
8.36
Samples taken from Unit 2 feeder discharge.
'Oxygen determined by difference.
31.51
39.50
2.68
0.64
0.05
0.73
12.91
11.98
TABLE 2. MASS TRAIN DATA
Run Load Temp.
No. (MW) (F)
1	580
2	568
3	566
380
385
385
SCA
(ft2/
kacfm)
171
163
165
Inlet Inlet	Outlet Outlet
loading percent loading percent
(gr/dscf) isokinetic (gr/dscf) isokinetic
6.22
6.37
7.32
105.7
109.8
109.9
0.111
0.0783
0.120
97.0
99.4
99.1
Effi- Opac-
ciency ity
(%) (%)
98.22 26.4
98.77 27.5
98.36 31.8
6-6

-------
Figure 2 shows the inlet and outlet particle size distributions deter-
mined from the impactor sampling. The curves shown were obtained by averaging
the spline fits (9) to a number of impactor runs; the points do not represent
the raw data, but rather the interpolated points generated by the spline fit.
The fractional efficiency curve derived from the inlet and outlet impac-
tor measurements is shown in Figure 3. This apparently shows a minimum in the
collection efficiency at just below 1 Mm. The curve appears to be qualita-
tively consistent with theoretically predicted fractional efficiency curves in
the range of 1 to 10 Um. The fractional efficiencies are also consistent with
the measured overall efficiency if one considers that half of the mass is
contained in particles larger than 14 ym. The cumulative efficiency for
collection of sub-5-lim particles is 97%.
S03 AND NH3 MEASUREMENTS
S03 samples were collected ahead of the air heater and downstream of the
SO3 injection system using the Cheney-Homolya sampling system (7). For SO3
and NH3 sampling at the ESP inlet and outlet locations, the probe and quartz
filter temperature was reduced from ~550F to the prevailing flue gas tempera-
ture. This modification of the sampling procedure was necessary to prevent
thermal decomposition of the ammonium sulfate particulate assumed to be pres-
ent in the flue gas at these locations. The system was further modified for
NH3 sampling by removing the condenser and substituting one of two Greenburg-
Smith bubblers for a Greenburg-Smith impinger. A dilute (0.02 N) I^SO^ solu-
tion was used as the collection medium for all NH3 samples.
S03 samples were analyzed by BatClO,^ titration with Thorin indicator
(7); NH3 samples with a specific ion electrode (5). The results indicate
that about 5 ppm of natural S03 was present in the flue gas ahead of the air
preheater. Downstream from the SO3 injection probes about 14 ppm of l^SO^ was
found. Therefore, it appears that approximately 9 ppm of	was added to
the flue gas by the injection system. Based on the flowmeters on the injec-
tion system, only 6.5 ppm of	should have been added to the flue gas. It
is possible that this difference is due to nonuniform distribution of the
l^SO^ across the duct, since traverses of the entire duct area were not made.
Less than 0.5 ppm of both NH3 and S03 were found at the precipitator
inlet, presumably due to adsorption on fly ash and reaction to form a sulfate
particulate. Based on the NH3 injection rate of 5 ppm, 2.5 ppm of l^SO^ would
have been consumed in the formation of ammonium sulfate, and the remaining
11.5 ppm adsorbed by the ash.
The adsorption of 5 ppm of NH3 onto the fly ash particles having a mass
loading of 6.64 gr/dscf (see Table 2) would be expected to produce an NHit +
concentration of 0.029% by weight in the ash, assuming complete adsorption.
The mean measured	concentration in the inlet mass train samples was
0.022%, or 76% of the expected value. The adsorption of 14 ppm of S03 would
be expected to produce a SO^.-2 concentration of 0.4% by weight in the ash,
again assuming complete adsorption. However, the measured concentrations of
SO4-2 in the inlet mass train samples ranged from 0.84 to 1.2%. The most
likely explanation appears to be that the ash adsorbed additional S03 while it
resided in the mass train thimble and flue gas was passed through it.
6-7

-------
CO

:f 103
O)
E
a>
o
S
<]
102
101 
10
10"
INLET
OUTLET
10
101
PARTICLE DIAMETER,/im
Figure 2. Differential mass distributions at inlet and outlet ESP.
6-8

-------
10
PARTICLE DIAMETER, jum
Figure 3. Fractional efficiency curve.

-------
ELECTRICAL OPERATION OF ESP
Voltage-current curves were obtained for all of the ESP 2-1 transformer-
rectifier sets. ESP 2-1 contains four fields: East inlet, West inlet, East
outlet, and West outlet. Voltage-current curves for all four fields are shown
in Figure 4. All of the curves are quite well behaved, indicating excellent
electrical operating conditions are achievable. As will be shown later, this
is a direct result of the low electrical resistivity of the conditioned ash.
LABORATORY RESISTIVITY STUDIES
These studies were done with a sample of fly ash taken from an outlet
hopper of ESP 2-1. Normally, a composite sample from both inlet and outlet
hoppers would be preferred for this work, but only the outlet hopper sample
was available. It was, of course, biased toward fine particles (e.g., 12% of
the sample by weight was smaller than 2 ym compared to the inlet impactor
samples vrtiich showed only 2% smaller than 2 Pm). This resulted in dispropor-
tionately high concentrations of	and SO^-2 in the outlet hopper sample
(e.g., 0.046% NHit+ and 1.54% SO^-2 in the outlet hopper sample compared to
0.022% and 1.01% SO^-2 in the inlet mass train samples). This is to be
expected since the finer particles have higher specific surface areas for
adsorption of NH3 and SO3. Thus, the effects observed with this particular
sample may be somewhat greater than the effects that would have been observed
with a composite sample consisting of all particle sizes. Nevertheless, a
composite sample should show the same trends.
In order to investigate the effect of the surface film that was assumed
to be present on the ash as a result of the conditioning, the electrical
resistivity was measured by two techniques: one in which the resistivity of
the as-received ash was measured as a function of increasing temperature
starting at 95C, and one in which the ash was first heated to 450C to
destroy the surface film, with resistivity then being measured as a function
of decreasing temperature starting at 450C. The results are illustrated in
Figure 5. It is obvious that the resistivity measured with increasing temper-
ature is substantially lower than that measured with the decreasing tempera-
ture after the sample was heated to 450C. This shift is apparently due to
the destruction of the surface film deposited by the NH3-SO3 conditioning
process.
Figure 5 also shows the predicted resistivity as a function of tempera-
ture. This curve was predicted from the ash composition, exclusive of NH^-4"
and SO^"2, using the technique of Bickelhaupt (10). The good agreement
between the predicted curve and that measured with decreasing temperature
suggests that there is no conditioning effect present in this case, or that
the surface film has indeed been removed. Thus, at a temperature of 150C,
for example, the surface film appears to be responsible for reducing the
resistivity by more than an order of magnitude.
It is important to note that the resistivity curves shown in Figure 5
were obtained without any NH3 or SO3 in the test environment. It would be
desirable to make a resistivity measurement after the ash in the test cell had
6-10

-------
90
CM
E
u
<
c
>
H;
So
z
cc
oc
D
O
50
40
30
20
20
EAST OUTLET
WEST OUTLET <
O
o
(P
~ m
a cf
i a 
a f>#
o o
~ o
a
"EAST INLET
WEST INLET
1
25 30 35 40 45
SECONDARY VOLTAGE, kV
4*1
50
Figure 4. Voltage-current curves for ESP 2-1 T-R sets.
6-11

-------
10
13
200	300
"J	MEASURED WITH |
DESCENDING
TEMPERATURE
TEMPERATURE, F
400	500
600
700
T
800
900
T
COMPOSITION
0
1
NJ
PREDICTED
WEIGHT PERCENT
Li20
0.02
Na20
0.35
k2o
0.79
MgO
3.5
CaO
15.7
Fe23
6.9
ai2o3
21.2
Si02
48.4
Ti02
1.7
p25
0.06
so3
1.6
LOI
0.4
MEASURED WITH
ASCENDING
TEMPERATURE
109o
J	L
80
90 100
150
200
TEMPERATURE, C
300
400
500
Figure 5. Resistivity measurements made with ascending and descending temperature
and predicted resistivity as a function of temperature.

-------
been allowed to equilibrate with the same levels of NH3 and S03 found in the
flue gas at the precipitator inlet. Unfortunately, however, the simultaneous
use of NH3 and SO3 gases in the test cell is not recommended due to contamina-
tion of the apparatus with ultrafine (NH4)2S01+ fume. Therefore, it was
decided to make additional resistivity measurements with the ash equilibrated
with S03 only. The results for the as-received ash and the ash that had been
heated to 450C (annealed) are indicated by the solid triangle and solid
circle in Figure 6. The equilibration with S03 reduced the resistivity of the
as-received ash to about 1 x 1010 ohm cm, compared to 5 x 1010 ohm cm for the
annealed ash. The response of the as-received ash to the S03 was much greater
than that predicted by the technique of Bickelhaupt (10) illustrated by the
open symbols in Figure 6. For example, the predicted resistivity with 0.3 ppm
ofi0SO3 is about 5 x 10** ohm cm compared to the measured value of about 1 x
10 ohm cm for the as-received ash with the same level of S03. Therefore, it
18 clear that the surface film has made the ash more responsive to S03 condi-
tioning. An interpretation of this is that the NH3 has been deposited on the
ash surface as an NH^+ ion, which serves as an additional charge carrier that
was not present in the case of S03 alone.
SUMMARY AND CONCLUSIONS
The surface film, which is believed to be deposited on the fly ash as a
result of the NH3SO3 conditioning process, appears to be contributing to the
low resistivity of the ash. The hysteresis between the laboratory resistivity
curves obtained with ascending and descending temperatures indicates at least
an order of magnitude effect of the surface film on the resistivity (i.e., the
descending curve was obtained after the ash had been heated to a temperature
sufficient to decompose the surface film, and the resultant resistivity value
at about 150C was about an order of magnitude higher than the corresponding
point on the ascending curve). The surface film also appears to have a syner-
gistic effect with a small amount of adsorbed sulfuric acid (i.e., the resis-
tivity of the as-received ash is attenuated by the equilibration with a given
level of S03 to a greater degree than is the thermally annealed ash in which
the surface film has been decomposed). The observations noted above suggest
that the addition of NH3 to the flue gas improved upon the efficacy of the SO3
conditioning, probably by the contribution of an additional charge carrier,
the NH^* ion.
The voltage-current curves for ESP 2-1 showed good electrical operation
in all of the fields. This is, of course, a direct result of the low resis-
tivity of the conditioned ash. A resistivity of 1 x 1010 ohm cm was measured
in the laboratory when the as-received ash was equilibrated with the level of
S03 found to be present in the flue gas at the ESP inlet. This value of
resistivity is a factor of 4 to 5 lower than in situ resistivity data obtained
by plant personnel. This sort of difference between laboratory and in situ
data is quite common. There may have been slight differences in ash chemis-
try* gas temperature, S03 levels, or other conditions as well. Equilibration
with NH3 was not possible due to possible contamination of the test cell, but,
if anything, this would reduce the resistivity still further. As indicated '
previously in Figure 6, the resistivity would have been on the order of
1012 ohm cm in the absence of any NH3 or S03.
6-13

-------
1014
10"
o
~
&
o
.o
4
w/o so3
WITH 0.3PPMS03
WITH I.OPPMSO3
WITH 4.OPPMSO3
WITH 10.0 PPM SO3
WITH 0.2 PPM SO3, LAB, ANNEALED ASH
WITH 0.3 PPM S03, LAB, AS-RECEIVED ASH
O0:
FIELD INTENSITY ASH COMPOSITION, %"
E = 12 kV/cm
FLUE GAS
h2o
SO3
12.00%
0.30 ppm
0.02
0.35
0.79
3.49
15.67
6.88
21.15
48.29
1.70
0.06
1.60
/
2.2	2.0
182	227
359	441
TEMPERATURE
1000/T K
Figure 6. In situ and laboratory resistivity measurements with SO3 and predicted
resistivity for various SO3 levels.
6-14

-------
Considering the small size of the ESP (SCA = 166 ft2ykacfm), its overall
performance appears to be quite good. The overall mass collection efficiency
for all particle sizes was found to be within the range of 98.22 to 98.77%,
with a mean of 98.45%. The mean value is in good agreement with the overall
collection efficiency predicted by a semi-empirical computer model of electro-
static precipitation (11), assuming that 10% of the flue gas bypasses the
electrified regions of the ESP and that the standard deviation on the gas
velocity profile is equal to 25% of the mean (rms) velocity. Without the
resistivity attenuation afforded by the NH3-SO3 conditioning process, however,
the predicted collection efficiency is well below 90%. This is based on a
resistivity of 1 x 1012 ohm cm without any conditioning. In reality there
would be some natural conditioning effect of the S03 formed in the boiler even
with the injection systems turned off. Therefore, the collection efficiency
with the conditioning systems turned off should be somewhat greater than 90%.
The combined NH3-SO3 conditioning process appears to be a feasible and
useful method of maintaining acceptable performance levels in a cold-side
electrostatic precipitator. Additional work is needed, however, to fully
understand the mechanisms by which the NH3 and SO3 act together to condition
the fly ash and affect precipitator performance. A more thorough investiga-
tion of the NH3-SO3 conditioning process is justified in view of the potential
for improving ESP performance, or reducing the size of a new ESP. There is a
need for field measurements to be made with and without NH3, SO3, and both
agents injected ahead of the ESP. There is also a need for measurements of
the effect of NH3-SO3 conditioning on rapping reentrainment. Further labora-
tory studies are also needed to elucidate the nature and chemical speciation
of the ash surface film produced by the NH3 SO3 conditioning process.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore the contents do not necessarily reflect the
views of the Agency and no official endorsement should be inferred.
ACKNOWLEDGMENTS
This work was supported by the U. S. Department of Energy under Contract
No. DE-AC18-80FC10225 and the Electric Power Research Institute under Contract
No. RP724-2. The DOE Project Manager was Dr. D. Richard Sears, and the EPRI
Project Manager was Dr. Ralph F. Altman.
6-15

-------
REFERENCES
1.	40 CFR 60, Appendix A, pp. 220-237.
2.	D. B. Harris, "Procedures for Cascade Impactor Calibration and Operation
in Process Streams." EPA-600/2-77-004, January 1977.
3.	ASTM Standards D3172-73 and D3176-73 (Reapproved 1979), in Annual Book of
ASTM Standards, Part 26, American Society for Testing and Materials,
Philadelphia, PA (1982), pp. 392, 404-407.
4.	ASTM Standard D3682-78, ibid., pp. 451-458.
5.	M. L. Eagan and L. DuBois, "The Determination of Ammonium Ion in Airborne
Particulates with Selective Electrodes." Anal. Chim. Acta 70, 157-167
(1974).
6.	R. Steiber and R. Merrill, "Application of Ion Chromatography to the
Analysis of Source Assessment Samples," in Ion Chromatographic Analysis
of Environmental Pollutants, Vol. 2, J. D. Mulik and E. Sawicki, Eds.,
Ann Arbor Science, Ann Arbor, Michigan (1977), pp. 99-113.
7.	J. L. Cheney and J. B. Horaolya, "Sampling Parameters for Sulfate
Measurement and Characterization," Environ. Sci. Technol. 13, 584-588
(1979).
8.	R. S. Dahlin, R. E. Bickelhaupt, G. H. Marchant, and J. P. Gooch, "On-
Site Field Tests for Study of Low-Rank Western Coal Fly Ash. Technical
Summary Report, Field Test No. 3. Big Brown Station Electrostatic
Precipitator." D0E/FC/10225-1584, February 1984.
9.	J. W. Johnson, B. E. Pyle, and W. B. Smith, "Extending Precision in a
Computer-Based Cascade Impactor Data Reduction System," in Proceedings of
the Second Symposium on Advances in Particulate Sampling and Measurement,
W. B. Smith, Ed., EPA-600/9-80-004, January 1980.
10.	R. E. Bickelhaupt, "A Technique for Predicting Fly Ash Resistivity," EPA-
600/7-79-204, August 1979.
11.	J. R. McDonald, "A Mathematical Model of Electrostatic Precipitation,"
EPA-600/7-78-111, June 1978.
6-16

-------
CONDITIONING OF POWER STATION FLUE GASES
TO IMPROVE ELECTROSTATIC PRECIPITATOR EFFICIENCY
Gemot Mayer-Schwinning, Manager, Research & Development
Lurgi GmbH
Frankfurt, West Germany
J. D. Riley, Director, Pre-Contract Operations
Lurgi Corporation
River Edge, NJ 07661
ABSTRACT
Adverse collection conditions in electrostatic precipitators for
cleaning flue gases from power stations can be improved by conditioning
the flue gas. In many cases it is possible to lower the emission values
from old plants or operate new plants at maximum efficiency. This
paper will describe conditioning by the addition of SO3 and conditioning
with water in an evaporation cooler, the latter applied by Lurgi in a
power station for the first time.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the views of the Agency and no official endorse-
ment should be inferred.
7-1

-------
INTRODUCTION
Electrostatic precipitators have been used to dedust power station flue
gases for many decades. By constantly improving their design and dimension-
ing them accordingly, they have been adapted to meet the prevailing pollu-
tion control standards, which are constantly becoming more stringent, all
over the world. For example, the most recent laws in the Federal Republic
of Germany require that all new plants must reduce emissions to 50mg/NM3
&.02gr/SCF) and all old plants must be re-equipped by 1988 to meet emis-
sions of 125 mg/NM^ (.05gr/SCF).
In general terms, the following reasons may favour the use of condition-
ing units:
EXISTING PLANTS
The emission values originally established may no longer be attainable
due to the use of different coal grades or a change in the boiler operating
method.
Tighter pollution control legislation may force improvements in collec-
tion efficiency. A sufficient increase in precipitator size may be prevented
by inadequate local conditions which could lead to considerable capital ex-
penditure and undesirable shutdowns.
NEW PLANTS
If collection conditions are particularly difficult, the possibility of
conditioning should also be looked at, even for new plants. It enables
smaller precipitators to be installed, resulting in annual savings in capital
and operating costs of typically 10 - 15%.
If, in new plants, the boiler is fired with various types of coal, it
is possible to optimize the precipitator by constructing them smaller for
the coals which permit easier dust collection and only conditioning the
flue gas when the more problematical coals are being used. This becomes
particularly attractive if the latter adverse situation only occurs rela-
tively rarely.
CONDITIONING WITH SULFUR TRIOXIDE
Lurgi's experience with SO3 conditioning dates back to the fifties
and early sixties. At that time, hardly any flue gas cleaning plant was
able to fulfill the required warranties in Australia,, when low sulfur coal
was fired.
The answer was the use of stabilized SO3 imported from the United
Kingdom, then available only within a narrow temperature range and impos-
sible to evaporate without residues. These residues, once in place, were
7-2

-------
nearly impossible to remove.
In following years, up to 1973, further tests were occasionally run
with SO3, but owing to the problems in connection with the SO3, they were
never pursued to the stage of a technically viable solution.
The subject of SO3 conditioning received renewed interest in 1973, when
because of high dust resistivities, the guaranteed emission was not achieved
at a 315 MW power plant in southwest Germany. After SO3 conditioning tests
were successfully completed at this plant in 1975, we designed our first
industrial conditioning plant on the basis of sulphur combustion. This
plant became the first one built in Europe and was very successful. Our
biggest problem was to convince the government, who was concerned about SO2
emissions, which we will discuss later in this paper.
This process was not new to Lurgi. It is based on the know-how of
about 70 plants built for short-term sulphurization of washing-active sub-
stances and on the erection of approximately 110 plants for the production
of S02j SO3 and H2SO4.
These existing processes had to be simplified in most points on the
way to becoming an SO3 conditioning plant since the high demands as to gas
quality, H2O/SO3 content and SO3 conversion were not imposed.
Other agents were simultaneously tested for their effectiveness and
economy as conditioning agents, but SO-j from sulphur proved to be the most
reliable and most economical of them all.
What is the experience gathered with conditioning plants between 1975
and 1984?
In an installation erected specifically for testing purposes, we have
tested the flyashes of more than 35 coals from over 30 power plants with
all significant firing methods.
Sixteen large-scale plants are now in operation or under construction
(one of them on the basis of liquid SO2 evaporation).
SO3 CONDITIONING TECHNIQUE
The aim of conditioning with SO3 or water is to reduce excessively
high dust resistivities to ranges which do not impair the collection effi-
ciency of the precipitator.
The conditioning effect of SO3 is based on the conversion of the water
present in the flue gas to sulphuric acid, the condensation of this acid in
the pores of the dust particles and/or the adsorption of the sulphur trioxide
onto the surface of the dust particles with formation of sulphates. While
the flue gas temperature remains constant, the dust resistivity is reduced,
in some cases considerably.
7-3

-------
The raw material is liquid sulphur, which is burned to S02 in a
furnace. The SO2 gas is catalysed in a converter to SO3, which is blown
into the flue gas ducts upstream of the precipitator via a system of lances
and nozzles, using hot air as the carrier. (See Figure 1.)
filter
silencei
sulphur
1 catalyst
orifice
conditioned
flue gas
S0:
fan
electric air
heater
furnace
dosing pump
sulphur
flue gas duct
flue gas downstream
of air preheater
aoa
bypass	
r	1
^damper
|	Oiiiiuc n:r
-H\|	1 hPla,e p1
sulphur storage tank
Figure 1. SO3 Conditioning Plant (Sulphur Combustion)
It is also possible to use liquid SO2, which is evaporated in a
water bath and fed to the converter. This variant is considerably lighter
in weight as there is no combustion furnace. It is recommended as a
stationary system for smaller power station units or for dealing with
occasional dust peaks in larger units. Mobile pilot plants also operate
on this principle.
In larger boiler units, however, liquid sulphur should be used as it
is 4 to 5 times cheaper than liquid SO2. Also, it burns exothermically
which means a saving of electric power which would otherwise be required
for heating the carrier air and evaporating the S02.
Sulphur feeding is normally regulated in accordance with the boiler
load; however, it can also be controlled according to other criteria, e.g.
opacity, and integrated into the complete plant by means of micro-computer
(e.g. "Precicontrol").
EFFECT OF CONDITIONING WITH SO3 ON PRECIPITATOR PERFORMANCE
Lurgi's experience with approximately 30 boiler units can be summa-
rized as follows:
 Dust resistivity can be reduced by 0.5 to 2 powers of ten. However,
as the SO3 dosing rate is increased, a state of saturation occurs,
which is dependent on the boiler type and the properties of the dust
7-4

-------
such as chemical composition (basicity), specific surface area and
particle geometry. (See examples in Figure 2.)
30 40 Vppm 50
S03- dosing *
1	mixed Soar coal, S = 1,0%
2	Ruhr-coal,	S=1,2%
3	Polish cool,	S=0,8%
4	Southafrican coal, S-Q.5%
5	US-coal,	S=0,7%
Figure 2. Dust resistivity vs. SC^-dosing.
This contradicts the occasionally stated assumption that the dust
resistivity can be reduced to any desired value.
There is no measurable increase in the SO3 concentration at the pre-
cipitator outlet, provided that saturation of the flyash is not
reached.
The voltage-current characteristics of the electrical fields provide
significant information on the collection behaviour of flyash. If
these have too steep a gradient or even slope backwards, back ioniza-
tion is taking place. If the curve is too flat, the dust resistivity
is also high and/or severe space-charge conditions are present. In
both cases, the addition of SO3 results in the characteristics being
shifted into more favorable ranges. (See Figure 3.)
7-5

-------
0.35

-------
Range of improvement
_ by cond. with S03	
Figure 4. Migration velocity vs. SO3 dosing.
As could be seen already from the compilation of some measuring
results, w-value increases of up to 100% can be achieved by SOt conditioning
although it should be admitted that these increases are dependent on the
initial values and is only possible if there is a relatively low migration
velocity initially.
This is demonstrated more clearly by Figure 5, where the w-value
increases are plotted versus the initial value of the migration velocity
in the unconditioned state.
7-7

-------
Initial w-Value
Figure 5. Improvement of w-Value vs. initial w-Value.
CHANGE OF S02/S03 EMISSIONS IN SO3 CONDITIONING
The additional emissions of sulfur dioxide and sulfur trioxide in the
waste gases during the conditioning remains to be a much discussed subject
in the light of the official restrictions on the sulfur dioxide emissions
of power plants. As pointed out before, however, the extra emissions are
low and well within the range of measuring tolerances as evidenced by theo-
retical estimations as well as by actual tests.
This can be supported by an arithmetic example based on the following
pessimistic assumptions:
	Assume a boiler with ash return (smelting chamber or cyclone boiler)
where no flyash-bound SO2 is withdrawn from the process as would be
the case with a boiler with dry ash removal;
	Assume all chemical compounds of the ash constituents with injected
SO3 gas are separated again in the combustion chamber of the boiler
and the gas volume corresponding approximately to the dosed SO3 gas

-------
quantity leaves the boiler as sulfur dioxide;
Assume the SO2 to SO3 conversion rate of the catalyst is only 85%
compared to the otherwise usual 90 - 95%.
This leads to a theoretical SO2 increase as shown in Figure 6.
(/)
0)
A
C
fc
C
0
1
o4
V)
2000
tooo
800
100]
	
S=0,B8
S=1,0
Germany/Saar
	

S=0.7

Poland



US/Ohio

S = 0.34
South Africa
theoretical



15
25 Vppm 35
S03-Coritent before Precipitator 	.
cfl
E j,
cf
 A
? />
-5
n
O 3
(/> 7T
St
in 11
Vppn
3
o
<->,



US/Ohij,^''

South Africa

{"^6ermany/Soa


5	15	25 Vppm
S03-Coritent before Precipitator 
35
Figure 6. S02 and SO3 content n flue gas
vs. SO3 dosing.
In collaboration with the appropriate authorities, Lurgi has now
conducted a series of measurements at a power plant with the aforementioned
less favorable type of firing.
The theoretical, calculated extra burden of S02 in the waste gas by the
SO3 conditioning as calculated according to the above-mentioned example is
plotted in the upper graph. The pictures indicate clearly that the SO2
emission depends exclusively on the type of coal charged. The shape of the
curves show that an increase of the S02 emissions by conditioning cannot be
detected or is within the range of measuring tolerances.
During the described measurements, the SO3 content in the flue gas at
the stack inlet with and without conditioning was also determined in
7-9

-------
addition to the inlet and outlet gas dust contents.
The determined sulfur trioxide contents in the flue gas are depicted
in Figure 6, again according to the amount of SO3 dosing. It must be
emphasized here that the SO3 measurements are generally very difficult and
afflicted by a great factor of uncertainty, such that the represented
values of measurement are rather to be considered as trend indications.
However, it can be seen that the SO3 emission rises in a low degree by
about 2 to 4 Vppm up to values of approximately 3 to 5 Vppm SO3, which
are normally found in flue gases behind pit-coal fired steam generators.
On the other hand, the SO3 conditioning achieves a substantial reduction
of the clean gas dust contents down to ranges of 50 to 100 mg/NM3 with three
of the coal types charged.
FLUE GAS CONDITIONING BY EVAPORATION OF WATER
The concept of conditioning waste gases by evaporation of water is
not new. It has been practiced in cement works downstream of preheater
kilns since the early sixties, using evaporation coolers. The first
industrial-scale system in a power station was constructed by Lurgi, with
2
Figure 7. Flue gas dedusting precipitator with
conditioning tower.
7-10

-------
financial assistance from the Federal German Ministry of Research and
Technology (BMFT), in the Mannheim superpower station downstream of a
melting cyclone boiler with a capacity of 150 MW.
At the right of the diagram in Figure 7 is a prededusting precipitator
in the power station building. Downstream of this precipitator is an evap-
oration cooler 10.5 m in diameter and 32.9 m high.
The fan is housed in the vertical duct at the left. The downstream
precipitator is installed on a support structure above the evaporation
cooler. This arrangement was selected of necessity due to existing jobsite
conditions. In a new plant, the prededusting precipitator would naturally
be dispensed with and the electrostatic precipitator constructed at ground
level directly following the evaporation cooler.
ano*16 evaPrat:1-on cooler is designed to cool the flue gases from 150C
Through the combination of the lowering of the temperature and
t e increase in the dewpoint of the waste gas, the dust resistivity Is
reduced by over two powers of ten. (See Figure 8.)
temperature t in C
Figure 8. GKW-Mannheim Dust resistivity.
7-11

-------
In addition to this effect, the breakthrough voltage of the gas
increases, which permits a remarkably large rise in the corona voltage.
Conditioning with water thus results in a surprisingly large improve-
ment in collection efficiency. Figure 9 shows the charts on the boiler
switchboard in simplified form.
160
temperature in C
0.25
sick
150
clean gas dust content (sick)
	T
HO
0,20
130
0,15
120
110
0,10
U in kV 2. field
100
0.05
0,00
time
Figure 9. GKW-Mannheim Operational readings.
Recorded over time are:
	temperature downstream of the evaporation cooler (at precipitator
inlet)
	opacity at the precipitator outlet (as a relative measure of the
clean gas dust content)
	high voltages present in the 1st and 2nd fields of the precipitator
At about 11:30, the water injection system in the evaporation cooler
is shut down: the temperature rises, the applicable high voltages fall and
the clean gas dust content increases with a lag of about 10 minutes. At
about 12:15, the evaporation cooler is brought back into operation: the
temperature drops, the clean gas dust content improves immediately while
the high voltages rise. At 13:00, the changes just described are repeated.
To demonstrate the true improvement, the complete prededusting precipi-
tator and half of the downstream precipitator were shut down. Thus only
25% of the installed collection area was in use. Conditioning with water
reduced the clean gas dust content from 600 mg/m3 (NTP, dry) to 30 mg/m3
(NTP, dry), while the high voltage rose from 40 kV to 60 kV. (See Figure
10.)
7-12

-------
600 -
500 -
400 -
>
Ol
300 -
o>
100 -
SO 90 100 110
flue gas temperature C
120
130
MO
Figure 10. Clean gas dust content/high voltage
vs. flue gas temperature.
The same improvements were achieved whatever the type of coal used,
whether from Australia, South Africa or the Ruhr.
This plant has been in operation for almost 4 years without major
problems occurring or effects of corrosion and condensation appearing.
As there is an evident trend in the power station field towards a
single-pass boiler with dry ash removal, it would be entirely reasonable
to imagine systems in which the vertical duct to the precipitator is
designed as an evaporation cooler and used for gas conditioning with
water and possibly for SO2 removal with the addition of a reagent.
In conclusion, it can be said that conditioning with SO3 and water
is in many cases an attractive option from the point of view of cost.
Furthermore - and this should be particularly emphasized in conclusion -
the collection efficiency of the precipitator becomes largely independent
of the coal being burnt.
7-13

-------
PILOT-SCALE STUDY OF A NEW METHOD OF FLUE-GAS CONDITIONING
WITH AMMONIUM SULFATE
Edward B. Dismukes,
E. C. Landham, Jr.,
John P. Gooch
Southern Research Institute
Birmingham, Alabama 35255
Ralph F. Altman
Electric Power Research Institute
Chattanooga, Tennessee 37411
ABSTRACT
A new process for flue-gas conditioning with ammonium sulfate was
investigated on a pilot scale at EPRI's Arapahoe Test Facility. The agent is
decomposed to ammonia and sulfur trioxide in a hot gas slipstream to simulate
bypassing the air heater in a full-scale plant. The process thus avoids loss
of sulfur trioxide (the effective chemical species) and corresponding
plugging of the air heater, which occur when the agent is decomposed conven-
tionally in the main fluegas stream ahead of the air heater and the decom-
position products recombine upon cooling in the air heater. Chemical aspects
of the decomposition process were characterized on the pilot scale and
compared with earlier laboratory data. Lowering of the electrical resisti-
vity of ash from low-sulfur coal was demonstrated by (1) direct measurements
of reduced resistivity in situ and (2) observations of electrical manifesta-
tions of suppressed back corona in an ESP. Comparable improvements in ash
collection efficiency in the ESP, from about 98.6 to 99.6%, were observed
with sulfur trioxide obtained by the decomposition of ammonium sulfate or by
the vaporization of liquid sulfur trioxide.
The work described in this paper was not funded by the U.S. Environmen-
tal Protection Agency and therefore the contents do not necessarily reflect
the views of the Agency and no official endorsement should be inferred.
8-1

-------
BACKGROUND
The use of ammonium sulfate (formula: (NH^ ^SO^ ) for flue gas
conditioning began to receive prominent attention some 15 years ago. Much
of the original interest in the compound stemmed from the belief that it
provided an optional source of the classical conditioning agent SO3, which
could be used without the hazards associated with other existing sources of
SO3. Generation of SO3 from (NHij^SO^ was assumed to occur as the result of
thermal decomposition when the compound was injected into a flue gas duct,
either as the finely divided solid or as a spray of a concentrated aqueous
solution. The thermal decomposition process can be written as follows:
(NH^ )2SOi+(s) 2NH3 (g) + S03 (g) + H20(g)
The earliest interest in (NH^^SO^ for flue gas conditioning was
exhibited in Great Britain, and the earliest technical papers on the subject,
from both practical and fundamental perspectives, were published by
investigators of the CEGB (1,2,3,4). Application of the compound in the
United States followed that in Britain and was promoted primarily by
commercial interests (5,6). At the third of these symposia in Orlando in
1981 (7), staff members of Southern Research Institute presented the results
of comprehensive studies of (NH^^SO^ conditioning at two U.S. power plants,
which had been performed as part of a research project financed by the
Electric Power Research Institute. Further details of this work have more
recently appeared in an EPRI report (8).
A key finding by SRI investigators is that SO3 resulting from the
thermal decomposition process is the key to the conditioning process because
of the usual effect of S03 on the electrical resistivity of fly ash. The
Corette plant of Montana Power employed (NH^^SO^ successfully with the most
common site of injectionin the backpasses of the furnace. The Gannon plant
of Tampa Electric was able to use the compound with injection either before
or after the air preheater. Success at Gannon with injection at the lower
temperatures downstream from the air heater arose in large degree because
these temperatures were well above 300F and at least partial decomposition
could still be effected.
Despite success at Corette with high temperature injection, the utility
industry has encountered repeated difficulties with high temperature
injection of (lSHi+^SO^. in the form of air heater pluggage. This phenomenon
is a result of the reversal of the decomposition process on the relatively
cool metal surfaces in an air heater. The frequency of this difficulty, plus
the quite limited decomposition of (NHi+^SOi* that occurs at prevailing air
heater exit temperatures, led SRI and EPRI to the formulation of a new
process for conditioning with (NHt+)2^0^namely, injection of the compound in
a slipstream of hot flue gas or hot combustion air that bypasses the air
heater. This stream can be sufficiently high in temperature to give complete
decomposition, while obviously avoiding the air heater pluggage problem.
8-2

-------
The first step to implement this concept was a laboratory study at
SRI supported by EPRI. The key findings in the laboratory were as follows:
	(NH^^SO^ injected as a concentrated aqueous solution can be
completely volatilized in a gas stream at 700 to 900 F with
product concentrations up to 2000 ppm of NH3 and 1000 ppm of
SO3. With an SO3 concentration of this magnitude attainable,
the slipstream need have a flow rate approximating only 1% of
the gas stream to be treated, assuming a typical treatment
level of 10 ppm.
	The decomposition process follows the desired course, rather
than the thermodynamically favored process in which the
products are N2 and S02 rather than NH3 and S03 (9). This
merely means that the reaction follows a thermodynaraically
predicted course but not the course with the greatest release
of energy (Ap). The presumed explanation is that NH3 and SO3
are released by flash volatilization and the reactants
required for N2 and S02v production do not remain in contact.
	The decomposition process nears completion in a reaction time
of about 2 sec in the stated temperature range, making for a
practical length of the gas duct in a power plant where
decomposition would be effected.
Certain major issues that could not be successfully resolved in the
laboratory were addressed on a pilot scale at EPRI's Arapahoe Test Facility,
where a pilot-scale ESP was available for study.
SCOPE OF THE PILOT-SCALE STUDY
The principal phases of the work in preparation for pilot-scale studies
and execution of these studies at Arapahoe were as follows:
	Fabrication and installation of a decomposer that could
provide up to 20 ppm of SO3 in a flue gas stream having
flow rates up to 5000 acfm at 300 F.
	Exploratory studies with the apparatus to determine its
performance capabilities with respect to:
achieving (NH1+)2S0it decomposition efficiencies comparable
to those in the laboratory,
accounting for the fate of the decomposition products
following their injection into the flue gas duct,
determining the effect of the conditioning agent on fly ash
resistivity, and
establishing the effect of the agent on the electrical
behavior of a pilot-scale ESP.
8-3

-------
Exit temperature from the decomposerminimum, 600 F; maxi-
mum, 700 F.
Solution feed rate to the decomposerminimum, 10 mL/min;
maximum, 60 mL/min. The most commonly used feed rate was
30 mL/min, which delivered sufficient agent from a 20%-by-
weight solution to produce 20 ppm of SO3 at the typical flow
rate of flue gas to the decomposer, 3500 acfm at 300 F.
EXPLORATORY STUDIES1
A number of experiments were conducted to determine relative concentra-
tions of NH3 and SO3 issuing from the decomposer outlet in the gaseous and
particulate states. In most of these experiments, the rate of addition of
(NHit)2S0i+ was selected to produce a concentration of AO ppm of NH3 and 20 ppm
of SO3 in the flue gas duct. The essential results from these experiments
were as follows:
	The percentages of the total amounts of NH3 and S03 collected
in the particulate state were approximately 0.01% for NH3 and
0.08% for S03. These results signified that the (NHt^SO^
was essentially all decomposed to the vapors. The somewhat
higher value for SO3 probably was a result of some vapor
adsorption on the sampling filter. A typical distribution of
the collected decomposition products is given in Table 1.
	Recoveries of both NH3 and S03 were considered satisfactory
in view of the handicap faced in the sampling operation.
There could be no direct physical connection between the
sampling line and the decomposer outlet; instead, the inlet
of the sampling line had to be placed as close as possible to
the outlet from the decomposition chamber, with the sheath
air used as a shield against dilution by the flue gas.
During injection of 40 ppm of NH3 and 20 ppm of SO3, the
recoveries of both decomposition products averaged 83%.
Further details are given in Table 2.
	No SO2 from the decomposer was detected. Hence, the
(NHit^SOit decomposed exclusively to NH3 and SO3, as desired.
Another set of experiments was conducted to determine the fate of NH3
and SO3 from decomposed (NHi+^SO^. Here, the sampling operation was at the
inlet of the ESP; the results are given in Table 3. Total recoveries of NH3
Although the decomposer was designed for use with either air or flue gas
as the transport medium in the decomposer, air was used at all times in the
work discussed in this paper.
8-4

-------
TABLE 1. DISTRIBUTION OF NH3 AND S03 IN THE SAMPLING TRAIN AT THE
DECOMPOSER OUTLET
Species
collected
Component of the
sampling train
Relative moles collected
nh3
SO 3
Particulates
Instack filter

0.0002
0.0008
Gases
Probe

0.90
0.81

Water bubbler

0.64
0.03

Condenser

0.36
0.15

Backup filter

0.04
0.01

H2SO^ bubbler

0.06




2.00
1.00
TABLE 2
. RECOVERY OF NH3 AND
SO^ AT THE DECOMPOSER OUTLET

Feed solution
Calculated
Recovery, %
Number of

concentration


experiments
Wt % ml/min
as SO3, ppm
nh3
S03
5
20 30
610-680
73-93
67-96
2
30 30
1060-1070
62-70
70-79
1
40 30
1480
42
56

TABLE 3. MATERIAL BALANCE AT THE ESP
 INLET

Experiment

1
2
3
4
ppm (NH^^SO^ injected 12
11
21
22
ppm NH3 found -




particulate
16
12
20
18
gas
total(recovery, %)
12
28(115)
9
21(95)
13
33(79)
17
35(80)
ppm SO3 found -
particulate
gas
total(recovery, %)
13
< 1
13(108)
10
< 1
10(91)
18
< 1
18(86)
13
< 1
13(59)
8-5

-------
Determination quantitatively of the improvement in fly ash
collection efficiency by this ESP unit upon treatment of
the flue gas with (NHii^SO^ and comparison of the effect of
(NH^^SO^ with that produced by SO3 vapor from stabilized
liquid SO3 (that is, without the simultaneous addition of
NH3).
FABRICATION AND INSTALLATION OF THE DECOMPOSER
The decomposer was constructed from parts fabricated in the machine shop
of SRI in Birmingham and certain other parts that were purchased from
commercial sources. A schematic drawing of the apparatus is shown in
Figure 1. The primary components were as follows:
An inlet to allow either ambient air or flue gas from an
existing duct at Arapahoe to be drawn into the system.
A blower to force the air or flue gas to a heater capable
of producing an exit air or gas temperature of 900 F.
A wye at the heater exit leading to two concentric vertical
channels, with the inner channel to serve as a chamber for
(NH^^SO^ decomposition and the surrounding annular channel
to provide insulation.
A spray nozzle for introducing (NH^^SO^ solution into the
inner channel.
A nozzle for directing the decomposition products into the
flue gas duct, with the sheath gas providing a protective
barrier between the inner stream of decomposition products
and the surrounding flue gas, to allow gradual cooling of
the decomposition products and gradual mixing with the flue
gas.
The apparatus was mounted at Arapahoe above a duct having an inside
diameter of 18 in., carrying flue gas from Unit 4 of the Arapahoe power
station to the pilot-scale ESP. It was thus located some 56 ft from the
entrance of the ESP and provided a residence time of about 1.5 sec between
the decomposer outlet and the ESP inlet.
Performance characteristics of the decomposer in actual use were as
follows:
Air flow through the decomposer at a heater exit temperature
of 900Fminimum air flow, 90 acfm; maximum, 210 acfm.
Gas residence time in the decomposerminimum, 1.3 sec;
maximum, 3.0 sec.
8-6

-------
SPRAY NOZZLE
CONCENTRIC
CHAMBERS
Ca. 11 FT
HEATER
AIR OR FLUE GAS
13 FT
BLOWER
INJECTION


FLUE GAS DUCT
Figure 1. Elevation view of decomposer.
8-7

-------
averaged 90% and of SO3 84%. Approximately half of the NH3 was found in the
particulate phase and half in the gas phase. All of the S03 found, on the
other hand, was in the particulate matter, indicating either partial
recombination of NH3 and S03 as NH^HSO^ particulate matter
NH3(g) + S03(g) + H20(g) * NH^HSO^(s or 1)
or selective adsorption of S03 by the fly ash.
Efforts to demonstrate the resistivity effect of NH3 and S03 from decom-
posed (NH^^SOit were handicapped by the absence of a consistent composition
of the coal being burned in Unit 4. The resistivity of unconditioned ash
ranged from 2 x 1011 to 3 x 1012 ohm cm. Part of this variation was
attributed to observed variations in the Na20 content of the fly ash;
however, there were apparently other contributing factors not identified.
Figure 2 is the result of an effort to show the effects of conditioning
in the face of large variations in the baseline ash. Here, the ratio of
resistivity of conditioned ash to that of unconditioned ash is plotted versus
the injected concentration of S03, the source material being either (NHi+^SOi^
or NHltHS01+. The effect appeared to have been most evident below 10 ppm of
S03 ; the reduction was around one order of magnitude from 10 to 20 ppm of
S03. The source compound, (NH^^SO^ or NHttHSOt+, apparently had no bearing on
the magnitude of the effect produced. NH^HSO^ (ammonium bisulfate) is an
analog of (NH^^SO^, which decomposes similarly to produce a similar effect.
The effects on the ESP electrical conditions of (NHi+)2S0lt injected at a
rate equivalent to 20 ppm of S03 are shown in Figure 3. Also shown for
comparison are the effects of 10 ppm of S03 from the stabilized liquid S03
source material. The plots of current density versus voltage without
conditioning show clear evidence of back corona, and those with conditioning
by either (NH^^SOit or S03 show suppression of back corona. A difference
between the effects of the two agents is higher voltages for given current
densities with (NH4 )zS0i+ than with S03; this may have been the result of a
space-charge enhancement by the NH1+HS0lf aerosol produced by partial
recombination of NH3 and S03, as mentioned above. There are other anomalies
in these plots that cannot be explained on an unqualified basis. Even with
conditioning, the maximum current densities that could be reached without
sparking were low (<10 nA/cm2). While this project was in progress, attempts
to raise the attainable current density by cleaning of insulators, wires, and
wire frames were unsuccessful; only later, after washing down the collection
plates, did the limitation disappear.
ESP PERFORMANCE TESTS
Four series of tests were conducted to quantify the performance of the
ESP with and without conditioning. Three series (referred to as Tests 1, 2,
and 3) were conducted with the addition of (NH^^SO^ through the decomposer
2
These data were from the performance tests subsequently described.
8-8

-------
> 0.20
0.02 
O (NH4)2S04
A IMH4HSO4
0.01
5	10	15
EQUIVALENT SO3 CONC, ppm
20
Figure 2. Effect of decomposed (NH4)2 SO4 and NH4H SO4 on fly ash resistivity.
8-9

-------
7.5
5.0
2.5
OUTLET FIELD
~
\
NO	}
CONDITIONING O
/
O

A
I
I
I
~Ai
/ /
I
SO 3 j
I I
/ /
D /
/ a(NH4)2S04
5.0
MIDDLE FIELD
O
\
2.5
NO	o
CONDITIONING j
O
S03/
O
/
 ~JC
5.0
/
A ^
"""* (NH4)2 so4
INLET FIELD
/
SO,
NO
CONDITIONING-
2.5
~	 O
/
/
A (NH4)2 so4
P^A'
22
26
30	34
VOLTAGE, kV
38
42
Figure 3. ESP electrical conditions without conditioning (circles) or with
(NH/p2 so4 conditioning (triangles) or SO3 conditioning (squares).
8-10

-------
at a rate equivalent to a nominal SO3 concentration of 20 ppm. The fourth
series (Test 4) was conducted with 10 ppm of S03 vapor from stabilized liquid
SO3. During Test 4, the SO3 was vaporized by heating a drum of the liquid
source material, and the vapor was fed through a flow meter and a pipe
mounted at the top of the decomposition chamber in the decomposer with air as
a carrier gas. The injected concentration of S03 from the source liquid, 10
ppm, was the maximum rate that could be maintained with this injection
system. Parameters measured during the ESP performance tests included:
inlet and outlet fly-ash concentrations, outlet opacity, electrical
resistivity, ESP electrical conditions, and NH3 and S03 concentrations.
So-called Energy coal was the customary fuel at the Arapahoe, but
another coal referred to as Arco-Somerset was burned during one of the four
tests. Both are subbituminous Western coals, typically with 0.6% sulfur.
Analysis showed that the two coals differed somewhat in the composition of
their ashes, as revealed by the results of analyses of hopper ash that are
given in Table 4. The most important differences are in components that have
an important bearing on electrical resistivity (Na20, K20, CaO, and Fe203).
The higher sodium and potassium contents of the ArcoSomerset ash, in
particular, lead to a lower predicted resistivity based on the Bickelhaupt
model (10): for example, 2 x 10** ohm*cm for Arco-Somerset versus 7 x 10
ohm*cm for Energy (calculated for 300 F, 8% water vapor, at 10 kV/cm).
Table 5 gives the results of measurements of the collection efficiency
of the ESP with and without conditioning in each of the four test series.
Except in Test 2, the reduction in penetration due to conditioning was around
60%; in Test 2, the reduction was only about 30%. There is no apparent
reason why (NHi^SO^ performed so differently with Energy coal in Tests 1 and
2 although, as will be shown, data on the effect of the conditioning agent on
the SO3 content of the ash reveal an abnormally low recovery in Test 2.
Table 5 suggests that 20 ppm of (NHtt)2S0i+ was capable of producing an
improvement in ESP performance comparable to that observed with 10 ppm of
SO3. it should not be concluded that (NHt^SO^ will generally be required at
higher concentrations than SO3. To the extent that resistivity reduction
determines the effectiveness of (NHi+)2S01+, one would expect that 10 or 20 ppm
of this agent would produce equivalent effects (see Figure 2). It was
unfortunate that a more comprehensive comparison of the two agents was not
possible at Arapahoe. The unexpected limitation in the output of the SO3
system, plus scheduling difficulties, forced postponement of more a
definitive comparison until later.
Determination of the effects of (NHi+)2S0lt on the opacity of the flue gas
at the ESP outlet revealed a significant suppression in both the time-average
opacity (from about 16% to 8%) and the intensity of rapping puffs (from about
60% to 30% of the opacity scale). The latter effect, it might be argued, is
a manifestation of increased cohesion of the ash deposits, an effect
previously observed with NH3 additive combined with naturally occurring
S03 (11).
8-11

-------
TABLE 4. COMPOSITIONS OF FLY ASH FROM ENERGY AND
ARCO-SOMERSET COALS

Weight
percent in ash

Component*
Energy
Arco-Somerset

Li20
0.03
0.03

Na20
0. 52t
1.92t

K20
1. 2t
1.5t

MgO
1.5
1.4

CaO
4.5T
5.21"

A1203
26.6
23.6

Fe203
3.5t
7.91*

Si02
57.0
54.6

Ti02
1.2
0.95

P205
1.2
0.95

S03
0.45
0.70

~Determined after ignition of ash at
1380F.


tSignificant differences for components have a
large impac t on
electrical
resistivity.



TABLE 5. ESP MASS
COLLECTION EFFICIENCIES

Test Series
SCA,
ESP
Relative
Coal Agent* ft2
/lOOOacfm
efficiency,%
penetration
1(Energy) None
290
98.86
1.14
(nh4)2so4
330
99.60
0.40
2(Energy) None
324
98.44
1.56
(NHit )2S04
320
98.93
1.07
3(Arco-Somerset) None
367
95.60
4.30
(NHi+ )2S0it
367
98.14
1.86
4(Energy) None
364
98.83
1.17
S03
364
99.63
0.37
*(NHlt)2SOit, 20 ppm. SO3, 10 ppm.




8-12



-------
The results of measurements of ash resistivity are summarized in
Table 6. The magnitudes of the effects of both agents on resistivity were
comparable reductions by one order of magnitude or more. The effects of
both agents on voltage-current behavior of the ESP were also similar, as has
been illustrated previously in Figure 3.
The results of determinations of NH3 and SO3 in the vapor state at the
ESP inlet and determinations of these substances in the ash deposited in the
ESP hoppers are summarized in Table 7. These experimental results are
compared with the increments expected from the rates of injection as
percentages recovered. Calculated recoveries of NH3 and SO3 are generally
similar to those obtained in the preliminary studies, except in Test 2 when
the recovery of S03 was abnormally low.
CONCLUSIONS
Ammonium sulfate can be thermally decomposed to NH3 and SO3 by spraying
an aqueous solution into a hot gas stream. Product concentrations range
upward to at least 2000 ppm of NH3 and 1000 ppm of S03 in a residence time of
2 sec or less in the temperature range from 700 to 900 F. Introduction of
the product stream into a flute gas duct upstream from a cold-side ESP, in the
ratio of about 1 volume part of product stream to 100 parts of flue gas,
effectively lowers ash resistivity and produces corresponding improvements in
the voltage-current performance of an ESP. The efficiency of ESP performance
is correspondingly improved; reductions of ash penetration to the ESP outlet
of 60% have been observed. The effect of ammonium sulfate conditioning on
ESP performance approximates that by SO3 introduced in the absence of NH3.
8-13

-------
TABLE 6. ASH RESISTIVITY DURING ESP PERFORMANCE TESTS
Resistivity	Relative
Test	Operating Mode*	(ohm cm)	Resistivity
1
Baseline

1.3
X
1012
1.00

Conditioning,
(NH^SO^
6.8
X
1010
0.05
2
Baseline

3.0
X
1012
1.00

Conditioning,
(NH^SO^
8.9
X
1010
0.03
3
Baseline


-T

	

Conditioning,
(NH^SO^
3.4
X
1010

4
Baseline

2.8
X
1012
1.00

Conditioning,
S03
3.0
X
1011
0.10
*Tests 1, 2, and 3 with 20 ppm of (NH^^SO^ or Test 4 with 10 ppm of SO3.
tNot successfully determined.
TABLE 7. RECOVERIES OF NH3 AND SO3 DURING ESP PERFORMANCE TESTS
NH3, ppm	SO3, ppm
Operating 	 	
Test	Mode	Vapor Solidt Recovery,% Vapor Solidt Recovery,%
1
Baseline
1.4
0.1
-
0.9
8.7
-

Conditioning
12.8
15.4
67
<0.1
17.6
40
2
Baseline
1.1
<0.1

0.8
6.8
_

Conditioning
20.6
5.7
63
<1.0
9.2
10
3
Baseline
6.8
0.1

<1.0
5.4


Conditioning
26.2
12.6
9.5
not det'd
13.3
40
4
Baseline
1.4
0.3

1.0
9.1


Conditioning
0.6
0.1
-
4.3
12.0
60
~Recovery: total found for vapor and solid during conditioning less that for
baseline operation, relative to injected value, 40 ppm of NH3 or 20 ppm of
S03.
tCalculated as the vapor concentration equivalent to that in the solid
collected in the ESP hoppers.
8-14

-------
REFERENCES
1.	Lowe, H.J. Reduction of Emission of Pollutants - Recent Advances in
Electrostatic Precipitators for Dust Removal. Phil. Trans. Roy. Soc.
Ser. A. 265(1161):301, 1969.
2.	Dalmon, J, and Tidy, D. A Comparison of Chemical Additives as Aids to
the Electrostatic Precipitation of Fly Ash. Atmos. Environ. 6:721,
1972.
3.	Halstead, W.D. Thermal Decomposition of Ammonium Sulfate. J. Appl.
Chem. 20:129, 1970.
4.	Dalmon, J. Research Division, Central Electricity Generating Board.
Private communication, 1976.
5.	Method of and Apparatus for Chemically Conditioning a Particle-Laden Gas
Stream. U.S. Patent 3,665,676 (Assignee: Koppers Company, Inc.).
1972.
6.	Method of Conditioning Flue Gas to Electrostatic Precipitator. U.S.
Patent 4,043,768 (Assignee: Apollo Chemical Corporation). 1977.
7.	Landham, E.C., Jr., Marchant, G.H., Jr., Gooch, J.P., and Altman, R.F.
Field Evaluations of Ammonium Sulfate Conditioning for Improvement of
Cold Side Electrostatic Precipitator Performance. In: EPA-600/9-82-
005b. U.S. Environmental Protection Agency, Research Triangle Park, NC,
1982. p 237.
8.	Gooch, J.P., Bickelhaupt, R.E., and Dismukes, E.B. Investigation of
Ammonium Sulfate Conditioning for Cold-Side Electrostatic Precipitators.
Volume 1: Field and Laboratory Studies, EPRI Report CS-3354, Volume 1,
1984. 59 pp.
9.	Kelley, K.K., Shomate, C.H., Young, F.E., Naylor, B.F., Salo, A.E., and
Huffman, E.H. Thermodynamic Properties of Ammonium and Potassium Alums
an Related Substances, with Reference to Extraction of Alumina from Clay
and Alunite. Technical Paper 688, Bureau of Mines, Washington, D.C.,
1946. 104 pp.
10.	Bickelhaupt, R.E. A Technique for Predicting Fly Ash Resistivity. EPA-
600/7-79-204. U.S. Environmental Protection Agency, Research Triangle
Park, NC, 1979. 15 pp.
11.	Dismukes, E.B. Conditioning of Fly Ash with Sulfur Trioxide and
Ammonia. EPA-600/2-75-015, U.S. Environmental Protection Agency,
Research Triangle Park, NC, 1975. 155 pp.
8-15

-------
POWER PLANT PLUME OPACITY CONTROL
J. Martin Hughes
Kai-Tien Lee
Civil Engineering Department
Virginia Polytechnic Institute and State University
Blacksburg, Virginia 24061
ABSTRACT
Achieving compliance with plume opacity standards remains a significant
problem in the electric utility industry. Even modern plants equipped with
electrostatic precipitators and scrubbers emit plumes that do not comply with
standards for visible emissions. The conversion in the presence of
catalysts, of sulfur containing flue gases to sulfate aerosols is the major
source of the problem. This paper deals with the control of plant operating
parameters and the application of control equipment and/or chemical additives
to bring about plume opacity compliance. In achieving compliance with
opacity standards a reduction in plant particulate emissions will be achieved
y the fact that particulate formation in the plume will be reduced.
9-1

-------
INTRODUCTION
More than forty Class Al"'" SIP and NSPS Power plants operate with plume
opacity in excess of opacity standards as measured by EPA Reference Method
9(1). Enforcement action against these plants consists of Federal Court
Orders, Federal Consent Decrees, Administrative consent Orders and State
Stipulation Agreements. The number of megawatts out of compliance due to
these plants operating amounts to 29870(1).
Table 1 indicates overwhelmingly that ESP controlled plants generate the
majority of the compliance problems. EPA Policy has been "... not to pursue
visible emission violations where we believe it is probable that the source
is in compliance with the mass standard"(2). Due to this EPA policy, almost
no visible emission enforcement action has been instigated against coal fired
power plants with FGD equipment.
Consultation with certified opacity observers in several states and
Lidar operators indicated that a number of FGD type coal fired power plants
do have plume opacity levels above the standards(4).
TABLE 1 ELECTRICAL UTILITY PLANT DATA*
Plants
Operating
#
Under
Construction
Control
Facilities
Collection
Efficiency
%
VE Violations
Reported
59
41
FGD or
ESP+FDG
45-99
4 (3)(4)
178

ESP
95-99+
41 (5)
31
11
FF
99.9
0 (6)
*	Plants Greater than 100 MWe Capacity
#	Reported number of utility power plants (not individual boiler units)
VE Visible Emissions
FGD Flue Gas Desulfurization
ESP Electrostatic Precipitation
FF Fabric Filter (Baghouse)
^"Emissions of more than 100 TPY of criteria pollutant from plant with
operating air pollution control equipment
9-2

-------
Therefore, the four FGD plants indicated in Table 1 depict only a
fraction of the FGD type plants with non-compliance status with reference to
plume opacity (VE) standards. The number of these units experiencing
problems in complying with plume opacity standards should be determined.
Some plumes are subject to a no visible plume standard (5% opacity) and some
a 10 or 20% standard for 3 minutes maximum depending on the state having
jurisdiction and any applicable PSD or SIP designations.
The factors affecting plume opacity in reactive plumes were not well
understood in 1977 when EPA promulgated NSPS for power plants(7). As a
result the enforcement of regulations was such that the power plant opacity
required by the standard could be overlooked after it had been demonstrated
that mass emission limits were being met and that steps were taken to operate
the source so as to minimize opacity emissions(8).
BACKGROUND
Oil fired power plants violating opacity standards typically comply with
mass emission standards while coal fired plants typically violate VE
standards when simultaneously violating mass emission standards(9). The
investigation reported here focuses on the atypical case when in-situ
transmissometer VE measurements show compliance but plume opacity as measured
by Reference Method 9 or LIDAR exceeds opacity standards. This case comes
about due to gas phase reactions that produce fine aerosols, vapor phase
condensation and physical agglomeration of sub micron sized clusters and
particles.
Stack sampling and monitoring facilities do not permit the determination
of the composition and aerosol size distribution changes in plumes. Sampling
of reactive plumes via ground based and airborne platforms presents
challenging problems to be solved(lO). Tethered balloons, helicopters and
aircraft have been used with some success. The plume opacity control
technology applicable to these aerosols which are created and/or grown in
power plant plumes is the main topic of this paper.
Other topics that need to be investigated in more detail are
1.	Particulate emissions inventory modification due to reactive
plumes as related to receptor and dispersion modeling
2.	The impact of near stack plume reactions on rain water quality
3.	Modifications of ambient air quality levels resulting from
plume reactions as related to existing ambient and proposed
fine particulate standards
9-3

-------
POWER PLANT PLUME OPACITY CONTROL
OIL FIRED POWER PLANTS
Over the past twenty years additives have been used for controlling
sulfur compound levels in oil fired plant flue gases(11). Table 2 summarizes
some results that relate to flue gas opacity control in that SO., at only a
few PPM in flue gas is strongly related to plume opacity.
TABLE 2 TREATMENT OF FLUE GAS FROM RESIDUAL 0IL(2-4.5%S) FIRED UNITS(11)

COMPANY
ADDITIVE
WT%
AS/IN
SO
% REDUCTION
Florida Power
Dolomite
0.1
Slurry/Fuel Oil
33
Florida Power
Dolomite
0.1
Dry Power/Flue Gas 42
Florida Power
Dolomite
0.2
Dry Powder/Flue
Gas 50
English
Dolomite
0.2
Dry Powder/Flue
Gas 80
English
MgC03
0.1
Dry Powder/Flue
Gas 75
Long Island
Lighting
MgO*
NS
NS
Significant
English
Mg(20-50 Mesh)
0.02
Dry Powder/Flue
Gas 60-100#
Consolidated
Edison
MMT**
NS
NS
45
* Low Excess Air (10% Range), Small Particle Size (10 Micrometer)
// Function of Injection Point
** Methyl Cyclopentadienyl Manganese Tricarbonyl
NS Not Specified
Recently Brookhaven National Laboratory conducted a study (12) of
sulfate formation in oil-fired power plant plumes and produced the following
empirical equation which the authors called a "somewhat speculative
correlation".
TPO - 1.67[MS04] + 3([H2S04] - 6)	(1)
9-4

-------
where
TPO is the total plume opacity in %
[MSO^] is the ppm on a molar basis of the metal sulfates
in the flue gas
[^SO^] is the ppm on a volume basis of the sulfuric acid
in the flue gas
The value of [MSO^] was predicted by
0 8
[MSCV ' (7.07 'o"2[02])(1- jZI5o)|[a]+0.38([VMMg]H10[H2S04][ (2)
where
[02] is the furnace oxygen in Vol. %
ESP is the ESP efficiency in %
[Na], [V], [Mg] are the ppm on a weight basis of sodium, vanadium, and
magnesium in the fuel oil. Eq. (2) was reported to correlate with measured
results with an overall relative standard deviation of 24%.
The sulfuric acid concentration in the flue gas was given by
1300L[0][S]
[H SO ]	1		 	[5J		O)
(7.07 + 0.32[02])2 ([V] + [Mg] + [Na] + [M] + 1/2[CJ)
where
[H2S0^] is the sulfuric acid concentration in ppmv in the flue gas
L is the operating fraction of full plant power output
[S] is the sulfur content of the fuel oil in wt. %
[M] is the wt. % of metal other than [V], [Mg], and [V] in the fuel
[C] is the unburned carbon in the fly ash as a fraction of carbon
in the fuel oil, wt. ppm
Since this opacity correlation is based on sparse data from various
sources with incomplete data sets, additional studies must be carried
ut to validate its range of applicability. For the Brookhaven data set
the data ranges investigated are listed in Table 3.
Plume opacity exceeds in-stack opacity for the cases of (A) high
emissions of sulfuric acid which condense (13,14) and (B) the presence in the
plume of metal sulfate particles with sizes less than the visible size range
(0.4-0.7 um) (15-17) which grow into the visual size range by hydration in
the plume as it cools.
9-5

-------
TABLE 3 RANGE OF PARAMETERS INVESTIGATED


PARAMETER RANGE
PARAMETER
MINIMUM
MAXIMUM
[H2S04], vol. ppm
0.1
39.5
L, fraction of full power
0.39
1.02
[02], vol. %
0.1
2.9
[S], wt. %
0.31
2.59
[V], wt. ppm
4.0
463
[Mg], wt. ppm
11
841
[Na], wt. ppm
32
179
*
[M], wt. ppm
50
100
[C], wt. ppm unburned
230
6100
Other metals, M, in the fuel ash.
This plume opacity model and the field data from investigations which
support it provide a composite picture of the significant factors that govern
plume opacity in oil fired power plant plumes. They are:
1.	Furnace oxygen
Plume opacity increases almost linearly with furnace oxygen above
the [HS0,] greater than 6 ppmV level up to 3% 0_ and then at a somewhat
lesser rate for higher furnace oxygen concentrations. For a given furnace
[0] operating level the firing of fuel oil with a higher vanadium content
will result in increased opacity levels.
2.	Sulfur and Vanadium content of fuel oil
To give an example of the effect of sulfur and vanadium content in fuel
oil on plume opacity the data in Table 4 was used with Eq. (1). Note that
for [H-SO.] less than 6 the second term on the RHS of Eq. (1) is dropped.
The plume opacity resulting from the operation of unit 1 was calculated by
Eq. (1) to be more than 4 times greater than for unit 2 based on Table 4
data.
For [S3 content of fuel oil below 2% the TPO is essentially all due to
metal sulfate particulate. For [S] content of fuel at the 4% level TPO would
be 36% with only 3% due to [MSO,] and the balance of 33% plume opacity would
be due to [l^SO^] .
9-6

-------
TABLE 4 EFFECTS OF FUEL OIL COMPOSITION ON PLUME OPACITY
Unit
[S]
[V]
t2]
[mso4]
[h2so4]
TPO* %
1
2.10
390
2.0
12.3
19.5
61
2
*
0.33
5.0
3.6
8.1
0.07
14
Calculated
3. Plant Load Level
The plant load, L, depicted in Eq. (3) was seen to be proportional to
[H2S04^* Below the 6 ppmV level [H-SO^] does not typically contribute to
plume opacity. For these cases [MSO^] levels determine TPO. Table 5 data
was taken from a plant (12) which operated at 1/2 load (176 MW output) for 5
hours (1500-2000) on 6/12/78 and then at nearly full load (320 MW) for 2 1/2
hours (1330-1600) the next day with the same fuel oil with [S] = 0.75 wt. %.
From the point of view of steady state plant operation for the case
examined in Table 5 doubling plant load will increase TPO by a factor ranging
from three to seven.
TABLE 5 PLANT LOAD EFFECT ON PLUME OPACITY
Load
[msoa]
[H SO ]
tsp3
*
TPO
MW
ppmV
L f
rag/m
%
176
2.69-3.04
1.82-2.17
29-34
4.5-5.1
320
*
9.70-20.17
3.80-3.82
177-293
16-34
Calculated
4. ESP Performance
Less than 5% of the total measured ^SO, was found adsorbed on the fly
ash (12). The concentrations of sulfuric acid vapor were measured at the
inlet and outlet of an operating ESP and found, as expected, to be
essentially the same (12). The ESP effect on TPO was indirect. As the
furnace oxygen level was decreased, an increase in particulate carbon
resulted. Therefore, for the more efficient ESP installations the furnace
[o2] can be reduced to a lower level to achieve lower levels of [I^SO^] in
the flue gas and reduced plume opacity than would be the case with less
9-7

-------
efficient precipitator installations. For [t^SO^] greater than 6 ppmV this
indirect effect is pronounced. For the case when [l^SO^] is less than 6ppmV
the plume opacity would drop an insignificant amount (from 20% to 19%) when
ESP efficiency was raised from 88% to 92% with an accompanying furnace [C^]
reduction from 2.7 to 1.0 vol %.
5. Combustion Conditions
The importance of good combustion achieved by efficient mixing of
combustion air and fuel oil aerosol is illustrated by data from a study
conducted at two units of an oil fired power plant (12). Measured results
which were compiled over the July-December operating period are given in
Table 6. Furnace oxygen concentrations were reduced in unit #3 in attempts
to achieve compliance with TPO standards. The attempts failed and the TPO
problem remained due to a flue gas sootiness resulting from poor combustion
conditions. Only after air-fuel mixing and combustion was upgraded to that
of unit #1 was TPO compliance achieved.
TABLE 6 COMBUSTION EFFICIENCY EFFECTS ON TPO
UNIT #1	UNIT #3
[H2S041 ppm
16 +
6
2.1 + 1.7
[o2] %
1.3 +
0.4 #
0.15
+ 0.09
COMBUSTION QUALITY
GOOD
POOR
[C]* mg/m3
3.6 +
17
45 +
18
ESP, %
77 +
13
65 +
15
**
0 -
15


TPO, %
30 -
50
^Higher furnace oxygen needed to control carbon build up in ESP
*
At ESP inlet
Estimated range (Unit 3 cited for TPO violation)
6. Chemical Additives
As seen in Table 2, Dolomite (a calcium magnesium carbonate), MgO, and
MMT have been used to reduce acid corrosion and neutralize the acid gas
fraction of flue gases from oil fired power plants. Liqui-Mag, a commercial
additive containing MgO, has been used by direct addition to fuel oil at many
power plants. Coaltrol M, a commercially available pulverized MgO additive
marketed by the Basic Chemical Division of Combustion Engineering has been
added in the economizer outlet area to scavenge t^SO^ by adsorption and/or
9-8

-------
reaction with subsequent removal of the dry material in the ESP.
Calculations (12) indicated that the use of an additional 200 lbs/hr of
Coaltrol at a typical oil fired power plant would reduce	by 26%.
However, decreasing furnace l^] from 0.3 to 0.2% would reduce [t^SO^] by 33%
at a lower cost.
For units firing high sulfur (2.5 wt %) and high vanadium (400 wt ppm)
fuel oil the following set of recommendations have been given from the
experience gained from an extensive study (12).
1.	Maintain furnace 02 no higher than 0.4 to 0.5%, preferably lower.
2.	Add Liqui-Mag or other finely dispersed sub-micron MgO additive to
the oil (front end addition) at 1.6 lbs Mg/lb V.
3.	Provide efficient fuel and combustion^air mixing (flue gas carbon at the
ESP inlet should be less than 40 mg/m ).
4.	ESP efficiency should be 75% or higher.
5.	Adjust fly ash recirculation injection location for minimum flue gas
carbon.
COAL FIRED POWER PLANTS
Coal fired power plants have historically used ESPs for plume opacity
control. ESPs normally achieved high removal efficiencies with relatively
low power consumption. With the imposition of S02 emission standards in the
early 1970's the shift to low (less than 1% S) sulfur coals began. ESP
performance decreased and plume opacity increased. Coal fired power plants
equipped with ESPs had a plume opacity problem to solve.
To burn the less costly high sulfur coals while meeting the stringent
SO emission standards required the use of flue gas desulfurization (FGD)
units. Some of these units were found to emit plumes with high opacity
levels. Nader and Conner (14), for example, reported plume opacities at FGD
equipped plants as high as 70-90%. Some FGD equipped coal fired plants
meeting NSPS particulate standards of 0.03 lb/MM Btu can still not meet plume
opacity standards.
Since the late 1970's baghouses have been applied for the collection of
particulate matter at coal fired power plants fueled by low-sulfur coal. No
plume opacity problems have been experienced to date. As applications of
baghouses to power plants firing high sulfur coal develop, opacity will
become a problem. The bag filter cake coupled with flue gas conditioning may
be a combination which will allow compliance with emission standards while
firing coal with a moderate sulfur content.
9-9

-------
Opacity Control With ESP Units
Properly operated and maintained ESP units produce plume opacity
problems when fly ash bulk particle resistivity is above or below the
10 -10 ohm-cm resistivity range. The natural resistivity of the par^iclj^
layer deposited on the ESP plates can be adjusted up or down to the 10 -10
ohm-cm range by (1) temperature change, (2) flue gas moisture change, (3)
raising or lowering SO^ concentration in flue gas, (4) selecting or spiking
coal with higher or lower sodium content. Ammonia has been used as a
conditioning agent but does not change resistivity. Ammonia brings about an
increase in space charge within the ESP which can raise collection efficiency
by increasing the electric field strength in the precipitation zone. Of
these alternatives SO^ has been used most successfully.
In one case a midwestern utility plant firing 3.5% sulfur coal to two
250-MW boilers was generating a detached "blue plume" of sulfuric acid
aerosol at a 40-60% TPO level (18). The problem was low resistivity due to
SO^ levels of 30 ppmV entering the ESP. The opacity was reduced to the 15%
level by spraying 224 lb/hr/unit of a Nalco chemical company liquid
containing soluble sodium upstream from the ESP which chemically reducedQthe
SO^ concentration and raised the fly ash resistivity to approximately 10
ohm-cm.
AEP's ESP equipped Tanners Creek plant firing 3.8-4.2% S coal
experienced a dense bluish-white plume with opacity exceeding the 20%
standard (19). MMT was sprayed into the primary combustion zone at 20 gm/ton
coal and did not reduce SO^ enough to effect the TPO level of 30-40%.
Injection of ammonia at 50 ppm just upstream from the ESP in conjunction with
the MMT injection reduced TPO to a consistent 10-18%. It's hypothesized that
the ESP efficiency increased in the sub micron size range resulting from the
ammonia conditioning. In addition the MMT vaporized to molecular form and
provided manganese molecules to form enough [MSO^] which was removed by the
ESP to bring about the TPO drop. No testing with ammonia alone was reported.
To acquire the combined benefits of ammonia and sulfur trioxide
conditioning Tampa Electric Company has injected ammonium-sulfate solution
upstream of the ESP. Opacity was reduced for two high sulfur coal fired
units at the Big Bend Station from 30% to 10% (20). Gannon Station Unit 6
(300 MWe) firing 1.3% sulfur 9% ash coal achieved ESP efficiencies of 99.9%
with ammonium-sulfate conditioning (injection both before and after the air
preheater). Prior to conditioning the high resistively of the low-sulfur-
coal ash had caused high TPO due to high spark rates which burned charging
wires and shorted out several ESP sections.
Opacity was reduced from 30% to 8% by sodium sulfate conditioning at
Gulf Power Company's Lansing Smith Station (21). Ash was collected with a
hot-side precipitator. ESPs are applied in this mode when used with boilers
firing low sulfur coal. One gallon of sodium sulfate was added per ton of
coal as a powder before the coal was pulverized.
9-10

-------
In a study of plants burning high sulfur coal it was found that the
number of ESP fields operating did not influence appreciably the sulfur
composition of the ESP outlet flue gas (22). This was in contrast to
concentrations of trace metals that decreased as more fields were energized.
These results support the hypothesis that the submicron sulfate aerosols and
thus the opacity are not effectively removed by ESP.
Opacity Control with FGD Units
The opacity problems encountered with high sulfur coal fired boiler
emissions treated with combined ESP and FGD units are mainly due to sulfate
emissions not fly ash. Concentrations of sulfate in combustion gases at
scrubber outlets have been reported to exceed scrubber inlet sulfate values
(23). It was found that on the average a wet limestone scrubber with
venturi-absorber will remove [H-SO,] and [MSO^] at the 29% efficiency level
while scrubbing SO^ at the 78% level.
This desparity in efficiency levels is explained by the fact that S0_
absorption is effective with the typical low energy (pressure drop) scrubber
operation whereas highly efficient removal of the submicron sulfate aerosol
requires high energy (pressure drop) scrubber operation. After passing thru
the scrubber the submicron sulfate aerosol is essentially unabated in its
pass thru the mist eliminator and forms an aerosol plume with its
characteristic bluish white hue. The submicron size range of the aerosol is
such that plume opacity may be well above the 20% opacity standard even when
mass emissions are at or below the 0.03 lb/MM Btu standard or lower.
A high sulfur (4% S) coal fired plant with two 240 MW units followed
by an ESP and a FGD unit of the limestone spray tower type typically had
residual plume opacity levels in the 30-50% range (24). Pilot (1MW) and full
scale studies were conducted in order to make recommendations for solving the
opacity problem among others. It was found that a venturi scrubber operating
at a pressure drop in excess of 30" W.C. would be required to reduce the
opacity to the required 20% level. The energy consumption to operate the
venturi unit would be enormous.
A better solution was found. The basic idea for solving the opacity
problem was to increase the size of the submicron sized sulfuric acid aerosol
upstream of the ESP and FGD units. A quencher unit was used to provide the
temperature and humidity conditions to promote growth in aerosol size. A FGD
spray tower was operated immediately downstream of the quencher unit,
Impactor measurements by Radian Corporation with low fly ash loadings
indicated tremendous mass growth in suspended aerosol in the 0.2 to 0.7 pm
range as the acid mist and fly ash passed thru the scrubber. In other words,
a negative removal efficiency for submicron fly ash and sulfate particles.
For high fly ash loadings the removal efficiency was generally positive for
size ranges of 1/2 micrometer and larger. It was evident that no useful
purpose was served by removing the fly ash prior to the acid aerosol. It was
also evident that to control opacity with reasonable scrubber energy
consumption that the submicron sulfate and fly ash particles must be
collected with a unit downstream of the scrubber. A wet tubular precipitator
9-11

-------
(first used commercially some 75 years ago to control sulfuric acid mist from
a sulfuric acid plant) was the obvious selection. It has the capability
while operating at low power consumption levels to remove submicron aerosols
at high efficiency. Its performance is virtually independent of coal sulfur
content and collected fly ash resistivity. The mist eliminator was
eliminated. By operating the WESP in the 50-60 KV range with specific
collection areas (SCAs) of 20-30 ft /1000 acfm, plume opacity levels were
found to be in compliance. The quencher pressure drop and the SCA of the
WESP may be adjusted to meet the 5% plume opacity (no visible plume) that
some state air quality agenices now require.
SUMMARY
Residual plume opacity results from sulfate and non-sulfate aerosols in
the 0.4-0.7/im size range. To reduce non-sulfate plume constituents that
cause opacity problems three means exist (1) maximize combustion efficiency
by use of additives and excess air adjustment, (2) increase energy input to
ESP and/or FGD units to reach peak efficiency, (3) replace ESP and/or wet FGD
with baghouse units.
For sulfate plume opacity control, (1) reduce formation of sulfate, (2)
increase sulfate aerosol size within the size distribution to enhance
collection efficiency, (3) apply WESP. To reduce plume opacity caused by
sulfate formation, use lower sulfur fuel containing less catalytic
elements such as vanadium. Flue gas additives reduce sulfate aerosol
formation in the plume by promoting the conversion of SO- to metal sulfates
which are efficiently removed by collection equipment whereas the acid
aerosols are not. Without additives SO- contributes to plume opacity by
becoming acid sulfate in the plume of ESP or FGD controlled plants.
The plume opacity caused by sulfate aerosol is best controlled by
hydration of the submicron sulfate aerosol in a quencher to increase aerosol
size. Following the quencher, use a WESP to remove the aerosol from the flue
gas prior to reheat and stack discharge.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not necessarily
reflect the views of the Agency and no official endorsement should be
inferred.
9-12

-------
REFERENCES
1.	Director, Division of Stationary Source Enforcement Attachment I In:
Compliance Analysis of Operating Class A SIP and NSPS Power Plants.
Environmental Protection Agency, Washington D.C., August, 1983.
2.	Bennett, K. M. Guidance on policy for enforcement of VE violations
against sources which are meeting an applicable mass emission
standard. U. S. Environmental Protection Agency, Washington D. C., May,
1982.
3.	Edison Electric Institute, Location of FGD Systems in North America,
Washington D. C., September, 1983.
4.	Personal Communication with Certified Opacity Observers & Lidar
Operators, 1984.
5.	McDonald, J. R., and Dean A. H. Electrostatic Precipitor Manual, Noyes
Data Corporation, Park Ridge, N. J., 1982 pp.	426-433,.
6.	Carr, R. C. and Smith, W. B. Fabric Filter Technology for Utility Coal
Fired Power Plants, Journal of the Air Pollution Control Association 34:
178-185, 1984.
7.	Federal Register, 40 CFR 60.11(e), May, 1977.
8.	Director, Division of Stationary Source Enforcement, Enforcement against
opacity violations from power plants, U. S. Environmental Protection
Agency, Washington D. C., June, 1981.
9.	Lappan, J., Factors affecting plume opacity from power plants and
associated enforcement problems. Environmental Protection Agency,
Washington D. C., 1981.
10.	Hughes, J. M., Aerosol Signatures For Receptor Modeling. DOE/PC/61254,
U. S. Department of Energy, Pittsburgh, Pennsylvania, 1984, 40 pp.
11.	Eliot, R. C. (ed.), Boiler fuel additives for pollution reduction and
energy saving. Noyes Data Corp., Park Ridge, N.J., 1978.
12.	Dietz, R.N. and Wieser, R.F., Sulfate Formation in Oil-Fired Power Plant
Plumes. EPRI EA-3231, Vol. 1, Palo Alto, California, 127 pp. 1983.
13.	Conner, W.D,, "A Comparison Between In-Stack and Plume Opacity
Measurements at Oil-Fired Power Plants." Proceedings of Fourth Annual
Conference on Energy and the Environment., Cinncinnati, Ohio, October
4-7, 1976.
14.	Nader, J.S. and Conner, W.D., "Impact of Sulfuric Acid Emissions on Plume
Opacity", In Workshop Proceedings on Primary Sulfate Emissions from
Combustion Sources, EPA-600/9-78-0206, August 1978, pp. 121-135.
9-13

-------
15.	E.P. Parry, A Study of Stack Effluents at an Oil Fired Utility Station,
Thousand Oaks, California: North American Rockwell, 1972.
16.	R.B. Engdahl, E.J. Schulz, and J.A. Gieseke, Summary Report on
Characteristics of Stack Effluents from the Northport Plant, Columbus,
Ohio: Batelle Memorial Institute, 1969.
17.	K.T. Knapp, W.D. Conner, and R.L. Bennett, "Physical Characterization
of Particulate Emissions from Oil-Fired Power Plants", Proceedings of
Fouth Annual Conference on Energy and the Environment. Cincinnati, Ohio,
October 4-7, 1976.
18.	Albanese, V., "Additive aids precipitator performance when fly ash
resistivity is too low", Electric Light and Power, 1982.
19.	Gray, M., The Effects of a Volatile Fireside Manganese Additive on
Coal-Fired Utility Boiler Operation." In: Workshop Proceedings:
Applications of Fireside Additives to Utility Boilers, EPRI WS-80-127,
Palo Alto, California, 1981, pp. 3-1, 3-13.
20.	Hudson, J.L., Flue Gas Conditioning Experience at Tampa Electric
Company." In: Workshop Proceedings: Applications of Fireside Additives
to Utility Boilers, EPRI WS-80-127, Palo Alto, California, 1981, pp
316, 3 28.
21.	Altman, R.F., "Fly-Ash Conditioning", In: Workshop Proceedings:
Applications of Fireside Additives to Utility Boilers, EPRI WS-80-127,
Palo Alto, California, 1981 pp 3-29, 3-40.
22.	Bennett, R.L. and Knapp, K.T., "Sulfur and Trace Metal Particulate
Emissions from Combustion Sources", In: Workshop Proceedins on Primary
Sulfate Emissions from Combustion Sources, Vol. II, EPA PB-287 437,
Research Triangle Park, N.C., 1978, pp 165-183.
23.	Homolya, J.B. and Cheney, J.L., "A Study of Primary Sulfate Emissions
from a Coal-Fired Boiler with FGD", JAPCA 29:1000-1004, 1979.
24.	Laslo, D., Bakke, E., and Chisholm, E. "Limestone/Adipic Acid FGD and
Stack Opacity Reduction Pilot Plant Tests at Big Rivers Electric
Corporation",In: Proceedings 77th Annual Meeting of the Air Pollution
Control Association, San Francisco, California, 84-97.4, 1984.
9-14

-------
PULSE ENERGIZATION SYSTEM OF ELECTROSTATIC PRECIPITATOR
FOR RETROFITTING APPLICATION
Senichi Masuda and Shunsuke Hosokawa
Department of Electrical Engineering,
University of Tokyo
7-3-1, Hongo, Bunkyo-ku, Tokyo, Japan 113
ABSTRACT
A cassette type pulser module has been developed for electrostatic
precipitators for upgrading the performance of existing plants. The module
is to be inserted between the existing dc power supply and corona electrodes.
No coupling means like a pulse transformer or a coupling condenser are used.
A saw-teeth voltage appears on the corona electrodes at a desired frequency,
which includes at its leading edge a transient LC oscillation with its very
sharp first peak and its amplitude decaying rapidly. The first peak behaves
as a submicrosecond pulse and produces very active negative streamers
uniformly distributed along the entire length of existing corona wires of the
conventional construction. The average of the saw-teeth voltage produces a
dc collection field between the corona and the collecting elctrodes. This
direct-coupled pulse energization" indicated in our laboratory test exactly
the same collection performance as the hitherto most effective submicrosecond
pulse energization using a coupling capacitor and a dc bias voltage. A
specific feature of the present pulse energization system is its simplicity
in construction and low initial and operating costs.
INTRODUCTION
The authors investigated the use of travelling wave coronas for pulse
energization (1). They recognized that this system could produce an ideal
uniformity in current distribution to produce a very high collection
performance. The pulse generator with a phasecontrolled rotating spark gap
and a ceramic capacitor bank proved to be quite cost-effective and reliable
(2). Moreover, this travelling wave scheme allows the use of pulse-
peaking (1) to regenerate the corona-eroded pulse peak voltage, enabling the
exploitation of the pulse energy to the full. The only problem which
remained was the necessity of a dc-biased long corona transmission line of
either a twin-electrode (wire-to-duct) or a tri-electrode (wire-third
electrode-duct) type which must be replaced to the existing corona
10-1

-------
electrodes.
The present pulse generator produces a high-frequency LC oscillation
when it is connected to the corona electrode of the conventional construction
as a lumped constant load. It produces a very active plasma at its first
peak over the entire length of the corona electrode. The saw-teeth like
voltage appears between the corona and collecting electrode at a switching
frequency, and a very sharp peak of the LC oscillation at its leading edge
to act as the submicrosecond pulse. The average of the saw-teeth voltage
provides a dc bias voltage. A very uniform current distribution could be
produced with a concurrent improvement of collection performance just as good
as that obtained by the travelling wave coronas.
The present direct-coupled pulse energization enables the use of the
existing dc power supply and corona electrodes, so that it can be applied in
the retrofitting applications for upgrading the existing precipitators.
Pilot plant test of the present pulser module will be carried out soon. In
this paper are reported the results of our laboratory tests made for
developing this pulser system.
CONCEPT OF PULSER MODULE
Fig. 1 shows the concept of the present cassette type pulser module. It
consists of a tank condenser CQ to be charged by the existing dc power
supply, voltage-doubling LC oscillation circuit with a thyrister switch Th>
and inductance L, a pulse forming capacitor C , and a rotating spark gap for
producing a sharp rising narrow pulse. The moSule is inserted between the dc
power supply and corona electrode of existing plants. C is charged up to
Vq = 2Vq - V2 from C with the aid of a switching LC oscillation in the
voltage-doubling circuit, where represents the voltage of CQ and the
remaining voltage of C just before Th is switched. Then, after Th recovered
its interruption abi?ity, the rotary spark gap fires to connect C to the
corona electrode. The pulse repetition frequency can be changed in the range
of 10 - 200 Hz by the controller which controls the servo motor and the gate
circuit of T, .
h
EXPERIMENTAL APPARATUS
TEST PULSER
The test pulser used throughout the present investigation is shown in
Fig. 2 where the voltage-doubling circuit is removed. L in Fig. 2
represents the overall circuit inductance of the connecting wires and spark
channel of the gap. The pulse rise time, T , is altered by inserting a coil,
L., in series to the gap. The phase of rotation of the spark gap is adjusted
so as to close the gap in the blocking half-cycle of the diode, D. Charging
of C is made by an ac voltage, which makes the charging efficiency very
high.'5 A pulse repetition frequency, f , is 50 Hz throughout the present
investigations where a twin-contact roter^is used in the spark gap at a speed
of 1500 rpm.
10-2

-------
coronas, and the current density becomes very large on the opposit of these
corona spots (Fig. 5(c) and Fig. 6(c)). Looking at the current distribution
in Figs. 6(a) and (b) the direct-coupled pulse energization is anticipated to
produce the same level of effect of back corona correction and performance
enhancement as the condenser-coupled travelling pulse energization.
PULSE WAVE FORM AND EQUIVALENT CIRCUIT
The appearing voltage on the corona electrode is of a saw-teeth form as
shown in Fig. 7(a). At its leading edge, it has a transient oscillation as
shown in Figs. 7(b) and (c). When the spark gap fires, the LC oscillation
starts to occur, and the voltage takes its first maximum, V^, at t = T . The
oscillation decays quickly and stops at t = T^ when the spark channel is
opened to disconnect C from the corona electrode. Approximate
characteristics of the oscSllation can be considered in consideration of an
equivalent circuit as shown in Fig. 8(a). The oscillation phase lasts for
some tens to hundreds microseconds, whereas the main negative streamers are
produced only in its first half cycle (4). The oscillation decays quickly by
the losses in the circuit, spark channel, and ion oscillation in corona
plasmas. Then, we get the following expression for the damped oscillation:
V (t) - VR + k (V0 - VR)	(1 -	Ut)
p EP	(1)
 < t < Td
where
k : loss factor, 0.8 < k < 1
V^ : residual voltage ~
ol = r/2L,	r: equivalent overall loss
co - 1 //LC, C = Cp CEp / (C + CEp)
and the equivalent pulse rise time, T , is given by
~ T / 2 = tt / to = rr/LC	(2)
In the oscillation period, the ion migration towards collecting electrodes is
negligible and Rgp is included in the loss factor of K, which represents the
streamer formation loss.
Finally, the spark channel turns off at t = T^ as shown in Fig.	7(b),
and the decay phase of the saw-teeth starts. It consists of a first	rapid
decay (decay phase I) and the second slower one (decay phase II).	The
initial voltage of the decay phase I, V^, will be :
VD " VR + k C-f-cZ 
-------
Phase 1 : VEP - Vd e 1	)
TI =	^EP
Viiase II : VEp . V	(J)
TII = (rep>iicep
Fig. 9 shows the ionic current wave form measured on the collecting
electrode with one of the ionic current probes (Fig. 3). It can be seen that
the ionic convection current lasts as long as several milliseconds. This
time corresponds to the duration time of the decay phase I (Fig. 4(a)). A
substantial fraction of energy stored in the inter-electrode capacity is
consumed by the ion convection current flowing in the decay phase I with a
concurrent sharp decay of V^p. In the decay phase II, V drops much more
slowly because the pulse-induced plasma has been depleted and only leakage
current and, in some cases, dc coronas are existing.
FACTORS AFFECTING VOLTAGE WAVE FORM
It is observed that among the factors affecting the overall voltage wave
form in the present pulse energization the following three are the most
dominant : the magnitude of the pulse peak voltage, V^; pulse rise time, T ;
and occurence of back corona.	r
Pulse Peak Voltage, V^:
Figs. 10(a) and (b) indicate the effect of V^j measured under a back
corona free condition with the test electrode as shown in Fig. 3. A higher
V^ produces a sharp drop of the voltage in the phase I to the dc residual
voltage which remains almost constant in phase II and is substantially lower
than that obtained at a lower magnitude of Vw (Fig. 10(b)). Whereas the use
of lower V^ produces a much less voltage drop in the phase I and a long
lasting gradual drop in the phase II (Fig. 10(a)).
This difference may be explained by the difference in the carrier
production in the pulse-induced corona plasmas. The lower peak voltage will
produce less carriers available for the discharge of the inter-electrode
capacity, resulting in a smaller voltage drop in the phase I, which will,
however, produce the dc corona current, small in magnitude, and a
concurrently slower, but long lasting voltage drop in the phase II.
Contrarily, the higher peak voltage, V^, will produce enough carriers to
discharge the inter-electrode capacity down to a level below the corona
starting threshold in the phase I.
Pulse Rise Time, T :
' r
Figs. 10(b) and (c) show the effect of T , which is identical to that of
V^j. The use of a shorter T produces the same effect as that of a higher
Vj^. The difference may again be understood when the difference in carrier
production in the corona plasma is considered. A shorter rise time produces
10-4

-------
a much more uniform and intense coronas with a greatly enhanced carrier
production. But it also results in a decreased level of the average dc
field.
Back Corona
Fig. 11 indicates the difference between the pulse waves with and
without back corona. The wave form without back corona (Fig. 11(a)) is
essentially the same as those (Fig. 10) observed with clean electrode system
f Fig. 3. But it indicates a drastic change when back corona takes place
(Fig. 11(b)). The -voltage continues to drop during an entire period of a
Pulse cycle, and the difference between the phase I produced by the
convection of pulse induced ions and phase II after depletion of these
initial ions disappears. Positive ions by back corona contribute to the
continuous discharge of the inter-electrode voltage in this particular case
f extremely high dust resistivity. Concurrently, the final voltage, V^, and
the average level of the dc field take much lower values to produce a
Performance degradation, aside from charge degradation of dust.
design parameters
As already indicated in the previous section, the production of ions by
a pulse voltage is enhanced by the use of a larger and a smaller T^. Vw
can be changed by C and V of the pulse forming capacitor. Assuming K = 1
and r = 0 in Eq. (1)? we get
'm ' vo - v* ' vo + - VV
2 C-'r/^'Pp)
(6)
1 + (vcep>
fig. 12 indicates the calculated value of \,/Vq as a function of C /Cp for
various values of	assumed. It can be seen that Cp/Cgp has to be
increased to increase the magnitude of Vw, whereas its increasing rate
becomes diminished beyond C /CEp = 1. Hence, in consideration of the cost of
the pulse forming condenser, the value of C should be restricted in the
range CEp < C < 1.5 CEp.	p
Fig. 13 shows the effect of VQ on the distribution of ionic current
density measured with the center probes (Fig. 3). In this particular case of
C = 2.99 nF and T = 200 ns, |V_]> 80 kV is needed to get an uniform current
distribution. Thus it is imperative in the present pulse energization system
to use a sufficiently high Vq to obtain a satisfactory uniformity in current
distribution.
Fig, 14 shows the effect of T on the ionic current density, again,
measured with the center probes (fig. 3). A smaller T gives a better
uniformity in the current distribution with its magnitude also increased. T^
is reduced by lowering L and C in Eq. (2). However, it is difficult to
reduce the overall circuit inductance, L (spark channel, feeding wires,
etc*)> below a certain level. Hence, the magnitude of CEp, i.e. the number
f parallel ducts, to be energized with a single pulse power supply should be
limited so as to keep T at a desired level.
10-5

-------
CORONA MEASUREMENTS
Various kinds of corona measurements are made using a wire-to-duct
system of full scale dimensions as shown in Fig. 3. The corona electrode is
of a conventional construction consisting 25 square wires, each having a 5 mm
x 5 mm cross section. The pulser output is directly connected to a bottom
corner of the corona electrode, and both ionic current density and its wave
form are measured using 39 probes attached on the collecting electrode. Each
probe is on the opposite of a corona wire, and has a 10 cm x 20 cm area. The
capacitance between the corona electrode and two plates is measured with a
LCR meter (K0KUY0 ELECTRIC CO., LTD. Type KC-534), and its value is Cpp =
MEASUREMENT OF PERFORMANCE ENHANCEMENT
Fig. 4 illustrates a laboratory test precipitator in a race-track gas
circuit used for measuring the performance enhancement produced by the
present pulse energization. For the purpose of comparison tests are also
made with a conventional dc-energization and a condenser-coupled
submicrosecond pulse energization applied on top of a dc-bias voltage. Two
fl^ijish sampl^ are used and their resistivity altered in the range of 5 x
10 - 1 x 10 ohm-cm by changing gas temperature.
Three collection fields are energized with a common power supply. Each
field consists of 2 parallel ducts, each being spaced at 0.3 m, 0.76 m high,
and Q>2^ m long in the gas flow direction. The total duct cross-section is
0.46 m ^ an^ the total collection area of a single field is 1.52 m with SCA
= 4.03 m /(m /s) at 0.82 m/s standard gas velocity. Three corona wires with
3 mm in diameter are installed in each duct with a 80 mm wire-to-wire
spacing. The carrier gas is air, and its temperature can be raised by an
electric heater up to 200 C. The inlet dust mass loading, W., is kept
constant by using a constant dust feeder, monitored by a dustmeter (Konitest)
based on tribo-electrification. The dust mass-loading is measured at the
outlet, Wq, by a sampling filter.
Back corona sevirity, expressed in terms of a ratio of the positive
ionic current density to the negative one, I /I_f is measured by the bi-polar
current probe (3).
EXPERIMENTAL RESULTS
IONIC CURRENT DENSITY DISTRIBUTION
Fig. 5 shows the light activity of coronas in the full scale duct. Fig.
5(a) is for the present direct-coupled pulse energization. For the purpose
of comparison, the cases of the condenser-coupled travelling pulse
energization and the conventional dc energization are shown in Figs. 5(b) and
(c), respectively. In the pulse energization of both types ((a) and (b)) the
uniformity in the light activity is satisfactory, resulting an uniform ionic
current density distribution on the collecting electrode as shown in Figs.
6(a) and (b). On the other hand, the dc energization produces spotty
10-6

-------
Under these constrictions the average current density on the collecting
electrode, a parameter determining the avoidance of back corona, can be
altered only by changing the pulse frequency, f .
TESTS OF COLLECTION PERFORMANCE
The collection performance by the present pulse energization is tested
in the laboratory precipitator (Fig. 4) at T = 800 ns using a high
resitivity fly-ash sample, and compared with the test results obtained with a
conventional dc-energization and a condenser-coupled submicrosecond pulse-
energization applied on top of the dc bias voltage. The test conditions are
indicated in Table 1. The collection performance, expressed by both the
mass-based collection efficiency, 1, and the Deutsch migration velocity, w,
are shown in Table 2. The back corona severity, I /I_, measured with the bi-
polar current probe is also indicated.
As can be seen from I /I_, the dc-energization causes back corona at all
of the three resistivity^levels, and the back corona severity in terms of
I+/I_ becomes larger, and the collection performance lower with the increase
in dust resistivity. The collection performance in the dc energization is
the worst among all of the three energization modes. On the other hand, a
distinct improvement is produced by both modes of the pulse energization.
However, practically no difference exists between the two pulse-energization
modes in the dust resistivity range tested. The performance enhancement
factor defined in terms,Deutsch migration velocity becomes H = 1.45, 1.35,
and 1.1 at Pd = 5 x 10 , 2 x 10 , and 1 x 10 ohm-cm, respectively. The
performance enhancement is pronounced in the first two cases where back
corona can be corrected by pulse energization. In contrast back corona can
nt be corrected by any mode of the two pulse energization in the case when
^d = 1 x 10 ohm-cm, and the performance enhancement is only H = 1.1. It
should be added that the use of a precharger (Boxer-Charger) in combination
with the pulsed collection fi;lds produces a substantial performance
improvement even at Pd = 1 x 10 ohm-cm (5).
Once back corona occurs, V, drops drastically in the present direj|
coupled pulse-energization as shown in Table 1 (from -30 kV at Pd = 5 x 10
ohm-cm to -9 kV at Pd = 1 x 10 ). The back corona extinguishing voltage,
ve. is found to be so low when Pd = 1 x 10 ohm-cm that back corona can be
hardly interrupted and even tends to propagate in the lateral direction to
cover the wide area of the collection electrode. Hence, a careful control of
corona current is required at this extremely high resistivity level so as not
to initiate back corona.
CONCLUSION
d). The pulser module comprizing a voltage-doubling circuit, pulse forming
condenser, and a rotating spark gap can be inserted between the existing dc
high voltage power supply and the corona electrodes of the conventional
construction.
(2). a very uniform distribution of ionic current density is produced by
10-7

-------
the present direct-coupled pulse energization, as good as that obtained by
the condenser-coupled travelling pulse energization.
(3).	A transient LC oscillation occurs when the spark gap fires, and its
sharp first peak acts as a fast rising narrow pulse voltage, producing
streamer coronas at its first rising phase. This LC oscillation decays
rapidly to disappear when the gap is off. At this instance there remains a
sufficiently high voltage between the corona and the collecting electrodes,
to separate negative ions form the corona plasmas and drive them toward the
collecting electrodes. The ion convection current lasts several
milliseconds, with a concurrent rapid decay in the voltage. This fast decay
of the voltage is succeeded by a much slower decay due to leakage and dc
corona current. As a result, a saw-teeth like voltage wave form is obtained.
(4)	The decay pattern is affected by Vq and T , through the difference in
carrier production, and also by the occurrence of back corona which produces
positive ionic current.
(5).	The optimum value of C lies in the range C^p < C <1.5 Cgp. As the
uniformity in corona curren? distribution is secured Bnly by keeping a
smaller T , the overall inductance of the pulse and feeder circuits must be
kept as low as possible. In addition, the number of parallel ducts to be
energized by a single pulser should be restricted. The level of average
current density should be controlled primarily by changing pulse frequency,
f .
P
(6).	The present direct-coupled pulse energization shows the same
collection performance for high resistivity fly-ash as the condenser-coupled
submicrosecond pulse energization applied on top of a dc bias voltage. The
performance enhancement factor in terms of Duetsch migration velocity is the
same in these two pulse-energization^ijiodes, which^re H = 1.45, 1-35. and 1.1
at dust resistivities of 5 x 10 , 2 x 10 , and 1 x 10 , ohm-cm
respectively. At an extremely high dust resistivity of Pd = 1 x 10 ohm-cm
back corona can not be corrected even with any of the two pulse energization
modes operated at f = 50 Hz, with a concurrently decreased performance
enhancement of H = 1.?.
In conclusion, the pulser module and the present direct-coupled pulse
energization system are expected to provide useful and economical means of
back corona correction, in particular for retrofitting applications.
The work described in this paper was not funded
by the U.S. Environmental Protection Agency and
therefore the contents do not neccessarily reflect
the views of the Agency and no official endorsement
should be inferred.
10-8

-------
Existing DC
Power Supply
EP
Rotating Spark
Gap
Voltage-Doubling
Circuit
Pulse
Forming
Condenser,
Tank
Condenser,
o
Servo
Motor
Gate
Circuit
Frequency Controller
I	1
L		1
Fig. 1 Concept of Direct-Coupled Pulse Energization.
R
Rotating
Spark Gap
	O^O	55T
EP
"W
3
D : diode
Cq : pulse forming capacitor
Ls : overall circuit inductance of connecting wires
and spak channel
: inductance of inserted coil
Fig. 2 Pulse Power Supply
10-9

-------
Ionic Current Collection Corona
Probe	Electrode	Electrode
B
&
&
b
b
Camera
B

S-
Top
B
b
b
&
B
B
o-
Q
O
B
Bottom
s
e
b
b
Pulse
Input
B
O
Q-
B
Left
Middle
Right
duct spacing : 250 mm	; wire-to-wire spacing : 100 mm
corona electrode : 25 wires ; wire : 5 x 5 mm cross section
Fig. 3 Single Duct Test Electrode System
HEATER
BLOWER
COLLECTION FIELD
U/l-
DUST FEEDER
0.82 m/s
15m/s /
7.5 m/s
: inlet dust loading
WQ : outlet dust loading
I+/I- : back corona sevirity measured by bi-polar ion current probe
pj : dust resistivity
Fig. 4 Test Precipitator in Race-Track System
10-10

-------
(a)
(a)	Direct-Coupled Pulse Energization;
Tr = 200 ns, Th = 280 ns, VQ = -70 kV
(b)	Condenser-Coupled Travelling Pulse
Energization; = -25 kV, E^c =
-2.0 kV/cm, Vp = -50 kV
(c)	DC-Energization; Vj. = -50 kV, Ej =
-2.5 kV/cm
Fig. 5 Light Activity of Coronas
-10 KV
3.91 nP
200 ns
280 ns
(mA/rn }
0.10
0.05
Right
liddle
'Pulse Inlet
(a) Direct-Coupled Pulse Energization
(sharp rising submicrosecond pulse)
( mA/n2 )
0.15
0.10
-50 kV
380 ns
Middle
End Terminal
Right
-Pulse Inlet
Top
(b) Condenser-Coupled Travelling Pulse
Energization (L  200 m)
k vdc " kV
( mA/nr )
0.4
0.3
0.2
0.1
\ j
i /
Kz
0.08 mA/m
Right
Middle
Top
(c) DC-Energization
Fig. 6 Distribution of Ionic
Current Density on
Collecting Electrode
10-11

-------
lOKV/div
Sjna/div:
IQkV/div
5ns/ Transient LC Oscillation	CO Magnified Oscillating Part
Sharp Initial Peaks
Fig. 7 Voltage Wave Form, V p, Measured at the Inlet Terminal of Discharge
Electrode (see Fig. 3;

	55$	**	


L = L +L.

cp
s-i

(a) Oscillation Phase
CEP

(b) Decay Phase
Fig. 8 Equivallent Circuit
*EP

Irat/oiv
Wave Form of
Ionic Current
Measured at
Collection
Electrode
10-12

-------
10 K.y/4iv
b m/&tv
lOkVAUv
Fig. 10 Effects of V and T on Pulse Voltage Wave Form (clean electrodes;
see Fig. 3)	r
0-
1991
0-
= -60 kV V0 - -60 kV
= -30 kV Tr = 700 ns
= -30 kV Tu * 900 ns
(a) Without Back Corona
(Pd  5 x 10^ ohro-cm)
- -62 kV
-50 k\
700 ns
= 900 ns
-53 kV
(b) With Back Corona
(Pd - 1 x 1()13 ohm-cm)
Fig. 11
Effect of Back Corona
on Pulse Voltage Wave
Form
10-13

-------
{mA/m )
C = 2.99 nF
P
T  200 ns
r
T. = 280 ns
n
(mA/m )f
Bottom	Top
Fig. 13 Effects of Vq on Distribution
of Ionic Current Density
Measured with the Center
Probes (see Fig. 3)
0.2 -
o.i
C = 3.91 nF
P
VQ = -70 kV
200 ns
500 ns
1 2 ys
Top
Fig. 14
Effects of T on Distribution

of Ionic Current Density
Measured with the Center
Probes (see Fig. 3)
Table 1 Test Conditions
Dust Resistivity
DC-Energization
(a)
Condenser-Coupled
Submicrosecond
Pulse Energization
Direct-Coupled ^
Pulse Energization
5 x 10 11 ohm-cm
(T - 130 C)
vdc - "V
I - 0.2 mA/m2
Vb - -35 kV
Vp - -20 kV 
I - 0.06 mA/m
V0 - -60 kV
VM - -59 kV
Vr - -30 kV
l' - 0.09 mA/m2
12
2 x 10 ohm-cm
(T - 50 C)
VDC " "38 kV
X * 0.2 mA/m^
Vfc - -30 kV
V - -32 kV
I - 0.04 mA/m2
Vo - -56 kV
VM - -65 kV
VR - -20 kV
I - 0.09 mA/m?
13
1 x 10 ohm-cm
(T - 100 C)
VDC " "32 kV
I  0.2 mA/m^
Vb - -22 kV
V - -32 kV
I - 0.04 mA
V0 - -50 kV
Vm - -62 kV
VR - - 9 kV
I - 0.08 mA/nr
(a)	T - 40 na, T. - 180 na, f - 50 Hz
r	n	p
(b)	T - 800 ns, T. - 900 ns, f - 50 Hz
r	n	p
10-14

-------
Table 2 Back Corona Sevirity and Collection Performance
Dust Resistivity
DC-Energization
Condenser-Coupled
Submicrosecond
Pulse Energization
Direct-Coupled
Pulse Energization
Enhancement
Facter
Back Corona
Sevirity
(V1.)
Collection
Performance
Back Corona
Sevirity
a+/o
Collection
Performance
Back Corona
Sevirity

-------
Session 7: ESP: PERFORMANCE ENHANCEMENT II
B. G. McKinney, Chairman
Electric Power Research Institute
Chattanooga, TN

-------
PRACTICAL IMPLICATIONS OF PULSE ENERGIZATION OF
ELECTROSTATIC PRECIPITATORS
H. Milde, J. Ottesen, C. Salisbury
Ion Physics Company
A Division of High Voltage Engineering Corporation
Burlington, Massachusetts 01803
ABSTRACT
In a conventional electrostatic precipitator an applied dc voltage
produces negative ions, charges particles and provides the electrical field
needed to transport particles to collecting plates. In the case of high
resistivity dust or fly ash, high dc voltages at the collecting plate produce
positive ions in a back corona discharge. This positive current effectively
reduces the negative charge on particles which reduces collecting efficiency.
Pulse energization, which consists of superimposing narrow high voltage pulses
onto a base dc voltage offers a way to improve collection efficiencies. Indepen-
dent control of pulse amplitude and pulse frequency allows for separate control
of negative ion generation and charging on the one hand and particle transport
fields on the other. Techniques for applying pulses and the pragmatic signifi-
cance of pulse energization are discussed along with benefits such as increased
collection efficiency at low SCA and improved control. Operating results and
cost data from coal-fired boiler installations are presented.
11-1

-------
INTRODUCTION
While ESP's operate, in most cases, very efficiently and reliably, the
collection of dust with a specific resistivity p > 1011 ficm poses very severe
problems. In such cases the product of current density J and resistivity p
can reach values in excess of the electrical breakdown strength of the dust
layer. The resulting breakdown produces, among other constituents, positive
ions which will travel towards the negatively charged particulates and thus
partially offset the negative charge on the particles. This phenomenon,
called back corona, is especially prevalent in ESP's at coal burning power
plants using coal with a sulfur content below 1%, since low sulfur coal usually
produces a high resistivity dust layer.
When back corona conditions exist, it is necessary to lower the base
voltage until the condition is under control. This reduction in voltage
results in a nonuniform or spotty corona emission along the wire electrodes,
insufficient current, and control difficulties, which in turn lead to poor
collection efficiencies.
Pulse energization, however, is an attractive solution to precipitator
problems stemming from high-resistivity dust. The frequent short high voltage
pulses provide the high electric fields needed to produce corona current
while the lower base voltage provides the collecting field. The relationship
of pulse voltages to the base voltage is shown in Figure 1. Pulse energization
systems work in like fashion to correct precipitator problems arising from
warped collecting plates and dust encrusted wires; and can be easily retrofitted
onto existing precipitators with little or no modification.
BACKGROUND
The origins of pulse energization of ESPs reach back as far as 1931, when
R. Heinrich et al (1) applied for a U.S. patent in which the application of
sharp electrical impulses to a precipitator is described.
In 1952, H. J. White and H. J. Hall (2) performed the first full scale
field tests. During the tests high voltage pulses of approximately 100 ys
duration were applied to a wire-plate type precipitator. The pulses were
applied by the action of a rotary spark gap, a pulse transformer, and blocking
diode which prevented the current from flowing back to the pulse generator.
In 1967, J. E. Luthi (3,4) described a three electrode arrangement whereby
the charge carriers were produced by a pulse voltage between the third
electrode and the anode. A more sophisticaed three-electrode arrangement
has been discussed by Penny and Gelfand (5), whereby the potential on the third
11-2

-------
PULSE ENERGIZATION CONCEPT
Time
Corona Onset
Voltage
Voltage
Interpulse
Period
Base
Voltage
Pulse
Voltage
Figure 1. Voltage-time characteristics of pulse
energization system.
11-3

-------
electrode was reduced at the same time that a pulse voltage was applied to the
corona wire. This method requires, however, handling of rather large displace-
ment currents.
A unique pulse energization system nicknamed "Boxer Charger" was described
by Masuda et al (6,7). The system operates as a precharger to an ESP. In the
latest version the electrodes consist of double-helices energized by a pulse
propagated along the double-helix. A separate charging field is provided by
an alternating voltage applied to adjacent double-helical arrangements.
Full-scale commercial pulse energization systems for the standard wire-
plate arrangement have been developed (8,9,19) and are beginning to find
application.
Ion Physics has been developing pulse systems since the early 1970's and
has recently introduced a second generation model, the PES-2000. The new system
is a compact, energy efficient, highly reliable, and inexpensive unit designed
to produce very short, rapid risetime pulses.
A schematic diagram of the IPC pulse energization system is shown in
Figure 2. It consists of two separate arms, the conventional base supply with
additional protection devices on one side and the pulse generating equipment
on the other side. The pulse generator consists of a high voltage pulse capac-
itor, transfer switch and transformer as well as low voltage pulse circuitry.
The pulse is applied to the precipitator by initiating switch closure at the
time the pulse capacitor has been charged to its desired pulse voltage. The
IPC system is designed to produce bursts of heavily damped sinusoidal signals
with a period of approximately 1 microsecond and a repetition rate on the order
of 150 pulses per second. The pulse repetition rate is controlled by varying
the dead time before recharging.
TECHNICAL FEATURES
For the following discussion it was assumed that narrow pulses were super-
posed over a dc base voltage at a frequency of approximately 100 to 300 pulses
per second (pps), as indicated in Figure 1.
INDEPENDENT CONTROL OF VOLTAGE AND CURRENT
In standard precipitators, the applied dc field is responsible for ion
production, particle charging, and particle transport. Performing these three
functions in a fixed geometry can lead to severe inefficiencies, especially
under back corona conditions where the current density must be limited. Pulse
energization offers here a unique advantage by allowing independent control of
voltage and current. The base voltage is adjusted for minimum spark rate
(^ 1 spark/minute) while the corona current is controlled by the pulse
repetition frequency and pulse amplitude.
11-4

-------
CIRCUIT DIAGRAM
Pulse	Protection
Capacitor	Components
Low Voltage
Capacitor
W
Base T/R
Controls
Transformer
Transfer
ESP
Switch
Figure 2. Schematic diagram of IPC pulse
energization system.

-------
UNIFORM PRODUCTION OF CORONA CURRENT
Since the electric field is far in excess of the corona onset voltage
during the pulse period, a bright uniform corona is produced over the entire
length of the corona wires during pulse application. This is in stark contrast
to the spotty corona observed for low voltage dc corona.
Recent measurements on current distribution on the collector plates, under
highresistivity conditions, by D.E. Rugg et al (11) have verified our own
findings and revealed that for dc energization and an average current density
of 5 nA/cm2 there was no current on ninety percent of the collector plate area
and high current densities on a few cells.
Uniform production of corona by pulse energization allows each precipitator
section to contribute equally towards particle collection. Pulse energization
also overcomes minor localized wire contamination, since the electric field
during pulse application is high enough to produce corona even around a
contaminated wire.
INCREASED CHARGING AND COLLECTION FIELD BY THE USE OF CORONA WIRES OF
LARGER DIAMETER
Because very high electric fields are produced around corona wires during
pulsing, it is possible to utilize corona wires of larger diameter and still
obtain adequate corona current. This fact can be utilized to produce a higher
and more uniform electric field in the proximity of the anode (12), where it is
most beneficial for particle transport and consequently, collection efficiency.
A conventionally energized precipitator, in contrast, requires a strong field
concentration around the corona wire to provide adequate corona current without
sparking.
LOWER SPARKING RATES
When pulse energization is added to the base precipitator the base TR
voltage is adjusted so that sparking frequency is minimized. Pulse systems
work best at spark rates of around one per minute versus several tens of sparks
per minute with dc operation alone. These low spark rates lead to improved
precipitator performance and reduce wire erosion thereby increasing the life-
time of discharge electrodes.
REDUCED CHARGING TIME
Due to its high instantaneous charge density, pulse energization is capable
of charging particles in a much shorter time and consequently, in a shorter
travel time (13,14). This feature of pulse energization is especially
advantageous in precipitators suffering from reentrainment or precipitators
which have a short overall length in the direction of gas flow. Additionally,
it has been shown that pulse energization increases the effectiveness of
sections near the precipitator inlet thereby increasing the contributions of
these sections to overall collection efficiency.
11-6

-------
BENEFITS OF SHORT PULSES
Short pulses offer some inherent advantages, partly because of the time
dependent character of corona emission (15,16), strong space charge bunching,
reduced severity of back corona at high resistivities and increased breakdown
strength of gases for short pulse voltages.
TIME DEPENDENT CHARACTER OF CORONA EMISSION
Very early in our experimental effort we determined that emitted charge
Per pulse was only moderately determined by pulse width (17). This finding,
which was recently confirmed by Professor Masuda et al (18), provided the
basis for selecting pulses of very short duration.
Besides allowing the production of ample current, the time dependent
character of corona emission also provides a means of converting nearly all
the energy stored in the pulse capacitor into corona energy. On the other
hand, for longer pulses it is reported (10) that the voltage on the pre-
cipitator is returned to the base value by special circuitry which recovers
some of the stored energy in the precipitator and returns it to the pulse
capacitor.
SEVERITY OF BACK CORONA IS INCREASED WITH LONG PULSES FOR HIGH RESISTIVITY ASH
Extensive back corona studies by Professor Masuda et al (19) revealed
that for high resistivity ash the severity of back corona, i.e. the ratio of
the positive ion current to negative ion current (i+/i_), increases with in-
creasing pulse width for equal overall current. Masuda et al speculated that
this pulse width dependence relates to the ratio of the ion transit times
across the electrode gap. However, Masuda varied the pulse width only between
50 Usee and 1 msec and did not include data for a pulse width of 1 ysec.
NARROW PULSES MORE SUITABLE TO LARGE DIAMETER CORONA WIRES
In order to realize the benefit of a larger diameter corona wire It is
Necessary to apply very high voltages to the corona wires to overcome the less
avorable emission geometry. Since the electrical breakdown strength of gases
 strongly influenced by the pulse width, narrow pulses have a distinct
a vantage in providing ample corona current without excessive sparking.
OVERCOMES EFFECTS OF WIRE BUILDUP
Relatively high allowable pulse voltages attainable with short pulses
ncrease the effectiveness of overcoming wire buildup. High pulse voltages
Produce more corona current than lower voltage pulses with both clean and
rty discharge wires. During laboratory tests of pulsers on a small ESP with
rty "156" diameter wires the PES system raised the efficiency from 75.38%
thout pulsing to 98.86% with pulsing. No pulse operation with clean wires
Save 95.7% versus 98.9% with pulsing. With clean .25" wires and no pulsing
11-7

-------
the efficiency was 94.3% and 98.3% with pulsing. Dirty .25" wires caused the
efficiency to drop to 92.7% with conventional energization while pulsing
brought the efficiency back to 98.2% under identical dirty wire conditions.
Both sets of data strongly suggest that short pulses are very capable of
nullifying the negative effects of thick ash deposits on discharge electrodes.
SIMPLIFIED CONTROL CIRCUITRY
Very narrow pulses can be applied randomly at any point on the TR voltage
waveform without initiating sparking. Complicated control circuitry to syn-
chronize pulse application is therefore not needed. Initially the base TR
voltage is set to a minimum spark rate (typically one spark/minute) and the
pulse voltage and frequency are then set to optimize precipitator performance.
Changes in pulse voltage and frequency alter the TR voltage only slightly as
shown in Figure 3. As the voltage and frequency are increased with short
pulses the base TR voltage remains nearly constant thereby providing an
optimum collecting field with very high corona producing pulse voltages.
Since the application of pulse voltages has only a modest effect on the
operation of the base supply it is, in most cases, possible in a retrofit
application to retain the existing base TR controls.
PRACTICAL BENEFITS
IMPROVED COLLECTION EFFICIENCY
Collection efficiency of an electrostatic precipitator can be expressed
by the modified Deutsch equation
n = 1 - exp [-(wkA/V)m]	(1)
where n is the collection efficiency, w^ the modified migration velocity in
ft/min, A the collection area in ft2 V the gas flow in acfm, and m an exponent
dependent on inlet particle size distribution (typically between 0.5 and 1).
From numerous measurements it was found that the ratio (frequently referred
to as enhancement factor) of migration velocities for pulsed versus dc ener-
gization varied between 1.3 and 2 depending on resistivity of the collected fly
ash, as shown in Table 1.
TABLE 1
ENHANCEMENT AS A FUNCTION OF ASH RESISTIVITY
Resistivity - ohm cm
1011
1012
1013
Enhancement H
1.3
1.6
2
An increase in migration velocity is of great economic significance, since
according to Equation (1) an increase in migration velocity is equivalent to a
corresponding increase in collection area. Consequently, an increase of w^
11-8

-------
	 30 kV Pulse
o	o 40 kV Pulse
a	o 50 kV Pulse
50--
DC Operation
40
ri
>
it
30- -
UJ
CD
<
O
>
20--
UJ
cn
<
QQ
200
150
100
50
PULSE REPETITION RATE [pps]
Figure 3. Base TR voltage as a function of pulse repetition
frequency.
11-9

-------
by a factor of two due to pulse energization has the same performance impact
as increasing a dc energized precipitator's size by a factor of two. An
enhancement of 1.5 typically implies a reduction in output loading by more
than a factor of two as can be seen from Figure 4. For the conditions tested
up to date a 4.5 fold reduction of outlet loading has been our best perform-
ance on a full scale installation, since the efficiency of the unpulsed pre-
cipitator was always below 99%. If outlet reduction in excess of 4.5 can be
achieved as Figure 4 would indicate for high unpulsed efficiencies is still
an open question.
FUEL SWITCHING AND FUEL FLEXIBILITY
Since pulse energization has the effect of increasing precipitator size
for high resistivity dusts a relatively small older precipitator can be
upgraded by addition of PES2000 to collect fly ash from a lower sulfur coal
than was originally burned while still allowing the boiler to meet emission
standards. Furthermore, pulsing normalizes fly ash resistivity differences
and treats dust from all fuels at near optimum precipitation conditions.
Boiler operators are thus able to avoid costly fuel management procedures or
switch to a low sulfur coal while maintaining high collection efficiencies
without resorting to chemical conditioners or additional collecting area.
Pulse energization also permits the use of higher ash content coal while
maintaining emission standards through enhanced collection efficiencies.
Such a step could potentially provide a substantial savings to the utility
since higher ash content coal is generally less expensive. For example, if
pulse energization would allow a cost reduction of one dollar per ton of coal,
a payback period of less than one year could be achieved for the pulser.
EASY TO RETROFIT
The IPC PES-2000 is a compact modular roof mounted system compatible with
V J?ndard	Installation can be accomplished with a minimum
of boiler outage, usually a day or less. All pulse energization system
components are external to the base precipitator hence no internal modifications
are required during installation. Controls for the PES-2000 are independent
allowing the base TR sets to operate when the pulse system is turned off.
Alternatively independent pulser units can be turned off while the remainder
are left in an operating mode. Pulser modules are designed to pulse precipi-
; "7 SiZ& UP t0 50,000 Sq ft PermittinS existing sectionaliza-
^ fc rfnain intact- Pulsers can be added to both cold and hot-side
precipitators to improve performance.
ALL-ELECTRIC APPROACH TO PERFORMANCE IMPROVEMENT
DPHnWfllU"elftriC nature of pulse energization systems means no unusual
peripheral services are required to make the PES-2000 system functional.
Performance improvements are obtained via improvement in dust charging
attendan^hieh^ ^	^ ^ injection of costly chemicals and the
attendant high operating costs.
11-10

-------
100--
O)
c
"O
<0
o
J
3
a
3
o
c
10--
$6
a
.001
.01
0.1
0
?o = 1-a0-^99.9%	99%	90%
Figure 4. Reduction in outlet loading as a function of unpulsed
efficiency for various pulsed enhancement factors.
11-11

-------
LOWER MAINTENANCE COSTS
Optimal operating conditions for the PES-2000 occur when base TR voltages
are adjusted for minimum sparking rates. Low sparking rates dramatically
decrease wire erosion and as a result, extend discharge wire lifetimes. The
high voltages associated with the pulse system also permit use of larger
diameter wires. The two factors singly or in combination lead to reduced
downtime and maintenance costs caused by broken wires. The high reliability
of the pulser sytem, as demonstrated by over 500,000 hours of pulser module
operation at full scale commercial installations, also leads to reduced
maintenance costs. Pulse systems work even if discharge wires become en-
crusted with dirt or collecting plates become warped.
REDUCED CAPITAL COST
By taking advantage of enhancement factors associated with pulse energi-
zation new precipitators can be designed at relatively low SCA thereby,
drastically reducing the size of the new precipitator and hence installation
costs. The PES-2000 can be added to a large precipitator for as little as
$1.47 per sq ft. The comparable charge for increasing the size of a precip-
itator by 50% which corresponds to an enhancement factor of 1.5 would be
about $6.00 per sq ft of base precipitator area. For large sized facilities
requiring millions of square feet of collecting area the capital savings can
be substantial. The same is true for retrofit applications where adding more
collecting plates is one proposed solution to poor collection efficiency.
Addition of pulse systems is much cheaper than adding new precipitator sections.
In many cases, increasing the size of existing precipitators is either
impossible or extremely expensive due to space limitations.
LOWER OPERATING COSTS
The highly energy efficient PES-2000 converts nearly all the pulse
energy into corona energy. The short pulses negate the requirements for
expensive energy recovery circuitry and elimination of all energy dissipative
components from the charging system ensures a high energy conversion effi-
ciency. Typically, the PES-2000 system uses 5kW of electrical power per
15,000 sq ft of collecting area. However, total energy requirements for a
pulse energized precipitator are generally slightly less than those of a
well-energized conventional precipitator. Since PES-2000 operates on
electricity only, there are no operating costs for chemicals or chemical
storage facilities.
ON-SITE DEMONSTRATIONS
The modular compact nature of the PES-2000 allows for on-site demon-
strations of the effectiveness of pulse energization. Connections are non-
disruptive and require no intenal precipitator modifications or elaborate
interfaces with existing equipment. Base TR sets are uneffected and
sectionalization remains intact.
11-12

-------
OPERATING RESULTS
Pulse systems manufactured by IPC have been tested at various demonstration
sites and at full-scale commercial installations. Table 2 is a summary of
results from various installations. In all cases, the observed enhancement
factors equalled or exceeded expectations based on laboratory data. The results
cover a range of resistivities from 5 x 1010 to 1013 ficm. In all cases, outlet
loadings were reduced by a factor of 2.0 or more when pulse energization was
added to the basic precipitator. Most of the units were cold-side precipitators.
One site, however, was a hot-side precipitator using .7% Na conditioning;
Pulsing achieved a 50% reduction in outlet loading which corresponded to
an enhancement factor of 1.46. These results illustrate and underscore the
effectiveness of pulse energization as a technique for dramatically improving
Precipitator performance.
COSTS
The installed costs of PES-2000 depend strongly on the size of the ESP
sections being pulsed. This feature is shown in Figure 5 where costs are
stated in $/ sq ft of total ESP collecting area. If the entire ESP is pulse
energized costs vary from about $6.00/sq ft of total area for very small
sections to about $2.50/sq ft for large sections. In many cases, ESP perfor-
mance goals can be achieved by pulsing only a fraction of the total ESP. The
lower curve in Figure 5 shows the costs when only 50% of the total precipitator
Is pulse energized. The installed costs shown in Figure 5 are only good
approximations since in actuality costs would also depend on total number
f sections, total size of the ESP and particulars associated with installing
the pulser at a specific site.
Pulser costs can be compared to costs for increasing collecting area to
achieve equivalent collection efficiencies. Retrofitting collecting area
costs about $20-25./sq ft of new area, while new ESP's cost about $15-20/sq ft.
Consequently, adding pulse energization systems is significantly less costly
than adding additional collecting area.
11-13

-------
TABLE 2
SUMMARY OF PULSER PERFORMANCE AT VARIOUS SITES
Site Data
Operation
ft2/1000
SCFM
P
fi cm
Precipitator
Efficiency
Inlet
Loading
gm/m3
Outlet
Loading
gm/m3
Opacity
H
30 MW Boiler-Cold Side
Precipitator
Coal 1.09%-l.16% Sulfur
Not Pulsed
Pulsed
353
370
5x1011
94. 7%
97.8%
6.548
6.82
.343
.145
26%
16.5%
1.52
35 MW Boiler-Cold Side
Precipitator
Coal .58-.77% Sulfur
Not Pulsed
Pulsed
173
173
5xl010
96.19
98.53
10.40
10.48
.407
. 154
33.2%
16.8%
1.51
420 MW Boiler-Cold Side
Precipitator
1000 hr. Test
1.12% Suflur Coal
Not Pulsed
Pulsed
280
282
5x1010
97.67
99.02
7.98
7.81
. 186
.076
43%
31%
1.50
4th Field Not Energized
5000 hr. Test
.57% Sulfur Coal
Not Pulsed
Pulsed
270
272
IxlO13
85.42
93.65
11.69
12.18
1.703
.773
84.5%
68.8%
2.03
200 MW Boiler-Hot Side
Precipitator
Coal .85% Sulfur
No Sodium Conditioning
Not Pulsed
Pulsed
326
316

96.67
98.43
15.781
16.227
. 526
. 255
53.5%
37%
1.49
.7% Sodium
Conditioning
Not Pulsed
Pulsed
300
307

98.24
99.24
15.472
16.341
.272
. 124

1.46

-------
6--
5--
4--
2 
100% ESP PULSED
50% ESP PULSED
1 --
t
t
t
t
r
o
10	20	30	40
SECTION SIZE 11000 Sq. Ft.I
50
Figure 5. Approximate installed costs for PES-2000.
11-15

-------
REFERENCES
[1]	R. Heinrich, W. Feldmann, Electrical Cleaning of Fluids, U.S. Patent
2,000,017, Applied April 4, 1931, granted May 7, 1935.
[2]	H. J. White, A Pulse Method for Supplying High Voltage Power for
Electrostatic Precipitation. Trans, AIEE, Nov. 1952, p. 326-329.
13] J. E. Luthi, Grundlagen zur elektrostatischen Abscheidung von
hochohmigen Stauben, Diss. Nr. 3924 ETH Zurich 1967, Prof. A. Guyer
[4]	B. Bohlen, J. Luthi, A. Guyer, Neues Verfahren zur elektrostatischen
Gasentstaubung, Chemie-Ing. -Techn. 39 (1967) Heft 15 910-913.
[5]	G. W. Penney, P. C. Gelfand, The Trielectrode Electrostatic Precipitator
for Collecting High Resistivity Dust, JAPCA, Jan. 1978, Vol. 28, 53-55.
[6]	S. Masuda, D. Akutsu, Boxer Charger - A Novel Charging Device for High
Resistivity Powders, Conf. Record of IEEE/IAS Annual Mtg., 1978, 16-22.
[7]	S. Masuda, A. Mizuna, H. Nakatani, H. Kawahara, Application of Boxer
Charger in Pulsed Electrostatic Precipitators, Conf. Record of IEEE/lAS
Annual Mtg. 1980, 904-911.
[8]	H. I. Milde, P. L. Feldman, Pulse Energization of Electrostatic Pre-
cipitators, IEEE/IAS (1978) Annual Mtg., Conf. Record, p. 66-70.
[9]	P. L. Feldman, P. Aa, Operating Results from the First Commercial Pulse
Energization System to Enhance Electrostatic Precipitator Performance,
Conf. Records, American Power Conference, Chicago, April 1981.
[10]	P. Lausen, H. Henriksen, H. M. Petersen, Energy Conserving Pulse
Energization of Precipitators, IEEE/lAS 1979 Annual Mtg., Conf. Record,
163-171.
[11]	D. E. Rugg, M. D. Durham, G. A. Rinard, L. E. Sparks, An Experimental
Study of Electrostatic Precipitator Performance with Pulse Excitation,
Atmospheric Environment, Vol. 16, No. 5, 1251-1256 (1982).
[12]	H. I. Milde, Electrostatic Precipitation, U. S. Patent 4,183,736, Applied
August 1972, granted January 1980.
[13]	H. I. Milde, Pulse Corona Discharge in Electrostatic Precipitators
IEEE Transactions on Electrical Insulation Vol. EI-17, No. 2, April 1982,
179-186.
[14]	L. N. Menegozzi, P. L. Feldman, The Physics of Pulse Energization of
Electrostatic Precipitators, Presented at Third EPA Sponsored Symposium
on Transfer and Utilization of Particulate Control Technology - Orlando,
March 1981.
11-16

-------
J. B. Thomas, T. R. Williams, Pulsed Corona Discharges, 1962 Gas
Discharge and the Supply Industry, London: Butterworth, p. 202-208.
D. B. Miller, S. E. Collier, J. L. Maitten, J. M. Niziolek, P. J. Phillips
Measurements of Pulsed Corona in Coaxial Electrode Structures, IEEE/IAS
1978, Conf. Records, Annual Mtg. 128-135.
H. I. Milde, Reduced Power Input for Improved Electrostatic Precipitation
Systems, U. S. Patent 4,133,649, Applied December 1973, granted January
9, 1979.
S. Masuda, H. Nakatani, K. Yamada, M. Arikawa, A. Mizuno, Production of
Monopolar Ions by Travelling Wave Corona Discharge, IEEE/IAS 1981
Conf. Records, Annual Mtg. 1066-1073.
S. Masuda, Y. Nonogaki, Bi-ionized Structure of Back Discharge Field
in an Electrostatic Precipitator, Record of IEEE/IAS Annual Meeting
1981, p. 1111-1119 (Oct. 1981, Philadelphia).
11-17

-------
LABORATORY AND FULL-SCALE CHARACTERISTICS OF ELECTROSTATIC
PRECIPITATORS WITH RIGID MAST ELECTRODES
H. Krigmont, R. Allan, R. Triscori and H.W. Spencer, III
Western Precipitation Division
Joy Industrial Equipment Company
Los Angeles, California 90039
ABSTRACT
There was an increase in development activities on rigid discharge
electrodes in the mid-seventies. Since that time, JOY Manufacturing
Company has carried out tests on a patented rigid mast electrode in the
laboratory and on full-scale retrofitted and new units. JOY's mast
electrodes are now in operation in 29 units around the world. The use of
this electrode has improved precipitator reliability while maintaining
particulate performance equal to or greater than that of conventional
weighted-wire designs. This paper describes the electrode and its
characteristics in various applications.
12-1

-------
1. INTRODUCTION
JOY has a number of on-going development programs concerning ways
and means of improving electrostatic precipitator performance and
reliability.
This paper discusses one of these programs, discharge electrode
design, the results of which have shown that significant improvement in
electrostatic precipitator reliability can be achieved while maintaining
or improving performance.
The summary of various development projects carried out under this
program are discussed. The results have led to the development of a
rigid mast electrode design that is rugged, durable and exhibits
electrical characteristics that surpass those of traditional wire
electrodes.
The discharge electrode discussed in this paper was described in the
patent as "an improved discharge electrode having a stiffener centrally
of planar side and corona edges in an electrostatic precipitator."(1) In
more simplified terms, the electrode consists of a single sheet that is
stiffened by means of a Sshaped profile. Corona points are created
along the length by providing notches of a size and spacing selected to
generate desired corona characteristics. This discharge electrode is
unique in that it is of unitized construction and fabricated by a process
that involves roll-forming and treating, to insure straightness and to
eliminate critical stress concentrations. The electrode, in its most
recent form, is shown in Figure 1.
Units are operating around the world that employ varied plate
spacing, plate height, field lengths and methods of rapping utilizing
this rigid mast electrode.
2. DEVELOPMENT OF THE MAST ELECTRODE
2.1 PRELIMINARY LABORATORY INVESTIGATIONS
The JOY patented mast discharge electrode has been under
development since 1974. The mast electrode as developed replaces
two wire electrodes in North American electrostatic precipitator
design. Initial configurations tested consisted of flat sheets with
saw-toothed edges, notched edges and flared edges. These tests were
conducted with center support tubes with flat plates attached.
Prototype mast electrodes are shown in Figures 2 and 3.
It was determined during these early studies that a notched
plate could be used to simulate electrical characteristics of two
wire electrodes. As shown in Table 1, it was possible to control
corona characteristics by adjusting the separation of spacing
between slots and the notch geometry.(2)
12-2

-------
SUPPORT CHANNELS
SHROUDS
ELECTRODE
SUPPORT ANGLES
SPACER BAR

ALIGNMENT
BARS
1COLLECTING
SURFACE
SYSTEM
MAST
ELECTRODES
Figure 1. Mast Electrode High Voltage Assembly
12-3

-------

Figure 2. Prototype Mast		o
Electrodes	Fl6Ure 3' "type Mast
Electrodes
12-4

-------
TABLE 1
Center-to-Center
Spacing
(Inches)
2-1/4
3-1/2
No. of Emitting
Points per
Foot per Side
5.34
4.0
3.43
Notch
Opening
(Inches)
0
1/4
1/2
0
1/4
1/2
3/4
0
1/4
1/2
3/4
Normalized
Milliamps*
1.00
1.56
2.16
1.00
1.56
1.90
2.00
1.00
1.22
1.60
1.87
*Normalized to 3" Center-to-Center at Four Emitting Points/Ft. and
40 kV.
12-5

-------
During these early studies, various member sizes were
investigated and moments calculated to minimize deflections under
different loads.
Flow investigations were also carried out at various velocities
to determine if vibration would occur. It was found the frequency
of shedded vortices did not correspond to the natural or primary
harmonic frequency of the mast. In short, flow induced vibration
did not occur.
The next development of the mast electrode was to replace the
tubular member with a S-shape form. Electrical characteristics of
this profile were extensively tested in the laboratory and a
high-bay test facility used to evaluate 37-foot masts. Voltage
current characteristics were obtained and compared to those of wires
in a 9-inch duct, typical values of which are presented in Figure 4.
This figure shows that the mast electrode exhibited
characteristics superior to those of wires in that the onset of
corona occurred at lower voltages and lower current was provided at
higher voltages in normal operating ranges. Higher sparkover
voltages were also achieved with the mast electrode.
A special heat chamber was constructed to maintain constant
differential temperatures, and movements of the mast electrode were
measured with the aid of a laser beam. It was determined during
these tests that the mast had excellent thermal stability. It
maintained straightness within suitable tolerances at those
differential and elevated temperatures which could be expected under
adverse conditions in most applications.(3)
The electrostatic stability of the mast electrode was also
investigated. The configuration of electrostatic forces considered
for rotation of the mast electrode is shown in Figure 6. This
investigation confirmed that the built-in stability was more than
sufficient to account for the electromechanical forces exerted.
2.2 SUMMARY OF PILOT SCALE STUDIES
Pilot investigations of the mast electrode were conducted by
Sumitomo Heavy Industries, JOY's licensee in Japan.(4) (5)
Electrical characteristics and performance studies were conducted
with cement finish mill dust for temperatures up to 350 C. A
comparison of normalized effective migration velocity for various
inlet dust concentration is presented in Figure 5.
Further studies of heat distortion, stability and voltage
current characteristics were also completed verifying laboratory
data.
12-6

-------
60-
50-
40-
30-
W IRES
10-
MAST
80
60
70
50
30
VOLTAGE kV
40
20
Figure 4. Typical Laboratory V-I Curves

-------
o
<:
cr
S2 >-
2 t
QX
uj O
NJ
_l
<
2
cc
0
z
u
>
* NORMALIZED TO A
WIRE OF 2.6 MM. DIA.
AT 5GRAM/Nm3
1.8
1.6
1.4
1.2
1.0
0.8
o o
t= 200C
v= 1m/s
i = 0.1 mA/m
2 5 10 20 50
(g/Nm3j
INLET DUST CONCENTRATION
Figure 5. Normalized Effective Migration Velocity Dust Concentration

-------
N5
I
VO
i
FORCE R
FORCE S
FORCE S
FORCE R
I
Figure 6.

-------
2.3 FULL-SCALE INVESTIGATIONS
Based on the laboratory and pilot studies, it was decided that
full-scale operating experience was required. Two existing
weighted-wire precipitators collecting flyash from a coal-fired
boiler were partially retrofitted with mast electrodes.(6) Air load
and operating voltage-current data were consistent with laboratory
and pilot results as shown in Figures 7 and 8. Performance results
were inconclusive, however, considerable knowledge was gained as to
how fabrication and construction techniques might be improved.
Based on the information obtained, continued studies were
undertaken in both the high-bay test facility and the field.
Acceleration measurements were obtained in conjunction with
development of improved methods of fabrication and installation.
Subsequently, a considerable number of new and rebuilt
precipitators were supplied worldwide using the mast electrode in
various applications.
3. OPERATING EXPERIENCE
Thirty-seven (37) installations incorporating mast electrodes are in
operation or final design around the world on a variety of applications.
These include the following:
-	Coal Flyash
-	Recovery Boilers (Direct and Non-Contact)
-	Cement (Dry Kiln, Wet Kiln, Cooler or Finish Hill and Preheat
Kilns)
-	Potash Dryer
-	Wood-Waste Boilers
-	Coal Drying
Two non-flyash applications are described and the results of
detailed studies on a flyash application are presented in the following
case studies:
3.1 CEMENT INDUSTRY
A two-field, two-chamber conventional weighted-wire JOY
electrostatic precipitator was installed after the dry process
cement kiln at the Genstar plant in Regina, Saskatchewan in 1960.
This unit maintained or exceeded the design efficiency of 97.5%
in all performance tests conducted between 1960 and 1980. However,
due to increasingly stringent emission regulations and increased
dust loadings due to kiln modifications, improved performance was
required.
12-10

-------
WIRES
< 200
MAST
16 20 24 28
SECONDARY VOLTAGE - kV
Figure 7. Current vs Voltage Wire and Mast Sections Airload

-------
CURRENT VS VOLTAGE WIRE AND MAST SECTION
WIRES
MAST
^ 400
cr
<
 350
O
lj 300
200 J
OJ
O
o
Q
rn
sf
IT)
cn
ro
n
SECONDARY VOLTAGE-KV
FIG. 8
Figure 8. Current vs Voltage Wire and Mast Section

-------
JOY was awarded a contract in 1980 to replace the discharge
system of the existing installation (primary unit) and supply a new
two-chamber precipitator in series (secondary unit) using the rigid
mast electrode design. The complete installation was designed for
99.87% while handling 145,000 ACFM at 625 F.
The primary and secondary units were placed in service in
November 1980 and June 1981 respectively. Since commissioning, the
installation has equaled or exceeded performance requirements.
During testing, the average gas inlet temperature was 716 F and
excursions in excess of 800 have been experienced without adverse
effect on mast alignment.
According to plant records, the primary unit operates at
increased voltage and reduced current compared to the original
configuration, and precipitator availability has improved
dramatically.
SALT CAKE RECOVERY BOILER
There were existing weighted-wire electrostatic precipitators
installed after Georgia Pacific's direct contact recovery boilers at
the Woodland, Maine plant. These units were initially installed in
1964, each of which was designed to handle 180,000 ACFM at 98%
efficiency. Although the units performed satisfactorily,
deterioration due to corrosion necessitated major retrofit of both
precipitator installations.
The client specified rigid discharge electrodes, anticipating
improved long-term reliability while maintaining performance equal
to or greater than that of conventional wire weight design.
In December of 1981, JOY was awarded the contract for materials
and engineering to rebuild both units utilizing the rigid mast
electrode design. Erection was scheduled such that the two separate
casings on each boiler were retrofitted sequentially so as not to
interfere with plant production. The first rebuilt chamber was
placed on line June 1982, and the entire project was completed early
in October 1982.
Subsequent performance testing resulted in emissions well
within design and state requirements on both units. Since startup
the units have maintained performance with nearly 100% availability.
COAL-FIRED BOILERS
In 1970, two JOY weighted-wire electrostatic precipitators were
installed after coal-fired boilers at a midwest location. Since
time, the performance of these units had become marginal due to
changes in fuel and In state emissions standards.
12-13

-------
Discussions were held and the operating conditions reviewed to
determine if precipitator performance could be improved. Studies
were undertaken to evaluate the influence of changes in discharge
electrode design, plate spacing and sectionalization using computer
modelling. The results of this investigation have been presented in
Table 2.
It was decided that the precipitator would be rebuilt utilizing
Alternate No. 5 from Table 2. The rebuild included split first
field, collecting surfaces, mast electrodes, rapper attachments and
the addition of transformer-rectifier sets and controls for the
first field.
In March 1984, a comprehensive test program was undertaken in
conjunction with compliance testing. The purpose of this program
was to enable a better understanding of variables affecting
precipitator performance using the mast electrode, and the data
could be used to evaluate computer-generated predictions.
A summary of the significant observations and trends for Boiler
No. 1 is as follows:
-	Modified Migration Velocity versus SCA, Figure 9
-	Modified Migration Velocity versus Power Consumption,
Figure 11
-	ESP Efficiency Measured versus
Computer Prediction, Table 3
-	Stack Opacity Measured versus
Computer Prediction, Table 4
A summary of a regression analysis of performance data obtained
for Boiler No. 1 is presented in Figure 9. During the tests, in
situ resistivities were measured and these varied due to changes in
the fuel composition, soot blowing and other factors. It was found
values approaching 1012 0hm/cm were experienced, and further
evaluations were carried out at these elevated resistivity levels at
which back corona was clearly occurring (see Figure 10).
Modified migration velocity versus power input has been plotted
in Figure 11 for those tests in which the unit was operating well
into back corona region. It can be seen that the characteristics of
the mast electrode are at a minimum consistent with those of wire
electrodes operating in this region. Increased power does not
Improve performance and, in fact, contributes to marginal
deterioration.
JOY's modified computer model was used to simulate performance
of the precipitator using measured applied voltage and current,
inlet particle size distribution and dust loading as input data.
Predicted performance was obtained and demonstrated that this
computer model can be used to predict performance of full-scale
units with a high degree of confidence.
12-14

-------
TABLE 2
ESP COMPUTER MODEL PREDICTIONS. SUMMARY
cowmw
(MMNGENFNT)
D.E., TYPE
FLUE TOS FLOW, WCFH
FACE VFWJCm, P1/S
SCA, <93. FT./KACFH
RESISTIVITY CHVtH
(Assuvd)
EFF.,
MIGRATION VEL. W, Ol/S
AUG. APPL. VOLT*fF, kV
AW3. OPR. nra I^FT2

KXISTING PPTO.
(9V6Y6*)
W/WIPE
 9" SP
31 G.P.
1
2* to'
97
9-53
35.5
36.
167
5.
187
2*10'
94.8
8.03
27
23
167
5.
187
2*10"
90.7
6.45
23.5
19
EXISTING PPTO.
(9V6'/6,|
"HOT"
9 9" SP
31 G.P.
167
5.
107
2*10*
96.6
11.6
43.
36
167
5.
187
2*10'
97.6
10.1
37.8
29.
167
5.
187
2*10
94.75
8.
28.7
19.5
EXISTING PPRT.
iV/V/V)
"MftSr
$ 10" SP
27 G.P.
167
5.
163
2*10^
97.6
11.6
44.5
38.7
167
5.
163
2M010
96.6
10.54
40
36
167
5.
163
2*1011
92.6
7.99
30
20
EXISTING PPT*!.
(9V6V6M
"KCT"
$ 11" SP
25 G.P.
167
5.
151
2M0*
97.6
12.55
46.67
39.8
167
5.
151
2*10*0
96.17
10.98
40.33
36.67
167
5.
151
2*101t
91.96!
8.48
31.37
19.75
SPLIT FIRST FLD.
(4.5'/4.5 V6V6')
"mast"
? 9" SP
31 G.P.
167
5.
187
2*10'
99.05
12.65
42.33
37.5
167
5.
187
2M0
98.
10.63
36.33
32.17
167
5.
187
2*10*1
95.56
8.46
27.87
19.Rt
SPLIT FIW?T FID.
(4.5'/4,5,/6,/6,>
"WAST"
* 10' SP
27 G.P.
167
5.
163
2*10^
98.22
12.56
44.33
37.8
1fi7
5.
163
2*1010
96.8
10.73
38.5
30.2
167
5.
163
2M011
92.16
7.94
28.42
29.9
~Split fields are both
wechaniea! fc electrical.
SPLIT FIRST Ft J").
(4.5V4.V/6V6M
"MAPT"
0 11" SP
25 G.P.
167
5.
151
tf>9
*8.2
13.5
44.A3
38.02
"Asswed WD * 12 *aw
Microns and 6 * 2.5

-------
TABLE 3
ESP COMPUTER MODEL EVALUATION OF SELECTED TEST RESULTS
Test
SCA
Inlet
Load-
ings
Measured
Efficiency
Predicted
Efficiency
Average
Current
Density
Average
Applied
Voltage
No.
Sq.Ft./
KACFM
GR/ACF
%
%
mA/K
Sq .Ft.
kV
13
214
1.92
99.30
99.26
22.13
34.41
14
224
1.85
99.31
99.38
21.29
34.02
15
225
1.92
99.24
99.35
21.14
33.71
16
243
1.74
99.30
99.3
32.
27.87
17
257
1.97
99.34
99.26
32.08
28.53
18
232
1.76
99.14
99.1
33.
28.72
23
245
2.46
99.28
99.35
17.30
30.13
12-16

-------
TABLE 4
OPACITY MODEL EVALUATION OF THE TEST RESULTS
Test
No.
ESP
Eff.,
%
Outlet
Loading
GR/ACF
Outlet Particle
Size Distribution
Opacity %
Measured
(Avg.)
Visually
Observed
Computer
Predicted
MMD

01
97.66
.0611
9.
3.75
18.27
-
19.54
03
97.02
.065
10.
3.85
21.86
0-25
19.32
04
92.88
.228
22.
3.79
42.5
20-75
30.96
05
93.78
.1792
17.
3.15
31.6
-
25.30
06
95.88
.1252
7.6
2.38
27.25
-
27.74
08
99.15
.013
4.6
2.88
3.
-
6.8
10
98.74
.0193
13.
3.42
1.5
-
3.4
11
99.62
.006
2.
3.75
1.5
-
1.01
13
99.3
.015
12.
4.29
4.4
-
4.5
16
99.3
.0134
6.8
2.83
5.4
-
4.67
17
99.34
.014
7.8
3.9
6.1
0-10
5.69
18
99.14
.0167
6.6
3.14
6.68
-
6.68
19
98.77
.0226
10.
3.33
7.75
-
6.23
20
99.1
.0186
9.1
3.37
8.88
-
5.73
21
99.14
.0177
6.6
2.93
9.5
-
6.54
22
99.12
.024
11.
3.06
11.5
5.
6.03
23
99.29
.019
7.
2.69
10.5
-
6.03
24
99.30
.018
10.
2.78
11.5
-
4.10
12-17

-------
MIGRATION VELOCITY
(Y/,Wk ), CM/S.











































4


Sip












\ '
|C




















SPEC
AREi
IFIC C
* SQ.F
-OLLEC
T./100
1
-T 1 ON
OACFM
0 40 00 120 160 200 240 280 320 360 400 440 480 520 560 600
Figure 9. Specific Collection Area (SCA) vs. Migration Velocity (WjW^)

-------
380
340
320
300
280
260
240
< 220
 200
t; 180
W 160
q:
a: mo
id
U 120
100
80
60
40
20



/









l



%

I
i



k
\



1
1

\ ^



4
\ $



I
\ ]



1
%
X i
1


\
V

LEGEND:

\
X
1 t

	O- FLD1

\
v  \

 -A- FLD2


\ ' <
I
-O- FLD 3

/
\ / /

	O- FLD4

 /









'dt?


0-	
srz^


10	20	30	40
VOLTAGE, kV
Figure 10. Typical V-l Characteristics, BLR1, 300,000#/HR

-------















MIGRATION \
JEL0CITY VS. TOTAL


POWER INFUT FOR THE CONSTANT

SCA. fBLR.l)

















ft
(EX






&








TOTAL
INPUTj
(as ME
AT TH
MARY
- . . 	 .4
50WER
KVA
A SURED
E PR I -
SIDE)
'
0	20	30	40	50 60 70 80
Figure 11. Migration Velocity vs. Total Power Input for the
Constant SCA. (BLR.1)
12-20

-------
The results of the efficiency test data as measured and as
predicted by the computer model are presented in Table 3. The
computer predicted efficiencies were adjusted to correspond to
measured results by adjusting non-ideal factors. The differences in
the non-ideal factors can be explained by the differences in
particle size distribution, gas temperature, slight changes in
sneakage and gas velocity distribution and variation of the rapping
losses.
In addition, the results of the computer model predicted
opacity are presented in Table 4 and are compared with the measured
and visually observed opacities. It can be seen that the measured
and calculated opacities are in agreement.
4. CONCLUSION
We have discussed the development of the rigid mast electrode.
It has been demonstrated that electrostatic precipitator reliability
is improved and performance equal to or greater than that of a wire
electrode can be achieved.
Additionally, field data were used to verify computer prediction
models for determining electrostatic precipitator performance and
in-stack opacity.
BIBLIOGRAPHY
1.	U.S. Patent No. 4,303,418.
2.	Merrill, S., R&D Report R146, December 1974.
3.	Goland, Y., "High Voltage Mast Electrode - Additional
"Investigations, Engineering Report E-37, April 1978.
4.	Spencer, H.W., Interoffice Memorandum Dated April 2, 1977.
5.	Sumitomo Heavy Industries, R&D Internal Report.
6.	Wennerholm, R., Jr., "Current Developments in Conventional
Electrostatic Precipitation Concepts", CSIRO Conference,
August 1978.
12-21

-------
FULL SCALE EXPERIENCE WITH PULSED ENERGIZATION
OF ELECTROSTATIC PRECIPITATORS
K. Porle
Technical Manager, Air Pollution Control Technology
Flakt Industri AB
Vaxjo, Sweden
K. Bradburn
Manager, Pre-Contracts
Flakt, Inc., Environmental Systems Division
Knoxville, Tennessee
ABSTRACT
One of the most promising new technologies for Electrostatic
Precipitators is pulsed energization. Extensive research and
development during the last 10 years has introduced several different
concepts.
Reduced outlet emissions have up to now been emphazlsed but
research by Flakt shows that remarkable power savings can be obtained by
using suitable pulse concepts.
This paper presents test results from full-scale installations
where up to 90% of the power has been saved in combination with reduced
emission compared to conventionally energized ESP's.
Two different technologies have been used. A special T/R-set has
been developed for the multiple pulse charging concept where thyristors
are incorporated in the high voltage circuit thus avoiding any pulse
transformer. Low energy losses are obtained and the pulse widths are in
the microsecond range.
The semi pulse charging concept uses modified conventional T/R-set
controls and produces pulse widths 1n the millisecond range. Plants
with both technologies are today in operation worldwide and the
equipment has shown excellent reliability and availability.
13-1

-------
INTRODUCTION
Recently developed high voltage thyristors have made the concept of
pulse energization a commercial reality. The transition from the
laboratory to successful application on operating precipitators has been
made In an economically viable manner. Flakt pulsing systems, and
others, are now operating on precipitator installations In many parts of
the world.
The ability of pulse energization to give Improved precipitator
collection efficiency offers an alternative to extremely large
precipitators which are required when high resistivity ash associated
with low sulphur coal must be collected. In addition, when upgrading
existing precipitators to comply with new emission regulations or to
accommodate more difficult ashes pulse energization becomes a
particularly attractive option. An alternative of adding additional
fields to the precipitator requires plant space which may not be
available, pulse energization does not. Flue gas conditioning is also
an option, but has both a chemical and operating cost over the life of
unit. Pulse energization offers a reduction in the power consumed by
the precipitator yielding an operating savings to the owner.
In addition, pulse energization can be installed with minimal
precipitator downtime and requires no more maintenance or supervision
when installed than conventional systems.
It must be understood that pulse energization is not a panacea for
precipitator problems. The full benefits of pulse energization will
only be realized if the ash is of medium high or high resistivity and
the precipitator is in good condition.
Today, we can note the development of three generations of
equipment for microsecond pulsing:
1.	The spark-gap method, which requires a high degree of
maintenance. The spark gap has to be contained in a gas-filled chamber.
2.	Low-voltage thyristors and a pulse-transformer which require
an extra base T/R set. The pulse-transformer is also a very costly
piece of equipment.
3.	High-voltage thyristors without a pulse-transformer which
supply base DC voltage as well as pulses (the Flakt method).
Years of research and development, together with test programs on
numerous precipitators installations, has culminated in Flakt's
development of two distinct pulsing systems.
1. The MultiPulse Concept (MPC)*: bursts of 50 to 100
microsecond pulses at 3 to 100 millisecond intervals.
13-2

-------
2. The Semi Pulse Concept (SPC): power-line half-waves at 30 to
150 millisecond intervals.
The MPC system does not require an additional conventional
transformer-rectifier coupled in parallel since it inherently provides a
DC base voltage. This makes the equipment physically compact and cost
effective when increased precipitator performance is required with
highly resistive ashes.
The SPC system is a software feature incorporated into the Flakt
microprocessor transformer-rectifier controller (EPIC). The EPIC
controller with SPC is physically the same size, if not smaller than
many controllers in use today, and can be incorporated into existing
control panels with little difficulty and at minimal expense. The SPC,
therefore, provides an opportunity to Increase precipitator efficiency
when operating on medium to high resistive ash, for a cost not
significantly higher than that required to upgrade older transformer
rectifier controllers with basic microprocessor controls.
Tests on commercial precipitators in many parts of the world
operating on different types of fly ash have demonstrated that these
pulsing systems cut energy consumption to such a dramatic extent that
they are a viable economic alternative even if only minor improvement in
precipitator efficiency is achieved.
Monitored operation of precipitators of different designs equipped
with MPC and SPC systems has shown energy consumption in some cases can
be reduced by up to 90 percent and stack emissions up to 75 percent.
BACKGROUND TO PULSED ENERGIZATION
Good precipitator performance requires that the corona current must
be evenly distributed over the entire collecting electrode surface. Ash
build-up, temperature stratification in the flue gas, misaligned
electrodes, etc., are examples of many causes of uneven current
distribution. Experience shows that a certain minimum current is needed
to ensure at least some flow to the most quiescent surface areas.
Each ash layer exerts a resistance to the flow of electrical
current. According to Ohm's law, the voltage drop caused by this
resistance across the dust layer is proportional to the magnitude of the
current. Many ashes have very high electrical resistivity yielding a
very high voltage  so high, that local breakdown occurs 1n the ash
layer. This is called back corona. It is a very critical factor 1n
precipitator performance. In order to lower the voltage drop to make
the back corona disappear, 1t is necessary to reduce the current or
change the ash resistivity by means of gas conditioning (NH3, SO3,
moisture, etc.). The first method is, of course, very simple but
results in the loss of current in some areas of the precipitator
lowering performance.
~Developed in close cooperation with Kraftelektronik AB Surte, Sweden.
13-3

-------
The precipitator designer has to study the current distribution
both in microscale (by detailed studies of electrode geometry) and in
macroscale (distribution between front and rear surface areas of the
precipitator).
Poor current distribution and back corona conditions can be
overcome to a certain extent by pulsed operation. Let us first examine
the electrical capacitance of the ash layer on collecting plates. If
the resistivity of the ash is R (ohmm), and the relative dielectric
constant is E, then the time constant is: T=8.85*E*R*10"'2 sec. This
1s a measure of the "electrical inertia" of the ash layer. For a
resistive ash (for example, R=10^ ohmm, E=2, T=1.8 sec.), T is in the
second range, which is a long time even compared to the 10 millisecond
half cycle of the power line. This also implies that the critical
current density, i'crit where back corona starts in the ash layer, is
very nearly equal to the time-average current density. Thus, high peak
currents from the discharge electrode normally have little effect on
back corona  only the average value does; it is possible to
momentarily increase the power for good current distribution at regular
intervals. By selecting peak power for good current distribution, as
well as interval length for a low average current, the risk of back
corona is minimized.
The quantitative improvement obtained by pulsed operation cannot be
evaluated theoretically. Extensive testing in actual operation is
therefore required. Since the current produced under pulsed operation
is more efficiently used, we obtain the double advantage of performance
improvements and power savings.
THE MULTIPULSE CONCEPT (MPC)
Figure 1 shows the principal circuit diagram of the Flakt pulsed
power supply. A capacitor after the high-voltage rectifier is charged.
By means of thyristors on the high-voltage side, the energy is
transferred to the precipitator through inductance to form an
oscillating circuit. The energy swings back and forth until an
essential part has been utilized. The thyristor switching is carried
out via transformers from ground potential. Since there 1s no need for
a pulse-transformer, internal energy losses are low. The conventional
base rectifier is not needed for DC since the pulse unit is designed to
supply an inherent base voltage which automatically tracks the corona
onset  a salient feature of the MPC system. Since third-generation
electronic components are utilized, the unit is small, lightweight and
cost-effective.
If a base rectifier is still used as a standby unit to provide
extra current for low-resistivity dust, a coupling capacitor Is not
required.
Figure 2 shows the high-voltage waveform for the MPC system.
Before each burst, the required pulse energy is stored 1n a capacitor.
When this energy is released, the first pulse in the burst is generated
13-4

-------
(conventional T/R)
oil tank
i	
base voltage leak
	1
diode chain
power
line
thyristor chain
I	
pulsing
Ctrl
current
Ctrl
Figure 1. Multi-Pulse Power Set
13-5

-------
EP voltage
Flashover limit
Time
Figure 2. Multi-Pulse HV Wave Forms
13-6

-------
and the energy is then free to oscillate between the capacitor and the
EP and through the inductance. While the oscillation proceeds, the
pulse amplitude will decline, since capacitor energy is progressively
used to form the corona in the precipitator. A train of pulse bursts is
created. Each burst consists of closely spaced multiple pulses which
decrease in size. Hence, the name "multipulse concept."
A precipitator becomes more prone to sparking as the corona forms.
Compared to DC, this tendency towards sparking decreases as the pulses
shorten. It takes some time before the precipitator's original sparking
sensitivity is restore. The time between two pulses within a burst is
not sufficient for full recovery but by decreasing the pulses in size,
we avoid provoking the precipitator and causing it to spark.
In the MPC system, energy is very efficiently utilized in the
precipitator. The favorable effect on current distribution from the ion
cloud produced by the first pulse in each burst will remain during
practically the entire duration of the burst. The average current can
be controlled by varying the burst repetition frequency and/or pulse
amplitude.
THE SEMI PULSE CONCEPT (SPC)
Figure 3 shows current and voltage from a conventional T/R set.
Here, the controlling thyristors are fired each half cycle of the line
power. Now it is possible to control the firing of the thyristors by
still using the same standard T/R set. This is accomplished in such a
way that only each third half wave is allowed to pass, resulting in the
waveforms shown in Figure 4.
By maintaining the same amplitude on each half wave, the average
current decreases to one-third the original. The voltage waveform in
Figure 4 has a more pronounced ripple than in Figure 3. It is obviously
possible to block any number of the half waves to make the ripple even
more pronounced. The voltage curve form now becomes a DC base level
with "pulses" superimposed on it. These pulses are of a much longer
duration  3 to 5 milliseconds ~ compared to those previously
described in the MPC. We call this semipulsing.
The total number of unblocked pulses in relation to the total
number of half cycles is called the charging ratio (CR). Since
semipulsing is achieved by means of software programmed into the already
powerful EPIC microprocessor unit, 1t represents a very minor investment
offering major benefits. Field tests have shown that power consumption
can be reduced substantially  and 1n many cases, the outlet emission
can also be lowered  without any modification to the standard T/R set
system other than the installation of the EPIC control units.
REVIEW OF OPERATING EXPERIENCES
Flakt pulse energization systems have been installed and tested for
13-7

-------
T/R current
T/R voltage
y* resulting
firing
corona onset voltage
Figure 3. Conventional Full Wave T/R Operation
r\
11
I i
I I
i I
I I
r\
I I
I I
I I
I I
I I
I I
XX
II II
I I I I
_L_1	I	L_
n
11
i I
I I
I I
I !
i\
i!
i i
i i
xk
i1 i1
J_j	Li_
r\
11
I i
I l
i i
I i
i i
A
I I
*><
T/R current
. resulting
EP voltage
corona onset voltage
Figure 4. Semi-Pulsed T/R Operation
13-8

-------
various precipitator applications in the United States, Sweden, Belgium,
Denmark, Japan, Taiwan, Finland, Australia, and South Africa.
In the United States, new Flakt precipitators are being installed
with SPC as part of the microprocessor transformer rectifier controllers
(EPIC). The first of these controllers installed 1n the U.S. were
supplied on Flakt precipitators on a small (29 MW) boiler firing 1%
sulphur coal. Performance tests conducted in conjunction with
compliance testing demonstrated that emission levels, as measured with
conventional charging, could be maintained with SPC while only consuming
15% of the transformer rectifier power normally required. Figure 12.
SPC test work has also been conducted on a large pilot precipitator
installation 1n the U.S. operating on high resistivity ash. Tests were
conducted with two different electrode systems with change ratios
ranging from 1:1 to 1:9. Despite some data scatter normally associated
with pilot test work, it was shown that significant power savings and
reductions emissions were achieved over the range of charge ratio's
tested.
Figure 5 summarizes the data for precipitators operating on flyash
presented in Figures 6 through 14. Flakt data confirms the findings of
others that the efficiency of pulsed operation increases the higher the
resistivity of the ash and the shorter the pulse width.
With a precipitator collecting high-resistivity ash, we have the
following options:
1.	Particulate emissions exceed the required limit. The main
interest 1s now to decrease emissions. We can expect to
achieve a reduction of approximately 50% with SPC and 75% with
MPC. Power consumption is also reduced.
2.	Particulate emissions comply with the required limit since a
very large precipitator was installed. The correspondingly
large power consumption  1000 kW is the order of magnitude
for a 500 MW boiler  can be reduced 50 to 90% without
affecting EP performance. The capitalized cost for power
evaluated against equipment suppliers 1s usually U.S. $3,000
to 6,000/kW. A saving of 500 kW is thus worth $1.5 to $3
million.
The corresponding expectation for medium-resistivity ash 1s:
1.	A reduction of 10 to 20% 1n emissions.
2.	A reduction of 50% 1n power.
Operating results are uncertain when MPC or SPC 1s used in
connection with low-resist1vity ash.
MPC and SPC have also been tested in other applications
(metallurgical, cement, recovery boilers). Similar results were
13-9

-------
obtained. The higher the resistivity of the dust, the greater the
improvement in EP performance when pulsed energization is used.
CONCLUSION
Flakt's world wide testing has demonstrated that both the
Multipulse and Semi pulse concepts are viable economic means to reduce
both precipitator emissions and energy consumption.
Pulsing permits the selection of smaller new precipitators for high
resistivity ash and offers significantly lower transformer rectifier
power consumption.
Existing precipitators required to operate with high resistivity
ash and meet new stringent emission regulations can be fitted with pulse
energisation. Both emissions and power consumption can be drastically
reduced. Additional plant space is not required and retrofitting can be
made with minimal unit downtime. No chemical additives or associated
operating costs are required.
Both MPC and SPC have their place in precipitator systems. The
choice of which system to install must be fully evaluated considering
all aspects of each specific project.
Pulsed energisation has been demonstrated to be both viable and
reliable technology for application to electrostatic precipitators.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the views of the Agency and no official endorsement
should be inferred.
13-10

-------
REFERENCES
Lausen, P. and L1nd, H. Application of Pulse Energization on Fly Ash
Precipitators. Tiz-Fachberichte, Vol. 106, No. 12, 1982.
Petersen, H. Hoegh. An Energy Conserving Pulse Energization System. EPRI
Conference on Electrostatic Precipitator Technology for Coal-Fired Power
Plants. Nashville, Tennessee, July 1982.
Feldman, P. L., AA, P. J., and Thanh, L. C. Present Status of
Research-Cottrell Pulse Energization Technology. EPRl Converence on EP
Technology for Coal-Fired Power Plants, Nashville, Tennessee, July 1982.
Masuda, S. Novel Particulate Control Technology. 4th EPA-Symposium on
Transfer and Utilization of Particulate Control Technology, Houston, Texas.
Stomberg, H. and Lundquist, S. Pulsed Radio Frequency Resonance Operation
of Electrostatic Precipitators. Inst. Phys. Conference 5er\ Nol 557
Session VI. Paper presented at Electrostatics, 1983, Oxford.
Lundquist, S. and Sekar, S. An Investigation of Continous and Pulsed High
Frequency Corona with Application to Electrostatic Precipitation.
University of Uppsala, Sweden, Uurie 140-81 (December 1981).
White, H. J. Industrial Electrostatic Precipitation. Addison-Wesley
Publishing Company, Inc. (1963).
13-11

-------
100% ~ performance with pulse unit off.
Plant
Outlet
mission
%
Power
consumption
%
Dust
resistivity
A
AUSTRALIA
66 MW
28
30
high
B
DENMARK
60 MW
34
80
60
60
high
medium
C
AUSTRALIA
660 MW
SO
100
60
100
60
60
medium
medium
high
D
DENMARK
600 MW
65
100
10
25
high
medium - low
E
TAIWAN
140 MW
45
20
high
F
U.S.
20 MW
100
15
medium-low
Figure 5. Summary of Present Experience Regarding Pulsed
Operation of EPs for Coal-Fired Boilers
13-12

-------
Dust emissions for different energization methods
CC (conventional), MPC (multipulse), SPC (semipulse).
% Dust emission
1:X  CR, charging ratio
160
CC
1M
100
1:15
o ^
10 ' '
40 
MPC
fi A/m' currant density
15 
in 2nd &
75
150
to
100
Arrangement of T/R-sets.
At 151u A/rrf a primary
kWh-meter gave a power
consumption of 1 W per
rrf collecting area with
MPC.
Figure 6. Plant ACoal-Fired Boiler, 66 MW, Australia
Case
1st
Raid
2nd
3rd
A
CC
CC
CC
B
CC
SPC
SPC
C
CC
MPC
MPC
currant
danKy
rtnys
150
w
tod
13-13

-------
Comparing outlet emissions for different
energization methods. The emission for
conventional operation at medium resistivity
is set" 100%.
Values shown  (emission %)/(emission enhancement).
Operating mode
Dust resistivity
100
70/1.42
50/2.00
medium
276
193/1.43
95/2.91
high
conv.
SPC
MPC

Plant configuration: 2 EP's, 2 fields each.
Pulsed operation in rear fields only.
Pulsed operation in front fields less effective.
Figure 7. Plant BCoal-Fired Boiler, 60 MW, Denmark
13-14

-------
Medium dust resistivity, long term testing.
Opacity %
30
20
15
10
8
. 1:9
o 1:7
charging ratio
CR: 1:x
TT
80 100
I	lll
200 300 400 600 800
mA/zon#
Data obtained from opacity meter recordings.
The table shows estimated power consumption
based on the opacity being kept constant 13%.
System
Currant
mA/bus Met.
Powar
kW
Energy saving
%
Conventional
600*800
565
0
EPIC w. peaksaakar
430
305
46
EPIC w. peaksaakar
and semipulM
160
128
77
Figure 8. Plant CCoal-Fired Boiler, 660 MW, Australia
13-15

-------
Blended S. African and Australian coals fired.
Ash resistivity high.
OTP %
200
150
100
X
1001
-75
50
X
X
X
CR
1:1	1:3	1:5	1:7	1:9 1:11
Energy saving
0% 50% 73% 80% 86% 89%
CR  charging ratio (1:1  conventional)
OutJet emission measured by gravimetric sampling.
Plant configurations: 4 cells, 5 fields20 bus sections.
20 T/R sets all with EPIC controls.
Figure 9. Plant DCoal-Fired Boiler, 600 MW, Denmark
13-16

-------
Ash resistivity low.
Semipuls*
Current density
Pwr consumpt.
Emission
mod*
|i A/m*
%
kW
%
mg/m* NTP
1:1
276
100
1066
100
58
1:3
90
32
350
33
47
1:5
55
20
211
20
50
1:9
33
12
126
12
69
"Normal"
65
24
269
25
54
The power consumption measured with kWh-meter
and includes T/R-set losses. Note that pwr. cons, is
prop, to current density.
"Normal": OT-1:3 for t, 2. and 3. fields.
CR 1:5 for 4. and 5. fields.
Plant configuration see fig 9.
Figure 10. Plant DCoal-Fired Boiler, 600 MW, Denmark
13-17

-------
Australian coals, resistivity high.
Emission	Rel. total
mg/m5 NTP	current %
conventional, peakseeker off
400
peakseeker on
X * emission
O  current
100
300
200
50
100
1:1 1:3 1:5 1:7 1:9 1:11 Semipuise
mode
Plant configuration: 3 fields, 3 T/R-sets.
All T/R-sets have EPIC controls.
Figure 11. Plant ECoal-Fired Boiler, 40 MW, Taiwan
13-18

-------
1% sulphur bituminous coal, medium resistivity
emission
mg/m' NTP 10
i ' i	I	1	1	1	1	1	1
20 40 60 60 100
Percentage
Relative T/R power comsumption
~ Convention charging
o SPC
emission
GR/DSCF
0.00375
0.0025
0.00125
No visible emission for any test.
Data spread within accuracy of testing
measurement.
Figure 12. Plant F--Coal-Fired Boiler, 29 MW, U.S.A.
13-19

-------
Low sulphur bituminous coal, high resistivity

Electrode A



Relative






Emission (%)







Charge ratio


test series
1>I
1:3
1:5
1 *.7
1:9
temp *C
1
too
82



120
2
100
90
68


120
3
100
99
80
75
54
120
4
100
77


79
120
5
100


62

120
6
100
98

95

160
7
100
75

71

160
Relative






Current (/
>.)





1
100
45



120
2
100
35
19


120
3
100
33
18
16
12
120
4
100
33


8
120
5
100
33

12

120
6
100
38
14


160
7
100
33

II

160
Figure 13. Plant GCoal-Fired Boiler, Large Pilot Precipitator
13-20

-------
Low sulphur bituminous cool, high resistivity
Electrode B
Relative
Emission (%)
Charge ratio
test series
IM
H3
is 5 1:7
1:9
temp *C
1
2
3
100
100
100
91
54
110
120
120
120



Relative
Current (%)



1
2
3
100 30
100 7
100 12
120
120
120



Figure 14. Plant G--Coal-Fired Boiler, Large Pilot Precipitator
13-21

-------
NEW LIFE FOR OLD WEIGHTED WIRE PRECIPITATORS:
REBUILDING WITH RIGID ELECTRODES
Peter J. Aa
Product Manager - Electronics
Maintenance & Service Division
Research-Cottrell, Inc.
Gary R. Gawreluk
Product Manager - Precipitators
Air Pollution Control Division
Research-Cottrell, Inc.
ABSTRACT
There are literally thousands of weighted wire precipitators currently
operating in the United States. Most of these units were designed in the
50's, 60's and early 70's. Since that time pollution control requirements
have become more stringent and electrostatic precipitator technology has
advanced considerably.
This paper will discuss how vintage weighted wire precipitators may
be upgraded to state-of-the-art rigid electrode designs while improving
performance, enhancing reliability and avoiding costly equipment replace-
ment.
Three case histories are examined where rigid electrode upgrades have
been undertaken.
	A black liquor recovery boiler
	An industrial power boiler
	A major utility generating station
Special attention is paid to major design modifications made and final
operating results. Also discussed are modifications made to electrical
control equipment in order to improve performance and reliability.
14-1

-------
INTRODUCTION
A recent article in the Journal of the Air Pollution Control
Association estimates that there were roughly 1400 electrostatic precipi-
tators in operation on U.S. coal-fired power plants as of 1980(1). Further,
it may be conservatively estimated that there are at least two or three
times that many ESP's on line in various other applications including pulp
and paper, petrochemical, rock products, and iron and steel.
Considerable numbers of these precipitators were designed and
installed in the timeframe prior to 1970. In the utility industry alone
this figure respresents more than 800 units, nearly all of which are
weighted-wire design. What this points to is that any program that a
utility or industrial user may be considering with regard to plant better-
ment or life extension would be remiss if it did not consider upgrade of
its electrostatic precipitators. Rebuilding these older, vintage weighted-
wire units, specifically rebuilding with state-of-the-art rigid electrode
designs, is one option that many users are currently exercising in order
to extend life, improve performance, enhance reliability, and minimize
maintenance requirements.
WHY REBUILD?
Rebuilding an electrostatic precipitator is not an easy task nor is
it inexpensive. In many cases, however, it may represent the most
technically viable and cost effective approach to achieve the desired
results.
There are many reasons why a user may wish to consider a rebuild
option, a number of which are interrelated:
	Fire/explosion - The useful life of a set of precipitator internals
can come to an abrupt end as a result of poor combustion efficiency
or a boiler flame-out. High LOI combined with lengthy hopper
storage and air inleakage have caused numerous damaging ESP
fires (2). Explosions have also been known to occur where a
boiler trip has sent large quantities of raw fuel and oxygen
into the ESP, being ignited with a single spark.
	Corrosion - The thin gauge electrodes in an electrostatic precipi-
tator are normally the first components to experience problems
in a corrosion prone unit. Typical causes of corrosion include
frequent start-ups/shutdowns, continuous operation below the acid
dew point, and cool Insulator purge air. Oftentimes the process
gas by itself is inherently corrosive as in the case of recovery
boilers for the paper industry. In a recent survey of owners
of precipitators on recovery boilers, for example, 30% reported
experiencing severe corrosion problems(3).
	Availability - Thin element discharge electrodes, whether weighted
wire or supported in a rigid framework, have relatively small mass.
As a result they are susceptible to failure from electrical arc
14-2

-------
erosion when subjected to high sparking rates, close electrical
clearances, and ash build-ups. Some earlier designs had inherent
failure modes due to poor shroud designs or electrode attachments.
On precipitators with limited sectionalization a number of broken
electrodes could have a negative impact on a plant's availability.
In a survey in the mid-70's of 174 ESP users, discharge electrode
problems were cited as the principle source of precipitator mal-
function (4) .
	Emissions - Precipitator deficiencies which result in emissions
in excess of original design are well documented in the literature.
Rebuilds can frequently address these deficiencies by improving
rapping effectiveness, increasing electrical sectionalization,
and correcting gas distribution. Additionally, when newer, more
stringent emission requirements, fuel switches, or process changes
cause a unit to fall out of compliance, a rebuild offers the
potential for improving performance through an increase in
collecting area, lower gas velocity, longer treatment time, and
higher migration velocities.
	Life extension - In the power industry in particular, many
utilities are considering comprehensive life extension programs
rather than contend with the problems, demands, and uncertainties
regarding new plant construction. Typical goals being set look
for adding another 20-30 years on top of a plant that is already
more than a quarter century old. A precipitator rebuild can
provide the means of achieving these goals by providing the
latest technology in an existing casing while avoiding the expense
of a new unit.
There is no one single type of rebuild. Each new rebuild opportunity
must be examined independently for what the client is trying to accomplish
and what the site-specific conditions will permit. In the past, the most
common type of rebuilds involved replacing the precipitator internals
in kind, essentially duplicating what was initially supplied. More
recently we are observing owners of weighted wire precipitators opting
for rebuilds incorporating new rigid electrodes.
A rebuild with Research-Cottrell's rigid electrodes generally involves
converting the original design from 9" gas passages to 12" gas passages in
order to accommodate the larger mass of the electrode with regard to
electrical clearances, as well as to take advantage of its higher operating
voltage potential and attendent higher migration velocities. This increase
in the interelectrode spacing permits some interesting things to be done
with component selection and resulting precipitator configuration. The
following discussion highlights a few of the more significant rebuild
aspects and addresses the benefits to be derived.
RIGID DISCHARGE ELECTRODES
The most obvious modification made in a rebuild involving rigid
discharge electrodes is the RDE itself. Compared to the thin diameter
14-3

-------
emitters characteristic of weighted wire arid European style rigid frame
designs its mass is many times greater (Figure 1). This yields an electrode
that, for all intents and purposes, will not break from electrical arc
erosion.
Research-Cottrell's Dura-Trode rigid electrode design (Figure 2) was
commercialized in 1978 after a 6-year-long research and development program.
Each electrode is shop fabricated of steel sheet at least as thick as the
collecting electrodes. Its roll-formed body provides the desired section
moment for structural rigidity and electrical stability in addition to
enhancing the electrical field strength. The scalloped edges serve to
establish corona generation at discrete points ensuring an even corona
distribution along its entire length. This feature is especially beneficial
on high resistivity applications since it reduces the potential for highly
localized current densities which intensify back corona. Additionally,
the design is based upon a single-pin top connection with a swaged end
(which acts like a hinge) affording a self-aligning feature in both planes.
At first glance it would seem that there might be a significant weight
differential involved in substituting rigid electrodes for weighted wires
but this is not normally the case since:
1)	One Dura-Trode replaces two weighted wires,
2)	the heavy cast iron weights are eliminated, and
3)	wider plate spacing actually reduces the total lineal feet
of electrode required with no compromise in particulate removal
efficiency.
On a typical 9'-wide by 36'-long field, for example, the total
additional weight for the Dura-Trode electrodes represents only about a
15% increase. After adding in the reduction in collecting area due to the
wider plate spacing, an overall decrease in weight is actually realized.
The benefits to be derived from utilizing rigid electrodes in a
weighted wire rebuild are not solely attributed to the improved reliability
and availability inherent in a virtually unbreakable electrode. A few
other points deserve mention:
 Lower power consumption - Past sizing practice for transformer-
rectifiers on weighted wire precipitators typically called for
relatively large current ratings in order to drive up the voltage
high enough to achieve desired space charge effects. The geometry
of the Dura-Trode electrode, however, results in considerably
higher voltages and consequently higher field strengths when
operating at comparable weighted-wire output currents in
equivalent gas passage widths (Figure 3). This means that in
most applications power consumption may be reduced with Dura-
Trode electrodes with no negative impact on outlet emissions.
14-4

-------
Weighted Wire
(shrouded)
Rigid Frame	Rigid Frame	Rigid Electrode
(bedspring)	(strung mast) (Dura-Trode)
Figure 1. DISCHARGE ELECTRODE TYPES
Bolted
"Connection
Main Structural Member
J Corona Generating Vanes
25 r
0 109"
Diam Wires
Dura-Trode
20 30 40
Precipitator Voltage-KVDC Avfl.
Note: (1) Qee load 330 to 3MF
(2) No t. by curvet denote precipitator uctlon
Figure 2. DURA-TRODE RIGID
ELECTRODE
Figure 3. DURA-TRODE AND 0.109*
DIAMETER WIRES IN 12"
GAS PASSAGES
14-5

-------
	Improved cleanability - One of the major headaches plaguing
numbers of weighted wire units has been keeping the wires free
of ash build-up. The difficulty of setting up a standing wave
in a wire that's part of a large bus section has been compounded
by the loose connection of wire to supporting frame. Dura-Trode
electrodes, however, behave more like a collecting electrode
than a discharge electrode with respect to rapping response.
Each^glectrode is firmly secured to its support frame with a
Huck bolt fastener (Figure 4). A consistent clamping force of
20,000 pounds is applied to each electrode insuring both an
excellent mechanical as well as electrical connection.
Additionally, design changes that have been incorporated in the
rapping assembly have improved system energy transfer by as much
as an order of magnitude compared to assemblies in use on older
weighted wire units.
	Longer electrodes - Older weighted wire units were typically
designed with electrodes no longer than 36 feet, the majority
of which were in the 24 to 30 foot range. In addition, many
units were designed with gas treatment velocities in excess of
6 feet per second which was in line with the design practice of
the times. Rebuilds with rigid electrodes, however, afford the
opportunity to consider extending electrode length to as much
as 48 feet. This increase in height would serve to bring down
the gas velocity in line with current sizing practices as well
as to increase both collecting area and treatment time. The
bonus is that these performance benefits may be realized within
the confines of the original equipment boundry limits (Figure 5).
COLLECTING ELECTRODES
Since a rebuild with rigid electrodes normally requires a widening
of plate spacing, it is appropriate to examine the replacement of the
collecting electrodes as well. More often than not, serious problems
with the collecting electrodes themselves Initially launch the client
into investigating a full scale rebuild in the first place. The benefits
to be derived from replacing the collecting surfaces are both reliability
and performance oriented due to changes that have been made in their
design In the past decade.
Barring the catastrophic, typical reliability-related failure modes
are corrosion and metal fatigue. Considering the relatively large numbers
of collecting electrodes involved on any given job, materials of con-
struction are typically limited to mild steel or high strength, low alloys
in order to be cost effective. Good control of the boiler or process goes
a long way in minimizing corrosion problems.
Metal fatigue, on the other hand, should be the responsibility of
the equipment manufacturer and is specifically related to the design of
his collecting electrode system and the manner in which rapping forces
are introduced into it. This issue is also very closely related to overall
performance when one considers the responsiveness of the collecting
14-6

-------
electrodes as a function of rapping input energy. A relatively "dead"
system may require high input energy or multiple level rapping in order
to effect good plate cleaning, and the harder and more frequent that a
plate is hit the more attention must be paid to addressing potential
structural failures.
As a result of a trend towards increasingly taller precipitators as
well as a desire to improve further the performance and reliability of
its current product, Research-Cottrell launched a major R&D effort in the
mid-70's aimed at collecting electrode redesign. The end product
plate that, on the surface, looks very similar to the original Opzel
design but upon closer examination contains some very significant design
modifications. These include;
1)	a rigid 10 gauge "stiff-back" which serves as a support for the
thinner gauge collecting surface,
2)	hyperbolically tapered mounting pads designed through finite
element analysis to gradually introduce rapping stresses into
the plate, and
3)	a new approach to roll-forming plate elements which, when
assembled, greatly enhances the dynamic responsiveness of the
plate surface.
TM
This new collecting electrode design, tradenamed G-Opzel (Figure 6),
can take considerably more abuse than its Opzel predecessor but yet in
comparative testing it has demonstrated superior response to rapping
forces, requiring roughly one half the input energy to achieve equivalent
"G" levels as compared to the earlier design. More than 35,000 of these
plates have been installed in over 75 precipitators since being introduced
in 1977. An excellent performance record has been established with not
a single plate failure reported to date.
This same electrode that is used in all new equipment offerings is
also well suited for most weighted-wire rebuild applications due to the
similarity in overall dimensions and method of support.
RAPPERS & CONTROLS
Both the rigid electrodes and collecting plates are typically
cleaned with a single impact rapping device (Figure 7). This affords
the opportunity to replace any electromagnetic vibrating devices and
consolidate control requirements and spare parts inventory.
Research-Cottrell's single impact rapper design employs two funda-
mental principles: electromagnetics and gravity. A low voltage D.C.
pulse is applied to the rapper coil causing electromagnetic lift of a
solid steel rod. Gravity causes the rod to drop as the coil is de-
energized, imparting as much as 24 foot-pounds of energy to the electrodes
below. No springs are used in the rapper as in a number of other top-
mounted designs.
14-7

-------
c v
Weighted Wire
ORIGINAL DESIGN
	SCA - 250 (9")
	GAS VELOCITY - 6.5 FPS
	TREATMENT TIME - 5.5 SEC.
GAS
FLOW
A
Dura-Trode
30' WEIGHTED
WIRES
42' RIGID
ELECTRODES
,"BELLY
BAND"
REBUILD
	SCA - 263 (12 )
	GAS VELOCITY - 4.6 FPS
	TREATMENT TIME - 7.8 SEC
Figure 4. DISCHARGE ELECTRODE
CONNECTIONS
Figure 5. IMPACT ON EXTENDING CASING
TO UTILIZE LONGER ELECTRODE
PHENOLIC TUBE
COIL COVER
HEAVY DUTY
SOLENOIO COIL
UPPER
CASING
SOLID ROD
HAMMER-
STATIONARY RAPPER
HAMMER ANVIL
SEALING
BOOT	
LOWER CASING
Top Stiffener
Mounting
Pad
Gas Flow
Roll-Formed
Element
Figure 6. ^-OPZEL^'COLLECTING	Figure 7. MIGI RAPPER ASSEMBLY
14-8

-------
Aside from the high reliability inherent in a rapper of such simple
design, rebuild situations can realize performance enhancements through
the addition of state-of-the-art microprocessor-based rapper controls.
A typical control used on many vintage weighted wire units tied all rappers
together on a single, motor-driven distributor switch. Turning at a
constant speed the switch would fire consecutive rappers at identical time
intervals, regardless of their location in the precipitator. Cycle times
of two minutes were standard practice on many of these units. Although
this may be satisfactory on inlet fields that build up quickly from high
dust loadings, the outlet fields suffer: the lower loadings coupled with
the fast rapping cycle typically result in higher than necessary outlet
emissions due to excessive rapping re-entrainment losses. Modern controls
can optimize cycle time on each field in the precipitator in order to
minimize these losses. Further optimization may be achieved through fine-
tuning individual rapping intensities at the control.
TRANSFORMER-RECTIFIERS & CONTROLS
As mentioned previously, rebuilding a weighted wire unit with rigid
electrodes normally involves changing the original gas passage spacing
from 9" to 12". In order to take advantage of the performance advantage
of the wider gas passages an upgrade of the high voltage power supplies is
often necessary. Typically this involves replacing the existing 70 KVp
transformer-rectifiers with units rated at 85 KVp output. Depending upon
the application, however, impact on plant load centers and field wiring
may be negligible. This is due to the lower current draw of the Dura-
Trode electrodes and whatever sizing conservatism which may have been
built into the original T-R size selection.
The benefits derived from going to wider plate spacing are generally
well recognized by most major suppliers and specifiers of electrostatic
precipitators. The primary motivation has been the reduced cost of
precipitator internals. On many applications SCA reductions proportional
to the increase in spacing have been achieved while maintaining or
improving design performance levels. This is widely attributed, for the
most part, to the improvement in migration velocity due to the higher
voltage and enhanced space charge effects.
With regard to transformer-rectifier controls, significant advance-
ments have been made in the last 5 years, let alone the past couple of
decades. Improvements in fundamental control functions such as spark
sensing, quenching and recovery have resulted in faster system response
and higher average power levels into the precipitator. Additionally, the
introduction of microprocessor-based control systems has permitted the
introduction of new integrated control strategies aimed at improving
performance, reducing power consumption, and assisting in equipment
troubleshooting and maintenance. It would be appropriate to say that any
unit purchased more than 10 years ago, whether a rebuild candidate or not,
should consider and investigate the benefits that may be derived from a
control upgrade.
14-9

-------
CASE HISTORIES
The foregoing discussion has outlined some of the basic modifications
involved in rebuilding a precipitator with rigid electrodes and their
impact on unit performance, reliability and maintainability. What follows
are a few case histories of actual installations where rebuilds have taken
place.
CASE HISTORY #1 - CONVENTIONAL RECOVERY BOILER
It is well known that conventional recovery boiler precipitators are
prone to internal corrosion. The first internal component to be affected
by this corrosion is usually the discharge electrode wire. Corrosion of
these wires causes them to break which results in grounded electrical
frames. Typically, when an electrical frame is grounded, the firing rate
of the boiler is reduced in order to maintain acceptable stack emissions,
resulting in reduced productivity.
Case History #1 is an example of an 8 year old recovery boiler
precipitator where the internals were corroded and needed replacement.
The customer chose to rebuild with Dura-Trode electrodes to increase
the overall reliability and availability of the precipitator.
The support steel and shell of the precipitator did not require any
replacement or repair. And since there was no change in operating
conditions, the rebuild was done within the confines of the existing
shell without any increase in plate height or treatment length.
As stated previously, the use of Dura-Trode electrodes necessitates
the use of 12" gas passage spacing and higher voltage transformer-rectifiers
(T-R's). Since the existing T-R's were the Askarel (PCB) filled type,
replacing them with oil filled 85 KVp T-R's was an added benefit of the
rebuild.
Due to the low current operating characteristic of Dura-Trode
electrodes, the new T-R's were sized substantially lower in current rating
than the old T-R's. The installed KW of the new T-R's is less than half
that of the old T-R's.
In addition to the T-R replacement, the out-of-^ate T-R controls and
rapper controls were also modernized with Powertrac T-R controls and
microprocessor rapper controls to improve the overall performance of the
precipitator.
Precipitator design conditions before and after the rebuild are
exactly the same since neither thfi boiler operation nor the necessary
precipitator efficiency required any changes (Table 1).
Table 2 highlights some of the precipitator design features before
and after the rebuild. Due to the increase in gas passage spacing from
9" to 12", the number of gas passages was reduced from 41 to 31 per
chamber. This gas passage reduction had a negligible impact on both gas
14-10

-------
velocity and treatment time as would be expected from an internal rebuild
within the confines of the existing shell. The SCA reduction reflects the
fewer number of collecting plates used with 12" gas passage spacing and
is compensated for by the increased migration velocity achieved with wide
plate spacing and higher field strength. A substantial reduction in the
number of rappers was achieved through the use of G-Opzel collecting plates
which are highly responsive and require only leading edge rapping.
As of the writing of this paper, only one half of this unitized
precipitator had been rebuilt. However, the customer was operating at
75% of normal boiler firing rate with all gases passing through the
rebuilt half of the precipitator and was maintaining a 30 to 35% opacity.
This opacity level is slightly less than that achieved under the same
conditions with the original internals. Based upon this observed per-
formance, there is every reason to believe that the unit will perform as
expected when both chambers are treating full gas volume.
CASE HISTORY #2 -INDUSTRIAL BOILER
Case History #2 represents a rebuild on a 12 year old precipitator
serving a pulverized coal fired boiler for an industrial user. The primary
reason for the rebuild was corroded internals, however, the plant also
desired to switch to a coal that produced ash which was more difficult to
collect than its present coal. Also, the plant wanted to update their
saturable reactor T-R controls to silicon controlled rectifier (SCR)
controls, and wanted to install a more reliable and flexible rapping system.
Except for the change in coal specification, the design conditions
before and after the rebuild did not significantly change (Table 3). The
guaranteed outlet emissions ranged from .01 to .025 grains, depending on
the sulfur-to-ash ratio.
High reliability and consistent emission levels were high priorities
of this customer. To satisfy the customer's concerns, Research-Cottrell
recommended rebuilding with Dura-Trode electrodes and G-Opzel collecting
plates. Since the structural steel and shell were in good shape, the
rebuild involved replacing internals within the confines of the existing
shell without increasing plate height or treatment length.
Table 4 shows the before and after rebuild precipitator design. The
number of gas passages was reduced from 23 to 17 due to the increased gas
passage spacing. Remaining within the confines of the existing shell
caused a slight increase in gas velocity with a corresponding decrease
in treatment time. These changes were considered to be insignificant
due to the increased migration velocity produced by the higher field
strength. This increased migration velocity also compensates for the
reduction in SCA from 361 to 249.
As part of the rebuild the electrical sectionalization was increased.
Due to the relatively low dust loading and moderate resistivity, Research-
Cottrell was able to reuse the three existing T-R's that had energized the
original five bus sections. These three 70 KVp T-R's were placed on the
14-11

-------
TABLE 1. CASE HISTORY #1 - CONVENTIONAL RECOVERY BOILER,DESIGN CONDITIONS
Before Rebuild (WW) After Rebuild (RDE)
Gas Volume (ACFM)
Temperature (F)
Inlet Loading (GR/DSCF)
Efficiency (%)
430,000
300
6.6
99.7
430,000
300
6.6
99.7
TABLE 2. CASE HISOTRY //I - CONVENTIONAL RECOVERY BOILER, ESP DESIGN
Before Rebuild (WW) After Rebuild (RDE)
Gas Passages (Total)
Gas Passage Width (in.)
Gas Velocity (FPS)
Treatment Time (sec.)
SCA
T/R Controls
Rapper Controls
No. of Rappers
82
9
3.88
9.28
412
Outdated analog
Cam Switch Type
146
62
12
3.85
9.35
311
State-of-the-art analog
Microprocessor
100
TABLE 3. CASE HISTORY #2 - INDUSTRIAL BOILER, DESIGN CONDITIONS
Before Rebuild	(WW) After Rebuild (RDE)
Gas Volume (ACFM) 110,000	110,000
Temperature (F) 325	350
Inlet Loading (GR/ACF) 2.8	2.8
Outlet Loading (GR/ACF) ?	0.01 - 0.025^
(1) Depending on Sulfur/Ash Ratio
14-12

-------
last three bus sections and two new 85 KVp T-R's were placed on the first
two bus sections. Each of the five fields now has its own dedicated T-R.
The three old T-R controls were also reused and were converted from satur-
able reactor controls to SCR-type with microprocessor control. The two
new T-R controls were SCR-type microprocessor controls.
The collecting and discharge electrode rapping system was replaced with
MIGI rappers and microprocessor rapper controller.
The rebuilt precipitator was brought into service in September 1983.
Unit performance to date has exceeded guarantee requirements with average
outlet emissions of 0.006 GR/ACF at design gas volume as compared to the
corrected guaranteed level of 0.025 GR/ACF.
CASE HISTORY #3 - UTILITY PULVERIZED COAL FIRED BOILER
The third case history is a little more involved than the previous two.
This 17 year old precipitator installation which was originally designed
for high sulfur coal had experienced severe corrosion of its internals.
Additionally the utility was also planning a switch to low sulfur coal.
The fuel switch would result in a 21% increase in volume to the precipi-
tator and the plant was contemplating a change from forced draft system to
a balanced draft system. Lastly, the T-R and rapper controls were outdated.
Typically the above requirements would dictate that a new precipitator
be built. Preliminary investigations, however, showed that the require-
ments could also be met with a rebuild plus SO^ flue gas conditioning. An
economic evaluation was performed and Table 5 shows the highlights. Even
with the gas conditioning system, the cost of rebuilding was less than a
new precipitator. And the turn-around time to complete the rebuild with
conditioning was roughly one-third the turn-around time for a new precipi-
tator. In addition, a new unconditioned precipitator would have taken two
to three times the real estate of the rebuilt precipitator.
The design conditions before and after rebuild are shown in Table 6.
The gas volume increased by 21%, the operating pressure changed from
+5 H90 to + 25 H^O, the inlet grain loading increased by 75%, the fuel
sulfur content decreased from 3% to 0.6%, and the outlet loading guarantee
was set at 0.10 lbs/MBTU.
Overall, the new precipitator design resulted in a reduction in gas
passages from 200 to 150 due to an increase in plate spacing from 9" to
12" (Table 7). If was obvious, however, that rebuilding without further
increasing the collecting area would not meet the utility's requirements.
In order to place additional collecting area within the confines of the
existing precipitator shell, the new collecting plates were made twelve
(12) feet taller than the existing collecting plates. This modification
served to reduce gas velocity by 14%, helping to minimize the re-entrain-
ment losses, as well as to Increase treatment time by 1.74. Even though
plate height was increased, total SCA was still less than originally
installed due to the increase in gas volume.
14-13

-------
TABLE 4. CASE HISTORY #2 - INDUSTRIAL BOILER, ESP DESIGN
Before Rebuild (WW) After Rebuild (RDE)
Gas Passages (Total)
Gas Passage Width (in.)
Gas Velocity (FPS)
Treatment Time (sec)
SCA
No. of Fields (Mech.)
(Elect.)
T/R Controls
Rapper Control
23
9
4.43
8.13
361
5
3
Saturable Reactor
Motorized Switch Type
17
12
4.69
7.46
249
5
5
Microprocessor
Microprocessor
Note: Original ESP was designed by Western Precipitation
TABLE 5. CASE HISTORY #3 - UTILITY P-C FIRED BOILER, EVALUATION FACTORS
New ESP (W/0 Cond.) Rebuild ESP (W/Cond.)
Cost
Schedule (P.O. to startup)
Space Constraints
$15M
92 weeks
2-3 times more
real estate
$12M
34 weeks
Existing real estate
TABLE 6. CASE HISTORY #3 - UTILITY P-C FIRED BOILER, DESIGN CONDITIONS
Before Rebuild (WW) After Rebuild (RDE)
Gas Volume (ACFM)
Temperature (F)
Pressure (in. W.C.)
Inlet Loading (GR/ACF)
Fuel Sulfur Content (%)
Outlet Loading
1,900,000
250-300
+5
2
3
0.01 GR/ACF
2,300,000
250-300
+25
3.5
.60 - 1.3
0.10 LBS/MBTU
14-14

-------
TABLE 7. CASE HISTORY #3 - UTILITY P-C FIRED BOILER, ESP DESIGN
Before Rebuild (WW) After Rebuild (RDE)
Gas Passages (Total)
Gas Passage Width (in.)
Collecting Plate Height (ft.)
Gas Velocity (FPS)
Treatment Time (sec.)
SCA
No. of Fields (Mech.)
4.3
190
7.0
200
30
4
4
9
167
150
6.0
5.0
42
12
4
6
(Elect.)
No. of Rappers
244
448
The electrical sectionalization of each precipitator was also increased
during the rebuild. The ten existing Askarel (PCB) filled, 70 KVp T-R's
per precipitator were replaced with twelve oil filled 85 KVp T-R's. The
four existing electrical fields were replaced with six new electrical fields
to increase reliability, produce more stable precipitator operation, and
improve power input.
Obsolete saturable core reactor type T-R controls and rotary distri-
bution switch type rapper controls were replaced with microprocessor
controls. Microprocessor energy management systems and a remote data
logger were also supplied with the rebuild to optimize power consumption
and provide a permanent automatic record of precipitator electrical oper-
ation.
Although outlet emission tests have not yet been performed, the
station has run at full load conditions with less than eight percent
opacity which is well below the 20% opacity limit that was guaranteed.
Experience to date with more than 8 precipitators has shown that
rebuilding old weighted wire units with state-of-the-art rigid electrodes
often represents a technically viable as well as cost-effective alternative
to replacement with a new precipitator. Rigid electrodes have demonstrated
both higher reliability and improved performance as compared to wire type
electrodes. Further improvements in both reliability and performance may
be realized by including more responsive collecting surfaces, more flexible
rapping systems, and more effective transformer-rectifier controls.
Whether precipitator upgrade is being considered due to corrosion,
poor performance, catastrophic failure, or simply as a key element in a
plant betterment/life extension program, the rebuild potential of the
existing equipment should be carefully examined before a new equipment
purchase commitment is made.
CONCLUSION
14-15

-------
The work described in this paper w
mental Protection Agency and therefore
reflect the vievs of the Agency and no
inferred.
s not funded by the U.S. Environ-
the contents do not necessarily
official endorsement should be
REFERENCES
1.	Lagarias, J.S., McDonald, J.R., Giovanni, D.V. Assessment of the
commercial potential for the high intensity ionizer in the electric
utility industry. Journal of the Air Pollution Control Association
31 (11): 1221-1227, November, 1981.
2.	Bump, R.L. Fires in electrostatic precipitators - survey results,
discussions and recommendations. Paper presented at the Joint
ASME/IEEE Power Generation Conference, St. Louis, Missouri, October
4-8, 1981.
3.	Henderson, J.S. Final survey results for direct contact evaporator
recovery boiler electrostatic precipitators. TAPPI 65 (3): 91-94,
March, 1982.
4.	Bump, R.L. Electrostatic precipitator maintenance survey. Paper
presented at APCA Specialty Conference on Operation and Maintenance
of Electrostatic Precipitators, Dearborn, Michigan. April 10-12,
1978.
14-16

-------
PULSING ON A COLD-SIDE PRECIPITATOR
FLORIDA POWER CORPORATION, CRYSTAL RIVER, UNIT 1
Joseph W. Niemeyer and Robert A. Wright
Lucidyne, Inc.
4235 Ponderosa Avenue
San Diego, California 92123
Wayne Love
Florida Power Corporation
3201 34th Street, South
St. Petersburg, Florida 33733
ABSTRACT
Lucidyne, Inc. and Florida Power Corporation conducted a
pulser feasibility study on the Unit 1 precipitator at Crystal
River Station. The test was to demonstrate pulsing as a
viable means of enhancing precipitator performance.
Unit 1 Crystal River precipitator is a cold-side Buell
design which was expected to operate below specification due
to switching to low sulfur coal. Testing was done in one
outermost cell of six cells across gas flow. Lucidyne's
pulsers were installed on the first five fields in the
direction of gas flow, the third field was grounded, reducing
the number of pulsers to four in series.
Testing unpulsed with one field shorted resulted in an
efficiency of 96.7%. Adjusting for power level and plate area
provided an unpulsed base efficiency of 98.1%.
TEST RESULTS:
One pulser; enhancement factor = 1.09
Two pulsers; enhancement factor = 1.22
Three pulsers; enhancement factor 1.26
Four pulsers in series did not provide significant
improvement.
15-1

-------
PULSING ON A COLD-SIDE PRECIPITATOR
FLORIDA POWER CORPORATION, CRYSTAL RIVER, UNIT 1
In July and August of 1983, Florida Power Corporation and Lucidyne,
Inc. conducted a pulser feasibility test on the Unit 1 precipitator at
Crystal River Station. The test was one element of a precipitator
enhancement study performed as a part of a Florida Power Corporation plan
to allow burning of low sulfur coal in the unit. Compliance with emission
standards would become a concern when Florida Power began to burn low
sulfur fuel in Unit 1. A fuel switch is required to lower the SO2
emission limits. This is an existing regulatory requirement to take
effect prior to commercial operation of a new unit at the same site
(CR-5). The precipitator was not designed for low sulfur fuel, and
without modification was not expected to meet the particulate emission
limit of 0.1 lb/MBTU. The pulser test was to determine if Lucidyne's
pulsing system was a viable means of providing the required enhanced
precipitator performance. Other means of enhancement were tested, but
this paper is concerned only with the pulsing.
Crystal River Unit 1 is served by a 1978, cold-side Buell
precipitator designed to meet particulate emission limits of 0.1 lb/MBTU
and less than 20% opacity with the unit firing high sulfur coal. The
precipitator has two chambers with three cells each. There are seven
fields: three fields each six feet long followed by four fields each
three feet long. Twenty one transformer/rectifier (T/R) sets are used in
a double-halfwave each powering two side-by-side cells. The first three
fields are powered by T/R's rated at 45 kVDC and 1400 mA. The remaining
fields employ units rated at 45 kVDC and 1100 mA.
Lucidyne's pulsers were installed on the first five fields on the
outermost cell of the eastside chamber. A broken wire eliminated the
third field in the tested cell, thereby reducing the effective collection
area and the number of pulsed fields to four.
The T/R sets associated with the pulsers were converted to fullwave
operation for the test. The remaining T/R sets were left in double-
halfwave configuration. To simplify installation, cable was used for the
connecting bus work. The pulsers were connected in series with the T/R's
by inserting the pulser high voltage terminals into the buses which
connect the T/R outputs to the precipitator discharge electrode terminals.
Unit 1 has a rated capacity of 400 MW, but to prevent boiler tube
leaks and provide reliability output was reduced to 370 MW. Soot blowing
was scheduled to minimize impact on particulate testing. During testing,
Unit l's pulverized-coal-fired boiler burned coal rated typically at 1.0%
sulfur, 9-10% ash and 12,000 BTU/lb.
Particulate testing consisted of gathering inlet and outlet samples
over three separate runs. The data samples for each set of inlet and
outlet measurements were averaged for use in calculating unadjusted
efficiencies.
15-2

-------
Initial testing with no pulsers energized established a base
efficiency for normal operating conditions. Pulser testing began with
four pulsers energized. As each particulate sampling set was completed,
one pulser was de-energized, sequentially reducing the number of pulsers
in use to zero. The affected pulser was de-energized at the end of the
sampling set and the new configuration was operated overnight to condition
the system prior to collecting the next set of data. Upon completion of
the sequential reduction testing, a check was made to verify the results
by retesting at two selected points. One of the two selected data points
was with two pulsers on line, and the other with no pulsers.
The effect of pulsing on the precipitator is presented in terms of an
enhancement factor which represents the precipitator size increase that
would achieve the same performance improvement. Pulsing produces an
increase in the average migration velocity and consequently in the
collecting performance of the precipitator. The enhancement factor is
defined as the ratio of the migration velocity after the addition of
pulsing to the migration velocity before.
There are two commonly used methods of determining migration
velocities and enhancement factors to ascertain improvement in performance
levels. The Deutsch-Anderson equation is an industry standard which does
not take into account variation in particle size. Since large particles
are easier to collect than small, as the ash laden gas stream proceeds
down the length of the precipitator a greater and greater percentage of
the remaining particles are small and therefore harder to collect. A
modified form of the Deutsch-Anderson equation takes particle size
variation into account. Therefore, enhancement factors derived from the
modified Deutsch-Anderson equation more accurately represent the real
world, and are normally used to evaluate the performance of pulsing.
The unpulsed base efficiency was determined to be 97.4% at full
boiler load and with all precipitator cells operating. The efficiency was
calculated from the measured data. Then, since the data was taken at
92.5% load (370 MW instead of 400 MW) and with 20% of the tested cell
inoperable (one six-foot field shorted out of a total of 30 feet), the
efficiency was adjusted to reflect normal operation at full load.
Pulsing increased collection efficiency to 99.01%. The modified
Deutsch-Anderson equation was used to adjust the data and to calculate the
migration velocities corresponding to the two efficiencies. The resultant
enhancement factor of 1.57 shows that pulsing produced improvement
equivalent to increasing the size of the precipitator by 57%.
There was no significant difference in enhancement between pulsing
three and four fields. Both configurations provided an enhancement factor
of 1.57. With two pulsers energized the enhancement factor was 1.49.
With one pulser it was 1.18. As previously mentioned, a cross check was
performed. The resulting pulsed efficiency of 99.13% indicates that the
quoted pulsing enhancement factors are conservative.
15-3

-------
A useful method of viewing performance is in relation to precipitator
penetration. Penetration is a measure of the particulate which exits the
precipitator. It represents system loss and is a measure of the pollutant
of concern. In this test pulsing three fields reduced penetration by a
factor of 2.6 (outlet grain loading by 61.5%).
The tests conducted at Florida Power Corporation's Crystal River
Station Unit 1 show that pulsing was reliable and significantly improved
the performance of the electrostatic precipitator. It was agreed (by
Florida Power Corporation and Lucidyne) that the enhanced precipitator
performance provided by pulsing would bring the Unit 1 precipitator up to
the required performance level.
The pulser tests were carried out concurrently with flue gas
conditioning tests. The pulsers were tested of one cell on the eastside
chamber while the flue gas conditioning tests were conducted on the west
side.
The testing allowed Florida Power Corporation to estimate the
equipment needed for the required precipitator enhancement and to fine
tune the potential capital and operating costs of each acceptable method.
It was found that less enhancement was required than first expected. For
the enhancement methods studied, this tended to make the systems with
lower first cost and higher operating costs more attractive economically
than systems with higher first cost and lower operating cost. Lucidyne's
pulsing system represented the latter. Florida Power Corporation
ultimately chose a 15 ppm capacity SO3 flue gas conditioning system based
on the precipitator requiring approximately 2 ppm SO3 injection rate. The
system is due to come on-line in mid-1984.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the views of the Agency and no official endorsement
should be inferred.
15-4

-------
Session 8: ESP: ADVANCED TECHNOLOGY I
Norman Plaks, Chairman
U.S. Environmental Protection Agency
Air and Energy Engineering Research Laboratory
Research Triangle Park, NC

-------
FIELD STUDY OF MULTI-STAGE ELECTROSTATIC PRECIPITATORS
Michael Durham, George Rinard, Donald Rugg,
Theodore Carney ,and James Armstrong
Denver Research Institute
Denver, Colorado 80208
Leslie Sparks
Industrial Environmental Research Laboratory
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
A multi-stage electrostatic precipitator (ESP) utilizes a configuration
of charging and collecting devices to provide optimum performance for a
particular set of operating conditions. Each component of the ESP can then be
designed to cope with the dust loading, particle size distribution, and dust
resistivity encountered. This technology is applicable for improving ESP
collection of fine, high resistivity dust from such sources as low sulfur
power plant flue gas, cement kilns, and metallurgical processes.
This paper describes the results to date of field tests on a multi-stage
ESP at the EPA/DRI Valmont Electrostatic Precipitator Test Facility in
Boulder, Colorado. The ESP utilizes alternate stages of Temperature-
controlled Electrode Prechargers (TCEP) and parallel plate collectors. The
precharger consists of corona wires suspended between anodes that are cooled
by 38-43C (100-110F) water which reduces the resistivity of the dust
collected in the precharger, allowing operation at very high field strengths,
greater than 5 kV/cm. The mean flue gas temperature is reduced by only 1.7C
(3f) by the precharger.
Tests were performed using different combinations of chargers and
collectors. Collection efficiencies as high as 99.4% were obtained on an ESP
with an SCA of 46 s/m (235 ft /1000 acfm). However, the collector sections
experienced problems with non-rapping reentrainment, requiring operation at a
velocity below 1 m/s (3 ft/s). The results are summarized and discussed in
terms of further application of the precharger as a retrofit device on
conventional ESPs.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
* Currently with S-Cubed in Albuquerque, New Mexico
16-1

-------
INTRODUCTION
The multi-stage ESP is an extension of the two-stage ESP concept in that
particle charging is performed in a separate section from particle collection.
The multi-stage ESP is produced by putting two or more two-stage ESPs in
series. This allows for the recharging and collection of particles exiting
the upstream stages and reducing the effects of reentrainment which has
limited the effectiveness of two-stage systems in the past. The multi-stage
ESP concept offers a solution for the problems involved with collecting high-
resistivity particles.
The problems produced by high-resistivity particles in an ESP have been
the subject of numerous research programs. When the electrical resistivity of
flyash is greater than 10 ohm-cm, the collected dust layer on the plate
resists the flow of electrical charges to the grounded plates. This leads to
a buildup of very high electric fields within the dust layer, resulting in a
breakdown of the dust layer producing a phenomenon known as back ionization.
Back ionization leads to lower electrical operating voltages, poor particle
charging, and subsequently greatly reduced particle collection.
Ibis paper describes the field evaluation of the Temperature-controlled
Electrode Precharger (Rinard and Durham, 1984) that was developed as part of
an EPA funded program to provide cost effective particulate control technology
for sources of high-resistivity dusts. Results of laboratory tests and small
scale field tests indicated that this concept warranted further evaluation. A
near full-scale evaluation of the Gooled-Electrode Precharger was performed at
the 7.1 m /s (15,000 acfm) EPA/DRI Valmont ESP Test Facility. The results of
this evaluation are presented here to demonstrate the potential of this
concept. It has the capability to greatly reduce the size of an ESP required
for high-resistivity dusts and, because of its small size, it can easily be
retrofit into existing ESPs.
BACKGROUND
TEMPERATURE-CONTROLLED ELECTRODES
The use of temperature-controlled anodes is a concept that was first
investigated in the 50's by Penney (1962) and White (1974). With this
technique, back ionization is controlled by taking advantage of the relation-
ship between resistivity and temperature. The resistivity of flyash peaks at
about 150C (300F) and decreases at higher and lower temperatures. There-
fore, the resistivity can be reduced and back ionization controlled by heating
or cooling the flyash on the collector electrodes.
It should be noted that the purpose of cooling the anode is not to cool
the gas stream and reduce the resistivity of the particles entrained in the
gas, but to cool only the dust collected on the surface of the anode. It is
only the dust on the anodes that causes the detrimental effects in an ESP.
Therefore, by cooling the anodes, the resistivity of the collected dust is
reduced which allows corona current to flow freely through the dust layer,
eliminating the effects of back ionization.
16-2

-------
Water-cooled electrodes have recently been successfully demonstrated in
Japan (Drehmel, et al., 1979), where the entire collector plate is cooled,
itie collected dust is then removed from the collection surfaces by means of
mechanical scrapers. The DRI Cooled-Electrode Precharger also uses water to
cool the surface of the anode, but applies the concept in a precharger config-
uration. This wire-pipe precharger configuration has several advantages.
In the first place it minimizes the surface area to be cooled. This not
only reduces the total amount of equipment needed to cool and circulate the
water, but it minimizes the effect of the cooled anode on the gas stream
temperature. With this precharger design, it is possible to reduce the
temperature of the dust layer by 56-83C (100-150F) and yet only decrease the
mean gas temperature by 2.7 C (5F). In addition, the precharger requires
only about 0.3 m (1 ft) of space for installation, which provides a design
that could be easily retrofit into an existing ESP. Also the wire-pipe pre-
charger design can be rapped by conventional means and does not require a
scraper system. As an added advantage, the wire-cylinder geometry produces
higher field strengths at the surface of the cylinder than the wire-plate
geometry which enhances particle collection.
The initial tests with temperature-controlled electrodes, using heated
electrodes, showed that back ionization could indeed be eliminated by
thermally controlling the resistivity on a small collector surface area
(Rinard, et al., 1981a; Durham, et al., 1982). Cooled electrodes were then
tested because cooling would be much more energy efficient. In addition, the
cooled electrodes would operate in a region of surface-controlled resistivity,
rather than bulk resistivity; therefore, the electrical characteristics would
be controlled more by the components of the gas stream than by the elemental
material in the flyash. This would prevent the deterioration in performance
that occurs in hot-side ESPs (Gooch, 1979). Besides these advantages, the
cooled electrodes operated at a higher field strength due to the higher gas
density near the surface of the anode.
Laboratory experiments showed that, in addition to producing very high
particle charge levels, a single 6 cm (2-3/8 in.) cooled-pipe precharger
operating at an electric field strength of greater than 6 kv/cm had a collec-
tion efficiency on the order of 50% (Durham, et al., 1982). To validate the
laboratory experiments, a small test device was built to evaluate the concept
on a power plant flue gas stream. This device demonstrated at two power
plants that back ionization could be controlled by cooling the anode. The
next step in the evaluation was the evaluation of the precharger in a two-
stage configuration at the EPA/DRI Valmont ESP Test Facility.
THE EPA/DRI ELECTROSTATIC PRECIPITATOR TEST FACILITY
The tests were performed at a pilot ESP test facility located at the
Public Service Company of Colorado Valmont Station in Boulder, Colorado
(Rinard, et al., 1981b). The test facility was designed for use in evaluating
novel ESP concepts on an actual source of high-resistivity flyash. The
facility operates on a slipstream of approximately 7.1 m /s (15,000 acfm) of
stack gas from the 180 MW boiler burning pulverized low sulfur Western coal,
"fris particular location was chosen because the high-resistivity ash produced
by the boiler is so difficult to collect that the existing plant ESP was just
replaced by a new baghouse.
16-3

-------
The ESP itself was designed specifically for testing a number of novel
concepts. It is constructed with alternate sections that can be used for
charging and collecting the dust. Thus it is ideal for evaluating two-stage
ESP concepts; however, the design is such that other concepts, including
conventional ESPs, can be tested. The ESP has doors on the collector sections
that can be opened to allow easy removal of internal components. Racks are
provided so that these components may be easily rolled in and out of the ESP.
The test unit was built to a scale so that its performance would be
representative of a full scale unit.
Besides the test unit, the facility includes the following auxiliary
equipment: an air-cooled heat exchanger for controlling the gas temperature
to the unit; a fan, valving, and venturi for measurement and control of the
flow rate of the flue gas; an instrumentation trailer housing a computerized
data processing system and high voltage controls; a gas sampling trailer
containing an automated gas sampling and analysis system; and a sampling house
providing access to the inlet and outlet ducts for analysis of flue gas volume
and temperature, mass concentration, particle size distribution, flyash
resistivity, particle charge, opacity, gas composition (H2O, O2, N2, CC>2, CO,
S02* N0X, and SO3), and acid dew point.
TEMPERATURE-CONTROT .T .ED ELECTRODE PRECHARGER
Figure 1 is a schematic of the Temperature-controlled Electrode Pre-
charger design used in the pilot scale evaluation. As can be seen, it is a
very simple system consisting of a manifold of welded 6 cm (2.375 in.) pipe,
with water entering the bottom header, rising through the vertical pipes, and
exiting the ESP through the upper header. Corona is produced on wires which
are suspended between adjacent pipes by a hanger system. Rapping is per-
formed by means of a single-impact pneumatic rapper connected to the bottom
header.
Cooling water is circulated through the prechargers in a closed loop
system. The circulation system consists of the following components: a pump
with a bypass to control flow; a water-to-water plate heat exchanger cooled by
plant water; and an automatic valve on a bypass of the heat exchanger to
control the inlet temperature to the prechargers. Each precharger has a flow
meter, thermocouples, and valves to monitor and control the flow through the
manifold. The closed loop system is used because it provides an easy means of
controlling the water quality within the pipes. Since corrosion or fouling on
the inside of the pipes would hamper heat transfer, water treatment oould be
provided in the closed loop. This is probably the design that would be used
in a full scale ESP. It is estimated that the cooling water required from the
plant for a full scale ESP with four prechargers added would be below 1% of
the total plant cooling water flow rate, and therefore would not appreciably
impact the cooling tower heat rejection capacity.
Four cooled-pipe manifolds similar to the one shown in Figure 1 were
constructed for the ESP at Valmont. The vertical pipes were 6 cm (2.375 in.)
in diameter and spaced 22.9 cm (9 in.) center to center which made six gas
passages and a total width of 1.43 m (4.70 ft). The height of the vertical
pipes was 3.66 m (12 ft) and the,cooled area of collecting surface in each
precharger was 4.16 m^ (44.8 ft ). The 3.2 mm (0.125 in.) diameter corona
wires were 3.5 m (11.5 ft) long and centered in each of the six gas passages.
16-4

-------
HIGH VOLTAGE INSULATOR
TRUCK
TROLLEY TRACK
RUBBER'
BUSHING
WATER
OUTLET
6.3cmSCH. 40PIPE
5.1 cmSCH. 40 PIPE
PIPE CAP
WATER
INLET
FIGURE 1. SCHEMATIC OF TEMPERATURE-CONTROLLED ELECTRODE PRECHARGER USED IN PILOT SCALE EVALUATION

-------
In each precharger, the total corona wire length was 21.0 m (69 ft). Hie
minimum corona-wire-to-cooled-pipe distance was 8.4 cm (3.31 in.) and the
minimum distance from pipe to pipe was 16.8 cm (6.625 in.). Hie gas velocity
at the center line of the pipes is 1.36 times the velocity of the average flow
through the ESP. Two of the pipe manifolds were manufactured of stainless
steel, while the other two were made of carbon steel to determine the effect
of pipe material on corrosion and performance.
Four Cooled-Electrode Prechargers were installed in the first four
precharging sections of the ESP and were each followed by parallel plate
collectors. Grounded plates, 1.83 m (6 ft) x 3.67 m (12 ft), were spaced 22.9
cm (9 in.) apart forming six passages. The high voltage electrodes also were
plates of the same size placed midway between the grounded plates and parallel
to them. The parallel plate collectors were selected for the tests because
they avoid any problems with the high-resistivity ash by operating at an
electric field strength high enough to collect the precharged particles, but
lower that that required to produce corona current. The prechargers and
collectors were alternated so that particles that were reentrained could be
recharged and then collected again.
RESULTS
COAL, FLYASH, AND FLUE GAS ANALYSES
During the evaluation of the Cooled-Electrode Prechargers at Valmont,
coal and ash samples were collected daily and then combined and analyzed on a
weekly basis. Oxygen and carbon dioxide were measured daily using a Fyrite
analysis. Sulfur dioxide was measured continuously using a Thermo Electron
Co. pulsed fluorescence unit. Water vapor concentration was determined from
the condenser catches obtained during the mass concentration measurements.
The resistivity of the flyash was measured in-situ using the DRI Field
Resistivity Unit. The average of all the analyses during this test period are
presented in Table 1.
TABLE 1
SUMMARY CF COAL, FLYASH, AND FLUE GAS ANALYSES
Coal Analysis
H20	10.8%
Ash	8.3%
Volatile	34.1%
Fixed Carbon	46.8%
Sulfur	0.6%
Heat Content	11299 Btu/lb
Ash Analysis
Gas Analysis
Si02
AIoOt
Tift
Fe2^3
CaO
MgO
Na^O
k2o
p2o5
SO-j
61.44%
24.36%
0.66%
3.72%
4.43%
0.90%
0.94%
1.51%
0.83%
0.22%
2
COo
h2o
so2
10.3%
10.0%
6.8%
150 ppm
Resistivity @ 350F
4.2 x 1012 ohm-cm
16-6

-------
ELECTRICAL CHARACTERISTICS AND PARTICLE CHARGE MEASUREMENTS
Electrical characteristics and particle charge were measured at
temperatures of 38, 43, and 56C (101, 110, and 132F) to determine the
operating range of the Cooled-Electrode Precharger. The V-I curves for 38C
and 43C were essentially the same except sparking occurred at a lower current
level for the 43C temperature. This shows that the resistivity of the dust
on the surface of the cooled electrodes increased as the cooling water
temperature increased. For a cooling water temperature of 56C, sparking
occurred at a current level of only 2 mA, which shows a further increase in
dust resistivity. However, the resistivity was not high enough to produce
stable back corona. The electrical characteristics were essentially the same
for all four prechargers, except that the first precharger had lower currents
than the three downstream prechargers. Since the first precharger experiences
the higher particle loading, this indicates a space charge corona suppression.
This was confirmed by measurements using the second precharger with and
without the first precharger energized.
Charge measurements made using a screen probe showed that the maximum
charging level occurred at the lowest water temperature, although there was
not much difference between the results at 38C and 43 C. The measurements
show that, at low corona current levels, the charge levels were the same for
the three temperatures, which indicates that back corona was not established.
In all three cases, the charge level increased as the current increased until
light sparking occurred. Any attempt to increase voltage and current further
resulted in heavier sparking and reduced charge levels.
Charge measurements were made using the DRI In-Stack Charge/Mass (Q/M)
Probe at an ESP temperature of 150C (300"F) and a gas volume flow rate of 7.1
m3/sec (15,000 acfm). The overall measured Q/M ratio was 4.1 uC/g at a
current level of 2mA at 38 kV. At a current level of 5 mA at 44 kV, the Q/M
ratio increased to an average of 8.6 uC/g. However, the average Q/M ratio was
only 5.1 uC/g at a current level of 7 mA. The lower Q/M ratio at the higher
current indicates that the precharger was operating in a heavier sparking
mode. This is also indicated by the variation from 40 to 45 kV in the
recorded voltages.
Additional charge measurements were made at various locations within the
ESP using an electrically isolated impactor stage to determine the charge on
particles 1.1 to 2.6 um in diameter. The ESP temperature was 150C (300F)
and in one series of measurements the gas flow rate was 4.7 m /s (10,000 acfm)
providing a gas velocity of 0.94 m/s (3.1 ft/s). In the other series, the
flow rate was 7.1 nT/s (15,000 acfm) and the gas velocity was 1.4 m/s (4.6
ft/s). A summary of the results, presented in Table 2, shows that the
particles were charged by the inlet precharger and recharged by the following
prechargers. The charging at both gas velocities appears to be about the
same. As the particles exit a collector section, the Q/M ratio for the 1.1 to
2.6 um particles was about 8 uC/g or about 70% of the charge level out of the
previous prechargers. The test at the higher flow rate after the first
collector was an exception, with a Q/M ratio at the outlet of only 2.7 uC/g.
The theoretical Q/M ratio for 1.1 to 2.6 um diameter particles was
calculated using 43.5 kV, which produces an average field strength of 5.2
kV/cm across the charging region. The current density on the pipes is 120
nA/cm for 5 mA corona current. The treatment time is about 0.05 s for a
16-7

-------
volume flow rate of 7.1 m^/s (15000 acfm). The calculated theoretical Q/M
ratio was 14 uC/g compared to the measured value of about 12 uC/g. The
cooled-pipe precharger charged the particles to 85 to 90% of the theoretical
unipolar ion charging level. This indicates that the prechargers are charging
with unipolar current.
TABLE 2
SUMMARY OF RESULTS OF CHARGE MEASUREMENTS
Probe Type: MRI impactor with isolated 4th stage.
Stack Temperature: 150C (300F)
Flow, m3/s (acfm) 7.1 (15,000) 4.7 (10,000)
Q/M	Q/M
(uC/g)	(uC/g)
Precharger 1 14.38	9.11
Collector 1 2.68	8.35
Precharger 2 12.97	14.08
Collector 2 8.22	8.42
Precharger 3 17.71	20.67
COLLECTION EFFICIENCY
During most tests, the current in each precharger-was 7 mA and the
current density on the cooled-pipe surface was 160 nA/cm . The voltage for
the first precharger was about 45 kV for a field strength of 5.3 kV/cm, and 39
kv or 4.6 kV/cm for the other three prechargers. The parallel plate
collectors operated at about 33 kV, and the field strength was 2.9 kV/cm.
The water temperature at the inlet of the prechargers was about 39C
(102F). The average flow rate through each manifold was 44 1/min (11.6
gal./min) and the average temperature increase of the water was 2.7C (4.8F).
Using this data the heat removed from the flue gas was calculated to be
500,000 J/min (465 Btu/min) which, at a gas velocity of 1.4 m/s (4.6 ft/s),
would lower the gas temperature 1.7C (3F) through each precharger.
The measured average temperature drop from the ESP inlet to outlet was
18C (33F) with four cooled-pipe prechargers and 10C (18F) without the
prechargers. Therefore, the gas was cooled about 2C (2.6F) by each
precharger, which is in close agreement with the temperature loss of 1.7C
(3F) calculated from the cooling water flow rate and temperature increase.
Table 3 shows a summary of the results of the performance of the
Temperature-controlled Electrode Prechargers. All of the results in Table 3
represent averages of three or four runs. In Test 1 only one precharger and
one collector section were energized. The penetration was 12.4% for an SCA of
12 s/m (60 ft /1000 acfm). In Test 2, another precharger and collector
section were energized. For an SCA of 23 s/m (115 ftV1000 acfm), the
penetration was 2.9%. The results of Tests 1 and 2 were about the values
expected considering the particle charging levels being achieved.
16-8

-------
TABLE 3
TEST RESULTS WITH THE COOLED-ELECTRODE PRECHARGERS
WITH PARALLEL PLATE COLLECTORS
Test No.
1
2
3
4
2A
3A
4A
2B
4B
Precharqers
1
2
2
4
2
2
4
2
4
Collectors
1
2
4
4
2
4
4
2
4
Baffles
no
no
no
no
yes
yes
yes
yes
yes
Velocity
(m/s)
1.4
1.4
1.4
1.4
1.4
1.4
1.4
0.7
0.9
SCA
(s/m)
12
22
46
46
23
46
46
46
67
Inlet
(q/dscm)
4.6
4.7
4.8
5.4
3.8
4.1
4.7
3.6
4.3
Outlet
(g/dscm)
0.57
0.13
0.11
0.10
0.12
0.085
0.055
0.023
0.018
Penetration
(%)
12.4
2.9
2.5
1.9
3.2
2.5
1.2
0.6
0.5
Efficiency
(%)
87.6
97.1
97.5
98.1
96.8
97.5
98.8
99.4
99.5
In Test 3, the first and third prechargers were energized as in Test 2
and all four collector sections were energized, increasing the SCA to 46 s/m
(235 ft /1000 acfm). Hie penetration was 2.5% which was about the same value
as in Test 2. This showed that additional parallel plate collector sections
following a precharger do not improve collection. This effect had been
encountered in previous precharger tests using parallel plate collectors.
The test results are plotted in Figure 2. The two-precharger ESP line,
determined by Tests 2 and 3, shows the effect of increasing SCA in an ESP
containing two prechargers. Hie one-precharger ESP line shows the projected
effect of increasing SCA in an ESP with one precharger at the inlet. Curve A
shows the decrease in penetration when the SCA was increased by adding a
precharger with each additional parallel-plate collector section. Curve B
shows the penetration decrease when a precharger was added for every two
collector sections.
In Test 4 all four prechargers and collector sections were energized, in
comparison to Test 3, the two additional prechargers reduced the penetration
from 2.5% to 1.9%. Considering the results of Tests 1 and 2, the penetration
in Test 4 was considerably higher than expected. This is apparent from the
abrupt break in the slope of Curve A of Figure 2.
16-9

-------
CURVE A - ONE COLLECTOR PER PRECHARGER
CURVE B- TWO COLLECTORS PER PRECHARGER
VELOCITY - 1.4 m/s (4.6ft/sec)
VELOCITY-0.94 m/s (3.1 ft/sec )
VEL0CITY-0.70 m/s (2.31t/see)
0.5
TEST 1
TEST 2a
TEST 2 jLfftECHAftGFP
TEST 3a
TEST 4
 TEST 4a
0.01
TEST 2b
TEST 4b
0.005
NUMBER OF COLLECTOR SECTIONS AT 7.1 m3A( I5000ACFM)
0.003
22.7
(115.2)
34.0
(172.8)
45.4
(230.4)
57.6
(288)
11.3
(57.6)
68.0
(346)
SCA, s/m ( ff2/1000 ACFM)
FIGURE 2. EFFICIENCIES OF VARIOUS ESP CONFIGURATIONS
16-10

-------
One possible reason for the lack of improvement in performance was
leakage around the prechargers. Baffles were added to prevent gas flow
between the wall of the ESP and the outer pipes of the oooled-pipe manifold.
Tests 2a, 3a, and 4a were repeats of Tests 2, 3, and 4, respectively, except
for the added baffles. The results for Tests 2a and 3a were essentially the
same as for Tests 2 and 3, showing that sneakage was not a problem at the
penetration levels for these ESP configurations. Although the penetration in
Test 4a was 1.2%, which was lower than in Test 4, it was still higher than
expected based upon the projection of Tests 1 and 2 results shown in Figure 2.
Reentrainment from the parallel-plate-collector sections appeared to be
the most probable factor that was limiting performance in Tests 4 and 4a.
Since reentrainment is a function of gas velocity, two tests were performed at
reduced gas velocities to determine if penetration would decrease as velocity
decreased. In Test 4b, all prechargers and collectors were energized. The
gas velocity was reduced to 0.94 m/s (3.0 ft/s) and the SCA was 67 s/m (350
ft2/1000 acfm). The penetration was 0,005 which was considerably lower than
the level that could have been achieved by adding two additional parallel-
plate collector sections and keeping the velocity at 1.4 m/s (4.6 ft/s). The
velocity was reduced to 0.7 m/s (2.2 ft/s) during Test 2b and only two
prechargers and two collectors were energized, so the SCA was 46 s/m (235
ftvlOOO acfm) which was essentially the same as in Tests 3, 4, and 4a. The
penetration was measured to be 0.006 and is plotted in Figure 2 for comparison
with the other test results. It is apparent that penetration decreased as
velocity decreased, which indicated that performance was being limited by
reentrainment from the parallel-plate collector sections.
The best example of the effects of velocity on reentrainment can be
demonstrated by combaring Tests 2b and 3. Both configurations contain two
prechargers and the identical plate area per volume of flow. The only
difference is that in Test 2b, the velocity is reduced. Data in Table 2 shows
that the reduced velocity does not significantly affect charging, so any
differences in the results would be due to the effect of velocity on
reentrainment from the plates. As can be seen from Table 3, the lower
velocity resulted in a decrease in emissions by a factor of 4.
DISCUSSION
The test results provide a considerable amount of information on two-
stage ESPs. The V-I curves for the cooled-pipe prechargers showed that back
corona was eliminated when the cooling water temperature into the oooled-pipe
manifold was 43C (110F) or less. The temperature of the water was increased
about 2.7C (4.8F), and the temperature decrease of the flue gas was about
2C (3.6F). Without back corona, field strengths of 5 kV/cm could be
maintained in the charging region ot0 the precharger, and the current density
on the cooled surfaces was 160 nA/cm. Q/M ratio measurements for particles
in the 1 to 2 um diameter size range showed that the particles were charged to
about 85% or 90% of the theoretical unipolar ion charging level. Again, this
showed the effectiveness of the cooled-pipe precharger in eliminating back
corona when charging dust which had resistivity of 4 x 10 ohm-cm.
Performance measurements on several two-stage ESP configurations revealed
the limitations of the parallel-plate collector sections. The two-precharger
and two-collector configuration had a penetration of about 3%. For the four-
16-11

-------
precharger and four-collector configuration the penetration was reduced to
1.9% which showed that the additional prechargers and collectors had little
effect upon performance.
The problem appeared to be caused by some non-ideal effect such as
rapping or non-rapping reentrainment or sneakage. The frequency of rapping
prechargers and collectors was varied and the resulting change in penetration
was not measureable. This indicated that rapping reentrainment was not the
limiting factor. Baffles were added to eliminate sneakage around the cooled-
pipe prechargers. The penetration was reduced to 1.2% for the four-precharger
and four-collector configurations but was unchanged in the other
configurations tested. This indicated that sneakage was not the main factor
that was limiting performance.
Non-rapping reentrainment from the parallel-plate collector sections
appeared to be the main factor limiting performance. This was indicated by
Q/M ratio measurements which showed that particles lost charge as they passed
through a collector section. Cascade impactor measurements which showed a
considerable amount of large particles at the ESP outlet also indicated
reentrainment. Since non-rapping reentrainment is dependent upon gas
velocity, performance tests were made at reduced flow rates which verified
that performance was limited by reentrainment.
In the two-precharger and two-collector configuration* the velocity was
reduced by a factor of 2 which made the SCA 46 s/m (235 ft /1000 acfm). The
penetration was 0.6% compared to 1.2% measured in the four-precharger and
four-collector configuration at just over twice the velocity and approximately
the same SCA. Again, this indicated that non-rapping reentrainment was the
main factor limiting performance.
The non-rapping reentrainment was not entirely unexpected. Without
current in the collector sections, there is no force holding the particles
onto the plate, other than adhesion. This is similar to the situation when a
conventional ESP is operating on a dust with very low resistivity, such as the
case of many stoker boilers with high carbon carryover. This problem of non-
rapping reentrainment can be solved by designing the ESP to operate at a
velocity below 1 m/s (3 ft/s).
CONCLUSIONS
Although the performance was limited by non-rapping reentrainment from
the parallel-plate collectors, the test results showed that two-stage
precipitators can meet the 43 ng/J (0.1 lb/10 Btu) ^emission level or the New
Source Performance Standard of 13 ng/J (0.03 lb/10 Btu). The ooal analysis
showed that a penetration of 1.9% would meet the 43 ng/J (0.1 lb/10 Btu)
emission level, and 0.6% penetration would meet New Source Performance
Standard. The four-precharger and four-collector configuration with an SCA of
45 s/m (230 ft Z1000 acfm) at a gas velocity of 1.4 m/s (4.6 ft/s) met the 43
ng/J (0.1 lb/10" Btu) requirement. The twp-precharger and two-collector
configuration with an SCA of 45 s/m (230 ft /O-000 acfm) at a gas velocity of
0.7 m/s (2.3 ft/s) met the 13 ng/J (0.03 lb/106 Btu) requirements.
16-12

-------
Lower penetration with an SCA of 45 s/m (230 ft^/1000 acfm) at a gas
velocity of 1.4 m/s (4.6 ft/s) could be achieved if reentrainment from the
collector sections is reduced. It is expected that this can be provided by
interfacing the Cooled-Electrode Precharger with a conventional wire-plate
ESP. The wire-plate sections would be operated at current levels below that
required to produce back ionization but sufficient to hold the collected dust
on the plate. If this is not possible with dc energization, then pulsed
energization could be applied.
The Temperature-controlled Electrode Precharger has potential as a retro-
fit device on an existing ESP that is not performing as designed or at an
installation that has undergone a process change that resulted in higher
resistivity particles than the ESP was designed for. Examples of applications
of the latter case would include changes resulting from acid rain mitigation
strategies such as coal switching, coal washing, Limestone Injection with
Multistage Burners (LIMB), and dry injection. The precharger can be installed
in less than 0.3 m (1 ft) of space and provides high collection efficiencies
even for extremely high resistivity dust. An economic analysis has shown that
it can be installed at half the cost of an SO3 conditioning system. The
precharger will be tested in the near future at the Valmont Test Facility
using the retrofit configuration.
REFERENCES
Drehmel, D.C., C.H. Gooding, and G.B. Nichols (1979). "Particulate Control
Highlights: Recent Developments in Japan," EPA-600/8-79-031a (NTIS PB 80-
148802), November.
Durham, M.D., G.A. Rinard, D.E. Rugg, and L.E. Sparks (1982). "The
Development of a Charging/Collecting Device for High Resistivity Dust Using
Cooled Electrodes;" J. Air Pollution Control Association, Vol. 32, No. 11,
pp 1132-1136, November.
Gooch, J.P. (1979). "Electrostatic Precipitator Performance," In: Symposium
on the Transfer and Utilization of Particulate Control Technology, Volume 1,
EPA-600/7-79-044a (NTIS PB 295226), pp 1-17, February.
Penney, G.W. (1962). "Industrial Precipitator with Temperature-controlled
Electrodes," U.S. Patent 3,026,964, March 27.
Rinard, G.A., M.D. Durham, D.E. Rugg, and L.E. Sparks (1981a). "The Use of
Heated and Cooled Electrodes for Charging and Collecting High-Resistivity
Dust;" 74th Annual Meeting of the Air Pollution Control Association, Paper
81-63.3; Philadelphia, PA; June 21-26.
Rinard, G.A., M.D. Durham, J.A. Armstrong, D.E. Rugg, and L.E. Sparks (1981b).
"The EPA/DRI Valmont Electrostatic Precipitator Test Facility;" J. Air
Pollution Control Association, Vol. 31, No. 8, pp 908-911; August.
Rinard, G.A. and M.D. Durham (1984). "Electrostatic Precipitator using a
Temperature Controlled Electrode Collector," U.S. Patent 4,431,434, February
14.
White, H.J. (1974). "Resistivity Problems in Electrostatic Precipitation,"
Air Pollution Control Association, Vol. 24, No. 4, April.
16-13

-------
OPTIMIZING THE COLLECTOR SECTIONS OF MULTI-STAGE ELECTROSTATIC PRECIPITATORS
George Rinard, Michael Durham, and Donald Rugg
Denver Research Institute
Denver, Colorado 80208
Leslie Sparks
Industrial Environmental Research Laboratory
Environmental Protection Agency
Research Triangle Park, North Carolina 27711
ABSTRACT
A companion paper, "Field Study of Multi-Stage Electrostatic Precipita-
tors," describes tests of parallel plate-plate collectors in multi-stage
ESPs. The tests described in that paper showed that with this configuration
the gas velocity must be reduced to avoid excessive non-rapping reentrain-
ment. The present paper describes further tests that were performed at EPA
laboratories in Research Triangle Park, North Carolina, utilizing
conventional wire plate collectors in a multi-stage ESP. The ESP had cooled
pipe chargers and conventional collectors in alternating sections. Both
conventional and pulsed excitation were performed utilizing 0.32 and 0.95 cm
(1/8- and 3/8-in.) diameter rod electrodes.
The results indicate that good collection efficiencies are possible in a
multi-stage ESP, with high resistivity dust using conventional wire plate
collectors. A particularly important application is for retrofitting an
existing ESP on power plants utilizing coal switching, dry scrubbers, or the
LIMB process to reduce sulfur emissions. In these cases the higher grain
loading and higher dust resistivity produced by the sorbent material may
cause a conventional ESP to no longer meet compliance. The use of the multi-
stage technology can potentially improve the cost effectiveness of these
strategies for mitigating acid precipitation caused by power plant sulfur
emissions.
This paper has been reviewed in accordance with the U. S. Environmental
Protection Agency's peer and adminstrative review policies and approved for
presentation and publication.
17-1

-------
INTRODUCTION
A number of electrostatic precipitators (ESPs), on electric power
plants, are on the borderline of non-compliance. Often, because of these
ESPs, plant operators are required to reduce load to maintain particulate
emission standards. There are number of different causes for these poor
performing ESPs. In the case of plants firing low sulfur coals, the predomi-
nate problem is high resistivity fly ash which produces back ionization and
greatly reduces ESP performance. In the cases where high sulfur coals are
utilized, ESPs may have problems which are due to undersizing, poor gas
distribution, high gas velocity, or other detrimental causes. The Tempera-
ture Controlled Electrode Precharger (TCEP) has been investigated as one
means of solving these problems. Tests on this device were performed at the
EPA/DRI Valmont ESP Test Facility in Boulder, Golorado.
The TCEP is attractive because of its simplicity and ease with which it
may be retrofit into an existing ESP. The major component of the device is a
manifold of 6.1 cm (2 in.) pipes with inner-spaced corona electrodes. The
precharger is installed upstream of one or more electrical sections of the
ESP with the pipe manifold normal to the gas flow. It requires as little as
30.5 cm (1 ft) of space between eletrical sections for installation. Coolant
may be passed through the pipes to cool the surface of the collecting elec-
trodes to a temperature sufficiently low to prevent back corona. All the
coolant is contained in the pipes; therefore, there is no mixing of the
coolant and fly ash. The temperature of the pipe is always above the dew
point so the fly ash remains dry. The precharger produces a high electrosta-
tic field and high current density which charges the particles to a level
which is higher than is possible in a conventional ESP. The result of the
higher charge level is an increased migration velocity in the ESP with the
net effect of increasing the effective specific collection area (SCA).
This device has been tested as the precharger in a two-stage ESP (Durham
et al., 1984). In these tests, the precharger was installed upstream of
parallel plate-plate collectors. Hie results of these tests showed that a
higher collection efficiency could be obtained with this SCA than could
normally be achieved with the 1012 ohm-cm resistivity fly ash that was being
collected. However, there is an upper limit to the efficiency obtainable
with even higher SCAs and higher efficiencies could be obtained only by
lowering the gas velocity.
To find a more optimum collector, which would be effective at
conventional gas velocities, the TCEP was tested as a retrofit device for a
conventional wire plate ESP. These tests were performed in the pilot scale
ESP at the Environmental Protection Agency's Industrial Environmental
Research Laboratory, Research Triangle Park, North Carolina. In addition,
tests were performed with the wires replaced with 0.95 cm (3/8-in.) diameter
corona electrodes. Since the TCEP provides nearly all particle charging,
anything that can provide higher field strength in the collector and still
produce sufficient corona current should also improve ESP performance. This
is the role of the large diameter corona wires. This is important for high
17-2

-------
resistivity applications since in this case very little corona current is
necessary. SoRI recently reported good results with large diameter
electrodes in a two-stage ESP (Bush et al., 1984).
TEST AT A TWO-STAGE ESP AT VALMONT TEST FACILITY
The TCEP was tested as the precharger in a two-stage ESP at the Valmont
Test Facility in Boulder, Colorado (Durham, et al.,1984). In the collectors
of the ESP, the corona electrodes were replaced with flat plates resulting in
parallel plate-plate collectors. This configuration was used since it was
desired to produce a high electric field and to completely eliminate corona
current in the collector sections. The results of these tests are given in
Figure 1 which shows the penetration (1-efficiency) plotted on a logarithmic
scale as a function of the SCA. The numbers next to each data point repre-
sent, respectively, the number of charger sections, the number of collector
sections, and the total gas flow in cubic feet per minute. The penetration
was reduced significantly in going from one charger and one collector to two
chargers and two collectors at 7 m3/sec (15,000 acfm). The projected
efficiency from these two data points would be greater than 99% when using
four chargers and four collectors. However, as can be seen, the penetration
was not significantly reduced when going to four chargers and four collectors
at 7 m /sec (15,000 acfm). Further testing showed that, if the gas flow was
reduced to 3.5 mJ/sec (7,500 acfm), the penetration was significantly re-
duced, and 99+% efficiency was obtained. The gas velocity for the reduced
flow was 0.7 m/sec (2.3 ft/sec).
The results shown in Figure 1 and further testing verified that the
primary cause of the inability of the two-stage ESP to obtain high collection
efficiencies at the higher gas flow was caused by non-rapping reentrainment
of the ash collected on the plates. In the case of the flat plate-plate
collector, reentrainment is caused by insufficient corona current to provide
the necessary clamping forces on the collected ash layer.
The testing at Valmont on the two-stage ESP showed that the TCEP
functioned well. Even with the flat plate-plate collectors, the two-stage
ESP was able to obtain 99.4% collection efficiency with an SCA of 45.5 sec/m
(23(1 ft /1000 acfm) when collecting fly ash with an electrical resistivity of
10 ohm-cm. Sufficient base line data were obtained on the TCEP to show
that it charged and collected the high resistivity fly ash as expected.
However, further work was needed to find a collector that would work
efficiently at the higher velocities normally found in conventional ESPs.
TEST ON TCEP AS A RETROFIT FOR A CONVENTIONAL ESP
The tests were performed in EPA's Industrial Environmental Research
Laboratory (IERL), Research Triangle Park, North Carolina. The ESP has four
electrical sections of wire plate collectors each 4 ft long with 0.32 cm
(1/8-in.) wires spaced 22.9 cm (9 in.) apart. The wires were removed from
the first and third sections of the ESP. The corona support was used in
these sections to power TCEP sections which were installed in the spaces
between the first and second sections and the third
17-3

-------
No. Charge Sections
No. Col lector Sections
Totol gos flow rote, m3/sec(ocfm)
1,1-7(15,000)
2-7{ 15000)
\ \ 2,2-3.5U 500)
PROJECTED
x 4,4-7(15,000)
. 2,4-7(15000)
4,4-7(15,000)
150 , 200 290
SCA.f t / 1000 ocfm	
~i	r
23.7 39.6
SCA, sec/m
9.9
19.8
49.9
9&4
69.3
FIGURE X. PENETRATION MEASURED IN VALMONT TESTS
17-4

-------
and fourth sections. In the second series of tests, the corona electrodes
were replaced with 0.95 cm (3/8-in.) rods. The test conditions and ESP
configuration are summarized in Table 1. The TCEP installation is shown in
Figure 2. This figure shows the ease with which the TCEP is installed
between sections of the ESP.
TABLE 1. TCEP TESTS WITH CONVENTIONAL ESP AT EPA LABORATORY
TEST CONDITIONS
Temperature	148.9C (300F)
Gas Flow	0.47 mVsec (1000 acfm)
Moisture	4% volume
Ash Resistivity	9 x 10^ ohm-cm
ESP CONFIGURATION
Channels of Flow	One
Plate Spacing	25.4 cm (10 in.)
Wire Spacing	22.9 cm ( 9 in.)
Wire Size	0.32 cm (1/8-in.)
0.95 cm (3/8-in.)
Gas Velocity	1.5 m/sec (5 ft/sec)
Number of TCEPs	Two
ESP SCA	12.7 sec/m (64 ft2/1000 acfm)
Hie ESP operating conditions and penetration measurements are shown in
Table 2. As can be seen from this table, with very low current (-10/i A) in
the wire plate sections of the ESP, the collection efficiency was approxi-
mately 96%. For the direct current tests with an average current of 610fi A,
the average efficiency increased to 97%. When the two collector sections
were pulsed (750 fiA average current), the average collection efficiency
increased to 98%. When the prechargers were both turned off and the ESP
operated under pulse conditions, the collection efficiency dropped to 91%.
The results with 0.95 cm (3/8-in.) corona electrodes show that even higher
efficiencies are possible because of the resulting higher field produced by
these electrodes.
17-5

-------
FIGURE 2. TCEP INSTALLATION AT EPA LABORATORY
17-6

-------
TABLE 2. RESULTS OF TCEP AND CONVENTIONAL ESP TESTS AT EPA LABORATORY
Tests with 0.32 cm (1/8-in.) Discharge Electrodes
in the Collector Sections
Direct Current	Pulsed
No. Tests
Ave. Collector Voltage, kV
Ave. Collector Current,^A
Penetration
Efficiency, %
Penetration with TCEP off
Efficiency with TCEP off, %
5	9	9
32	35	32
-10	610	750
0.042	0.031	0.017
95.8	96.9	98.3
		0.093
		90.7
Tests with 0.95 cm (3/8-in.) Discharge Electrodes
in the Collector Sections
No. Tests	4	4
Ave. Collector Voltage, kV	32	38
Ave. Collector Current, fiA	-10	200
Penetration	0.016	0.008
Efficiency, %	98.4	99.2
The nature of the dust deposited on the collector plates of the last
section during high efficiency operation is of interest. This is shown in
Figure 3. The dust deposit was smooth and densely packed. As can be seen
from Figure 3, rapping removed the dust in patches. These patches evidently
separated from the plate in flake-like layers with little dust reentrainment.
It appears that one role of corona current is to produce forces that help
pack the deposited dust into a dense layer. When flat plate-plate collectors
were used, the dust was more loosely deposited on the plate and reentrainment
was more noticeable.
17-7

-------
FIGURE 3. DUST DEPOSIT IN LAST STAGE OF ESP
17-8

-------
ANALYSIS OF TEST RESULTS
The tests at Valmont, in the two-stage ESP utilizing flat plate-plate
collectors, indicate that with high resistivity ash very little current is
necessary to provide good clamping forces for the ash on the plates. This is
indicated by the fact that 97% efficiency was obtained with a 120 SCA
collector when utilizing two TCEPs. The clamping force in this case was
provided solely by the charge on the particles themselves. This current is
less than that found in a conventional ESP by two orders of magnitude.
However, the ash resistivity is two orders of magnitude higher than that
which is normally considered to be precipitable. Since the holding force is
a function of the product of current density and ash resistivity, this very
low current is sufficient for the high resistivity ash. However, as the dust
is collected in the ESP, the current carried by the charged particles drops
by another order of magnitude because of the lower dust concentration near
the outlet section of the ESP. In the latter stages of the ESP, dust no
longer carries sufficient current to provide the necessary clamping forces.
These results show, that if the TCEP is used as a retrofit on a conventional
wire plate ESP, a very low corona current needs to be maintained to provide
good collection. In most cases this current can be provided without
producing back corona; therefore, the TCEP followed by a conventional wire
plate ESP should be very efficient at collecting high resistivity fly ash.
The tests at EPA utilizing a TCEP as a retrofit in a conventional ESP
verify the predictions based on the Valmont test. The following calculations
were made using the modified Duetsch equation that is used to size
conventional ESPs for power plant applications:
. 1/2
Eff = 1 - e"(w VV)	(1)
where: Eff = Efficiency
w = Migration velocity, m/sec
A = Collector area, m
V = Volume of gas treated, nr/sec.
It is convenient to discuss the penetration, or amount of ash that exits the
ESP, rather than the efficiency of the amount collected. Therefore, it is
customary to define penetration as
. v 1/2
P = 1 - Eff = e"(w A/V)	(2)
17-9

-------
The 1/2 power in the exponent is used to account for different particle sizes
and other non-ideal effects. The migration velocity is proportional to the
charge on a particle and the electric field strength:
wttqE	(3)
where: w = Migration velocity, m/sec
q = Charge, C
E = Electric Field, V/m.
The predominate means of charging particles is due to a process which is also
proportional to the electric field strength:
qCCE.	(4)
In a conventional ESP, the field which produces charging is the same
field that forces the charges to the plate. When a TCEP is retrofit into an
ESP, two things are different than in a conventional ESP: first, the TCEP
acts like an ESP and collects a portion of the ash; and second, the TCEP
charges the particles which are then collected in the wire plate part of the
ESP. Because the field in the TCEP is higher than that in the wire plate
section, there is an increase in migration velocity as given by Equation (3).
The net result is enhanced collection over that possible in the conventional
wire-plate ESP. Table 3 summarizes the net effect of this enhanced collec-
tion for the operating conditions which were obtained at EPA. In this table
"Wire-plate ESP following TCEP" refers to the collector section of the two-
stage ESP in which collection is enhanced because the incoming particles are
already charged.
TABLE 3
SCA	E	W	Eff.
sec/m (ft /1000 acfm) kV/cm	cm/sec %
Unenergized ESP 12.7 64				50.0
Conventional Wire-plate ESP 12.7 64	2.76	45	90.7
TCEP Alone (2 sections) 1 5	5.16	157	71.0
Wire-plate ESP Following TCEP 12.7 64	2.76	84	96.1
Combination of TCEP and ESP 13.7 69					98.9
Conventional ESP Required 45.5 230					98.9
for Above Efficiency
The higher field in the TCEP is due to the higher operating voltage and
closer spacing of electrodes. The higher fields result in higher migration
velocities, both in the TCEP and in the rest of the ESP. With the TCEPs, the
collection efficiency of the wire-plate section is increased from 90.7% to
96.1%. The TCEPs themselves account for 71% collection. The combined
efficiency after retrofit is one minus the product of the corresponding
penetrations. The result is 98.9%. This is very close to that actually
measured, as shown in Table 2. The results are consistent with ESP theory.
The last line gives the SCA of a conventional ESP that would be required to
17-10

-------
obtain the same collection efficiency as the retrofit ESP. Thus, it can be
seen that approximately one-third as much collector plate area is required in
the retrofit ESP as would be required for the conventional wire-plate ESP
without the TCEPs.
The test results to date are for high resistivity fly ash. Similar
migration velocity enhancement can be expected for normal ash resistivity.
In this case, the TCEPs may or may not require cooling, depending on the SO3
level in the gas. Other operational problems such as poor rapping can reduce
the enhancement obtainable in practice. However, it is clear that the TCEP
offers the potential for relieving many of the problems that are presently
experienced with poorly performing ESPs.
CONCLUDING REMARKS
The temperature-controlled electrode precharger (TCEP) has been referred
to in previous articles as the Cooled Pipe Precharger or the Cooled Pipe
Charger/Collector. These names were used to emphasize the main geometric
features of the device and also to emphasize that, not only does the device
charge particles but due to the high electric fields and high current
densities, it is also a good collector. The new term TCEP was chosen because
the temperature of the collector electrodes may be either above, the same as,
or below that of the gas stream, depending on the application.
Figure 4 shows the means for retrofitting the TCEP in a conventional
ESP. The pipe manifold is mounted in the space between electrical sections.
Nearly all ESPs have sufficient space between sections for installing the
manifold. The pipes are suspended from the top of the ESP, and appropriate
high voltage feedthroughs are provided for suspending the corona electrode.
The TCEP is installed through an opening which is cut into the side wall of
the ESP so that the pipe manifold can be inserted between collector sections.
When cooling is required, coolant is pumped through the pipes. A simple
throttle valve maintains proper flow, and the coolant can be passed through
the plant's evaporative cooling tower. The added heat load on the cooler is
a fraction of 1%. No modifications need to be made on the rest of the ESP in
many applications. However, additional enhancement of the effective SCA of
the ESP may be obtained in many cases by using larger rod electrodes in the
collector section. This is particularly true when collecting high
resistivity fly ash.
REFERENCES
1984, Bush, P.V., Pontius, D.H. and Sparks, L.E., "Performance of Large-
Diameter Wires as Discharge Electrodes in Electrostatic Precipitators," The
Fifth Symposium on the Transfer and Utilization of Particulate Control
Technology, Kansas City, MD, August 27-30.
1984, Durham, M.D., Rinard, G.A., Rugg, D.E. Carney, T., Armstrong, J., and
Sparks, L.E., "Field Study of Multi-Stage Electrostatic Precipitators," The
Fifth Symposium on the Transfer and Utilization of Particulate Control
Technology, Kansas City, MD, August 27-30.
17-11

-------
FIGURE 4. PROPOSED FULL SCALE TCEP
17-12

-------
CERAMIC-MADE BOXER-CHARGER FOR PRECHARGING APPLICATIONS
Senichi Masuda, Shunsuke Hosokawa, and Shuzo Kaneko
Department of Electrical Engineering,
University of Tokyo
7-3-1, Hongo, Bunkyo-ku, Tokyo, Japan 113
ABSTRACT
A Boxer-Charger comprizing ceramic-made, high-performance, and compact-
size ionizer units has been developed for back corona free percharging of
high resistivity dust in electrostatic precipitation and electrostatic
augmentation of fablic filters. The ionizer units consist of high purity
alumina plates with tungsten corona electrodes on their surface and a
tungsten-film induction electrodes embedded inside the plates (IS-panel ; ion
source panel). A sharp pulse voltage with 200 - 500 ns duration time and -4
- -7 kV peak voltage produces very active surface streamers on the ceramic
plates as an active planelike ion source. The cost of a power supply for
ion generation is greatly reduced owing to a reduced pulse voltage required,
and yet a very satisfactory charging performance can be obtained.
The basic characteristics of ion generation and charging efficiency are
confirmed to be very satisfactory for percharging applications. Tests are
made jg a laboratory precipitator with high resistivity flyash with up to 1.4
x 10 ohm-cm resistivity. Their performance enhancement is especially
remarkable in the two stage type precipitator when the ceramic-made Boxer-
Charger is combined with the pulse-energized collection,fields where no back
corona takes place in the collection fields up to 2 x 10 ohm-cm.
INTRODUCTION
The authors developed Boxer-Charger (1-4) for back corona free
precharging of high resistivity dust, and novel ionizer panels made of high
purity alumina (IS-panel ; ion source panel) (5). These two technology are
combined to provide an improved type of Boxer-Charger. A pair of IS-panel
ionizer units is spaced at 150 mm across the charging zone, and supplied
with an a.c. low frequency voltage to produce a charging field. In
synchronous to this a.c. voltage, a pulse voltage with 200 - 500 ns duration
time and -4 - -7 kV peak voltage is alternately applied to one of these
181

-------
ionizer units between its corona and induction electrodes at an instant when
the ionizer unit takes the negative crest of the a.c. voltage. Then the
surface streamers are induced to produce an active plane-like ion source on
the surface of the ionizer unit, from which copious negative ions are
supplied. Thus, particles are charged in an a.c. field with negative ions
from both sides, alternately. No back corona takes place as the negative
charge accumulated on the dust deposit is eliminated by the streamer plasmas
produced each cycle of the a.c. voltage. The thickness of the alumina layer
between the corona and induction electrodes is only 0.5 - 1.0 mm. Hence the
required pulse voltage for streamer generation is only -4 - -7 kV compared to
-40 kV required in the double-helix type Boxer-Charger (3). Moreover, the
surface streamers have a cleaning effect to keep the active surface of the
ceramic plates free of dust contamination. As the result, a stable and
satisfactory charging performance is obtained up to the dust resistivity as
high as rd = 10 ohm-cm. A two stage type laboratory precipitator
comprizing this ceramic-made Boxer-Charger is tested, where comparison is
made between the d.c. and pulse energization in the collection fields. A
remarkable enhancement of the collection performance is obtained by the
present Boxer-Charger as far as the pulse-energized collection field is used
so that no back corona takes place. In this paper are reported these test
results.
BASIC CHARACTERISTICS OF CERAMIC-MADE BOXER-CHARGER
IS-PANEL
Three different types of ceramic made IS-Panels as shown in Fig. 1 are
used in this experiment. A high purity alumina ceramic sheet (92% purity)
with only 0.5 - 1.0 mm thickness can provide enough electrical strength
between the corona and the induction electrode even at an elevated tempera-
ture. Types I and II have fence-like corona electrodes. The capacitance
between the corona and induction elctrodes are 300 pF (Type I) and 200 pF
(Type II) in the non-corona condition. Type III has a zig-zag shaped corona
electrode with 825 mm length, forming a transmission line against the film-
like induction electrode with a surge impedance of approximately Zq = 60 ohm.
But Zq differs very widely with the discharge condition.
Fig. 2 shows the light activity of pulse-induced surface streamers of
these IS-Panels. A very uniform surface discharge occurs from the both sides
of the strip-like thin corona electrode.
CIRCUIT DIAGRAM AND PULSE VOLTAGE WAVE FORM
Fig. 3 shows the circuit diagram of the ceramic-made Boxer-Charger. An
a.c. main voltage, V , is applied between the two ionizer units, facing each
other and spaced at d = 15 cm, and pulse voltage, V , is applied alternately
to each unit between its corona and induction electrBdes at an instance 1 ms
ahead of its negative crest of V . This timing gives the best charging
performance.
Fig. 4 shows the wave forms of the pulse voltage at the input of the IS-
18-2

-------
Corona Electrode
'(50 ym high x 1 mm wide)
Induction Electrode
'(10 nm thick.)
(a) Cross Section of IS-Panel
Type I and III : a = 0.5 ram, b = 0.5 mm
Type II	: a = 1.0 mm, b = 1.0 mm
50 mm
130
mm
iC
ic
iC
IC
(b) Corona Electrode of
Type I and II
(c) Corona Electrode of
Type III
h
(a) Type I and II
(b) Type III
Fig. 2
Light Activity of
Surface Streamers
Fig. 1 IS-Panels Used
Spark Gap
T-O Or
d 15 cm
 * *
Spark Gap
Vm	: A.C. Main Voltage Applied between Two Ionizer Units
Vp	: Pulse Voltage Applied between Corona and Induction Electrodes
Cf	: Pulse Forming Capacitor
Rf	: Pulse Forming Resistor
Fig. 3 Circuit Diagram of Ceramic-Made Boxer-Charger
18-3

-------
panel and the light signal from the surface streamers detected by a photo-
multiplier. In Fig. 4(a) only a single IS-panel of Type I is used and the
pulse voltage obtained is in a form of a triangle. This light signal starts
at an instant when the pulse voltage reaches -3.6 kV. The light signal has a
long tail due to the after glow of plasma. In Fig. 4(b), 6 IS-panels of Type
I are used in series and the oscillating is enhanced in the voltage wave
forms. The decay speed of the voltage oscillation can be altered by a pulse
forming resistor, Rf, shown in Fig. 3. An intense light pulse in the rising
phase of the first voltage wave is followed by three successive voltage
waves. The surface streamer produced in the first voltage rise will stop
when the local field at its front becomes too small to maintain its
propagation. The successive weak streamers will be triggered at each phase
of voltage change when the field induced by the accumulated charge and
external field becomes strong enough to produce electron avalanche. At any
rate the first streamer plays an essential role in producing enough ions (6).
IONIC CURRENT DENSITY
The ability of ionic current supply of the ceramic-made Boxer-Charger is
measured at 50 Hz of the a.c. voltage using an apparatus of Fig. 5(a) which
comprizes one IS-panel of Type I and represents one of the two ionizer units.
The results obtained are shown in Fig. 5(b) where the ionic current density
J, is plotted as a function of the space-averaged charging field strength,
at its crest field strength, E~ = V	/ d. The ionic current density
obtained is about one-half of tnat obtained with the double-helix ionizer
unit (3), with a concurrently longer charging time constant. But, it is
still adequate for practical charging applications. The current density can
be increased easily by using either a higher frequency of the a.c. voltage or
multiple pulses instead of a single pulse at each half cycle.
SATURATION CHARGE
The saturation charge is measured using a steel ball with 3 mm in
diameter suspended by an insulating string in the charging field between two
IS-panels of a Boxer-Charger spaced at 15 cm. Measurement is made in room
air with clean IS-panels, so that no back corona takes place. The steel ball
after charging is brought into inside of a Faraday cage and its saturation
charge, Q , is measured. Fig. 6 shows the magnitude of Q as a function of
the average field strength at its crest Eq. Figs. 6(a), (8^, and (c) are for
the Types I, II, and III of IS-panels, respectively. The Pauthenier's
theoretical curve is also depicted. The steel ball is located at the center
and near one of the IS-panels (about 10 mm apart from its surface). It can
be seen that the magnitude of Q near the panel is always greater than the
one at the center. This is because the local field near the IS-panel surface
is enhanced by the ionic space charge. The saturation charge, Q , follows
the theoretical curve up to about EQ = 3.2 - 4 kV/cm, but it indicates a
sharp drop beyond this threshold. Fig. 7 shows the light activity of the
pulse-energized IS-panel of Type I operated at Eq = 4 kV/cm. Many bright
spots are seen to appear at the edges of the panel. These are the positive
parasitic coronas caused by the a.c. inter-electrode field in a half-cycle
when this IS-panel is not excited by the pulse. Copious positive ions are
emitted from these spots into the charging zone to neutralize the negative
18-4

-------
if mm
m


*

(a) 1 IS-Panel of Type I
(Triangle Pulse Voltage Wave Form)
4 / \ /
%
(b) 6 IS-Panels of Type I
(Oscillating Pulse Voltage Wave Form
with Slow Decay)
Fig. 4 Pulse Voltage Wave Form and Light Signal Induced
by Surface Streamers
50 Hz
P. G.
Probe
75 mm
(a) Ionic Current Measuring System
Fig
0.075 -
&
u
C

-------
Theoretical
/
Near IS-Panel
Center Region
Charging Field Strength, EQ ( kV/crtl )
(a) Type I
Theoretical
Near IS-Panel
e 200
Center Region
Charging Field Strength, EQ ( kV/cm )
(b) Type II
Theoretical
Near IS-Panel
Center Region
4	I	I	
1 2 3 4 5
Charging Field Strength, Eq ( kV/cm )
(c) Type III
Fig. 6
Effect of Eq on Saturation
Charge of 3 mm Steel Ball
18-6

-------
charge of the ball. The IS-panel of Type I with a thin ceramic plate and
sharp edges of the corona electrode at its corners easily subjects to this
charge limitting effect than the Types II and III. Hence, precautions are
needed to avoid these parasitic coronas in the design of the IS-panel itself
and the method of its assembly to form an ionizer unit.
It is discovered that the effective capacity between the corona and
induction electrodes increases from its original value when streamers propa-
gate as the effective area of the corona electrode increases. Hence, the
inter-electrode capacity indicates a complicated time-dependent variation.
In addition, the discharge produces a loss represented by a non-liner paral-
lel resistance. Concurrently, Types I and II are the lumped constant pulse-
loads with a non-liner capacitance and a non-liner discharge resistance.
Type III is a distributed constant pulse-load with a non-liner surge imped-
ance, and it allows the application of the pulse peaking technique (7) for a
better exploitation of pulse energy. In this sense, Type III may have an
inherent advantage as the ionizer unit of the pulse-excited Boxer-charger.
TESTS IN LABORATORY PRECIPITATOR
LABORATORY PRECIPITATOR
Fig. 8 shows the laboratory test precipitator in a race-track system
used to test the charging performance of the ceramic-made Boxer-Charger for
high resistivity flyash and its effect in enhancing an overall collection
performance. The precipitator has the Boxer-Charger in two ducts (Fig. 9)
and three two-duct collection fields. The carrier gas circulated is air and
heated up to 150 C. Two flyash samples I and II -^re used and their
resistivity changed in the range of 9 x 10 - 1.4 x 10 ohm-cm by changing
gas temparature. The charge-to-mass ratio, Q/M, of the flyash sample is
measured at the downstream of the Boxer-Charger by a suction type Faraday
cage. The dust mass loading at the precipitator inlet, NL, and the outlet,
M , are also measured. Each collection field has 0.3 m m plate-to-plate
spacing, 0.76 m in height, and 0.5 m2in length in the gas flow direction.
The total duct cross section is 0.^6 m , and the tot^l collection area of a
single collection field is 1.52 m with SCA = 4.03 m /(m /s) at 0.82 m/s
standard gas velocity. Each duct has three corona wires with 3 mm in
diameter and 110 mm wire-to-wire spacing. The effective length of collection
field is altered by energizing the first, the first and the second, and all
of the three fields successively. Tests are made with two different
energization modes of the collection fields : one being,the dc-energization
and another the pulse-energization on top of the dc bias voltage through a
coupling condenser. The back corona severity in the collection fields are
detected by bi-polar current probe (8), and expressed in term of a ratio of
the back-corona-induced positive ionic current density to the negative one.
BOXER-CHARGER INSTALLED
Fig. 9 shows the ceramic-made Boxer-Charger in two ducts with 4 ionizer
units in total, each having 6 IS-panels of Type II. The ionizer unit is
carefully assembled to avoid the parasitic coronas. The light of the pulse-
18-7

-------
HEATER
BLOWER
Fig. 7
Parasitic Coronas
at the Panel Edges
7.5 m/s
COLLECTION FIELDS
(I+/I-)
DUST FEEDER
0.82 m/s
15 m/s
./
( Q/M )
BOXER -
CHARGER
Wi : inlet dust loading, WQ : outlet dust loading
Q/M : charge-to-mass ratio of flyash
rj : dust resistivity
I_/I+ : back corona severity measured by a bi-polar ion probe
Fig. 8 Race-Track System of Test Precipitator
GAS FLOW
Fig. 9 Ceramic-Made Boxer-Charger Ins trolled in front of Collection Fields
18-8

-------
induced surface discharge is very uniform on all of the panels, as shown in
Fig. 9(c).
Fig. 10 shows the charging performance of the B^xer-Charg^ measured at
three different dust resistivity levels, r, = 9 x 10 , 1 x 10 , and 1.4 x
10 ohm-cm obtained at T = 25, 80, 150 C, respectively. The magnitude of
Q/M rises at first proportionally with Eq, following the same curve independ-
ent of r,. However, a sharp., drop in Q/M begins to occur at E.~. = 2.7 kV/cm at
T = j.|0 'C and r, = 1.4 x 10 , and at Eq = 3.4 kV/cm at T = 80 C and r^ = 1
x 10 ohm-cm. No back corona i^observed to occur on the ionizer units even
at T = 150 C and = 1.4 x 10 ohm-cm, in contrast to the double-helix
type Boxer-Charger. Hence, the drop in Q/M in the ceramic-made Boxer-Charger
should be due to the temperature-enhanced parasitic coronas, and it could be
corrected by an improved design of the panels to be reported separately.
Fig. 11 shows the dust contaminated surface of the IS-panels after a
couple of hours operation without rapping or scraping. The dust deposit is
remaining on the corona electrodes, but not on the region of the ceramic
surface exposed to the surface streamers. This is because the surface
streamers wipe off the dust deposit to keep the active region always clean.
PRECHARGING EFFECTS ON COLLECTION PERFORMANCE
Fig. 12 shows the collection performance in terms of dust penetration
measured as a function of the effective length of collection field (number of
series fields energized). The detailed test conditions are given in Table 1.
The test results of collection performance in terms of Deutsch migration
velocity, w, performance enhancement factor H = w / Wq (w = base line value
with dc-energization alone), and back corona severity, I./i_ are tabulated in
Table 2.
In Fig. 12(a) for r^ = 5 x 10^ ohm-cm the curve A represents the base-
line data with the collection fields dc-energized, and the precharger off.
Back corona is observed to occur in the collection fields with its severity
1	/I = 8.2 %, and the collection performance is the worst (wq = 12.1 cm/s).
Back corona disappears by the pulse-energization of the collection fields,
with a concurrent performance enhancement as shown by curve B (H = 1.5). The
precharging in combination with the dc-energized collection fields produces
an enhancement as shown by curve A', where the Boxer-Charger is operated at
En =2.7 kV/cm and V = -7 kV with no parasitic coronas being produced and
the collection fieldi with Vnp = -42 kV. The enhancement is remarkable in
the first field ((B) in Table 2 ; H = 2.0), but it is lost in the succeeding
fields with curve A' becoming parallel to curve A as a result of back corona
((A) in Table 2 ; H = 1.4). It can be seen that the combination of a single
precharging with the dc-energized fields produces almost the same enhancement
as that obtained by the pulse-energization alone. The highest performance
enhancement can be obtained by the combination of the precharging and the
pulse-energization of the fields (curve B' ; H = 2.0), where the enhanced
charge of the particles are preserved, not subjecting to back corona. A
very high enhancement obtainable in the first collection field ((B) in Table
2	; H = 2.7) suggests the effectiveness of a successive use of the
prechargers at every collection field. It should be noted that the repeat
18-9

-------
2
O*
1.5
1.0
t>0
U
CO
6
0.5
Fig.

1

Fig. 11 Dust Diposition
on IS-Panels
X .+ XiU
ohm-cm
T= 80 C
=1x10 ohm-cm
12 3 4
Charging Field Strength, EQ (kV/cm)
10 Effect of Eg on Charge-to-Mass
Ratio of Flyash
Flyash Sample I
T = 25 C g
rj = 9 x10 ohm-cm
18-10

-------
Flyash Sample I
 5 x lOH ohm-cm
T - 130 C
Keys:
A :	DC-Energization
A';	B.C. +DC-Energization
B :	Pulse Energization
B':	B.C. +Pulse-Energization
Test Conditions : see Table 1
Number of Collection Fields
(a) High Resistivity Flyash
Flyash Sapmle II
I'd " 2 x 10 ohm-cm
5 -
2 -
Keys:
A :	DC-Energization
B :	Pulse-Energization
B :	B.C. + Pulse-Energizatio
Test Conditio
see Table 1
_L
I
1	2	3
Number of Collection Fields
(b) Very High Resistivity Flyaah
Flyash Sample II
Tjj  1 x 10* ohm-cm
T - 100 *C
Keys:
A	DC-Energization
B :	Pulse-Energization
B';	B.C. +Pulse-Energization
Test Conditions : see Table 1
Number of Collection Fields
(c) Extremely High Resistivity Flyash
Fig. 12
Effects of Precharging on
Collection Performance
18-11

-------
Table 1 Test Conditions
Dust Resistivity
DC-Energization
*
Pulse-Energization
5 x 10^ ohm-cm
(T = 130 C)
V = -42 k.V
DC ,
I = 0.2 mA/m
V	= -35 kV
V	= -20 kV -
I^= 0.06 mA/m
12
2 x 10 ohm-cm
(T = 50 C)
V = -38 kV
DC ,
I = 0.2 mA/m
V	= -30 kV
V	= -32 kV ,
1^= 0.04 mA/m
13
1 x 10 ohm-cm
(T = 100 C)
VDC = "32 W2
1=0.2 mA/m
V	= -22 kV
V	= -32 kV 
I^= 0.04 mA/m
* Pulse Voltage : rise time = 40 ns
half-tail = 180 ns
pulse repetion frequency = 50 Hz
Pulse voltage is superposed on top of a dc base voltage, Vb.
Table 2 Tests Results
Dust Resistivity
(ohm-cm)
Operation Mode
Deutsch Migration Velocity
Back Corona
Severi ty
'tS-
without B.C.
(cm/s)
^A^with B.C.
(cm/s)
(B)with B.C.
(cm/s)
5 x 1011
(T = 130 C)
DC-Ene rgization
wn = 12. 1
17.0 (H - 1.4)
23.6 (H = 2. 0)
8.2
Pulse Energization
18.0 (H- 1.5)
24.1 (H = 2.0)
32.4 (H = 2.7)
0
2 x 1012
(T = 50C)
DC-Energization
Wq = 8.6


2-15
Pulse-Energizat ion
11.6 (H= 1.4)
16.5 (H = 1.9)
19.5 (H = 2.3)
0-2
1 x 1013
(T = 100 C)
DC-Energizat ion
wn = 6.6


27
Pulse-Enwegi zat ion
7.6 (H = 1.2)
9.5 (H = 1.4)
14.1 (H = 2.1)
24
Wq : base line Deutsch migration velocity in the de-energized fields
H :	enhancement factor in terms of
(A)	:	overall Deutsch migration velocity enhanced by B.C.
(B)	:	Deutsch migration velocity in the first collection field enhanced by B.C.
Table 3	Performance Enhancement by Double-Helix Type Boxer-Charger
Dust Resistivity
(ohm-cm)
Operation Mode
Deutsch Migration Velocity
without B.C.
(cm/s)
(A)with B.C.
(cm/s)
^with B.C.
(cm/s)
11 12
10 - 10
(T = 100 C)
DC-Energization
w0-11.7
18.0 (H = 1.5)
28.6 (H 2.4)
Pulse-Energization
19.1 (H  1.6)
26.9 (H = 2.3)
34.1 (H = 2.9)
18-12

-------
mode of precharging in the dc-energized collection fields and a single
precharging mode in the pulse-energized fields provide the same enhancement
factor (H = 2.0).
12
When rd = 2 x 10 ohm-cm (Fig. 12(b)), a weak back corona (I /I = 0 -
2 %) takes place even in the pulse-energized fields. Concurrently, the
performance enhancement by the pulse energization becomes diminished (curve B
; H = 1.4). However, a combination of the Boxer-Charger produces a great
improY^ment (curve B'; H = 1.9) which is about the same as in case of r, = 5
x 10 ohm-cm.. The performance by pulse-energization alone is not good in
this case (w = 11.6 cm/s). However, when combined with precharging, the
magnitude of w in the first collection field becomes as high as 19.5 cm/s
((B) in Table 2 ; H = 2.3).
In the case of an extremely high resistivity, r , = 1 x 1013 ohm-cm (Fig.
12(c)), a severe back corona occurs not only in the dc-energized fields,
(I+/I_ = 27 %), but also in the pulse-energized ones (I /!_ =24 %). As a
result curves A and B come close to each other. Again in this case, the
ceramic-made Boxer-Charger produces a substantial enhancement in the first
collection field when combined with pulse-energization (curve B* ; H = 2.1).
However this is lost in the second and the third fields. The overall
enhancement is H = 1.4. The precharger must be inserted in front of each
pulse-energized field in th^case when a performance comparable to the base
line Wq of r^ = 5 x 10 ohm-crn is required in this extremely high
resistivity region.
The performance enhancement by the doublehelix type BoxerCharger is
given in table 3 (2).It can be seen that the enhanacement obtained by the
ceramic-made Boxer-Charger is the same as that by the double-helix one is
achieved in the dust resistivity range of rd = 10 - 10 ohm-cm.
CONCLUSION
(1)	A ceramic-made IS-panel (ion source panel) excited by a submicrosecond
pulse with -4 	7 kV peak voltage can be used as a satisfactory ionizer.
The cost of the pulse generator is greatly reduced by the low pulse voltage
required.
(2)	The surface streamers take place at an instant when the pulse voltage
reaches a certain threashold, and produce copious monopolar ions.
(3)	The saturation charge given by the ceramic-made Boxer-Charger follows
the Pauthenier's theory up to the a.c. field level when positive parasitic
coronas begin to occur from the panel edges. They provide positive ions to
the charging zone to greatly degrade the charging performance.
1 Q
(4)	No back corona is observed to occur even at r^ = 1.4 x 10 ohm-cm. The
performance degradation is only due to the parasitxc coronas which occurs at
a lower a.c. field level with increasing temperature.
(5)	A dust deposition occurs only on the corona electrode, but not on the
18-13

-------
active surface region subjecting to the surface streamers. As a result, this
region is self-cleaned to maintain the discharge activity.
(6) A single-stage ceramic-made Boxer-Charger can provide the dc-energized
precipitator with as good performance enhancement ^ = 1.4) as the pulse-
energization (H = 1.5) in the case when r^ = 5 x 10 ohm-cm and back corona
severity-^t dc-energization is less than I+/I_ = 10 %. In the case when r,
= 2 x 10 ohm-cm with the medium level back corona severity (I /I =2 -15
% for dc-energization), a substantial performance enhancement = 1.9)
occurs by the combination of the present Boxer-charger with the pulse-
energized collection fields where I+/I_ = 0-2	In the case of an
extremely high dust resistivity level of r^ = 1 x 10 ohm-cm, performance
improvement is very small by the pulse-energization alone (H = 1.2). Even in
this case, the present Boxer-Charger combined with the pulse-energized fields
produces a substantial performance improvement (H = 1.4), especially in the
first collection field (H = 2.1), but its effect becomes lost in the
succeeding fields.
The work described in this paper was not funded
by the U.S. Environmental Protection Agency and
therefore the contents do not necessarily reflect
the views of the Agency and no official endorsement
should be inferred.
REFERENCE
1.	Masuda, S., Mizuno, A., and Nakatani, H. Application of Boxer Charger
in Electrostatic Precipitators, Conf. Rec. IEEE/IAS Annual Meet. 1979,
pp.131-138
2.	Masuda, S., Mizuno, A., Nakatani, H., and Kawahara, H. Application of
Boxer-Charger in Pulsed Electrostatic Precipitator, Conf. Rec. IEEE/IAS
Annual Meet. 1980, pp904-911
3.	Masuda, S., Nakatani, H., Yamada, K., Arikawa, M., and Mizuno, A.
Production of Monopolar Ions by Travelling Wave Corona Discharge, Conf.
Rec. IEEE/IAS Annual Meet. 1981, pp.1066-1073
4.	Masuda, S., and Nakatani, H. Distortion of Pulse Voltage Wave Form on
Corona Wires due to Corona Discharge, Journal of Electrostatics 14(1983)
pp.319-337
5.	Masuda, S., and Kiss, E. Ceramic-Made Electric Curtain Devices and
Their Applications, Invited Talk, Int. Conf. on Industrial
Electrostatics (May 17 - 18, 1984. in Budapest, Hungary.)
6.	Kaneko, S. Master Thesis, University of Tokyo, 1984.
7.	Masuda, S., Hosokawa, S., and Nakatani, H. Corona Transmission Line
Energized by Very Short Pulse Voltages as Applied in Electrostatic
Precipitors, Conf. Rec. IEEE/IAS Annual Meet. 1983, pp.966-972
8.	Masuda, S., Itagaki, T., Nohso, S., Tanaka, 0., Hironaga, K., and
Fukushima, N. Bipolar Current Probe for Diagnosing Full-Scale
Precipitators, to be presented at EPA-EPRI 5th Symp. on Transfer and
Utilization of Particulate Control Technology
18-14

-------
PRECIPITATOR PERFORMANCE ENHANCEMENT WITH PULSED ENERGIZATION
E. C. Landham, Jr.
J. L. DuBard
Southern Research Institute
Birmingham, Alabama 35255
W. Piulle
Electric Power Research Institute
Palo Alto, California 94303
L. E. Sparks
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
ABSTRACT
Research type pulsers were installed on each of three electrical fields of
the EPRI pilot-scale precipitator at the Arapahoe Test Facility. Either fast
risetime (1.5 ys) pulses or longer duration (140 ys) pulses of -25 kV amplitude
could be superimposed on -50 kVDC at variable repetition rate up to 200 Hz.
The precipitator collection efficiency, with and without pulsing during a
single day, was measured on 35 days over a period of 3 months at both 9-inch*
and 12-inch plate spacing. The fly ash resistivity was varied in the range of
5E10 to 5E12 ohm-cm. Omega-k enhancement factors for fly ash precipitators of
SCA from 150 to 400 ft2/kACFM, both full-scale and pilot-scale, correlate more
closely with the baseline collection efficiency than with fly ash resistivity.
~Readers more familiar with metric units are asked to use the conversion
factors at the end of this paper.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
19-1

-------
INTRODUCTION
A number of techniques exist for improving electrostatic precipitator
(ESP) performance when collecting high-resistivity fly ash. The options
include modification of ash resistivity by chemical means, enlarging the size
of the ESP, and the use of an unconventional ESP energization scheme. The last
of these options is potentially the most convenient and cost effective but is
also the least well understood.
One of the ESP electrical energization systems of interest at the present
time is termed pulse energization. This technique involves superimposing a
high voltage pulse on top of the conventional DC voltage waveform. Improved
performance of utility fly ash precipitators with pulse energization has been
demonstrated by several equipment manufacturers. In a retrofit application,
the installation of pulse power supplies requires minimal time off line. The
continuous operation of the pulse power supplies requires minimal supervision
and maintenance. The attractive features of this new technology have been
shadowed, however, by a limited ability to predict the performance of any
particular ESP. Therefore, an independent study of precipitator pulsing has
been sponsored by the Electric Power Research Institute and the U.S.
Environmental Protection Agency.
Southern Research Institute was commissioned to design and construct pulse
power supplies, and to test the effects of pulsing a pilot-scale ESP at the
EPRI Arapahoe Test Facility in Denver, CO. The tests were designed to acquire
systematic data on the effects of ESP pulsing, with wide variations in the fly
ash resistivity and wide variations in the pulse power supply operation. A
large quantity of pilot-scale ESP performance data are summarized and
interpreted in this paper, and compared with a small quantity of full-scale
ESP performance data.
EXPERIMENTAL FACILITIES
Commercial high voltage pulse power supplies fall into two categories.
One type of pulse power supply uses spark gap switching in the high voltage
circuit and produces a narrow pulse with a fast risetime. The other type uses
a string of thyristors to switch the pulse and produces a wide sinusoidal pulse
waveform. One of the project objectives was to test both types of commercial
pulse waveforms. The pulse power supply that was constructed for the pilot-
scale precipitator at the Arapahoe Test Facility uses a thyratron switch in the
high voltage circuit. In a research application, this circuit design has the
advantage of a widely variable pulse waveform (by switching resistors and
inductors in the pulse-forming network) in order to simulate the physical
effects of both types of commercial pulse waveforms. The research pulse power
supply can produce a narrow pulse with a fast risetime (1.5 ps) or a slower
pulse of longer duration (140 ys). Pulses of -25 kV maximum amplitude can be
superimposed on a maximum DC voltage of -50 kV, at pulse repetition rates
continuously variable up to 200 Hz. A separate pulse power supply was
installed on each of the three identical electrical fields of the pilot-scale
19-2

-------
precipitator. The pulse waveforms were recorded with calibrated, frequency-
compensated voltage dividers and a digital oscilloscope.
The pilot-scale precipitator receives flue gas at about 300F from
Arapahoe Unit 4 of the Colorado Public Service Company. The precipitator
handles a nominal gas flow of about 3500 ACFM, at a nominal specific collecting
plate area (SCA) of 400 ft^kACFM with 9-inch plate spacing. The collecting
plates can be set at 9-inch spacing with four parallel gas passages, or at
12 inch spacing with three parallel gas passages. Both configurations were
tested. The discharge electrodes were 0.115-inch diameter wires at 9-inch
spacing on a harp frame.
Various low-sulfur western subbituminous coals were burned in Arapahoe
Unit 4 on varying schedules during the test program. Therefore, the precipi-
tator collection efficiency was measured both with and without pulsing on each
of 35 test days over a period of three months. The precipitator electrodes
were not rapped during tests. Measurements of collection efficiency were
correlated with in situ measurements of fly ash resistivity and with the pre-
cipitator electrical operating points. In addition to continuous monitoring of
the outlet flue gas opacity, the inlet and outlet mass loadings were measured
with mass trains and with cascade impactors. A point-plane probe was used for
in situ measurements of the fly ash resistivity by the spark method. An SO,
injection system was installed on the inlet flue gas duct in order to vary the
fly ash resistivity over the range of 5E10 to 5E12 ohm-cm.
PULSER OPTIMIZATION
Some controversy exists over the optimum DC voltage level for pulsed
operation. One theory holds that a DC voltage which is below corona onset
may serve to reduce back corona and improve ESP performance. As a test of this
theory, the ESP outlet opacity was used as a measure of ESP performance while
the pulse rate was varied for two levels of DC energization. The results are
displayed in Figure 1. In this case, the high DC level performed much better
than the low. For the low level, the DC voltage was set immediately below
corona onset; while for the high level, the DC voltage was adjusted to the
maximum value without sparking. All performance tests of the ESP were per-
formed with a high DC level to maximize the ESP collection efficiency.
Further measurements of the outlet opacity were used to evaluate the
optimum pulse rate for both pulse types. The results, displayed in Figures 2
and 3, indicate that the best ESP performance with both high and moderate
resistivity fly ash occurred at repetition rates below about 50 Hz. The ESP
performance either degraded or did not improve beyond that frequency. The
figures also demonstrate that there was no apparent difference in the effect of
the two pulse types, with either high or moderate resistivity ash.
Outlet opacity was measured again to determine if a relationship existed
between the location in the ESP of the pulsed electrical field and the pulse
rate at which optimum ESP performance was obtained. Figure 4 shows the results
of a test where the pulse rate in only one field (either inlet or outlet) was
19-3

-------
varied, while the other fields were pulsed at 50 Hz. The data suggest that a
higher pulse rate in the inlet field and a lower rate toward the outlet might
be advisable. Since the range of optimum pulse rates for both inlet and outlet
fields coincided at 50 Hz in the pilot-scale ESP, the performance tests were
conducted with that rate in all three fieltls. Although high DC voltage levels
and fairly low pulse rates produced the best performance at Arapahoe, these
parameters should be optimized for each installation.
ESP PERFORMANCE
Extensive tests of the performance of the ESP with and without pulsing
were performed at the optimum operating conditions. The tests were conducted
with several coals and several levels of ash resistivity to simulate various
coal types and levels of ESP performance degradation.
OPACITY
The most dramatic illustration of the effects of pulse energization is
observed with the outlet opacity measurements. Figure 5 shows the result of
pulsing when the baseline ESP performance was very poor. The ash resistivity
at this time was 2.8E12 ohm-cm, and the baseline efficiency of the ESP was
82.7%. When the ash resistivity was reduced and the baseline efficiency was
improved with S03 conditioning, the results were significant but less
impressive. Figure 6, taken with 3.3 ppm S03 added, indicates that some
improvement was gained with an ash resistivity of 2E11 ohm-cm and a baseline
efficiency of 97.1%. When the ash resistivity was reduced to 2E10 ohm-cm with
the addition of 15 ppm S03, no change in opacity or efficiency was observed due
to pulsing.
COLLECTION EFFICIENCY
The primary method of evaluating the change in ESP performance due to
pulsing was measurement of the total mass collection efficiency. The measure-
ments were made by sampling simultaneously at the ESP inlet and outlet with
EPA Method 17 mass trains. Measurements were performed at various levels
of ash resistivity and baseline collection efficiency. The data have been
averaged into six groups according to the baseline efficiency, as shown in
Table 1, Comparison of the collection efficiency data with and without pulsing
suggests a trend. As the baseline efficiency drops, the pulsed efficiency
drops as well, but the improvement in performance due to pulsing increases. In
order to interpret the ESP performance, the data are analyzed with the Matts-
Ohnfeldt equation.
The Matts-Ohnfeldt equation provides a simple empirical method of
describing the performance of an ESP. The equation takes the form
P - 100 exp [-(mk SCA)k],	(1)
19-4

-------
where
P = ESP total mass penetration (%)
to. = effective migration velocity (m/s)
SCA = A/V (m2/m3/s)
k = 0.4 to 0.6 (assumed to be 0.5 here)
V = gas volume flow (m3/s)
A = total ESP plate area (m^)
From (1) the effective migration velocity can be calculated. This provides a
measure of the ESP performance that is approximately independent of ESP size.
tok = (V/A) [ In (100/P)]1/k	(2)
The performance improvement due to pulsing can be expressed as a ratio of the
effective migration velocities with and without pulsing. This is termed the
effective migration velocity enhancement, or omega-k enhancement, given by
0)jc(pulsed)/u5jc(baseline).	(3)
Application of this technique to the ESP performance data can be seen in the
last column of Table 1. From this analysis it is apparent that, as the base-
line efficiency decreases, the omega-k enhancement due to pulsing increases.
It would be desirable to develop a relationship between the resistivity
of the fly ash being collected and the omega-k enhancement expected with
pulsing. An attempt to do this is illustrated in Figure 7, which plots
the omega-k enhancement against the measured values of fly ash resistivity.
Several problems are observed with this presentation of the data. The scatter
in the data increases rapidly with increasing resistivity, and the data
obtained at 12-inch plate spacing lie in a separate group. Furthermore, the
fitted curve, Correlation A, does not reflect the fact that the ESP performance
improvement due to pulsing approaches zero at fly ash resistivities in the 2E10
ohm-cm range. It appears that factors other than the in situ fly ash resis-
tivity must be taken into account.
A tighter and simpler correlation of the mass train data is found upon
plotting the omega-k enhancement against the baseline percent penetration.
The pilot-scale data are plotted in this way as the solid symbols in Figure 8.
In this method of data correlation there are no systematic differences between
the data obtained at 9-inch and 12-inch plate spacing, nor between the data
obtained with a fast pulse and a long pulse. A linear least-squares fit and
its 90% confidence interval are superimposed on the pilot-scale ESP data. The
fitted line is denoted Correlation B.
The open symbols in Figure 8 represent full-scale precipitator performance
data gleaned from the literature. The full-scale data come from tests of the
pulse power supplies of several different commercial manufacturers in the
United States, over a time period of 7 years. Each of 17 full-scale data
points is detailed in Table 2. The calculations of omega-k and the omega-k
19-5

-------
ratio, in Table 2, use k = 0.5 in the Matts-Ohnfeldt equation. It is
apparent that the full-scale data exhibit a simple correlation with baseline
percent penetration similar to that of the pilot-scale data, but with a lower
slope. A least-squares fit to the full-scale data is denoted Correlation C.
Correlation B (Pilot-Scale Data)
V'Vdc " (1-34 * -32) * '-25 logio P<*<='	(4)
Correlation C (Full-Scale Data)
Vk.dc " (1'1A 1 '13) + '78 lo810 PdC	(5>
The lower slope for the full-scale data correlation probably results from
rapping reentrainment and a variety of electrical and mechanical defects that
are likely to exist in a full-scale precipitator  defects which pulsing
cannot fix. By comparison, the pilot-scale precipitator was put in good oper-
ating condition for this project, and the electrodes were not rapped during
tests.
Another relationship between pulsed and baseline ESP performance can be
seen in Figure 9. In this figure the baseline penetration is plotted versus
the pulsed penetration (reduced penetration). Careful inspection of the full-
scale data indicates that, at baseline efficiencies below 98%, pulsing should
be expected, on the average, to produce a factor of 2 reduction in the penetra-
tion. This assumes, of course, that the only problem with the ESP performance
is poor electrical operating conditions due to high-resistivity fly ash. In
using Figure 9, care should be exercised to note the upper 90% confidence limit
when estimating the improvement in performance of a particular ESP.
ESP ELECTRICAL CHARACTERISTICS
For an electrical energization technique to modify collection efficiency,
it should change the electrical conditions in the ESP. With AC coupling, the
pulses provide no time-average electrical power input to the ESP. Therefore,
the influence of the pulser is observed primarily in its effect on the rela-
tionship between the DC voltage and current. Figure 10 illustrates how the ESP
secondary voltage and current change with varying pulse rate. As the pulse
rate is increased, higher currents are observed at the same voltage. This
effect could allow the voltage and current values to be customized to a
particular application. However, this is not believed to be the cause of the
improved ESP performance. Other data from this study indicate that the primary
mechanism of performance improvement with pulsing is improved distribution of
corona power. More uniform distribution of the ESP current tends to reduce the
"hot spots" of high current density which produce intense back corona sites in
a high-resistivity ash layer. Support for this explanation is obtained from
Figure 11, a pair of voltage-current curves taken with and without pulsing
during operation with high-resistivity ash. The curve obtained without pulsing
indicates that back corona began around 0.5 nA/cm2, The corona current input
to the ESP above that level probably did not contribute greatly to improved
19-6

-------
performance. In comparison, the curve measured with pulsing indicates that
severe back corona was not present, and that all the current input was useful
in charging and collecting particles. Even though the average voltage was
reduced somewhat during pulsing, the increase in useful current density more
than compensated for that fact. Precipitator performance data obtained at the
same time as the V-I curves in Figure 11 showed that the collection efficiency
increased from 93.8% to 98.1% with pulsing. The outlet flue gas opacity
dropped from 40% to 21%.
CONCLUSIONS
Pulse energization improved the performance of all precipitators studied,
with the exception of the pilot-scale ESP when using high levels of SO^ con-
ditioning. The performance improvement appears to be related to the baseline
efficiency of the ESP, with more benefit gained at lower baseline performance
(< 98%). However, it does not appear that pulsing can improve efficiency more
than other techniques such as SO 3 conditioning. The evidence is that the use
of a pulser in addition to flue gas conditioning would not be appropriate. In
fact, pulsing cannot improve many conditions which may limit the performance of
a full-scale ESP. Use of pulsing should be based on a clear understanding that
an ESP is in good mechanical condition, and that the major limitation to the
performance of the ESP is degraded electrical operation caused by high-
resistivity fly ash.
The data on both pilot-scale and full-scale fly ash precipitators that
have been studied in this project indicate that the primary mechanism of
precipitator performance improvement with pulsing is simply the more uniform
corona discharge. The technology appears to be equally effective on both cold-
side and hot-side precipitators and at both 9-inch and 12-inch plate spacing.
The two pulse waveforms tested on the pilot-scale ESP were equally effective.
No systematic differences were observed between the performance of the various
commercial pulsers on the full-scale precipitators.
19-7

-------
METRIC EQUIVALENTS
Readers more familiar with metric units may use the following to convert
to that system.
Nonmetric
Multiplied By
Yields Metric
F
ft2
ft3
in.
5/9 (F-32)
0.0929
28.32
2.54
C
liters
cm
TABLE 1.
BASELINE AND PULSED ESP
PERFORMANCE AVERAGED
INTO SIX RANGES
Baseline
Baseline
Pulsed
Omega-k
Effic iency
Efficiency
Efficiency
Enhancement
Range
%
%

99+
99.49
99.61
1.09
98 - 99
98.45
99.51
1.65
97 - 98
97.45
99.25
1.80
95 - 97
96.10
99.16
2.21
90 - 95
93.01
98.40
2.43
80 - 90
84.79
97.55
3.88
19-8

-------
TABLE 2. PERFORMANCE TESTS OF FLY ASE PRECIPITATORS
WITH COMMERCIAL PULSE POWER SUPPLIES
Test	Pulse SCA, ft2/ Fly Ash Baseline Omega-k,	Ratio
Date	Type kACFM ohm-cm Pdc' ^ dc * ^t/s ^k^k dc
19 76 a
Fast
370
5E11
4.30
0.45
1.65
197 9b
Fast
183
5E10
3.00
1. 12
1.50

Fast
176
5E10
3.39
1.08
1.70
1980c
Fast
272
IE 13
13.68
0.24
1.99

Fast
270
IE 13
15.40
0.22
2.11
1980d
Fast
326

3.33
0.59
1.49

Fast
312

1.76
0.87
1.46

Fast
329

1.82
0.81
1.23

Fast
304

0.80
1.28
1.05
1980d
Long
347

3.67
0.52
1.53

Long
336

0.88
1.11
1.18

Long
330

2.25
0. 73
1.41

Long
347

1.88
0.76
1.29

Long
339

2.01
0.75
1.26
1983e
Long
230
IE 12
3.43
0.82
1.49

Long
230
1E12
3.43
0.82
1.62

Long
230
IE 12
3.43
0.82
1.61
a)	P. L. Feidraan and H. I. Milde. Pulsed Energization for Enhanced Elec-
trostatic Precipitation in High-Resistivity Applications. In:
Symposium on the Transfer and Utilization of Particulate Control Tech-
nology. Volume I. EPA-600/7-79-044a (NTIS PB295226). February 1979.
b)	K. S. Kumar, P. L. Feldman, H. I. Milde, and C. Schubert. The Results
of First Full-Scale Utility Demonstration of Pulsed Precipitation.
IEEE/IAS Annual Meeting. Cleveland, OH, 1979.
c)	P. L. Feidraan and P. Aa. Operating Results from the First Commercial
Pulse Energization System to Enhance Electrostatic Precipitator Perfor-
mance. American Power Conference. Chicago, IL, 1981.
d)	J. P. Gooch, G. H. Marchant, Jr., J. L. DuBard, and J. R. McDonald. An
Evaluation of Pulse Energization on a Hot-Side Precipitator. EPRI CS-
2634. Final Report of Project 1868-2. November 1982.
e)	Private communication from Florida Power Corporation.
19-9

-------
100
70
40
SO
4J
40
HIGH OC LCVCL
110
1*0
140
100
0
40
Pulser Repetition Rate, Hz
Figure 1. Outlet flue gas opacity, measured with 9-inch plate spacing and fly
ash resistivity about 2E12 ohm-cm, using fast pulses with high and
low dc voltage levels.
A SHOA T PULSC
A LONG rvisc
 SHORT ntLSf
OlONC FULU
6^
o
a
o

OfdCITT
'owe*
V				 _ 	
o-0-  p,
CM
e
n
5
a)
>
Q>
t-i
QJ
J
O
100
tio 1(0 1*0 N w>
Pulser Repetition Rate, Hz
Figure 2. Time-average electrical power input and outlet flue gas opacity,
measured with 9-inch plate spacing and fly ash resistivity about
2E12 ohm-cm.
19-10

-------

u
re
a.
o
0*CITV
powih
A SHOMT PUlSt
A LONG PULK
 SMOftT rum
O IONC PULK
00 110 140 1(0 110 200
CM
e


0)
~J

-------
100
to 
4J
 ^
40
30
Time, hrs
Figure 5. Chart recorder trace of outlet opacity, with and without pulsing,
with no SO^.
100
0
0
70
Time, hrs
Figure 6. Chart recorder trace of outlet opacity, with and without pulsing,
with 3.3 ppm S03 injected.
19-12

-------
\o
M
U*
oo
0>
I
=r
03
O
0
1
ft)
rr
*
1	1I I II I T |
O'AST PULSE. MUCH DUCT
 LONG PULSE. INCH DUCT
A FAST PULSE. 12-INCH OUCT
A LONG PULSE, 12-INCH OUCT
1	1 I I I I I |
iimrr
/
21
O
22
O
11
 o
/
13
/
|12
41
'co!!fil22S' .
" q27 O'  __,A	-	'^40
			AA45 A42
9*	'
-18
I I I I I I I I
10"
J	I	I 1 1 I I I I
10
12
X
I  I I 111
io"
Fly Ash Resistivity, ohm-cm
Figure 7. Calculated omega-k enhancement (k = 0.5) versus resistivity.

-------
c
a>
B
a>
0
c
rd
JS
C
U2
&
1
cd
bC
a;
to
 mot scale *ntcihtatoa
O FUIL tCAlf MICIHTATOAJ
y
s'
ov^'
 I  r. i  j
04 ..!	2	<  I 10
Baseline Penetration, %
Figure 8. Omega-k enhancement, calculated using k = 0.5, versus baseline
penetration of the pilot-scale precipitator and six full-scale fly
ash precipitators.
/_
//'-
r/ /
 PILOT SCALE PKEC^iTATOA
O 'Ult 5CALE *CI"TAT0S
o
4J
fl}
u
u
CO*ELATlON I
Q>
Baseline Penetration3 %
Figure 9. BaseLine penetration aftd reduced penetration of the pilot-scale
precipitator and six full-scale fLy ash precipitators.
19-14

-------
h r
CNJ
e
o
<
c
U *
 rA
U)
c
a)
o
c
a)
p
3
O
4 0 1 x
0

 i/yy
o
./
J
~
c
r f CLD
2sa
D
c
r  CLD
i mm

c
r 1 CLD
tmm
X
c
r 1 CLD
mm

c
r 1 CLD
2C
o
c
r 1 CLD

 I i I I I I I I I 1 I I I I I I i . I . I . . i . I I . t . I t . 1 1 I
t	II IS 11 IS 31 K HI
Secondary Voltage, kV
Figure 10. Change in the V-I curves for C Field, at 9-inch plate spacing and
with 3.5 ppm S03 injection, with different repetition rates of
25 kV long pulses.
i r
CM
e
o
<:
c
tn
C

-------
AEROSOL PARTICLE CHARGING IN A PULSED CORONA DISCHARGE
J. L. DuBard
Southern Research Institute
Birmingham, Alabama 35255
W. R. Piulle
Electric Power Research Institute
Palo Alto, California 94303
ABSTRACT
The charge acquired by individual aerosol particles, in both steady and
pulsed negative corona, has been measured with a Millikan cell. Measurements
have been performed on a pilot-scale fly ash precipitator and a wire-cylinder
laboratory precipitator. The laboratory experiments were designed to isolate
the particle charging mechanism from other electrical mechanisms that might
be affected by pulse energization. The laboratory data were compared with
particle charging computations, using a numerical simulation of the variations
in space and time of the electric field and ion density. The laboratory
measurements and computations showed that the periodic increases in electric
field and ion density occur too briefly for particle charging to be a dominant
mechanism of precipitator performance improvement. During tests of the pilot-
scale precipitator with pulse energization, omega-k enhancements of 1.7 to 1.9
were observed when the Millikan data showed no increase in fine particle
charge.
20-1

-------
INTRODUCTION
Tests of aerosol particle charging in a pulsed corona discharge have been
sponsored by the Electric Power Research Institute. This work was carried out
in conjunction with pilot-scale precipitator performance tests at the Arapahoe
Test Facility. One objective of the investigation of pulse energization was to
determine the way in which pulsing works to improve precipitator performance.
This paper describes the interpretation of data leading to the conclusion that
the improved uniformity of the corona discharge is the dominant mechanism of
performance improvement.
The precipitation process can be improved by increasing the charge on
individual dust particles and by increasing the interelectrode electric field.
Both experimental and theoretical determinations of particle charge values have
been thoroughly tested in recent years (1,2). Therefore, the investigation of
pulse energization included extensive tests of the hypothesis of increased
particle charge. The tests were performed on the pilot-scale precipitator
at the Arapahoe Test Facility and in the Birmingham laboratories of Southern
Research Institute. The laboratory tests were designed to isolate the particle
charging mechanism from other interactive electrical mechanisms in a precipi-
tator.
LABORATORY INVESTIGATIONS
EXPERIMENTAL APPARATUS
A wire-cylinder precipitator was used for the laboratory investigations of
particle charging. The discharge electrode was an 8-32 threaded steel rod,
along the axis of an aluminum pipe 8.125 inch (20.7 cm) inner diameter. The
active length of the corona discharge was 41.5 inches (105.4 cm). The total
length of the cylinder was about 70 inches. Laboratory air was pulled through
the cylinder at a velocity of 5.0 ft/s; so the aerosol treatment time in the
precipitator was 0.69s.
A threaded discharge electrode was used to achieve uniform spacing of the
tufts of the negative DC corona discharge. A liquid drop aerosol was used to
avoid any electrical effects of a high-resistivity dust layer on the collecting
cylinder. An artist's air brush was used to generate an aerosol of di-(2-
ethylhexyl)-sebacate, a light-weight, non-volatile oil. The aerosol included
fine particles that spanned the entire size range (0.5 to 2.5 um diameter) that
could be measured in a Millikan cell. In each experiment, the air brush was
adjusted to limit the aerosol particle concentration so that there was no mea-
surable effect of particulate space charge. The size and charge of individual
fine particles escaping the precipitator were measured in a Millikan cell.
The DC high voltage corona threshold in the precipitator was about -29 kV.
All laboratory measurements and computations of particle charging in a pulsed
corona discharge were performed with the DC voltage set at -29 kV. There was
20-2

-------
no steady corona discharge. However, the time-average current with pulsing was
read from the ammeter of the DC power supply. For a given pulse height, the
time-average current was strictly proportional to the pulse repetition rate.
The same behavior was observed for the AC-coupled pulse power supplies in-
stalled on the pilot-scale precipitator at the Arapahoe Test Facility when the
DC voltage was set below corona threshold. The high voltage pulse acted as a
switch to turn on a momentary corona discharge.
The laboratory pulse power supply was AC-coupled to the DC power supply
and to the discharge electrode through a specially designed and fabricated LC
network. The maximum pulse height was about -13.5 kV. The pulse risetime was
about 0.9 us, and the width was about 2 ys. A fast pulse was used in order to
simulate one of the two pulse waveforms tested on the pilot-scale precipitator
and to simplify the computations of particle charging.
The pulse repetition rate was continuously variable up to a few kilohertz,
but limited to 200 Hz during the experiments in order to have a single burst of
ionic space charge within the cylinder at any time. Both the pulse height
and the pulse repetition rate were monitored with an oscilloscope, using a
frequency-compensated voltage divider. Because the pulse width was negligible
in comparison with the ion transit time of a few milliseconds, the only signi-
ficant electrical parameter in the measurements and computations of particle
charging was the total ionic space charge (per unit length of the cylinder)
created by each high voltage pulse. That parameter was determined by the ratio
of time-average current to pulse repetition rate.
PARTICLE CHARGING IN STEADY CORONA
A series of experiments with ordinary DC energization of the precipitator
was performed to test the reproducibility of particle charge measurements
and to test the mathematical simulation of particle charging in cylindrical
geometry. Figures 1 and 2 show Millikan data obtained with two slightly
different electrical operating points, chosen to give current density values
that could be achieved with pulsed corona and no steady corona. The two
electrical operating points were (34.5 kV, 89 pA, 13.0 nA/cm2) and (35.5 kV,
134 pA, 19.5 nA/cm2).
The dashed curve shown on each figure was computed using a conventional
theory of unipolar ionic charging in a steady corona discharge (1). The compu-
tation assumed that fine particles were uniformly distributed over any cross-
sectional area of the cylinder, due to fully-developed small-scale turbulence.
The effect of turbulence on particle charging and particle sampling was simu-
lated by averaging the charge values computed for particles moving through the
precipitator at the area-median radii of seven equal increments of cross-
sectional area. In spite of the typically large scatter in the Millikan data,
the theoretical curves agree well with the mean values of measured particle
charge. Furthermore, a close comparison of Figures 1 and 2 shows that both
measured and computed particle charge values are slightly different for the
two different electrical operating points.
20-3

-------
PARTICLE CHARGING IN PULSED CORONA
A second series of experiments was performed with a pulsed corona
discharge and no steady corona. Three data sets are shown in Figures 3, 4,
and 5. The data in Figures 3 and 4 were obtained with different values of
pulse repetition rate and time-average current density, but the same ratio of
those two parameters. The ratio gives a total ionic space charge per pulse of
0.42 yC/m. The data in Figure 5 were obtained with 2/3 that ionic space
charge, 0.28 yC/m.
A mathematical model of particle charging in a pulsed corona discharge was
developed to assist in the interpretation of data. The model performs a de-
tailed numerical simulation of the variation in space and time of local values
of electric field and ion space charge density. The model is simplified by the
cylindrical geometry, the narrow pulse width, and the absence of steady corona
in the precipitator.
The computation begins with the experimental value of total ionic space
charge created by each high voltage pulse. The initial radial distribution of
charge is estimated from a numerical simulation of the electron attachment in
laboratory air. Then, working with a specified initial charge distribution,
eN(r), Gauss' law in cylindrical geometry gives
Finally, the values of Qand eN(r) are substituted back into Gauss' law to
j j	^wire
determine the electric field E(r). Then, time is incremented, and the radial
coordinate of each increment of space charge is changed by (B E(r) At) where B
is the ion mobility. A new space charge distribution is calculated, and the
entire procedure is repeated until all molecular ions reach the cylinder.
Because of the dense space charge around the corona wire after each high
voltage pulse, there are large local variations in the electric field. The
space charge cloud moves very fast in the first few hundred microseconds and
actually blows itself apart into two bunches, as illustrated in Figure 7.
Figure 7 shows the computation of eN(r) at 0.3, 0.6, and 1.2 ms after the high
voltage pulse, for total ionic space charge of 0.42 yC/m. By way of compari-
son, the ionic space charge density in a steady corona with the same current
density is of the order of 0.05 x 10-4 yC/m. This number would be barely
visible on the vertical scale of Figure 7.
Figure 8 shows the corresponding electric field enhancement in front of
the expanding space charge cloud. Local values of E(r) are plotted in pro-
portion to the electrostatic field for 29 kV on the corona wire. That ratio is
The unknown value of charge on the corona wire, Qwre is determined by
substituting the expression for E(r) into the radial path integral
20-4

-------
constant before and after each space charge cloud. As time elapses after each
high voltage pulse, the negative charge on the corona wire increases to the
value that the DC power supply maintains between pulses, and the electric field
at the corona wire increases to the electrostatic field.
Given the detailed numerical map of eN(r,t) and E(r,t), the model computes
the accumulation of charge on aerosol particles at seven area-median radii in
the cylinder. The average charge value for each particle size gives the solid
theoretical curves superimposed on the data in Figures 3, 4, and 5. The mean
values of measured particle charge agree well with the results of the numerical
simulation. A cLose comparison of the three figures shows that theory and
experiment give the same result for the two tests with 0.42 yC/m, Figures 3 and
4. However, the results are clearly different for the test with 0.28 yC/m,
shown in Figure 5. Figure 6 is the same as Figure 4, but with the addition of
the dashed theoretical curve from Figure 1. The data sets for particle charge
acquired in steady corona and in pulsed corona are comparable only in regard to
the same current density, 13.0 nA/cm2. However, Figure 6 illustrates that,
within the range of Millikan data, the computed difference in particle charge
with steady corona or pulsed corona is much less than the scatter in the data.
The model was used to compute trends in the charging of particles outside
the range of Millikan data. Table 1 shows the computed data for 0.2s exposure
time, or one foot of travel into the laboratory precipitator. In a pulsed
corona discharge with no underlying steady corona, submicron particles have
reduced charge levels. In spite of enhanced values of electric field and ionic
TABLE 1. COMPUTED PARTICLE CHARGING IN THE LABORATORY PRECIPITATOR
Particle	Saturation	Accumulated Charge	Q/QSAT
Diameter	Charge Units	0.2s Exposure	0.2s Exposure
Steady Corona
34,5 kV on the	wire
13.0 nA/cm2 on	the cylinder
0.2	3 9 3.0
0.6	19 47 2.5
1.3	82 163 2.0
2.8	366 603 1.6
6.1	1715 2399 1.4
Pulsed Corona (200 Hz)
0.4227 uC/m per pulse
13.0 nA/cm2 on the cylinder
0.2
0
0
-
0.6
26
27
1.0
1.3
113
154
1.4
2.8
508
652
1.3
6.1
2378
2751
1.2
20-5

-------
space charge density, the charging of submicron particles by thermal diffusion
of molecular ions is restricted by the brief exposure time during each burst of
ionic space charge. This behavior is illustrated by the numbers in the right-
hand column of Table 1. Q/QSAT is a measure of continued particle charging by
thermal diffusion. This behavior is illustrated also by the intersecting
theoretical curves in Figure 6, where the crossover point of higher particle
charge with pulsing (and 0.69s exposure time) occurs at about 1.7 v1111 diameter.
Particles larger than a few microns can be charged to higher values in a pulsed
corona discharge, by bombardment of ions driven by the electric field. The
momentary increase in local values of electric field leads to higher values of
saturation charge.
PILOT-SCALE PRECIPITATOR TESTS
PARTICLE CHARGE MEASUREMENTS
A Millikan cell was set up at the pilot-scale precipitator outlet during
performance tests at 9-inch plate spacing. The Millikan cell was used to
determine the size and charge of individual fly ash particles escaping the
precipitator. The charge data were acquired during 4 to 5 ppm SO, injection,
when the fly ash resistivity was measured in situ to be 2E11 to 4E11 ohm-cm.
Typical Millikan data, obtained during precipitator operation with
ordinary DC energization, are shown in Figure 9. The mean measured particle
charge and its standard deviation are shown in Figures 10, 11, and 12, for
precipitator operation with DC energization, fast pulses, and long pulses,
respectively. The same theoretical curve is shown on each figure, to guide
the eye. That curve is the computed particle charge for DC energization of the
precipitator. During all tests with pulse energization, the DC voltage was set
at spark limit in each field. The pulsed corona discharge was superimposed on
a steady corona discharge.
The Millikan data show no statistically significant increase in the charge
on fine fly ash particles with pulse energization. These data are consistent
with the Millikan data obtained during tests of full-scale, pilot-scale, and
laboratory precipitators (3,4,5). A large variety of DC and pulse energization
parameters have been used in the various tests, but the Millikan data are
unanimous: no statistically significant increase in the charge on individual
particles in the diameter range 0.5 ym to 2.5 ym. Nevertheless, during the
particle charge measurements compared in Figures 10 and 11, the precipitator
collection efficiency increased from 97.2% to 99.1% with pulsing, and the
outlet flue gas opacity dropped from 27% to 18%.
FRACTIONAL EFFICIENCY MEASUREMENTS
The pilot-scale precipitator collection efficiency, as a function of
particle size, was measured with cascade impactors. Data are shown in Figure
13, with and without fast pulses at 50 Hz. These data were obtained simul-
taneously with the Millikan data compared in Figures 10 and 11. Similar
cascade impactor data were obtained with and without long pulses at 50 Hz.
20-6

-------
Across the entire particle size range, the collection efficiency increased with
pulsing.
The effective migration velocity (to) for particles in a small size band
can be calculated from the cascade impactor data, using the Deutsch-Anderson
equation (6),
PERCENT EFFICIENCY =100 (1 - e ~ w A/V).
A/V is the specific collection area (SCA) given by the ratio of precipitator
plate area to volume gas flow. The simple difference in values of migration
velocity with and without pulsing (calculated from pairs of curves as shown in
Figure 13) is shown in Figure 14. The two data sets in Figure 14 suggest that
the long pulse worked better than the fast pulse. However, the difference is
well within the scatter of all of the precipitator performance improvement
data.
The experimental values of effective migration velocity can be compared
with the theoretical migration velocity for particles of diameter D,
oj = (constant) QE/D,
where Q is the particle charge and E is the interelectrode electric field at
the collecting plate (6). The theoretical increase in migration velocity with
pulsing can be written as
A(D =* (constant) A(QE),
for particles of a given diameter.
The left-hand side of Figure 14 covers the same particle size range as the
Millikan data. Since the Millikan data show that AQ = 0, it follows that the
nearly uniform increase in migration velocity of particles smaller than about
2 pm results from an increase in the plate-area-average value of collecting
electric field. In a typical fly ash precipitator, the collecting electric
field is controlled more strongly by the interelectrode space charge than by
the discharge electrode voltage. It follows that the increase in migration
velocity of fine particles is caused by a more uniform distribution (over the
plate area) of interelectrode space charge.
The Deutsch-Anderson collection efficiency equation assumes ideal electri-
cal and aerodynamic operating conditions in the precipitator. In particular,
the equation does not account for the reentrainment of collected dust. The
right-hand side of Figure 14 covers the size range of most reentrained par-
ticles. The pilot-scale precipitator performance tests were conducted without
rapping the electrodes. Therefore, the increase in migration velocity of
particles larger than about 2 y results from a suppression of non-rapping
reentrainment, as well as increased values of Q and E. The suppression of
reentrainment is caused by an increase in the plate-area-average value of the
electrical holding force on the collected dust. This increase is caused by a
more uniform deposition of ionic charge on the surface of the collected dust.
The suppression of reentrainment is assisted by the rapid recharging of re-
entrained particles larger than about 2 ym.
20-7

-------
CONCLUSIONS
The data from the investigation of precipitator pulsed energization
indicate that the effects of a more uniform negative corona discharge fulLy
explain the improved precipitator performance obtained with pulsing. The
improved plate-area-average uniformity of the negative corona in a pulsed
precipitator has several important effects. The effects can be summarized in
terms of 1) improvements in the charging and collecting of dust particles, and
2) suppression of non-ideal precipitator operating conditions.
1)	A uniform, controlled corona discharge leads to
a uniform interelectrode ionic space charge, which causes
	increased average collecting electric field, and
	increased initial rate of particle charging.
The data indicate that pulsing has little effect on the primary charging of
fine particles, but is more effective in the rapid recharging of large re-
entrained particles.
2)	A uniform deposition of ions on the collected dust layer leads to
a uniform electric field within the dust layer, which causes
	suppression of back discharge, and
	suppression of reentrainment.
The data indicate that pulsing is most effective in suppressing the conditions
of back discharge and reentrainment that degrade precipitator performance.
REFERENCES
1.	McDonald, J. R., M. H. Anderson, and R. B. Mosley. Charge Measurements of
Particles Exiting Electrostatic Precipitators. Report EPA-600/7-80-077.
U.S. Environmental Protection Agency. Research Triangle Park, NC. April
1980.
2.	Anderson, M. H., J. R. McDonald, and R. B. Mosley. Particle Charge Mea-
surements. International Conference On Electrostatic Precipitation.
Monterey, CA, October 1981. APCA. Pittsburgh, PA, 1982.
3.	Sparks, L. E., G. H. Ramsey, and R. E. Valentine. Comparison of Electro-
static Precipitator Performance With Pulse And DC Energization. Ibid,
4.	Sparks, L. E., G. H. Ramsey, R. E. Valentine, and J. H. Abbott. Particle
Charging in an Electrostatic Precipitator by Pulse and DC Voltages. EPA
Third Symposium on the Transfer and Utilization of Particulate Control
Technology. Orlando, FL, March 1981.
5.	Gooch, J. P., G. H. Marchant, Jr., J. L. DuBard, and J. R. McDonald. An
Evaluation of Pulse Energization on a Hot-Side Electrostatic Precipitator.
EPRI Report CS-2634. Palo Alto, CA, November 1982.
6.	White, H. J. Industrial Electrostatic Precipitation. Addison-Wesley
Publishing Co. Reading, MA, 1963.
20-8

-------
2000
1000 
coo 
na

"i
o
f'
n>
ro
0
1	o
VD P-
P>
n
00
4
4
1	2
Particle Diameter, ym
Figure 2. Mean values of particle charge acquired
in steady corona at 35.5 kV and 19.5 nA/era2

-------
3000
N5
0
1
PULSED CORONA. TEST CONDITION I
133 Hi. 60 jjA. 2/18/83
29 fcV DC 81 AS
SHOWING THE CALCULATED MEAN AND
STANDARD DEVIATION (68% CONFIDENCE
INTERVAL) OF MEASURED CHARGE VALUES
K

PartLcLe Diameter, um
Figure 3. DES particle charging in pulser test
at 133 Hz and 8.67 nA/cm2.
2000
PULSED CORONA. TEST CONDITION I
200 Ht. 89 jjA. 1-5-83
29 kV DC BIAS
SHOWING THE CALCULATED MEAN AND
STANDARD DEVIATION (68* CONFIDENCE
INTERVAL) OF MEASURED CHARGE VALUES.
1000
rt
400
o

300
100
0.8
Particle Diameter, um
Figure 4. DES particle charging in pulser test
at 200 Hz and 13.0 nA/cm2.

-------
2000
	1 I	1	1	I	1
PULSED CORONA. TEST CONDITION II
200 Hi. 60jA. 2 16-B3
29 hV DC BIAS
SHOWING THE CALCULATED MEAN AND
STANDARD DEVIATION (68% CONFIDENCE
INTERVAL) OF MEASURED CHARGE VALUES.
1000
Particle Diameter, um
Figure 5. DES particle charging in pulser
at 200 Hz and 8.67 nA/cra^.
PULSED CORONA. TEST CONDITION I
200 Ht. 89 mA. 15-83
29 kV DC BIAS
SHOWING THE CALCULATED MEAN AND
STANDARD DEVIATION (Eg* CONFIDENCE
INTERVAL) OF MEASURED CHARGE VALUES
Particle Diameter, um
Figure 6. DES particle charging in pulser test
at 200 Hz and 13.0 nA/cm^.

-------
0 3 mi
0kV HVDC
0 4227 pC/m PER PULSE
0 6 mi
0 3 mi
6 m* 12 mi
RADIUS OF MEDIAN
CROSS SECTIONAL AREA
1.2 mi
2	3	4	S	6	7
Distance Wire-to-Cylinder, cm

Figure 7. Local values of ionic space charge density, after an
instantaneous corona discharge.
2.S
1.6
OS
17mi
0 6 mi
0 3 mi
29 0 kV HVDC
0.4227 pC/m PER PULSE
*
r
/
y


X
RADIUS OF MEDIAN
CROSS SECTIONAL AREA
I
3	4	6	6	7	B
Distance Wire-to-Cylinder, cm
10
Figure 8. Local values of electric field enhancement, after an
instantaneous corona discharge.
20-12

-------
2000
T
T
i
ARAPAHOE
6-15-63 5 ppm SOj
NO PULSE
1000
O
O
COO 
T3
-i
o
ZT
0>
-t
0C
<0
G
3
400
200
too

o o O
o o
1%
1
o
it
so
tol	I	I	I	I	L
0.4	0.6	1	2
Particle Diameter, uni
Figure 9.
Measured particle charge data
with DC energization.
2000
ARAPAHOE MSU
6-15.16.17-63
MEAN PARTICLE CHARGE WITH DC ENERGIZATION
1000
600
rt
O
n
;r
o>
 200
OQ
n>
G
p
rt
CO
100
60
40
0.4
0.6
Particle Diameter, u
Figure 10. Measured mean charge and standard
deviation, with DC energization.

-------
ARAPAHOE
6-16, 17 83
MEAN PARTICLE CHARGE WITH FAST PULSE
1000 
600 
400 
$
A
200
100 
60 
A
T 1/1
u
0_
J	I	L
I
I
'0.4	0.6	1	2
Particle Diameter, ym
Figure 11. Measured mean charge and standard
deviation, with fast pulses at 50 Hz.
3000
ARAPAHOE
615 83
MEAN PARTICLE CHARGE WITH LONG PULSE
1000
600
to 400
1
n
200
>-
en
100
60
0.4
06
Particle Diameter, ym
Figure 12. Measured mean charge and standard
deviation, with long pulses at 50 Hz.

-------
>1
o
c
0)
<4-1
w
t-l
c
a
o
t-i
ID
Oh
99.99
99.96: r
99.9 ^ L
99.8- -
"6::
99^ '
98
95--
W: r
80^
70--
.'"i
l5lij ,
Sv
V
I .
1 1 '
5 5 '
I ,!H,
I
i
i
2
i
o.oi
:: 0.05
=^0.1
--0 2
:r0-5
- - 2
--5
--10
i
rt
n
4-1
0)
a

IX
60
10-1
-II I < lift)
^20
$30
40
I Hill H|	1I I  HH|
10	101	io2
Particle Diameter, y
Figure 13. Size-fractional collection efficiency, with and without
fast pulses at 50 Hz.
i	1I i I I I	1	1 *fML 11 ' | rJ
w 2.0 
0
o
*4 1.6 
3
I
C
o
ahanmoc lomq *jiii
AHUM IMlClNCV7
AAAf AHO( FAIT Milt
AtiLINI tFFICIINCV lm
 I '	1 I I I I
06 0. 1	i	4	  M	
Particle Diameter, m
Figure 14. The difference in effective particle migration velocity,
with and without pulsing.
20-15

-------
Session 9: ESP: ADVANCED TECHNOLOGY II
Walter R. Piulle, Chairman
Electric Power Research Institute
Palo Alto, CA


-------
performance of large-diameter wires as discharge electrodes in electrostatic
PRECIPITATORS
P. Vann Bush and Duane H. Pontius
Southern Research Institute
Birmingham, Alabama 3 52 55-53 05
Leslie E. Sparks
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle I&rk, North Carolina 27711
ABSTRACT
Conventional practice in the design of corona discharge electrodes for
electrostatic precipitators (ESP*s) tends toward those having a relatively
low corona inception voltage. The theory indicates that maximal electric
field strength is of more importance to the precipitation process than is
corona current, since it is beneficial to both charging and collection of
particles. Only those particles smaller than about 0.2 micrometer diameter
are more readily charged by increasing the ion current density instead of the
electric field strength. Laboratory tests of a small pilot ESP (1000 acfm*)
and field tests of a large pilot system (30,000 acfm) comparing the perfor-
mance of 3/8-inch diameter wires with more nearly conventional 1/8-inch wires
have borne out the theory and resulted in substantially superior performance
for the larger electrodes.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
*To convert to metric equivalents, please use the factors provided at the end
of this paper.
21-1

-------
INTRODUCTION
The first successful implementation of electrostatic precipitators
(ESP's) in a commercial application is credited to Frederick Cottrell in
1907. The capacity of the mechanical rectifiers limited operating voltages
to 10-15 kV. This limitation forced Cottrell to adopt an electrode which
would develop corona discharge at these relatively low voltages. His choice
was a wire coated with cotton filaments which produced a uniform corona dis-
charge at these conditions. It was W.A. Schmidt who introduced the innova-
tion of a fine wire to produce a vigorous corona discharge at the available
voltage levels. The advances in transformer-rectifier capacity were rapid
and, long before Deutsch produced a mathematical expression describing parti-
cle collection in ESP's (1 922), voltages in excess of 100 kV were attainable
(1 )
In spite of the implications of the Deutsch model (discussed in the
following section) and advances in ESP energization methods, the philosophy
which led Schmidt to choose a fine wire for the corona discharge electrode
has persisted in ESP design practice. It is true that a broad variety of
discharge electrodes have been applied in commercial installations. Among
these are weighted wires, wires on frames or masts, barbed wires, and rigid
electrodes. Small radii of curvature and sharp edges or projections are
commonplace (1,2). Endemic to this set is the design philosophy of Schmidt:
the discharge electrode must be capable of producing a very dense ion current
for particle charging. Certainly other criteria are included in the selec-
tion of electrode design, such as ruggedness, electric field strength, and
corona current distribution.
It is known prima facie that a fine wire will produce a more vigorous
discharge at a given voltage than a wire having a larger diameter. The
larger diameter wire would have to operate at higher voltages to produce the
same level of ion generation. The assumption that ion production is of fore-
most importance has led designers to overlook or constrain other criteria,
such as electric field strength. The theoretical description of particle
charging and collection indicates that a modest level of ion density is ade-
quate, and an increase in the electric field strength yields substantially
improved charging and collection. In practice, the ion current density in an
ESP must be limited if the electrical resistivity of the dust is high so that
performance will not be degraded by back corona. This is not an uncommon
condition.
There are other considerations which favor fine wires over wires of
larger diameter. The discharge from fine wires is relatively uniformly dis-
tributed over the length of the wire. The degree of uniformity at a given
total ion flux diminishes as the wire diameter increases (3). This permits a
larger number of particles to pass the wire without encountering ions for
charging. Also, the operating range between corona inception and sparkover
diminishes as the wire diameter increases. These are serious considerations.
Laboratory and field experience with wires of different diameter have reveal-
ed that the larger diameter wires can provide superior performance in spite
21-2

-------
of these features. Tests indicate that the degree of performance benefit is
equal to the predicted enhancements to particle collection which result from
higher electric field strengths.
THEORETICAL CONSIDERATIONS
Two charging mechanisms are incorporated in theories describing particle
charging: 1) particles are bombarded by ions which are driven by the force
of the electric field; and 2) ions are attached to the particles by ion dif-
fusion. For the first mechanism, Field Charging, the saturation charge level
of the particle is directly proportional to the electric field strength. In
Diffusion Charging the rate of charge acquisition depends on the thermal
energy of the ions and the exposure time. It is only for particles less than
~0.2 ym diameter that ion diffusion becomes the dominant charging mechanism
(4). Typically, this size range of particles constitutes less than 0.5% of
the mass of the fly ash entering an ESP. The electric field strength has a
dominant role in charging 99.5% of the mass entering the precipitator.
The Deutsch-Anderson equation has been shown to provide a reasonable
estimate of ESP collection performance. It is especially useful as a tool
for comparing the effects of changes in the operating parameters of an ESP.
A simple form of the equation is
P = exp (-wA/V)	(1)
where P is the overall penetration of the ESP (P = 1 - collection efficien-
cy), w is the precipitation rate parameter, A is the total collecting surface
area, and V is the gas volume flow rate. The precipitation rate parameter
for particles of radius a is of the form
w = aE2/2ir6	(2)
where E is the average electric field strength and 6 is the gas viscosity.
Combining Equations (1) and (2) and taking the logarithm results in
JlnP = aAE2/2rrV6 .	(3)
If we wish to determine the effect of variations in any of the parameters we
may use Equation (3). Specifically, the effect that changing the electric
field strength from Eq to E produces in the penetration Pq, with all other
parameters constant, can be derived easily from the result
nP/nP0 - (E/E Q) 2
Solving for P,
P = exp [(E/E0)2&nPQ] .	(4)
A family of curves resulting from this expression is shown in Figure 1.
21-3

-------
Figure 1. Theoretical predictions of the effect of electric field strength variations
on penetration.
21-4

-------
Of more interest than the values of penetration presented in Figure 1
are the trends that are indicated. Relatively small increases in the elec-
tric field strength result in a notable penetration reduction.
FIELD TEST RESULTS
The advantage of high electric field strength for collection of charged
particles led to the selection of large diameter wires as discharge elec-
trodes in the EPA/SoRl large pilot demonstration ESP. The pilot ESP was
constructed to demonstrate two-stage ESP technology for high resistivity fly
ash collection. The system was designed for 30,000 acfm of flue gas capac-
ity. The collector stage of the two-stage system consisted of four electri-
cal sections of commercial ESP design with the exception of the discharge
electrodes. The total specific collection area (SCA) as tested was 287
ft2/kacfm. Three-eighth inch diameter wires were installed in the Lodge-
Cottrell mast suspension. The wires were mounted 3-3/4 inches apart. There
were a total of 60 wires (10 masts) in each of the 13 lanes of the collector.
The lanes were 10 inches wide by 10 feet-3 inches high. Other details of the
pilot ESP design have been presented elsewhere (5,6).
The specific diameter of the discharge wires was selected based on
limited tests in the EPA small pilot ESP at Research Triangle Park, North
Carolina. The criteria for selection were combined high electric field
strength and low current operating regime, and feasibility of implementation.
These criteria were applied to the selection because the electrodes were
intended to be used in a collector stage downsteam of a particle charging
device. In this application it was expected that the absence of current
uniformity and low total current levels could be tolerated.
TEST CONDITIONS
The large pilot ESP was evaluated over a period of 2-1/2 years. For the
first two years of study the principal objective was to evaluate the two-
stage technology with the 3/8-inch diameter discharge electrodes in the col-
lector stage. The final phase of tests conducted in the pilot demonstration
program evaluated the precharger as a retrofit to existing ESP's. For this
study the collector stage was fitted with 1/8-inch diameter wires in the
Lodge-Cottrell mast suspension. Measurements of particle penetration of the
collector stage alone were made routinely, as baseline performance data.
These data permit direct comparisons to be made between the two discharge
electrodes under identical conditions.
The tests were conducted at a flue gas temperature of 325F. This pro-
vided an ash resistivity of 2-4 x 10*2 ohm-cm. Inlet fly ash mass loadings
for the tests were ~4.6 gr/acf. The mass median diameter of the ash was ~20
Vim.
21-5

-------
RESULTS OF COMPARATIVE TESTS
Voltage-current (V-I) characteristics for the 3/8-inch diameter wires at
the test conditions are shown in Figure 2. The operating range between onset
of corona and the onset of sparking and back corona was from 41 to 49 kV
applied voltage (3.2 to 3.9 kV/cm average electric field strength), and up to
7 uA/ft2 current density. During the tests which were used for the compari-
sons, the ESP with 3/8-inch diameter wires operated with applied voltages in
the range 42 to 48 kV, producing average electric field strengths of 3.3 to
3.8 kV/cm. The average current density was 4.7 pA/ft2. This corresponds to
a total power consumption of ~1700 watts.
The V-I curves for the 1/8-inch wires are presented in Figure 3. The
voltage range attainable for these electrodes was 36 to 41 kV, giving an
average electric field strength of 2.8 to 3.2 kV/cm. The corresponding aver-
age current density was 6.8 yA/ft2, giving a total power consumption of ~2100
watts.
The electrical operating conditions and resulting collection efficien-
cies for the two electrode tests are summarized in Table I. For the 3/8-inch
diameter wires the average of total mass collection efficiency measurements
was 99.64%. This high efficiency was typical of the performance of these
electrodes for all tests at these conditions. The average total mass collec-
tion efficiency for the 1/8-inch diameter wires was 98.58%. The difference
in penetration for the two cases was a factor of four (.0036 to .0142).
TABLE I. ESP PERFORMANCE
3/8-inch wires 1/8-inch wires
Electric field strength, kV/cm
Current density, yA/ft2
Power consumption, watts
Collection efficiency, %
3.3-3.8
4.7
1700
99.64
3.0-3.2
6.8
2100
98.58
The magnitude of the difference in collection efficiency for the two
wire diameters is great when compared to the differences in electrical oper-
ating modes. The larger wires functioned with 30% less current density, 20%
less power consumption, and 14% greater average electric field strength. The
difference in penetration is plotted in Figure 4, superimposed on the theo-
retical projections which were shown in Figure 1. The data points fit the
theoretical relationship between penetration and electric field strength.
This confirms the argument that electric field strength plays the dominant
role in ESP performance, and small changes in the strength of the electric
field have a large effect on the penetration.
21-6

-------
140
120
COLLECTOR SECTIONS
\
\
\
\
\
\

A
20	30
VOLTAGE, kV
Figure 2. Voltage-current curves for 3/8-inch diameter wires at 325F.
21-7

-------
40
SPARKING
COLLECTOR SECTIONS
<
E
h-"
z
LU
cc
cc
D
u
40
50
VOLTAGE, kV
Figure 3. Voltage-current curves of 1/8-inch diameter wires at 325F.
21-8

-------
90
5.0
2.0
1/8 IN. WIRE
O
l-
<
QC
h-
LU
Z
LU
a.
99
u.
0.5
99.5
3/8 IN. ROD
0.2
99.8
0.1
99.9
1.5
1.3
1.4
Figure 4. Measured penetrations at two values of electric field strength superimposed on
theoretical curves.
21-9

-------
LABORATORY CONFIRMING EVIDENCE
Tests were conducted on a 1000 acfm pilot ESP at the EPA Industrial
Environmental Research Laboratory in Research Triangle Park, North Carolina.
The tests were designed to provide an independent assessment of the
performance of the large diameter electrodes versus the conventional dis-
charge electrode design. The pilot ESP was tested with 3/8- and 1/8-inch
diameter electrodes installed in a 9-inch wide duct. Wire-to-wire spacing
was 9 inches. There are four electrical sections in the single-lane pilot
ESP, each with 32 ft2 collecting surface area. The SCA for the ESP is
therefore 128 ft2/kacfm.
Fly ashes from two sources were used in the tests. The resistivities of
the two ashes were ~2x1012 and 9x1O10 ohm-cm at 300F. Comparative perfor-
mance for 3/8- and 1/8-inch diameter discharge electrodes was determined from
cascade impactor and Method 17 mass concentration measurements. Operating
points for the two electrode configurations were chosen to given the lowest
possible outlet emissions as monitored with an opacity meter. In general,
both electrodes operated with very limited sparking.
The comparative performance of the discharge electrodes in the presence
of high resistivity fly ash (low 1012 ohm-cm) was, within the accuracy of the
measurements, equivalent to the results from the field measurements on the
large pilot ESP at Bull Run Steam Plant. The required low current density
operation and the large difference in the electric field strength favored the
large diameter discharge electrode. The 3/8-inch diameter wires were approx-
imately four times as effective in reducing penetration as the 1/8-inch
wires. These results were anticipated due to the theoretical support as well
as the field test results.
The performance benefit of large diameter wires in the case of moderate
resistivity fly ash was not as straightforward to predict. The large diame-
ter electrodes operate near the sparking limit with low current emission.
The operating current level is constrained by the wire-plate separation and
not the precipitated ash. The relatively small diameter electrode has a much
broader range of current emission, and the operating level is typically con-
strained by the ash resistivity. In the limit, reducing the ash resistivity
has the effect of reducing the difference in electric field strength between
the large and small diameter wires, and increasing the difference in operat-
ing current density. In fact, the difference in the electric field strength
increases to a maximum value and then decreases as the resistivity of the ash
decreases. The trends are illustrated in Figure 5.
Tests were conducted in the laboratory with fly ash having a resistivity
of 9x1Q10 ohm-cm at 300F. The average electrical operating values for the
pilot ESP (averaged over the four electrical sections and several tests) are
presented in Table II. The percent differences between the 1/8-inch diameter
electrodes and the 3/8-inch diameter electrodes are also given in the Table.
In this case, there were large differences in current and voltage. The meas-
ured fractional penetrations are shown in Figure 6 and listed in Table III.
The percent difference in penetration as a function of particle diameter is
plotted in Figure 7. The mean difference in penetration was ~21%.
21-10

-------
TABLE II. LABORATORY ESP OPERATING DATA FOR MODERATE RESISTIVITY FLY ASH
Averages;
3/8-inch
1/8-inch
Units
% Difference
Voltage:	52.510
Current:	0.302
Current density:	9.438
Power:	0.485
31 .167
0.899
28.099
0.877
kV
mA
UA/ft2
watts
-68.48
66.41
66.41
44.64
TABLE III. PENETRATION VS. DIAMETER MEASURED IN LABORATORY TEST WITH
MODERATE RESISTIVITY FLY ASH

3/8-inch wire
1/8-inch wire
%
Particle
average
average
Difference in
diameter
penetration
penetration
penetration
0.2
0.406
0.360
-12.778
0.3
0.655
0.535
-22.430
0.4
0.369
0.288
-28.278
0.5
0.255
0.225
-13.497
0.8
0.247
0.285
13.333
1
0.182
0.286
36.510
1 .5
0.155
0.301
48.571
2
0.169
0.305
44.479
2.5
0.163
0.288
43.417
4
0.089
0.172
48.245
5
0.058
0.112
47.611
7.5
0.019
0.1 75
89.413
10
0.029
0.041
27.753
15
0.060
0.052
-14.776
20
0.045
0.048
4.870
Averages:
0.193
0.231
20.830
21-11

-------
SPARK
1/8 MAX
1/8-INCH
LU
SPARK
3/8 MAX
3/8-INCH
VOLTAGE
Figure 5. Typical voltage-current characteristics of 3/8-inch and 1/8-inch diameter
discharge electrodes.
21-12

-------
98
95 |		 3/8-INCH WIRES
1/8-INCH WIRES
90
80
70
60
50
40
0.1
0.01

V \
30 	b\ J-*- ""-"
< 20	Nx-	i
\
\
2 10 1		".
51-
2 
1
0.1	0.2	0.4 0.6 0.8 1.0	2.0	4.0 6.0 8.010.0 20.0
GEOMETRIC MEAN DIAMETER, micrometers
Figure 6. Measured fractional penetration for small pilot ESP with 3/8- and 1/8-inch
diameter discharge electrodes. Temperature = 300 F; ash resistivity =
9x 10^0 ohm-cm.
21-13

-------
0.3 0.4 0.5 0.7 1.0	2.0	4.0	10.0 20.0
PARTICLE DIAMETER, micrometers
Figure 7. Percent difference in fractional penetration between 1/8- and 3/8-inch
diameter discharge electrodes. Temperature - 300 F; ash resistivity =
9x 1010 ohm-cm.
21-14

-------
CONCLUSIONS
Electrodes with emitting surfaces having small radii of curvature have
dominated commercial ESP designs. This feature has been preferred because of
the uniformity of current emission, high capacity for current emission, and
large operating range of voltage between corona onset and sparkover. These
characteristics are all desirable. However, in practice the operating cur-
rent level is restricted to a moderate-to-low level in many cases to prevent
back corona formation. Low current levels are adequate to charge particles
to saturation within a few feet in the ESP. These facts, combined with the
theoretical descriptions of electrostatic precipitation, support the practice
of maximizing the electric field strength in the design of electrodes.
Tests in a large pilot ESP have confirmed the accuracy of the theoreti-
cal models. Tests with 3/8- and 1/8-inch diameter discharge wires in the
same ESP under the same flue gas conditions enabled a direct comparison of
the effect of electrical operating conditions on ESP performance. The
average electric field strength differed by less than half the difference in
current density. The change in the strength of the electric field completely
accounted for the very large change in penetration.
Laboratory tests with high resistivity fly ash in a small pilot ESP
yielded results in agreement with the field test data. Tests in the labora-
tory with moderate resistivity indicated a reduced level of performance bene-
fit for the 3/8-inch diameter electrodes.
Conventional practice in the design of corona discharge electrodes has
been based on historical precedent and incomplete understanding of the impli-
cations of the theoretical models of electrostatic precipitation. Each pre-
cipitator installation should be designed in light of the pertinent con-
straints on performance, such as ash resistivity and loading. In general the
bias in the design of electrodes should be toward those which operate at the
highest electric field strength. This technology is definitely suited to
Retrofit applications. Significant performance benefits may result in many
instances from the use of large diameter wires as discharge electrodes.
AC KNOWLED GEMENTS
Much of this work was performed under EPA Contract No. 68-02-2 683. The
Measurements of the performance of the large pilot scale ESP were supervised
by Todd Snyder of Southern Research Institute. The laboratory experiments
were completed under the direction of Geddes Ramsey of EPA. Their efforts
are gratefully acknowledged.
21-15

-------
REFERENCES
1.	White, H.J. Industrial electrostatic precipitation. Addison-Wesley,
Reading, Massachusetts, 1963. pp 363-364.
2.	Lagarias, J.S. Discharge electrodes and electrostatic precipitation.
J. of APCA 10(4): 271-274, 1960.
3.	Bush, P.V. and Snyder, T.R. Laboratory analysis of corona discharge
electrodes and back corona phenomena. Paper presented the The Fifth Sym-
posium on the Transfer and Utilization of Particulate Control Technology,
U.S. EPA and EPRI, Kansas City, Missouri. August 27-30, 1984.
4.	Pontius, D.H., Felix, L.G., McDonald, J.R., and Smith, W.B. Fine parti-
cle charging development. EPA-600/2-77-173 (NTIS PB271727). U.S.
Environmental Protection Agency, Research Triangle Park, North Carolina,
1977. 240 pp.
5.	Bush, P.V. and Pontius, D.H. Pilot demonstration of the precharger-col-
lector system. In: Proceedings of the Third Symposium on the Transfer
and Utilization of Particulate Control Technology: Volume I. Control of
Emissions from Coal Fired Boilers. EPA-600/9-82-005a (NTIS PB83-149583).
U.S. Environmental Protection Agency, Research Triangle Park, North
Carolina, 1982. pp 157-164.
6.	Bush, P.V. and Pontius, D.H. Pilot demonstration two-stage ESP test
results. Presented at the Fourth Symposium on the Transfer and Utiliza-
tion of Particulate Control Technology, U.S. Environmental Protection
Agency, Houston, Texas. October 11-15, 1982.
CONVERSION FACTORS
Readers more familiar with metric units may use the following to convert
to that system.
Nonmetric	Times	Yields Metric
F	5/9(F-32)	C
ft	30.48	cm
ft2	0.093	m2
ft3	28.3	liters
gr	0.065	g
in.	2.54	cm
21-16

-------
TECHNICAL EVALUATION OF PLATE SPACING EFFECTS
ON FLY ASH COLLECTION IN PRECIPITATORS
Ralph F. Altman
Electric Power Research Institute
Chattanooga, TN 37411
Gerald W. Driggers, Ronald W. Gray
Combustion Engineering
Birmingham, AL 35243
James L DuBard
Southern Research Institute
Birmingham, AL 35255
E. C. Landham, Jr.
Southern Research Institute
Denver, CO 80210
ABSTRACT
A joint program between EPRI and C-E is being conducted to evaluate
the effects of increasing plate spacing and applied voltage in preci-
pitators collecting coal combustion fly ash. Plate spacings will range
from nine (9) inches to twenty-two (22) inches during the experimental
program being conducted at the EPRI Arapahoe facility pilot plant.
Spacings up to 31.5 inches (800 mm) have been reported on non-fly ash
applications and up to 15.7 inches (400 mm) on coal fly ash outside the
United States with apparent success. This is the first major experimental
program in the U.S. to systematically examine the effects on collection
efficiency of various plate spacings, applied voltages, and fly ash
resistivities. Total mass and fractional collection efficiencies will be
measured under the various operating conditions, and these data will be
examined by various analytical techniques to determine the best approach
for comparing performance under the varying conditions and to facilitate
scale-up to full size precipitators.
This paper reports on the results thus far obtained and their
interpretation.
22-1

-------
INTRODUCTION
There are more than one thousand electrostatic precipitators in
operation at utility plants today; and although few utilities have major
construction programs underway, more utility precipitators will
undoubtedly be built in the future. Further, the performance of many
existing precipitators will have to be improved to comply with changing
particulate emission regulations or to cope with changing coal supplies
and operating conditions. As a result, the electric utility industry,
through the Electric Power Research Institute, is sponsoring a number of
research projects to improve electrostatic precipitator performance and
reduce cost.
The goal of one of the principal projects in this area is the
determination of the effect of plate spacing on precipitator performance.
The Environmental Systems Division of Combustion Engineering is a
cosponsor and the prime contractor for the work. The interest in this
topic derives from international experimental and operating evidence that
ESP performance is relatively insensitive to plate spacing and that
required total collecting plate area can be reduced due to the higher
voltages possible at wider spacings.
A survey of precipitator installations in other countries has
identified over 200 industrial or utility units with plate spacings
exceeding 305mm (12 inches) either in operation or under construction.
Most of these have a spacing of 400 mm (15.7 inches) with some in Japan as
great as 800 mm (31.5 inches). Although primarily industrial units, there
are also several flyash installations either in operation or under
construction. Most data on these units is proprietary and not in the open
literature. It is particularly notable, however, that there are twelve
(12) 600 MW or larger units on order or under construction in South Africa
with AO0 mm spacing, reflecting the convincing nature of the available
data. This aggressive approach is based on economic considerations since
the wide spaced ESP is lighter and less costly to build. In fact, early
cost estimates made as a part of this project indicate a potential capital
cost reduction of 10-15%. This analysis is illustrated on Figure 1.
Some of the early work on the effect of plate spacing was by G.
Heinrich using data supplied by Dr. R. F. Heinrich (1). The work
suggested that precipitator performance was related to specific corona
power and independent of specific collecting area. This work resulted in
a patent on the subject which has since expired (1) (2). In fact, White
(3) derived an expression that relates precipitator particle loss, and
hence collection efficiency, to corona power and volume flow rate only.
Although this approximate expression was not specifically derived to
justify increased plate spacings, the statement is made that "the details
of the precipitator design do not enter into the basic relation between
precipitator loss and corona power."
22-2

-------
BASIS: FLANGE TO FLANGE COSTS
1,500,000 ACFM ESP
99.8% EFFICIENCY
10% LONGER BOX
AT WIDE SPACINGS
STANDARD BOX
10
11
12
13
14
15
16
17
18
19
20
21
22
PLATE SPACING, INCHES
Figure 1. Normalized Cost - vs - Plate Spacing for a Large Precipitator
22-3

-------
More recent work conducted in Japan (4) has provided experimental
evidence to support the connection between the electric field strength in
the vicinity of the collection plate and precipitator collection
efficiency. The thrust of this work is that effective migration velocity
is proportional to this electric field and that the field near the plate
increases as plate spacing increases. However, other work indicates the
effect of plate spacing is sensitive to operating parameters (5) or to
other details of construction (6). For further background reading on this
subject, see reference (1) and references cited therein.
EXPERIMENT PLAN
In view of the potential benefit to be gained from increasing
precipitator plate spacing (reduced weight and cost and simplified
construction), a project has been initiated to develop a definitive
understanding of the effect of plate spacing on the performance of
precipitators in utility applications. The first phase of the work Is
being conducted at the Arapahoe Test Facility where EPRI conducts many of
its pilot scale tests of environmental control systems (7). In phase one,
a 5000 ACFM pilot precipitator will be operated with plates spaced from 9
to 22 inches. Flue gas will be drawn from unit four of Public Service
Company of Colorado's Arapahoe Power Plant for all of these tests. The
tests are exploratory in nature; no attempt will be made to optimize
precipitator design for each plate spacing. If results from small scale
pilot tests are encouraging, then verification tests will be conducted at
a much larger scale with more attention paid to details of construction.
The pilot precipitator at Arapahoe has three independent fields in
the direction of gas flow. Each field is energized by a transformer-
rectifier (TR) set with maximum ratings of 150KV and 40mA. Plates are
rapped on the trailing edge and discharge wire frames are rapped at both
leading and trailing edges. Before commencing the performance tests,
certain modifications were made to the precipitator to ensure a uniform
gas velocity profile at the inlet and to minimize electrical section
bypassage (sneakage). Other modifications were made to provide the
clearances necessary for operation with 22 inch plate spacing. A one-half
scale clear acrylic model of the precipitator was constructed, and air
flow tests were conducted at Combustion Engineering's model facility in
Birmingham, Alabama to select the baffles and inlet and outlet flow
control devices. At the conclusion of the flow tests, it was determined
that maximum electrical section bypassage should be less than 5% per field
and that RMS deviation in the gas flow distribution at the precipitator
inlet was less than 10%.
In addition to the flow model study, a limited number of tests were
conducted to determine a suitable discharge electrode spacing for
different plate spacings. These tests were performed in air at ambient
temperature using a simple wire frame and a segmented collecting
electrode.
22-4

-------
The current distributions at the plate measured with this apparatus
indicated that a single wire spacing of 6" would produce reasonable
voltage/current curves and a relatively uniform current distribution at
the plate for all plate spacings (9 through 22 inches). Although this
result was unexpected, performance tests conducted at Arapahoe thus far
indicated the decision to use a single wire spacing was justified. The
data gathered at 9, 12, and 18 inches (the 22 inch performance tests were
not completed in time to be reported in this paper) have yielded
relatively high efficiencies under all test conditions. More work must be
done to optimize discharge electrode spacing and design; however, the
single wire spacing has produced satisfactory data for purposes of
comparison.
Following the completion of the flow tests, discharge electrode
spacing study, and precipitator modifications, performance testing was
begun at Arapahoe in January, 1984. Southern Research Institute is a
major subcontractor and their personnel are responsible for these tests
and subsequent data analysis. The test conditions are summarized in
Table 1. Inlet flue gas temperature was held constant at 300F, and the
precipitator was operated 24 hours a day at constant flow during all
performance tests. All three fields were kept energized. Two flue gas
flow rates were used to provide variation in gas treatment time and SCA.
Sulfur trioxide (SOg) conditioning is used to adjust fly ash resistivity
from its normal level of approximately 10 ohm-cm to about 10
ohm-cm so that data that can be applied to both high and low sulfur coal
applications are obtained.
TABLE 1. PLANNED AND TESTED CONDITIONS FOR PILOT SCALE TESTS AT ARAPAHOE
Test	Plate Spacing SCA (Nominal) Flue Gas	SO^
Condition Inches	ft /1000ACFM Flow Rate ACFM Conditioning
9 - 1
9
370
3,630
not used
12 - 1
12
275
3,690
not used
18 - 1
18
185
3,600
not used
22 - 1
22
185
3,600
not used
9-2
9
260
4,900
not used
12-2
12
195
5,220
not used
18-2
18
130
4,670
not used
22-2
22
130
4,900
not used
9-3
9
370
3,520
used
12-3
12
275
3,630
used
18-3
18
185
3,600
used
22-3
22
185
3,600
used
9-4
9
260
5,000
used
12-4
12
195
4,720
used
18-4
18
130
4,900
used
22-4
22
130
4,900
used
22-5

-------
Comprehensive sets of performance data are being gathered at each
test condition. These data include total mass collection efficiency
measurements made using mass trains (EPA method 17) and particle size
efficiency measurements made with impactors. Flue gas composition
including K^O, SC^, SOg, and O2 is determined on a periodic
basis. Fly ash resistivity is measured daily using an in-situ probe with
a point-to-plane geometry. Ash samples are gathered periodically for
chemical analysis. In addition to these performance measurements, a
modified back ionization probe developed by Dr. Senichi Masuda (8) in
cooperation with Kanomox Corp. of Japan is being used to measure the
electric potential near the collection plate for selected operating
conditions at each plate spacing. A knowledge of the electric field which
ca*ri be derived from the potential data should help provide an
understanding of the efficiency data since the migration velocity of fly
ash particles to the plate should be roughly proportional to this field.
At the time of preparation of this paper, the potential probe data are
being analyzed, and these results will be reported elsewhere.
EXPERIMENTAL RESULTS
At the beginning of the tests conducted at each plate spacing,
voltage-current data are taken at ambient conditions and plotted. These
plots, presented in Figures 2 and 3, were used to verify alignment,
compare electrical behavior, and to determine the voltage at which
sparkover occurs. Optimum ESP performance cannot be achieved if power
input to any field is limited by sparking from a discharge element to any
grounded surface except the collection plates. Voltage limited by
sparking to a support beam or baffle is an indication of inadequate
clearance. The clearances needed for 9" and 12" plate spacings are well
established and can be incorporated into the design of a small pilot
precipitator without difficulty. The clearances needed for operation at
18 and 22 inches were selected at the beginning of the program based on
experimental data (provided by Combustion Engineering) for sparkover
voltages as a function of high voltage and grounded shapes. It was more
difficult to provide the larger clearances without increasing gas sneakage
to an unacceptable level in this small pilot, but a satisfactory design
was developed; and the pilot ESP was modified to provide the necessary
clearances. Observations made at the beginning of the tests at each
spacing verified that voltage was limited by sparking to the collection
plates, and voltage-current curves demonstrated that sparkover voltage was
roughly proportional to the plate spacing.
After the alignment and clearances were verified at each spacing,
flue gas flow through the pilot ESP was started, and the precipitator was
operated for several days. When the unit reached reasonably stable
operating conditions, performance data gathering was begun. As stated
earlier, both total mass concentrations and particle size distributions
are measured at the inlet and outlet of the precipitator during the
performance tests.
22-6

-------
0
4
A FIELD
<%
O B FIELD	^
 C FIELD

*>
e
$


20 25	30	35 40	45 50	55 60 65
SECONDARY VOLTAGE, KV
Figure 2. Clean Plate Air Load, V-I Data for the Three Electrical
Sections Taken at Ambient Conditions
22-7

-------
100
90
9 - INCH
12 - INCH
18 - INCH
80
70
UJ 50
20
20
30
40
50
60
70
80
90
SECONDARY VOLTAGE, kV
Figure 3. Clean Plate Air Load V-I Curves for the Three Spacings
22-8

-------
At the time this paper was prepared, only the mass concentration data
had been analyzed, so only these tests with plates spaced at 9 and 12
inches and the high resistivity tests with plates spaced at 18 inches were
complete. However, these preliminary data are both interesting and
encouraging.
Before presenting the mass efficiency data, it is necessary to
discuss the in-situ resistivity data taken during the tests since these
data are needed to interpret the performance results. The resistivity
data are plotted in Figure 4. The inherent resistivity of the fly ash at
Arapahoe is high, usually above lxlO1^ ohm-cm. In fact, the resistivity
was greater than this value during all of the tests with no gas
conditioning except one. During the low flow, high resistivity test with
the plates spaced at 9 inches (condition 9-1 in Table 1), the average
resistivity was approximately 6X1011 ohm-cm, which is one-half order of
magnitude lower than the resistivity measured during the other test
conducted without flue gas conditioning. This change in resistivity is
attributed to a change in the coal supply that took place during the 9
inch tests. During the tests conducted with SOg conditioning, it is
possible to exercise better control over fly ash resistivity, and the
resistivity has stayed in the 3 to 5x10 ohm-cm range.
The average collection efficiency data from the tests completed so far
are plotted in Figure 5. These data show the expected trends with flue
gas flow rate and fly ash resistivity. This fact is apparent if the data
collected with the plates spaced at 9 inches is examined. When the
resistivity and the gas flow rate were both high, the measured collection
efficiency was 98.5%. When the flow rate was low, the efficiency
increased to 99.4% (but recall the resistivity was lower than normal
during these tests). When the resistivity was lowered to 4x10 ohm-cm,
the collection efficiency was determined to be 99.7% at the high flow rate
and 99.9% at the low flow rate. The same trends are evident in the data
gathered with the plates spaced at 12 inches. During the high resistivity
tests the collection efficiency was 98.4% at high flow rate and 99.0% at
the low flow rate and increased to 99.8% and 99.9% at the high and low
flow rates respectively when the resistivity was lowered by S03
conditioning.
The most interesting data in Figure 5 are the two average efficiency
points for the high resistivity tests with the plates spaced at 18 inches
(the only 18 inch tests completed when this paper was being prepared).
The efficiencies are unchanged from those measured with the plates spaced
at 12 inches, i.e. 98.4% at the high flow rate and 99.0% at the low flow
rate. In fact, the collection efficiency changed very little during all
of the high flow, high resistivity tests at 9, 12, and 18 inches. The
changes that did take place are comparable to the day to day fluctuation
in the efficiency data at a fixed spacing. The exception to this trend is
the low flow rate, 9 inch efficiency data which was taken during a period
of time when the resistivity was lower than usual.
22-9

-------
101
,13
101
>12
HIGH
FLOW
LOW AND
HIGH FLOW
LOW AND
HIGH FLOW

X
o
>-
t
>
I-
e/j
V)
LU
oc
10
LOW
FLOW
LOW AND
HIGH FLOW
LOW AND
HIGH FLOW
101
,10 1_
12
DUCT SPACING, INCHES
18
Figure 4. Resistivity Measured In-Situ During Various Phases of
the Test Program
22-10

-------
100.0
ui
2
fr 99.0
ui
8
98.0
LOW FLOW
2 - 5x1010OHM - CM
+
HIGH FLOW
LOW FLOW
6x10" OHM - CM
LOW FLOW
^IxlO^OHM - CM
- - +

HIGH FLOW
2 - 3x1012OHM - CM
-X
12
18
DUCT SPACING, INCHES
Figure 5. Comparison of Efficiencies Obtained Under Different Flow
and Resistivity Conditions
22-11

-------
The efficiency data are compatible with the operating voltages
measured during the performance tests. These data, which are summarized
in Table 2, show that the average voltage for each field and for the
precipitator as a whole was roughly proportional to the plate spacing.
This proportionality occurred even though spark rate was used to control
the voltage in all fields during all of the tests. These data and the
data of others suggest that, for high resistivity applications, operating
voltage must be at least proportional to plate spacing if collection
efficiency is to be maintained as the plate spacing is increased.
TABLE 2. ELECTRICAL OPERATING VOLTAGES FOR VARIOUS CONDITIONS
Spacing Resistivity	A FIELD B FIELD C FIELD ESP AVERAGE
(inches) (ohm-cm)	kV	kV	kV	kV
9
1012
33
32
30
32

o
t
o
36
36
34
35
12
1012
46
45
38
43

1010
51
51
50
51
18
1012
64
68
65
66
CONCLUSIONS
Since all of the tests and data analysis from the program have not been
completed, the conclusions drawn for the existing data analyses must be
considered tentative. However, it is true that, so far, the total mass
collection efficiency has shown very little sensitivity to increased plate
spacing and reduced SCA. This point is illustrated most clearly by the
data gathered with a high flue gas flow rate and high fly ash resistivity
with the plates spaced at 9, 12, and 18 inches and by the data gathered
with low resistivity with the plates spaced at 9 and 12 inches. The
preliminary economic analyses indicate that there is a potential capital
cost saving for a new precipitator of from 10 to 20% if the plate spacing
is increased to a range of 16 to 20 inches. If the results from the tests
that have not been completed continue to show little sensitivity to plate
spacing, then further work at a larger scale will be needed to begin to
optimize performance at wider plate spacings.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the views of the Agency and no official endorsement
should be inferred.
22-12

-------
REFERENCES
1.	Heinrich, Dieter 0. Review and some practical aspects of wide duct
spacing. In: Proceedings of the First International Conference on
Electrostatic Precipitation. Air Pollution Control Association,
Pittsburgh, PA 15230, 1982. p. 638.
2.	Darby, K. The use of pilot precipitators for process and design
development. Paper presented at second CSIRO Conference on
Electrostatic Precipitation, August, 1983.
3.	White, Harry J. Industrial Electrostatic Precipitation , Addison-
Wesley Publishing Co., Inc., 1963.
4.	Misaka, T., Sugimoto, K., and Yamada, H. Electric field strength and
collection efficiency of electrostatic precipitator having wide plate
pitches. Paper presented at CSIRO Conference on Electrostatic
Precipitation, August, 1978.
5.	Weber, E. and Wiggers, H. Investigations in a laboratory precipitator
concerning the influence of duct width and gas velocity on the
collection efficiency and the dust concentration distribution. Staub-
Reinh. Luft 40, 1980. p. 469-502.
6.	Pearson, K. and Puttick, D. Operating experience with narrow spaced
multi-rapped flat plate precipitators treating fly ash from low sulfur
coals, and a reconciliation of wide versus narrow spaced collecting
plates. Paper presented at Second CSIRO Conference on Electrostatic
Precipitation, August, 1983.
7.	Altman, R. and Carr, R. EPRI sponsored research at the Arapahoe Test
Facility. DOE/METC/84-13 Vol. 2. U.S. Department of Energy, Grand
Forks, ND, February 1984. p. 783.
22-13

-------
ELECTRICAL CHARACTERISTICS OF LARGE-DIAMETER
DISCHARGE ELECTRODES IN ELECTROSTATIC
PRECIPITATORS
Kenneth J. McLean
University of Wollongong
Wollongong, NSW, 2500, Australia
Leslie E. Sparks
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
ABSTRACT
The paper discusses some of the electrical characteristics that arise in
an electrostatic precipitator with large-diameter discharge electrodes
collecting high resistivity flyash. These electrodes are effective because,
once they are contaminated, corona current flows at voltages well below that
which conventional theory predicts. The voltage level however is greater than
that of the conventionally sized electrodes, and this produces higher charging
and collecting electric fields. The individual tuft discharges that produce
this corona current are randomly distributed along the discharge electrode,
are often unstable, and produce areas of high current density which can easily
charge a particle to saturation. The instability and distribution of these
tuft discharges affect the formation of back corona and reduce its deleterious
effects. The optimum diameter is about 9.5 mm [3/8-in.] for the test
condition.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
INTRODUCTION
One of the most interesting recent proposals for improving the performance
of conventional electrostatic precipitators collecting high resistivity fly-
ash is to use large-diameter discharge electrodes. Their potential advantage
was first observed at the EPA/SORI test facility at the TVA station, Bull Run,
where the conventional 3.2 mm (1/8-in.) wires were replaced with 9.5 mm
23-1

-------
[3/8-in.] diameter wires with the result that both the voltage levels
increased and the collection efficiency improved (1).
This effect was further investigated at the EPA in-house Pilot Electro-
static Precipitator at Research Triangle Park, N.C., where it was confirmed
that the voltage levels increased and efficiences improved with the larger
diameter discharge electrodes.
The purpose of this paper is to identify the special electrical and corona
discharge characteristics that occur in electrostatic precipitators with
large-diameter electrodes collecting high resistivity flyash and to speculate
on how they affect the collection mechanism.
EXPERIMENTAL PROCEDURES
The fundamental mechanisms of corona discharge formed on contaminated dis-
charge wires of different sizes were investigated using the EPA Optical Elect-
rostatic Precipitator. This unit comprises a single gas lane with plates 1.2 m
square, made of plexiglass and operated at ambient temperature. The plates
are covered with strips of aluminium tape separated from one another by a
small gap that provides a means of measuring the current distribution. These
strips can be arranged to run parallel or at right angles to the discharge
wires. The unit is surrounded by a dark curtain so the nature of the electri-
cal discharges can be observed through the plexiglass frame and panels.
The individual tuft discharges were simulated by a small metal projection
on the discharge wire. This produced a stable single discharge which could
be easily investigated. For other tests, the discharge wire was covered with
flyash, and the characteristics of the naturally produced tuft discharges were
identified and measured.
Measurements at elevated temperatures and identification of back corona
effects were made on the EPA's Pilot Electrostatic Precipitator. This unit
comprises a single pass with four zones comprising collecting plates 1.2 m
square. For these tests, gas flow was 0.47 m3sec at 150 C and the ash in-
jected with an air blast gun to give a dust concentration at the inlet of
2.3 g/m3. The electrical measurements recorded were made at the inlet zone
and current distribution measured by means of a special plate comprising a
matrix of isolated 50 mm square segments covering an area of 610 mm by 406 mm.
ELECTRICAL CHARACTERISTICS
VOLTAGE-CURRENT CHARACTERISTICS
The manner in which large diameter electrodes operate can best be under-
stood from the voltage-current curve of a contaminated electrode system as
shown in Figure 1.
The perfectly clean electrode characteristic starts at the critical volt-
age, Vc, and can be modelled using Poisson's Equation. This is possible
23-2

-------
-
z
UJ
cc
cc
3
o
/v.
Vt
Vbc /
VOLTAGE, V
Figure 1. Typical voltage-current characteristics for
large diameter-electrodes and high resistivity
flyash.
A clean characteristic
B tuft corona in full corona region, V > Vc
C tuft corona in partial corona region, Vt < V < Vc
D back corona characteristic
because the discharge appears to be evenly distributed along the length of
the discharge wire and is mathematically similar to positive corona. The
discharge however is much more turbulent, and the physics of the discharge
is obviously different.
As the discharge wire becomes contaminated, the visual appearance of the
corona discharge changes, and in the steady state, it appears as a number of
discrete tufts along the wire's length. The net effect is to,
(i) shift the full corona characteristic to the left, and
(ii) produce a number of randomly located corona tufts
between Vc and the tuft corona onset voltage Vt.
This means that corona current flows in this partial corona region at
voltage levels well below Vc- The tuft corona onset is approximately 0.6
of Vc, well below the value at which corona current is usually assumed to
23-3

-------
flow even after taking into account the roughness factor used in the
traditional approach. These effects are more obvious with the larger
electrodes than with the conventional 3.2 mm electrodes.
The main characteristic of the discharge tufts in this partial corona
region is that they switch on at arbitrary points along the discharge wire
in such a way that their current fields do not overlap. As the voltage is
increased, additional tufts are generated and, at Vc, they are evenly
distributed along the full length of the discharge electrode so that their
current fields just touch at the plate. Each individual tuft is unstable
in this region and often changes its position on the discharge wire.
The equation that reasonably models the average current in this region
is given by (2):
0.314 0fJL V.2
r v - Vt 1
VC -Vt
, for Vt * V * Vbc
(1)
= 0, for V < V,
where
0
dielectric constant of air
ion mobility
critical voltage
tuft onset voltage
distance from discharge electrode to plate
If the particulate on the collecting plate is of high resistivity, back
corona is formed, and positive ions are injected into the gas space and flow
to the discharge electrode. This will occur at some voltage Vbc depending
on the flyash resistivity. It affects the voltage-current characteristics
by causing a rapid increase in current as shown by line D in Figure 1.
The main operating voltage range of the elctrostatic precipitator with
large diameter electrodes is from V = 0.6 V to a voltage just above V^.
Ideally it should be slightly above this voltage.
WIRE DIAMETER VARIATION
A series of measurements were made at 150C and a range of discharge
electrode diameters using the EPA Pilot Electrostatic Precipitator
collecting a high resistivity flyash. The results are shown in Figure 2.
Since the voltage waveshapes of the output of the power supply are not
steady and vary significantly with the load current, the following voltage
measurements were made.
(i) At tuft
(ii) At onset
(iii) Maximum
corona onset.
of back corona,
average voltage.
23-4

-------
40

>
K
UJ
O
<
H
-I
O
> 20
ELECTRODE DIAMETER, mm
Figure 2. Variation of critical voltages with electrode
diameter. Temp = 150C.
A	tuft corona onset voltage
B	back corona onset voltage
C	maximum average voltage
D	maximum peak voltage
23-5

-------
(iv) Instantaneous peak voltage at maximum average voltage.
For measurements (i) and (il) the output voltage was approximately
constant; but once back corona began to flow, the ripple increased signifi-
cantly. The main measurements are made at the end of an 8 hour test run
by which time the discharge electrode and plates were well contaminated.
The absolute magnitudes are very dependent on the flyash resistivity
and to some extent are sensitive to the power supply characteristics. The
overall trends and relative values in Figure 2 are probably representative
of those in full scale electrostatic precipitators.
INDIVIDUAL TUFT DISCHARGES
The main characteristics of the individual tuft discharges in the
partial corona region are discussed in detail elsewhere (2), and only a
summary is included in this paper.
PROJECTION
2.00mm 0.95mm
20
<
a.
i-
2
LU
0.45mm
cc
DC
3
o
ACTUAL
CHARACTERISTIC >
/9.50mm
VOLTAGE, kV
Figure 3. Voltage-current characteristics of micropoints
on different diameter wires. Plate spacing = 230mm,
ambient temperature.
	 Diam. 3.2 mm K = 0.9-1.2 fiA/KV
	 Diam. 6.4 mm K = 1.4 fjA/kV
 Diam. 9.5 mm K = 1.5 (iA/kV
Some typical voltage-current characteristics for ambient temperature
are shown in Figure 3. These are approximately linear with slopes of
0.9-1.5 jiA/kV at ambient temperature which increase to approximately
1.4-2.4 nA/kV at 150C, for the range of wire diameters from 3.2 mm to
9.5 mm. The approximate expression for the total current from an individual
tuft discharge is:
23-6

-------
It = K(V - Vt), Vt S V S Vbc	(2)
where K slope
Vt tuft corona onset voltage
The pattern of current density distribution on the plate as measured
by the parallel aluminium strips on the plates is shown in Figure 4. The
distribution at right angles to the discharge electrode is shown in Figure
4(A) and is given by:
J/J0 = cosna	(3)
where J0 is the peak current density and n varies between 5 and 6, and is
reasonably independent of electrode diameter, voltage, and current density.
The main variation in the current density distribution occurs in the
direction parallel to the discharge electrode as shown in Figure 4(B). As
the total current from the tuft discharge increases, the width of the
current density curve increases. Tails may extend up to 140 mm either side
of the discharge. The maximum current density is determined from the
saturated current density equation (3):
 n V2
Jo - -V-	<4>
L
This can result in high current densities. At 150C, the current den-
sity is 260 nA/cm2 for a voltage of 35 kV with a plate spacing of 230 mm.
NUMBER OF TUFT DISCHARGES
If the total current is known either from equation (1) or from measure-
ment, the average spacing between the tufts can be calculated. The basic
assumption is that, once the tuft discharges are formed, they are of the
same intensity. Hence their number is:
N = 	r , Vt * V S Vb	(5)
K(V-Vt) r	DC
where I is the measured or calculated total current. The average spacing
between them, for a total electrode length of L^., is given by:
Zirji
S "
In many instances, some of the discharges are intermittent, and this
has the effect of reducing the effective value of K and hence of increasing
the number of discharge points.
23-7

-------
45
100
200 300
200
100
DISTANCE FROM CENTRE, mm
300
Discharge Tuft
200 100 0
DISTANCE FROM CENTRE OF MICROPOINT, mm
Figure 4. Distribution of current on plate produced
from a single micropoint. Plate spacing
230 mm, V = 30 kV, discharge electrode diam.
6.4 mm. Current normalized to peak value.
(A)	At right angles to the wire direction
(B)	Parallel to the wire direction
23-8

-------
BACK CORONA DISCHARGES
Some characteristics of the individual back corona discharges were
investigated by means of the special matrix plate located on the collecting
electrode which made it possible to isolate individual back corona
discharges and to observe and measure their characteristics. The following
observations were made:
(i) The individual back corona points on the plate tended to
move more frequently from one location to another as the
discharge wire diameter increased.
(ii) The number of back corona points in a given area was
reduced with increase in discharge wire diameter. In
the extreme case, with 12.7 mm diameter, one and
sometimes no back corona points could be detected on
the measuring plate.
(iii) The intensity of the individual back corona discharge
points increased with electrode diameter.
The reason for this is uncertain at this stage but is probably due to
the instability of the corona discharge on the wire electrode. One part-
icular streamer may initiate a back corona point, and this is sustained
for some time by the reverse flow of positive ions back to the discharge
electrode. The process is essentially unstable, and the tuft corona dis-
charge shifts to another location on the discharge electrode; this affects
the back corona discharge, causing it to 'jump' from one location to another
on the plate, or else to switch off altogether.
EFFECT ON PERFORMANCE
The main electrical factors that affect the operation and performance
of electrostatic precipitators with large-diameter discharge electrodes are:
(i) The existence of corona discharges at voltages well
below the critical voltage Vc, even after taking into
account reasonable values for the roughness factor.
The discharges in this partial corona region tend to
be unstable and are randomly spaced along the discharge
electrode in such a way that their current fields do
not overlap.
(ii) Since the tuft corona voltage onset, Vt, is higher for
the large diameter wires than for the conventional 3.2
mm diameter wires, the voltage levels for particle char-
ging and collection are correspondingly increased. The
9.5 mm diameter electrodes have an operating voltage
about 1.7 times higher than that of the 3.2 mm diameter
wire, and the E2 terms in the Deutsch Efficiency Equat-
ion are increased by a factor of 2.8. This can be
expected to improve collection performance of the
precipitator.
23-9

-------
(iii) Each isolated discharge tuft produces a current field
that covers a relatively large area. It may extend a
maximum of 600 mm in the direction of the gas flow for
a single wire; and for a multiwire system, to a distance
equal to the wire to wire spacing because of field
truncation. The width may vary up to a maximum value
of 280 mm. The maximum current density is very high and
can be calculated by equation (4). Values of 50 per cent
and above this maximum value occupy about a quarter of
the total area of the current field.
(iv) The separation between the individual tufts will reduce
the probability of particles being charged. However,
if there are a sufficient number of wires in series with
the gas path, this probability of charging will
increase.
(v) Since the current density under an individual tuft is
high, if a particle does move into the current field,
the NQt is sufficiently high to charge the particles
to saturation.
(vi) As the individual tuft discharges are often intermittent
and move about the surface of the discharge electrode,
the formation of stable back corona on the collecting
plate becomes more difficult. Hence, if back corona
is formed and the primary source of negative ions is
switched off, and since there are no other tuft dis-
charges sufficiently close by to support the back
corona discharge, the discharge decays. This makes it
possible for there to be a reasonable working voltage
range betwen Vt and Vbc.
(vii) If a back corona discharge does become firmly establish-
ed, and since they are well spaced from other dis-
charges, its volume of influence in the gas stream is
small. Providing it does not cause excessive sparking,
the back corona will not have a significant detrimental
effect on the overall performance.
(viii) There is a rapid improvement in the voltage level as
the electrode diameter increases from 3.2 mm; but for
values above 7.9 mm, the increase is less significant
and the operating voltage range is reduced. This
suggests that the optimum diameter is in the range of
7.9 - 9.5 mm [1/8 - 3/8-in.], although this will vary
with flyash resistivity and operating conditions.
23-10

-------
ACKNOWLEDGEMENTS
The author is grateful to the staff of the Particulate Technology
Branch, IERL-RTP, U.S.EPA for their assistance in carrying out the test
programs described in this paper.
REFERENCES
1.	Bush, P., Pontius, D.H., and Sparks, L.E. Performance of large-diameter
wires as discharge electrodes in electrostatic precipitators. Paper
presented at The Fifth Symposium on the Transfer and Utilization of
Particulate Control Technology, Kansas City, Missouri, August 27-30,
1984.
2.	McLean, K.J., Lawless, P.A., Sparks, L.E., and Ramsey, G. Negative
corona in wire-plate electrostatic precipitators. To be published.
3.	Sigmond, R.S. Simple approximate treatment of unipolar space-charge-
dominated coronas: The Warburg law and the saturation current. Jn.App.
Phys, 53, 1982, p.891.
23-11

-------
LABORATORY ANALYSIS OF CORONA DISCHARGE ELECTRODES AND BACK CORONA PHENOMENA
P. Vann Bush and Todd R. Snyder
Southern Research Institute
Birmingham, Alabama 35255-5305
ABSTRACT
An experimental research program was conducted to characterize corona
and back corona generation with different geometries in various environments.
A wire-parallel plate device and a wire-cylinder apparatus were utilized to
monitor the spatial distribution and temporal stability of corona and back
corona discharges. Conditions that were varied in the experiments included
temperature, diameter of corona electrode wire, corona polarity, and type of
corona electrode (smooth versus barbed) and its condition (rough versus pol-
ished). Back corona discharges were induced by depositing fly ash on the
grounded electrode. Visual observations of corona and back corona phenomena
were made with a sensitive video camera. The measured distributions of coro-
na and back corona current and statistical analyses allow a quantitative
assessment of the impact of corona electrode design on electrostatic precipi-
tator performance. This paper describes the laboratory apparatus and gives a
survey of experimental results. A discussion of the implications of this
study for new and retrofit ESP design completes the scope of this paper.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
24-1

-------
INTRODUCTION
A laboratory research program was conducted to evaluate the character of
corona generation from several sources in a variety of environments. Under-
standing the nature of corona generation, current distribution, and back
corona behavior is critical to optimum ESP design and modeling. The program
utilized two electrode configurations: wire-plate and wire-cylinder. Condi-
tions that were varied in the experiments included temperature, diameter of
corona electrode wire, corona polarity, and type of corona electrode (smooth
versus barbed) and its condition (rough versus polished). Back corona phe-
nomena were examined utilizing the wire-plate device with fly ash deposited
on the plate.
APPARATUS
The wire-plate corona system incorporated a segmented electrode as one of the
two plates. This electrode was made from a high-temperature copper-clad
board used for printed circuitry. The plate was segmented into parallel
strips by a PC board etching process, with strips 0.50 inches by 6.0 inches,
giving each strip an area of approximately 0.02 ft2. The dimensions of the
plate electrodes were 6.0 inches wide by 24.0 inches long. Wire-to-plate
spacing for this setup was 4 to 5 inches for all tests. The complete wire-
plate apparatus was installed in a convection oven, and the 32 segments were
connected to a dual 16-channel electrometer/multiplexer.
Figure 1 shows a diagram of the wire-cylinder apparatus. Components
included 32 isolated rings for collecting and measuring the distributed cur-
rent from a corona electrode positioned along the central axis of the array
of rings. The inside diameter of the rings was 9.776 inches, each ring hav-
ing a surface area of 0.107 square feet. Each ring in the cylinder was con-
nected via shielded cable to the electrometer/multiplexer.
The electrometer was designed to interface with a 16 channel, 12 bit A/D
converter installed in a DEC PDP 11/34 minicomputer used for data acquisi-
tion, manipulation, storage, and presentation. Data acquisition consisted of
800 passes through the A/D resulting in 400 sequential readings of the cur-
rent distribution along the 32 plate segments (or rings in the cylinder appa-
ratus). The total elapsed sample time was approximately 66 seconds. This
system allowed for rapid acquisition of data showing the current distribution
at any given sample time, the change in the current measured on a given seg-
ment through time, and statistical quantities describing the current distri-
bution.
The most useful statistical quantity was the Coefficient of Variation
(COV), which quantified the corona uniformity. COV was calculated with the
expression C0V=( cr/x) *100, where a is the standard deviation of the average
value of the current measured on the segment from the mean of the 32 seg-
ments, and x is the mean value of the distribution. This simply expresses
the standard deviation of a distribution as a percentage of its mean.
24-2

-------
N3
I
U>
CORONA ELECTRODE
CORONA ELECTRODE HOLDER/
AERODYNAMIC SHIELD
-MACOR INSULATING/POSITIONING RODS
HIGH VOLTAGE

STRUCTURAL SUPPORT
SLAT
MICA INSULATION
31 CURRENT MONITOR/MICA RING
SANDWICHES
MICA INSULATING RING
-CURRENT MONITORING SEGMENT (0.5 IN. THICK)
-MICA INSULATING RING (0.03 IN. THICK)
 FACE PLATE
Figure 1. Wire-cylinder current distribution monitor.

-------
CORONA UNIFORMITY
A test plan was designed to cover a broader scope of measurements than
the earlier studies of Tassicker (1) and Durham, et al (2). Studies of the
corona exhibited by smooth, polished electrodes of different diameters
revealed that the uniformity of corona emission increased as the diameter of
the corona electrode wire decreased. This effect can be seen in Figure 2.
Data in Figure 3 show that increasing the negative polarity corona current
emitted by a smooth wire discharge electrode gives a reduction in the value
of COV, indicating a more uniform current distribution on the plate. These
results are also apparent in the data taken with the wire-cylinder apparatus.
Figure 4 shows the distributions of current at two different voltage levels.
The end effects caused by the physical design of the wire-cylinder device are
clearly visible in the graphs. At typical operating levels the currents read
on collectors 7 through 28 were essentially independent of end effects.
Also shown in this Figure is the relationship of total current to the COV for
channels 7-28 for a 1/8 inch diameter wire at 68F in the wire-cylinder appa-
ratus .
Another comparison made was between the smooth wire data and data from a
barbed electrode. The fundamental difference noted was the way in which
corona current was increased with the two electrode designs. For the barbed
electrode a corona voltage was reached where each barb tip became a corona
source. Raising the voltage from this condition merely intensified the coro-
na activity at each of the barbs but did not raise the number of active coro-
na spots. In the smooth wire case, however, the dominant method for raising
the total current was an increase in the number of active corona sites on the
wire rather than an amplification in the activity of any given corona spot.
The data indicate that more uniform current distribution is reached by pro-
viding a higher density of corona points along the electrode.
Negative polarity high voltage excitation generated two very different
corona discharge patterns, depending on the condition of the wire surface.
One mode of discharge, referred to as Ideal Corona, was achieved by carefully
polishing the wire surface. The discharge appeared as a rapidly moving,
continuously distributed array of very fine streamers, with the motion trav-
eling along the length of the corona wire. This phenomenon was transient,
though it persisted for many minutes. Ideal Corona produced a very low COV,
which indicated a very uniform distribution of current. Typical negative
corona had the appearance of intense, isolated spots of corona activity which
sometimes moved along the wire surface, but were commonly stationary. These
discharge spots were persistent and produced much higher total current and
local currents than the Ideal Corona. Figure 5 plots COV versus Voltage for
the Ideal and typical negative corona. There was a sudden transition from
the Ideal Corona mode to typical negative corona which involved the develop-
ment of individual intense corona spots surrounded by a quiescent zone. The
change did not occur over the entire length of the wire simultaneously,
though the region affected by local transitions ranged from one inch to the
entire length of the wire. After typical corona was observed, Ideal Corona
could not be reproduced without re smoothing the wire surface.
24-4

-------
COV = 8.8
WIRE DIAMETER = 0.062 IN.
VOLTAGE =-21.3 kV
<
a.
tt
cc
3
O
COV = 29.1
WIRE DIAMETER = 0.125 IN.
VOLTAGE = -36.5 kVr
12 3 4 5 6 7B 9 IB 1112 1314 15 IB 17 IB IS 28 21122 23 24 29 2B 2728 29 3B 31 32
SEGMENT NUMBER
1 2 3 4 a S 7 8 9 IB 11 12 13 14 15 IB 17 IB 19 2B 21 22 23 24 2S 26 27 2B 29 3B 31 32
SEGMENT NUMBER
<
3.
-*
m l- 25
cc
cc
3
o
COV = 43.1
WIRE DIAMETER = 0.250 IN.
VOLTAGE = -50.0 kV
1 2 3 4 5 B 7 B B IB 11 12 13 14 IS IB 17 18 19 28 21 22 23 24 2S 26 272829 3B 31 32
SEGMENT NUMBER
Figure 2. The effect of wire diameter on current distribution at negative polarity and
200F with 100 \xA total current.

-------
*. 50t
<
a.
Z 2. SS-
CC
cc
o
A. COV = 155.3
TOTAL CURRENT 25 JJA,
VOLTAGE -32.1 kV
innnnnnfinnnrinnnn
t T1 h1 t* t t1
nfLn
<
a.
i-"
z
UJ
cc
cc
o
o
B. COV = 66.5
TOTAL CURRENT 50 fjA,
VOLTAGE -34.2 kV
mi
flUHflnnn
tail
I 2 3 4 5 6 7 B 9 18 11 12 13 14 IS 16 17 18 19 28 21 22 23 24 25 26 27 28 29 38 31 32
1 2 3 4 5 6 7 B 9 IB 11 12 13 14 IS 16 17 IB 19 2B 21 22 23 24 25 26 27 28 29 38 31 32
SEGMENT NUMBER
SEGMENT NUMBER
<
a.
UJ
cc
cc
3
o
C. COV = 29.1
TOTAL CURRENT 100 flA,
VOLTAGE -36.5 kV
<
a.
h-~
Z
UJ
cc
cc
D
o
1. 13
D. COV = 23.4
TOTAL CURRENT 200 yA,
VOLTAGE -39.2 kV
1 2 3 4 5 6 7 B 9 18 11 12 13 14 IS IB 17 IB 19 28 21 22 23 24 25 2fi 27 28 29 3B 31 32
1 2 3 4 5 6 7 B 9 IB U 12 13 14 15 16 17 18 19 28 21 22 23 24 25 26 27 28 29 SB 31 32
SEGMENT NUMBER
SEGMENT NUMBER
Figure 3. The effect on current distribution of increasing total current from a 1/8-inch
diameter wire at negative polarity and 200F.

-------
COV = 22.61
LU 2. 0B
1. 00
npnn
COV = 6.77
5. 25
240
160
80
Ui
CURRENT, (lA
g[lol]

0. 00
SEGMENT NUMBER
Figure 4. The effect on current distribution of increasing total current from a 1/8-inch
diameter wire in the wire-cylinder apparatus at negative polarity and 68F.

-------
100
300F CLEAN PLATE
z
o
-
<
cc
< 60
u.
O
K-
TYPICAL
Z
Ul
40
o
u.
u.
LU
IDEAL
O
o
20
ai
30.0
40.0
50.0
60.0
APPLIED VOLTAGE, IcV
Figure 5. Coefficient of variation vs. voltage for negative corona discharges.
Data from positive polarity experiments consistently showed a more uni-
form current distribution than negative polarity data. Even at low current
densities positive corona produced a very smooth distribution of current on
the grounded electrode. A graphic comparison of the current distributions of
positive and negative corona is given in Figure 6. The bars indicate the
average value of current to each collector.
Raising the temperature in the experimental setups had the effect of
shifting the voltage current (V-I) curves to the left. Other trends noted
were not substantially altered by the change in temperature. Figure 7 demon-
states the rapid changes that occurred in the amount of current flowing to a
specific collector. The change shown here indicates that the corona activity
moved from one barb to an adjacent barb.
BACK CORONA
Additional laboratory experiments were performed to gain a qualitative
understanding of the nature of back corona and to measure the effect of back
corona on the distribution and intensity of current. These studies were
designed to expand upon the work done by Spencer (3) and Masuda, et al (4).
Back corona was produced by spreading a smooth layer of ash on the plate
electrode. The corona electrode used for these observations was a 1/8 inch
diameter wire. Conditions varied in these studies included primary current
density, polarity of the applied potential, ash resistivity (induced by dif-
ferent operating temperatures), and other ash layer properties, such as
24-8

-------
<
a.
H
Z 1.5
Ltl
E
OC
D
O
0. 3
A. COV = 28.3
TOTAL CURRENT 50 /JA
VOLTAGE -26.6 kV
<
3.
OC
tr
z>
u
B. COV = 4.1

TOTAL CURRENT 50 jUA
VOLTAGE +29.5 kV
2 4 6 B IB 12 14 16 IB 20 22 24 26 28 30 32
2 4 6 S 10 12 14 16 18 20 22 24 20
N>
-P~
I
VO
<
=L
H-*
Z
ui
oc
DC
D
O
C. COV = 11.3
TOTAL CURRENT 200 JJA
VOLTAGE -33.2 kV
!
2 4 6 B IB 12 14 10 18 20 22 24 26 20 30 32
SEGMENT NUMBER
<
a.
H
Z
UJ
OC
OC
o
o
D. COV = 3.6
TOTAL CURRENT 200 A
-------
6.0
<
3.
g 4.5
cc
cc
D

-------
600
500
400
<
a.

Z 300
cc
cc
o
u
200
DIRTY PLATE
LU
CLEAN PLATE
100
50.0
60.0
40.0
30.0
APPLIED VOLTAGE, kV
Figure 8. Comparison of clean plate and dirty plate V-l curves with negative polarity
and 300 F.
This pattern is shown in the photograph of back corona in Figure 10. Major
defects in the ash layer, such as craters formed by arcs, provided stationary
sites for back corona. These sites sometimes became very intense, extin-
guishing nearby activity. Often these intense spots of back corona did not
produce radial streamers, but had plate to wire streamer propagation typical
of positive corona.
Several ash layer thicknesses between 1/8 and 3/8 inches were studied.
There was no observable difference in the distribution or character of the
back corona produced in the layers. There was also little difference in the
applied voltage at which back corona onset was detected.
The effect of back corona on the current distribution was more pro-
nounced for the positive polarity corona. Figure 11 is a plot of the current
distribution of clean-plate and dirty-plate conditions. The current with
back corona was an order of magnitude greater than the current generated from
the wire in the clean system. The plot of the COV versus Voltage for these
measurements (Figure 12) emphasizes the degree to which the back corona
degraded the current distribution on the plate. The back corona sites caused
corona streamers to develop from the discharge wire, severely limiting the
operating voltage range by leading to sparking.
The back corona associated with positive primary corona was different in
appearance from the negative corona product, as was expected. The back coro-
na sites typically extended over 1/4 to 1/2 of the area and were more numer-
24-11

-------
7.00
VOLTAGE-38 kV
I	1 BACK CORONA
MCLEAN
n n n n n n n
o.oo
12 3 4 5 6 7 8 9 1011121314151617181920212223242526272829303132
SEGMENT NUMBER
Figure 9. Current distributions for negative corona with and without back corona at 300F.
ous than the back corona sites induced by negative corona. There were no
detectable streamers from each site along the ash layer surface, but rather a
cloud of ionization appeared to penetrate within the layer.
IMPLICATIONS FOR ESP DESIGN AND OPERATION
CORONA DISCHARGE ELECTRODES
Utilization of the quantitative data on the relative corona uniformity
exhibited by electrodes of various designs and the integrated data contained
in the voltage-current curves provide a means to evaluate the impact of dis-
charge electrode design on ESP performance. The electrostatic precipitation
process has been effectively modeled as a statistical phenomenon, beginning
with the Deutsch equation. The coefficient of variation fits in with the
statistical description of the process. It is a measure which is as
characteristic of an electrode geometry as is the V-I curve.
The conventional practice of employing a universally applicable elec-
trode design may severely constrain the level of collection efficiency
attainable with a given collection surface area. For instance, an applica-
24-12

-------
Figure 10. Photograph of back corona on a dust layer (~10^ Sl-cm resistivity fly ash)
under negative polarity corona wire.
24-13

-------
7.00
5.25
fr. 3.50
CE
E
D
O
1.75
0.00
VOLTAGE 40kV
[ I BACK CORONA
CLEAN
 nnBoWlTf
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
SEGMENT NUMBER
Figure 11. Current distributions for positive corona with and without back corona at 300PF.
200
o
H
<
DC
<
>
150
DIRTY 300F
DIRTY 200F
O 100
(
z
111
o
LL
U.
O
o
50
CLEAN 200F
CLEAN 300F

	O
30
40	50
APPLIED VOLTAGE, kV
60
Figure 12. Coefficient of variation ks. voltage for dean and dirty plate with positive
polarity corona discharges at 200 and 300F.
24-14

-------
tion with heavy inlet dust loading would call for electrodes with high cur-
rent emission capacity to overcome the effect of space charge suppression of
the corona. For high resistivity ash, a low current level would be selected
to prevent back corona formation, yet high electric fields would be desired
for good collection performance. This latter case favors large diameter
wires. The criteria for choosing an electrode should include the operating
voltage level, the corresponding current emission level, and the uniformity
of the discharge under the expected conditions.
One reason large diameter wires have been avoided in conventional ESP
design is that they provide a less uniform current distribution. This
disadvantage is offset by the fact that a few microamps per square foot is
adequate corona current to ensure saturation charging of the particles within
a small fraction of their residence time in the ESP. The combined effect of
having n wires in series is to increase the probability that particles will
encounter ions for charging. This is shown in Figure 13 which illustrates
the relationship
C = COV/(n)1/2
where C is the effective coefficient of variation for a lane of n wires and
COV is the coefficient of variation of a single wire of the given diameter at
the specified operating condition. Since there are typically more than fifty
z
o
H-
<
GE
<
>
u.
O
I-
z
40
UJ
o
u.
U.
Ul
o
o
70
50
30
NUMBER OF WIRES
Figure 13. The coefficient of variation versus the number of corona wires.
24-15

-------
wires in an ESP lane there is essentially an equal probability with small or
large diameter wires that a particle will pass through the required number of
charging regions to reach saturation charge levels.
BACK CORONA
The effect of back corona in altering V-I curves and degrading particle
charging is well known. It has been assumed that the amount of increase in
current evident in V-I curves is equal to the current emitted from back coro-
na on the ash layer. This model neglects the effect of ionic space charge on
the primary corona. Back corona activity stimulates primary corona emission
by reducing the space charge. The apparatus developed for these experiments
has permitted measurement of the effects of back corona on the magnitude and
distribution of current. Quantifying the conditions for, and the effects of,
back corona will permit proper selection of operating voltage and current
levels in ESP's.
CONCLUSIONS
The test apparatus developed for these measurements was extremely versa-
tile for studying various phenomena and conditions found in field applica-
tions. The rapid measurement of the distribution of and changes in corona
current allowed the calculation of the coefficient of variation, a statisti-
cal quantity that describes well the corona uniformity and is therefore use-
ful in the comparison of electrodes for ESP design. Corona uniformity was
shown to increase as corona wire diameter decreased and negative polarity
corona emission was shown quantitatively to be less uniform than positive
polarity corona. These measurements were enhanced by the visual observations
of the nature of corona and back corona emissions made with the video camera.
Additional measurements quantified the large increases in corona current and
changes in current distribution due to the presence of back corona. Further
measurements with this test device will allow the determination of conditions
for back corona onset and the effect of back corona on precipitator electri-
cal operation.
ACKNOWLEDGEMENTS
This work has been sponsored by the Industrial Environmental Research
Laboratory of EPA under the direct guidance of Dr. L.E. Sparks (Cooperative
Agreement No. CR-808973-01). The authors appreciate the work of Mr. D.K.
Armstrong and Mr. W.G. Kistler in the supporting laboratory experiments and
the consulting expertise of Dr. D.H. Pontius.
24-16

-------
REFERENCES
1.	Tassicker, O.J. Measurement of Corona Current Density at an Electrode
Boundary. Electronics Letters. 5 (13): 285-286, 1969.
2.	Durham, M., G. Rinard, D. Rugg, and L. Sparks. Measurement and
Interpretation of Current Distribution and Charge/Mass Data. In:
Third Symposium on the Transfer and Utilization of Particulate Control
Technology, Vol. II. EPA-600/9-82-005b (NTIS PB83-149591), 1982.
pp. 54-65.
3.	Spencer, H.W., III. Electrostatic Precipitators: Relationship Between
Resistivity, Particle Size, and Sparkover. EPA-600/2-76-144 (NTIS PB
257130), U.S. Environmental Protection Agency, Research Triangle Park,
North Carolina, 1976. 68 pp.
4.	Masuda, S. and A. Mizuno. Initiation Condition and Mode of Back Dis-
charge. Journal of Electrostatics, 4: 35-52, 1977.
5.	Loeb, L.B. Electrical Coronas: Their Basic Physical Mechanisms.
University of California Press, Berkeley, California, 1965. 189 pp.
24-17

-------
Session 10: ESP: FUNDAMENTALS I
Grady B. Nichols, Chairman
Southern Research Institute
Birmingham, AL

-------
THE ONSET OF ELECTRICAL BREAKDOWN IN DUST LAYERS
R. P. Young
J. L. DuBard
Southern Research Institute
Birmingham, Alabama 35255
L. E. Sparks
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
ABSTRACT
The onset of electrical breakdown in 0.5 cm thick layers of fly ash and
cement dust has been determined by slowly increasing the applied voltage and
observing the oscilloscope trace of the first microspark in the dust layer.
The electrical properties of the sample are not altered by this sensitive
measurement technique; so multiple experiments can be performed on each sample.
Parallel plate electrode geometry is used to define precisely the average
electric field in the sample and to permit conventional measurements of the
sample resistivity, in addition to the novel measurements of the breakdown
voltage, with variations in sample temperature and environmental water vapor.
The measured breakdown voltage is systematically influenced by the water vapor
concentration, in the temperature region of particle surface conduction, and by
the particle size distribution and packing density of the sample.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
25-1

-------
INTRODUCTION
The negative corona current in an electrostatic precipitator induces a
large voltage drop across the dust layer on the collecting electrode. Re-
stricted electrical conduction through this dust layer leads to enhancement
of the local electric fields in the gas-filled voids of the layer until these
fields become large enough to initiate electrical breakdown across the void.
Sustained breakdown within the dust layer then leads to a back discharge of
positive ions into the precipitator gas stream, resulting in deterioration of
precipitator performance. Thus, an understanding of the factors contributing
to the onset of electrical breakdown in these dust layers is important for
maintaining and improving the performance of a precipitator.
In this study, a dust layer is hand-deposited between parallel plate
electrodes and the applied potential is slowly increased until the first
microspark in the dust layer is observed on an oscilloscope. The parallel
plate geometry provides a well-defined, uniform average electric field across
the dust layer, allows standard resistivity measurements to be performed, and
avoids large space charge distortions which can occur with other geometries.
In addition, the method used is not destructive to the dust layer, allowing
multiple experiments to be performed on each sample.
The onset of breakdown of the dust layers was found to be affected by
sample temperature, water vapor in the test environment, dust species, average
particle size, degree of mechanical compaction of the surface, and duration of
electrical stress across the sample. The breakdown data can be explained by
ordinary electron avalanche breakdown across the gas-filled voids in the dust
layer.
EXPERIMENT
The components of the apparatus used in these experiments are shown in the
schematic diagram of Figure 1. The dust layer was hand-deposited between the
parallel plates of the test cell to a uniform depth of 0.5 cm, using a constant
surface areal loading of 10 g/cm2. Negative polarity high voltage was applied
to the top electrode and the guarded electrode was connected, by means of a
switch (SW), to either the electrometer (A) for resistivity measurements or the
oscilloscope (OSC) for electrical breakdown measurements. The loaded test cell
was placed into an environmental chamber and baked overnight at 250C in dried
air. The following morning, humid air was introduced, and the system was
allowed to equilibrate for at least 1 hour before beginning measurements.
PROCEDURE
The resistivity measurement was made according to procedures listed in the
IEEE Standard 548-1981. While increasing the field across the sample from zero
to 4 kV/cm, the oscilloscope was monitored to be sure breakdown did not occur.
After the current was measured for the resistivity determination, the voltage
was immediately removed from the dust layer. In no case did breakdown occur
25-2

-------
before achieving a field of 4 kV/cm. Moreover, comparison of resistivity
measurements made with this test cell and those made with a standard resis-
tivity test cell under equivalent conditions showed no difference between the
values measured.
The breakdown measurement was performed by slowly increasing the voltage
across the dust layer while observing the trace on the oscilloscope until the
breakdown signal (1) appeared as a single pulse. The voltage reading on the
power supply meter (the average breakdown voltage, VgD, across the dust layer)
was recorded and the applied voltage was immediately removed from the sample.
The data obtained were found to be reproducible for different samples of a
given dust species under identical conditions. In addition, resistivity
measurements made with this test cell on a virgin dust sample were the same as
those obtained after breakdown measurements on the same sample. Thus, this
procedure for breakdown measurements did not alter the average electrical
properties of the dust layer, making possible multiple experiments using the
same sample.
The temperature of the environmental chamber was stabilized at a lower
temperature and these measurements were repeated. In this way, data were
obtained at approximately 15 degree increments from 250C down to 100C. If
tests for a different level of environmental humidity were to be performed on
the same sample, the entire process, beginning with the overnight bake, was
repeated. After all tests on a particular sample were completed, the dust
layer was removed for chemical analysis and scanning electron microscope (SEM)
photographs.
SAMPLES
Three dust species were chosen for this investigation: two coal fly ashes
(Sundance and Wabamun) and a Portland cement dust (Lehigh). The two fly ashes
are virtually identical in nature except for their sodium content, with
Sundance fly ash having five to six times the amount of sodium as Wabamun fly
ash. Since sodium is thought to be the dominant conductor of ionic charge in
fly ash, data for these two species were compared to provide insight into the
role of charge carrier concentration in the breakdown process. The fly ash
particles are predominantly spherical in shape and amorphous like a glass,
while cement dust is a more crystalline material and has very irregularly
shaped particles. Cement dust is very low in sodium and high in CaC03. Thus,
cement dust should present quite different behavior than that of fly ash.
RESULTS
The onset of electrical breakdown was measured with variations in
temperature, atmospheric moisture, particle size, degree of compaction, and
duration of electrical stress. It was found that, for most test conditions,
the three dusts behaved qualitatively the same. Therefore, this discussion
will focus on the results of tests with Wabamun fly ash, with any species-
dependent characteristics pointed out viien appropriate.
25-3

-------
BREAKDOWN DATA
The variation of average breakdown voltage, Vgp (or average electric field
at breakdown, EgD^'	temperature and test atmosphere humidity for Wabamun
fly ash is shown in Figure 2. Vgp is the power supply reading at the occur-
rence of the first microspark on the oscilloscope and is for negative polarity.
EgD is obtained by dividing VgD by the layer thickness, 0.5 cm. The indicated
water levels in the atmosphere are in volume percent.
Due to the statistical nature of breakdown, the data are somewhat
scattered. However, several features of the breakdown behavior are evident.
First, the magnitude of VgQ for a given humidity level depends only weakly on
temperature, with the greatest variation occurring in the range of 180 to
250C. This is the volume conduction region of the resistivity curve shown in
Figure 3. In the region where surface conduction dominates (100 to 140C), the
magnitude of Vg^ is relatively constant for a given water level. This is in
marked contrast to the rapidly changing resistivity characteristic of the fly
ash in this same region, as shown by Figure 3. Generally, increasing the water
vapor in the test environment decreases Vgjj. In addition, there is a much
larger difference between the magnitudes of Vg^ for the dry and moist ash
layers than between samples at different levels of moisture in this surface
conduction region. Finally, all Vg^ data, regardless of humidity level, tend
to converge at the high temperature (volume conduction) limit.
The greatest variation of the magnitude of Vgp at identical test con-
ditions between the three dust species occurs in ary air where differences of 1
to 1.5 kV are sometimes seen. However, the magnitudes o f VBD for humid samples
are quite similar, as illustrated in Figure 4 for 9% water vapor. In contrast,
the resistivity characteristics of the three dust species under the same
conditions are shown by Figure 5 to vary by over two orders of magnitude.
SIZE EFFECTS
The influence of particle size on the breakdown behavior of Wabamun fly
ash is shown in Figure 6 for 9% water vapor. The results are similar for other
humidity levels. "Ordinary distribution" refers to the as-received ash while
the two size-fractionated distributions are labeled according to their mass
median diameter (MMD), The ordinary distributions of the two fly ashes contain
particles trtiich are predominantly spherical in shape with only a few relatively
large, irregularly shaped particles. The fine fraction (MMD < 5 im) samples
are almost totally composed of spherical particles \diile the coarse fraction
(MMD > 50 pm) samples have an abundance of the larger, irregularly shaped
particles. Both the fine sample and the coarse sample for the two fly ashes
exhibited breakdown voltages higher than the ordinary distribution, with the
coarse sample producing the largest increase. The corresponding resistivity
characteristic for the Wabamun ash is shown in Figure 7. The coarse distri-
bution is about an order of magnitude higher in resistivity than the ordinary
distribution; while the fine distribution coincides with the ordinary dis-
tribution.
25-4

-------
The cement dust is composed of particles having no symmetry. The fine-
fraction particles of this dust are relatively uniform in size with only a few
larger particles; while the coarse fraction is dominated by the larger par-
ticles. Although the resistivity for this dust behaved similarly to that of
the fly ashes, essentially no difference in the breakdown behavior of the three
distributions was observed for the cement dust at 9% water vapor.
COMPACTION EFFECT
The effect of compaction on the breakdown behavior of the ordinary distri-
bution of the three dusts was studied by increasing the surface areal loading
of the samples to about 2500 g/cm2 in a laboratory press. In the case of fly
ash, this had essentially no effect on the breakdown behavior. However, a
slight effect was noted in the breakdown of the Lehigh cement dust of Figure 8.
Here the increased compression of the dust layer resulted in a lowering of the
breakdown potential, with a corresponding lowering of the resistivity curve in
Figure 9.
STRESS TEST
A sample of Wabamun fly ash was subjected to continual electrical stress
of 4 kV/cm at 152C and 9% water vapor for many days and its breakdown behavior
was monitored. The result is shown in Figure 10. In general, the magnitude of
VBD decreased steadily until, after about 20 days of stress, a somewhat abrupt
transition to continuous breakdown occurred. The oscillatory behavior discern-
ible in this data seemed to be typical of fly ash since it was also observed in
stress tests of other fly ash samples. In contrast, the resistivity of this
sample (Figure 11) showed no change corresponding to continuous breakdown at
about 20 days, but continued its steady increase.
DISCUSSION
It is generally accepted that electrical breakdown in a precipitated dust
layer occurs in the gas-filled voids between particles. (See, for example,
Mosley, et al. (2).) The local electric field in such voids must be enhanced
over the average value across the sample because the average field values at
which breakdown is observed are not large enough to initiate the breakdown
process. In the discussion vrtiich follows, these two points will be elabo-
rated.
CONDUCTION MECHANISM
A direct relationship between the onset of electrical breakdown in a dust
layer and its resistivity is not supported by the data. Figures 2 and 3 show
that, in the volume and surface conduction regions, the resistivity changes
exponentially with temperature rtiile the breakdown voltage varies only slight-
ly. Figures 4 and 5 illustrate how similar the breakdown behavior can be for
dust samples of widely different resistivities. Moreover, the size-fraction-
ated data of Figures 6 and 7 also show different trends. The fine particles
appear to control the resistivity of the ordinary sample since the resis-
tivities of the fine and ordinary samples are the same. However,	is
25-5

-------
consistently higher for the fine sample than for the ordinary distribution.
Further, the resistivity of the coarse sample is about an order of magnitude
higher than the ordinary and fine distributions, but its breakdown voltage is
only moderately higher than the other two samples. Therefore, the onset of
electrical breakdown in these dust layers and their resistivity are not
directly related.
Allowing for the scatter in the data of Figure 2, the addition of water
vapor to the test environment results in a large reduction of in the
surface conduction region compared to that of a dry sample. However, the
actual amount of water vapor has only a small effect on Vgp in this region.
This suggests that the important parameter is the presence of water vapor
rather than the actual humidity level. Furthermore, Figure 4 indicates that
when water vapor is present, the breakdown voltage does not depend on dust
species. Combining these results with the convergence of all data of Figure 2
at the volume conduction limit, it appears that the electrical breakdown of a
dust layer depends very sensitively on the electrical conduction mechanism
prior to breakdown.
ELECTRICAL STRESS
The increase of resistivity with time of applied electrical stress shown
in Figure 11 is consistent with the gross movement of charge carriers (the
alkali metal ions) and the creation of zones depleted of such charge carriers
(3,4). Lack of an abrupt change in the resistivity characteristic corre-
sponding to the transition to continuous breakdown (Figure 10) points out
the insensitivity of this macroscopic, average measurement to microscopic
variations in the dust layer. On the other hand, breakdown measurements are
inherently microscopic in nature. Thus, it is not surprising that breakdown
and resistivity cannot be directly related.
The oscillating behavior of VBD with time of stress is probably due to the
motion of the charge carriers in the bulk, as they change the local values of
the electric fields in the voids. It is also possible that this behavior may
reflect the existence of a trapping and detrapping mechanism suggested by
Mosley, et al. (5). However, the data do not permit conclusions on this
matter.
The steady decline in breakdown voltage with increasing duration of
electrical stress has obvious implications for electrostatic precipitator
operation. If a thin layer of ash remains on the collecting plates and is not
effectively removed by rapping, then continual electrical stress of this layer
will eventually lead to continuous breakdown in the layer and a degradation in
precipitator performance.
BREAKDOWN ACROSS VOIDS
The electron avalanche breakdown of a gas-filled gap is conventionally
interpreted using Paschen's law, which states that the breakdown potential
applied across the gap is a function of the product of pressure and gap height.
Typical data obtained by Earhart (6) for an air gap at different temperatures
are shown in Figure 12. In general, the minimum in the curve broadens and
25-6

-------
shifts to the right as the temperature of the gap is raised, accompanied by a
decrease in the slope of the linear portion to the right of the minimum.
Earhart also found that when the breakdown potential was plotted as a function
of gas density between the electrodes a single curve for all temperatures is
obtained.
As previously mentioned, the potential applied across the dust layer at
the onset of breakdown, VgD, is too low to initiate the breakdown process.
Thus, it is the enhanced local electric potential across the void, V^QC, which
corresponds to the breakdown potential of Paschen's law. However, the two
potentials should be directly related (through an effective dielectric
constant) so that Paschen's law can also be applied to V^p. To explain the
difference between the breakdown voltage magnitude of this study and that of
Earhart, it should be pointed out that Earhart obtained the data of Figure 12
using parallel plate geometry. The void in a dust layer will have nonuniform
geometry, and Raju and Hackam (7) and Heisen (8) have shown that the breakdown
potential increases dramatically as the geometry departs from that of parallel
plates.
By assuming the laboratory pressure to be 1 atmosphere, the gap distance
corresponding to the minimum can be calculated from Figure 12 for the three
temperatures. This results in distances of about 11, 14, and 17 ym for 22,
155, and 260C, respectively. Computer simulations for mixtures of spheres
(9,10) indicate that the largest void dimensions within the mixture correspond
to about 4 mean sphere diameters. Thus, a fly ash layer is likely to have
voids of dimension ranging up to several tens of micrometers. As the potential
applied across the dust layer is slowly increased, the first microspark will
occur at the lowest voltage consistent with the void dimension where breakdown
takes place. Since the variation of VBD with temperature is rather slow
(Figure 2) the onset of breakdown must correspond to positions on the Paschen
curve at or near the minimum and to the right of the minimum.
The general trend of decreasing VBD with increasing levels of humidity
(Figure 2) is opposite to the behavior of breakdown in an air-filled gap
between two electrodes, where it is observed that breakdown potential increases
with increasing water vapor. In the latter case, the increase is attributed to
an electron attachment and conversion process which occurs when water vapor is
present (11). However, the data for the present work indicate that the primary
interaction of water vapor is with the surface of the dust layer particles (and
consequently with the conduction mechanism) rather than with the gas molecules
in the void. This is in agreement with studies showing the dramatic effect of
the presence of water vapor on resistivity.
Paschen's law aids in interpreting the effect of particle size on VBD for
the two fly ashes. The coarse-fraction sample consists of particles much
larger, on the average, than those of the ordinary distribution, resulting in
larger average void dimensions in the coarse sample. Paschen s law then
predicts a larger applied potential at breakdown for this coarse sample than
for the ordinary distribution. The case for the fine-fraction sample is not as
simple because the porosity must be taken into account. Porosity of a dust
sample is conventionally expressed as:
25-7

-------
V - V
% Porosity = 	 x 100	(1)
Vc
where Vc is the volume of an empty standard cup (known), and is the volume
of dust contained in the standard cup when full. The ordinary distribution
of Wabamun fly ash has a porosity of about 53%; while the fine distribution
has a porosity of about 68%, largely due to the cohesiveness of the smaller
particles. This suggests that the fine sample also contains larger, on the
average, voids than does the ordinary distribution. Thus, Paschen's law
predicts higher breakdown potentials. Finally, the data of Figure 6 indicate
that the average void dimension of the fine sample is less than that of the
coarse sample. The relative insensitivity of . VBD to particle size fractions of
the cement dust suggests that the void dimension over which breakdown occurs
must be the same for the three distributions.
The mechanical compression used to change the surface areal loading of
the Wabamun ash had no effect on its breakdown voltage. Apparently, the
predominantly spherical particles of this dust pack so efficiently that the
mechanical loading could not further pack the sample and change its void
dimensions. The cement dust was affected by the extra compression (Figure 8).
In this case, the void dimension of the relatively inefficiently packed sample
was decreased and the breakdown voltage was lowered.
APPLICATION OF PASCHEN'S LAW
In this section, the data will be used to justify the application of
Paschen's law to breakdown in a gas-filled void, in the volume and surface
conduction limits.
One form of the expression for Paschen's law is (12):
Jpd
'loc " ~ ( Apcf
v.-	- 	(2)
ln(1+y"
In (_ZZe3Z_) '
ln(l+Y~l )
where V^oc is the local enhanced voltage across the void at breakdown; p, the
pressure; d, the void dimension; y, the Townsend secondary ionization coeffi-
cient; and A and B, constants related to the gas. Since temperature was the
variable in these experiments, it is preferable to write this equation in terms
of number density (n, molecules/cm3) using the ideal gas law:
Cnd
Vl~ " m ( D-d j '	U)
ln(l+y 1)
where C and D are new constants associated with the change in variable, y and
the primary ionization coefficient, a, are related by:
Y
Then
-1 = ead - 1 .	(4)
25-8

-------
Cnd
Vlc " In (SB.) '	<5)
l" (lr) " (r - ) - i (Is- - J)2 ~ t (ir" ')3 . (6)
for 0 < < 2.
Neglecting all but the first term, as justified below,
r ~ Cnd
loc ~ Dn
a - 1
or
where
(7)
vioc " rr=	<8)
n
m = a	h = Cgd
m rr	b ncr
Assuming that a and d remain essentially constant for the conditions of this
experiment, so that m and b are constants, equation (8) can be put into the
form of a straight line:
vloc *  A22-) + b 	<>
If the average applied voltage at breakdown is proportionally related to ViQC
(as through an effective dielectric constant) then the form of equation (9; can
be applied to VRD as
VBD * 	+ b' 	<10>
where b" accounts for the proportional factor. Figure 13 shows the result of
plotting equation (10) using the breakdown data of Wabamun fly ash for all
moisture levels in the two temperature limits. A straight line for each
temperature limit results. Moreover, the slope of these lines results in
values of (Dn/a) between 0.8 and 1.1, making the neglect of all but the first
term in the logarithmic expansion of equation (6) a valid approximation.
Figure 14 shows the same two lines with all available data. The data corre-
sponding to temperatures between the purely volume and surface conduction
regions fall between the two limits of equation (10). The data are not
complete enough to conclude that the slopes of the two lines are equal, but
they are very close. If the slopes were, in fact, equal, then the ionization
coefficient would be constant. This would lead to the conclusion that, since
the intercepts of the two lines are different, breakdown occurs in voids of
different dimensions in the two limits. Before full use of equation (10) is
possible, however, reliable values of C and D or A and B are required.
25-9

-------
CONGLUSIONS
The foregoing discussion leads to the following conclusions concerning
electrical breakdown in dust layers:
1.	Electrical breakdown is an ordinary electron avalanche process
originating in voids within the dust layer and obeying Paschen's law. The size
of voids in the 0.5 cm dust layers used in this work appear to be in the range
of 10 to 20 vim.
2.	The electrical breakdown mechanism is connected with the conduction
mechanism prior to breakdown. Water vapor strongly affects the breakdown
voltage by its presence or absence, but only weakly by its abundance.
3.	Both a fine (MMD < 5 ym) and a coarse (MMD > 50 jim) distribution of
fly ash have void dimensions greater than an as-received distribution and
exhibit correspondingly higher breakdown voltages.
4.	Moderate mechanical compression (2500 gm/cm2) has no effect on fly
ash, but does decrease the void dimensions in cement dust, resulting in lower
breakdown voltages.
REFERENCES
1.	Budenstein, P. On the Mechanism of Dielectric Breakdown of Solids. IEEE
Trans. Electr. Insul. EI-15:225-40, 1980.
2.	Mosley, R., Cavanaugh, P., McDonald, J., and Sparks, L. Measurements of
Electrical Properties of Fly Ash Layers. In: Third Symposium on the
Transfer and Utilization of Particulate Control Technology: Volume II,
EPA-600/9-82-005b (NTIS PB83-149591), July 1982.
3.	Bickelhaupt, R. Electrical Volume Conduction in Fly Ash. APCA Journal
24:251-5, 1974.
4.	Bickelhaupt, R. Surface Resistivity and the Chemical Composition of Fly
Ash. APCA Journal 25:148-52, 1975.
5.	Mosley, R., McDonald, J., and Sparks, L. Mathematical Modeling of Ionic
Conduction in Fly Ash Layers. In: Third Symposium on the Transfer and
Utilization of Particulate Control Technology: Volume II,
EPA-600/9-82-005b (NTIS PB83-149591), July 1982.
6.	Earhart, R. The Discharge of Electricity Through Gases at Various
Temperatures. Phys. Rev. 31:652-61, 1910.
7.	Raju, G. and Hackam, R. Note on Paschen Law and the Similarity Theorem at
the Minimum Breakdown Voltage. IEEE Trans. Plasma Science PS2:63-6,
1974.
25-10

-------
8. Heisen, A. Note on Paschen Law and Similarity Theorem in Gaps with
Cylindrical and Plane Geometry. IEEE Trans. Plasma Science PS4:129-33,
1976.
9. Nakano, Y. and Evans, J. The Relationships Between Pore Classes in a
Solid Computed by Monte Carlo Technique. Powder Tech. 35:115-8, 1983.
10.	Suzuki, M. and Oshima, T. Estimation of the Co-Ordination Number in a
Multi-Component Mixture of Spheres. Powder Tech. 35:159-66, 1983.
11.	Verhaart, H. and van der Laan, P. The Influence of Water Vapor on
Avalanches in Air. J. Appl. Phys. 55:3286-92, 1984.
12.	Husain, E. and Nema, R. Analysis of Paschen Curves for Air, N2 and SFg
Using the Townsend Breakdown Equation. IEEE Trans. Elect. Insul.
EI17:350-3, 1982.
HIGH
VOLTAGE
STAINLESS
STEEL
SV
100ka
ll
GLASS-FILLED
TEFLON
Figure 1. Schematic diagram for electrical breakdown measurements.
25-11

-------
D
> 3
DRY AIR
O
6% H20 *
 9%Q
O
&
103/T(Kl 2.7
C 97
2.6
112
2.5
127
2.4
144
2.3
161
2.2
182
2.1
203
2.0
227
1.9
253
Figure 2. Electrical breakdown of Wabamun fly ash.
DRY AIR
* h>o
10" -
10"-
10" -
10"
.10".
I03/T(K|	 2.8
C 		 M
Figure 3. Resistivity of Wabamun fly ash.
1 1
III'
O SUNDANCE
A WABAMUN
O LEHIGH 	
0
o
o 0
6
O  O o
0 A A 
* 8g
1 i
1 1 1 1 .....
61
T IK) 2.7 2.6
C 97 112
2.S
127
2.4
144
23
161
2.2
162
2.1
203
2.0
227
1.9
263
1013 
1012
'	Ir,  I
LEHIGH
X x
10" 
/
WABAMUN
^ \
\

O SUNDANCE	1
V
Figure 4. Electrical breakdown of at! dust species for
9% water vapor.
1>10|
103/T(K) 2.8
C 84
2.6
112
2.4
144
2.2
182
2.0
227
1.8
283
Figure 5. Resistivity of all dust species for
9% water vapor.
25-12

-------
i r
MMD > 50pm
AMMD<5Pm
O ORDINARY
DISTRIBUTION
103/T(K) 2.7 2.6
C 97 112
2.6
127
2.4
144
2.3
161
2.2
182
2.1
203
2.0
227
1.9
253
1013

V
| 1012
>
K
O 	o '
MMD > 50/im
1011 
1010l_
10^/T(K) 2.8
OC 84


ORDINARY
DISTRIBUTION
\
-to
V
2.6
112
2.4
144
2.2
182
2.0
227
1.8
283
Figure 6. Electrical breakdown of size-fractionated
Wabamun fly ash.
Figure 7. Resistivity of size-fractionated Wabamun
fly ash.
i r
10 gm/cm2_
> 4
Q
> 3
 
A A #
* 6
~2500 gm/cm^
J	I	L
12
10
8 i
6 s
ID
103/T(K) 2.7 2.6 2.5 2.4 2.3 2.2 2.1 2.0 1.9
C 97 112 127 144 161 182 203 227 253
Figure 8. Effect of compaction on the electrical
breakdown of Lehigh cement dust.
10""r
V 1013(.

>
p
M
w 1012 
qj 'w r
OC
10 gm/cm^
A A A
A 2500 gm/cm2 ^
10"l_
103/T(K) 2.8
C 84
2.6
112
2.4
144
2.2
182
2.0
227
1.8
283
Figure 9. Effect of compaction on the resistivity
of Lehigh cement dust
25-13

-------
DAYS:
Figure 10. Effect of continual electrical stress on the breakdown
behavior of Wabamun fly ash (4 kV/cm stress; 152C;
9% water vapor).
HRS:0 60 120 180 240 300 360 420 480 640
DAYS:0 2.5 5 7.5 10 12.5 15 17.5 20 22.5
Figure 11. Effect of continual electrical stress on the resistivity
of Wabamun fly ash (4 k V/cm stress; 152C; 9% water vapor).
25-14

-------
22 C
I
a> 400
15	20
pd, mmHg x mm
Figure 12. Paschen curves for breakdown in air from Reference 6.
iii i r
SURFACE CONDUCTION
REGION (100^115C)
 VOLUME CONDUCTION
REGION (220*-270C)
1-1 1.3 1.6 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1
^BD_ (10.16 v*3 )
molecule
6
6
>4
Q
3
2
1
- 1 1 1 1 1 1
1 1 1
SURFACE CONDUCTION
-a


_ VOLUME CONDUCTION
1 1 I 1 1 1
1 l l
1 1 1.3 1.6 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3 1
VBD .	volts-cm^
<10-
16
molecule
)
Figure 13. Plot of equation (10) in the volume and surface Figure 14. Plot of equation (10) for all Wabamun fly ash
conduction regions for all humidity levels of	data.
Wabamun fly ash.
25-15

-------
BIPOLAR CURRENT PROBE FOR DIAGNOSING FULL-SCALE PRECIPITATORS
Senichi Masuda and Toshifumi Itagaki
Department of Electrical Engineering, University of Tokyo
7-3-1, Hongo, Bunkyo-ku, Tokyo, Japan 113
Shigeyuki Nohso and Osamu Tanaka
Sumitomo Heavy Industries, Ltd.
2-1-1, Yato-cho, Tanashi-city, Tokyo, Japan 188
Katsuji Hironaga and Nobuhiko Fukushima
Nihon Kagaku Kogyo Co., Ltd.
2-1, Shimizu, Suita-city, Osaka, Japan 565
ABSTRACT
A bipolar current probe for measuring negative and positive ionic cur-
rent densities in a back corona field and its space potential distribution is
modified so that it can be applied to a full-scale precipitators for their
diagnosis. The sensor at a high voltage and the control unit at grounded
potential are separated by optical links. Special precautions are made so
as to prevent the probe from dust contamination and protect the sensor from
sparking damage. Tests are made at a laboratory precipitator and the modi-
fied probe is going to be used in a large pilot plant under practical oper-
ating conditions. The detailed construction of the modified probe is
presented, together with the test results.
1. INTRODUCTION
Measuring current densities of positive and negative ions in a precipi-
tator field is one of the most effective methods for the analysis of the ESP
performance under back corona. The ratio of the positive ionic current
density, i+, to the negative one, i_, in the bi-ionized ESP field theoreti-
cally gives the degradation in particle charge, and, hence, the drop of
apparent migration velocity of dust and collection efficiency. Therefore,
i and i_ should be regarded as key parameters to detect the severity of back
corona and to prevent its harmful effects.
The separate measurement of i and i_ is possible with a bipolar ionic
current probe developed by the authors (1,2). This probe was tested in a
laboratory ESP to confirm its theory and gave a satisfactory result: the
decrease in particle charge and dust apparent migration velocity estimated
from the measurement of i and i agreed with those measured by experiments.
Also other important informations concerning back corona were obtained with
26-1

-------
the probe. In particular, it can be used for measuring the space potential
in a bi-ionized field.
Based on these laboratory results, the authors attempted to modify the
prove for use in diagnosis of a full-scale ESP under practical operation.
The detailed construction of the modified probe and the laboratory test
results are presented in the following paragraphs.
2. CONSTRUCTION OF THE PROBE SYSTEM
Figure 1 shows the block diagram of the modified probe system. The
system consists of a probe to be inserted into the ESP field, a current
measuring unit, optical links for transmission of the current signal, and a
control unit where the magnitude of the ionic current is displayed. Each
unit has the following features.
2.1	PROBE
Figure 2 shows the configuration of the probe and its supporting pipe.
Alumina ceramic is used as the base of the probe. Tungsten electrodes are
printed on its surface, and electro-plated with nickel to prevent oxidiza-
tion. This configuration allows the probe to be used in an elevated temper-
ature. Mechanical strength of the probe is enough with its 6.2 mm diameter.
The probe is attached to the supporting pipe made of stainless steel.
When the electrodes are contaminated with dust, the probe is pulled into the
supporting pipe through a cleaning head by means of a geared motor which is
operated manually. As the probe passes through the cleaning head, the dust
on the electrodes is wiped off with ceramic wool packed in the head. The
cleaning can be done even if the probe is at high voltage, thus assuring a
long term continuous operation.
Figure 3 shows the effect of the probe contamination and its cleaning.
The probe is inserted into a laboratory ESP at a point 2.1 m downstream of
the inlet. The current I from the probe electrode 1 facing to the
discharge electrode drops with time owing to the accumulation of dust. The
dust concentration^at the inlet of the ESP, the gas temperature and the gas
velocity are 7 g/Nm , 140 deg.C and 1.1 m/s respectively. The recovery in
the magnitude of I by cleaning can be clearly seen.
2.2	CURRENT MEASURING UNIT
In the current measuring unit, one of the I/V converters (operational
amplifiers) connected to three probe electrodes is selected by a selection
signal from the control unit. The ion current entering the selected elec-
trode is converted into a digital optical signal so as to be transmitted by
an optical fiber. Figure 4 shows the terminals of the optical fibers.
Since the current measuring unit is at a high voltage while in opera-
tion, solar batteries are used to supply power to its electronic circuit.
They are charged through air gap by the light from a halogen lamp at grounded
26-2

-------
Probe
H. V.
A
Current Measuring Unit
HTlH
Control Unit
Lx channel He/o r
Sphere Gap | lRpwHnn 0*J ZZZ, !
[e/o [-
I dpm Hf/v h/e h
^ap j
.MPX Channel
* Selection
i
i
i
I/V: Current-Voltage Converter
V/F:	Voltage-Frequency Converter
F/V: Frequency-Voltage Converter
E/0: Electrical-to-Optical Converter
0/E: Optical-to-Electrical Converter
MPX: Multiplexor
DPM: Digital Panel Meter
H.V.: High Voltage
Figure 1. Block diagram of the modified probe system.
1800
110
30
1-10

,Cleaning Head
Part A
Probe
li
/Connection Pins
sProbe Supporting Pipe
Guard Electrode
,Electrode 1
-X" Guard Electrode
r fcfca
Electrode 2
Electrode 3
Detail of Part A
(Dimension: mm)
Figure 2. Configuration of the probe and its supporting pipe.
26-3

-------
- 1000
I -500
t-t
0
Figure 3. Time-dependent change of the probe current.
	1	r
'
Probe
Probe
Cleaning
^ i
Cleaning
1
T : 140 deg.C
VEP: -32 kV
I : -0. 32 mA
vg: 1.1 m/s
V : -4.8 kV
C^: 7 g/Nm3
1	L
x : 40 mm
I _ 1		
0	5	10 15 20
Time (min)
Figure 5. Solar batteries on the current measuring
unit and the halogen lamp for their
excitation.
Figure 4. Optical fiber connections to the
current measuring unit.
26-4

-------
potential (Figure 5). Figure 6 and 7 show the photographs of the current
measuring unit and the probe supporting pipe.
2.3	OPTICAL LINKS
Three optical fibers are used to separate electrically the current
measuring unit at a high voltage from the control unit at the grounded
potential in signal transmission. One is for the transmission of the probe
current signals, and the others for the transmission of the probe selection
signals from the control unit to the current measuring unit.
2.4	CONTROL UNIT
Figure 8 shows the photograph of the control unit. There are two
switches, one being to manually select one of the three electrodes of the
probe. The selection signal is transmitted to the current measuring unit,
and the ionic current entering the selected electrode is optically trans-
mitted to the control unit. Then, it is converted again to the electrical
signal and displayed on the front panel.
Another switch is for starting the cleaning of the probe electrodes.
By pressing the button, dust on the electrodes begins to be wiped off, and
the probe moves back to the initial position to restart measurement.
2.5	SPHERE GAP
The probe is subjected to a sparking caused by two different reasons.
One is the sparking between the discharge electrode and the probe occuring in
the case when the probe potential is set too much lower than the original
local potential of its position, or the probe is located too close to the
discharge electrode. The other is the sparking between the probe and the
grounded portion of the ESP. The former is much more likely to cause damage
to the current measuring unit or high voltage power supply of the probe
because of a large capacity of the high voltage power supply of the ESP.
A counter measure is taken by using a sphere gap in parallel to the high
voltage power supply, which keeps the voltage of the probe below the gap
spark voltage even if the sparking takes place. The spheres used are 20 mm
in diameter and the length of the gap is adjusted to a desired value: for
example 6 mm gap corresponds to about 20 kV spark voltage. In addition, 100
megohm resistors are connected in series to the electrode wires at the input
of the current measuring unit to suppress surge current. The combination of
the sphere gap with the resistors can completely prevent the spark damage
which would otherwise obstruct the operation of the probe.
3. OPERATION OF THE MODIFIED PROBE SYSTEM
3.1 MEASUREMENT OF LOCAL POTENTIAL AND ION CURRENT DENSITIES
The probe measurement begins with establishing the balanced condition at
which the probe voltage becomes equal to the original local potential Vp at
26-5

-------
Figure 6. Current measuring unit and
probe supporting pipe.
Figure 7. Housing of the current measuring
unit and probe supporting pipe-

Figure 8. Control unit.
26-6

-------
the probe position. The probe potential is varied, and the currents I , I ,
1^ detected by the measuring electrode 1 facing to the corona electrode, the
center electrode 2, and the measuring electrode 3 facing to the plate are
measured. Then, the ionic current densities i_ and i are calculated from
the values of I and I at the balanced potential, I and I as follows
(1):
i_ = -I /4al cos(d/2a)
i+ = I3p/4al cos(d/2a)
where a = radius of the probe (3.1 mm), 1 = length of the electrode 2 (10
mm) and d = width of the electrode 2 (2 mm).
The verification of these theoretical equations have already been made
by experiments (1), but tests are performed also for this particular modified
probe in a mono-ionized corona field of a needle-to-plane system (Figure
9(a)). Figure 9(b) shows the values of 1^, and I measured as functions
of the probe potential V. In the mono-ionized field the balanced potential,
V , is theoretically given by the crossing point, P, of the extrapolation of
the linear region of the	curve with the horizontal axis. He^ce, we get
I = -70 nA and I = 0 nA giving i =60 nA/cm and i+ = 0 nA/cm
3.2 INFLUENCE OF THE RIPPLES IN CORONA VOLTAGE
In a practical ESP the corona voltage always contains more or less
ripples to cause fluctuations in the local potential in the field. But even
in this case, the probe measurement is possible, so far as the probe poten-
tial be supplied through a voltage divider from the corona voltage. In this
case, the balanced condition is maintained at every instant, once it has been
established. The local potential Vp and the ionic current densities i_ and i+
thus obtained indicate the averages of the fluctuating values.
If only the measurement of space potential is required, there is no need
to use a voltage divider even when the ESP voltage fluctuates. For example
Figure 10 shows the comparison between the curves obtained with a voltage
divider and with a separate DC high voltage source in a corona system shown
in Figure 9(a) now energized by pulsating corona voltage. Although the
corona voltage fluctuates between -13 and -25 kV, no difference between the
two cases can be seen in the linear regions, giving the same value of Vp.
However, as the probe potential increases (in negative direction), the
curves with a DC probe voltage deviate to a higher level, possibly giving
rise to the i .
3.3 EXTENSION OF THE RANGE OF PROBE POTENTIAL
For accurate measurement of Vp, it is essential to get a linear portion
of the	curve i-n a wide range of the probe potential. But when a DC
power source is used for setting the probe potential, this is sometimes
difficult because there exists a lower limit of the probe potential. This
is caused by the corona current flowing into the whole surface of the probe
and its supporting pipe which increases at a decreased probe potential.
26-7

-------
Corona Current
Detection Probe
(1 cm2) \
Needle
Probe
m
(Dimension: mm)
Figure 9(a). Needle-to-plate corona system.
-150
-53 nA/cm'
-100
-50
0
Probe Potential V (kV)
Figure 9(b). Graphical method to determine the original
local potential.
-150
<
c
-100
-50

i 	r
Probe Voltage

X : DC Voltage

Supply 

# : Divider


\l2

\
\

-A-SI*. 1
-300
<
c
u
3
u
0	p -5	-10
Probe Potential V (kV)
Figure 10. Probe characteristics in a pulsating
potential field. (Needle voltage is
pulsating between -13 kV and -25 kV.
The average corona current density under
the needle Is about -80 nA/cm^.)
-100
"8 -10 "12 -14
Probe Potential V (kV)
Figure 11. Probe characteristics in EPRI-Alapahoe
pilot ESP. VEP: -43 kV, I: -38.5 mA,
x: 30 mm, T: 150 deg.C at the inlet
(by the courtesy of EPRI, CE and SRI)
26-8

-------
This current flows across the high internal resistance of the probe DC power
supply to produce a substantial voltage drop. The probe potential cannot be
lowered below this level of voltage drop.
In this case, connecting a resistance in parallel to the probe DC volt-
age source helps extend the range of the probe potential. Of course the
magnitude of the resistance should be determined in accordance with the type
of the DC power supply, ionic current densities to be measured, probe posi-
tion in the field, etc., but the use of a 120 megohm resistance doubled the
range of the linear portion in the measurement made at a pilot ESP.
3.4	CORONA STABILIZATION
In a clean wire-to-plate system corona points usually move around along
the corona wires and cause irregular changes in corona current in time and
space when smooth wires are used. This makes the probe measurement diffi-
cult, requiring a means of stabilizing corona current. Making notches or
attaching small needles on the wires is effective for this purpose, so far as
the attention be paid to keep the changes in the current-voltage characteris-
tics and the space potential distribution negligibly small.
But with an ESP in practical operation, no such movement of corona
points occur because the wires are covered with dust. Figure 11 shows an
example of the result obtained with a pilot ESP in the EPRI-Alapahoe Test
Station collecting fly ash. It has 9 inch plate-to-plate spacing and 6 inch
wire-to-wire spacing. The probe is inserted in parallel to the wire from
the top of the ESP. The center of the probe is 30 mm away from the collec-
tion plate. A DC high voltage power supply is used to apply voltage to the
probe. The 2.6 mm thick discharge wires without notches or needles but
covered with dust provide stable corona current, so very smooth current-
voltage curves are obtained.
3.5	AUTOMATIC MEASUREMENT WITH A PERSONAL COMPUTER
Although no computer is used so far in the operation of the modified
probe system, it is furnished with terminals to process signals with a
personal computer for determination of Vp, calculation of i_ and i+, and
control of the probe cleaning. The DC high voltage power supply or the
voltage divider for adjusting the probe potential is also to be controled by
the computer.
4. MEASUREMENT OF SPACE POTENTIAL IN A LABORATORY ESP
The space potential is also measured by the present probe in a laborato-
ry ESP which has 250 mm plate-to-plate spacing and 150 mm wire-to-wire
spacing (Figure 12(a)). There are 16 discharge wires each with 3.2 mm in
diameter. Each of the three neighboring discharge wires out of 16 has a
small metal ring wound on it to stabilize a corona point. As shown in
Figure 13(a) the rings are made of a 0.6 mm thick copper wire wound in two
turns around a notch on the discharge wires and covered with soft solder.
Before the experiments, the influence of the ring on the discharge character-
26-9

-------
Current
Detection Unit
Control
Unit
DC H.V. ^
H.V.
Probe
Collection
Plate
m
Ring
o
3.2$ Discharge
Wire
m
TTT
X
125
(Dimension: mm)
Figure 12(a). Experimental set up of the probe in a laboratory ESP.
T: 140 deg.C
1
-50
-30
-40
-10
-20
Voltage (kV)
Figure 12(b). Current-voltage characteristic of the laboratory ESP.
26-10

-------
3.2 rnmtji
Discharge	,
Wire	1*0
Ring
Ow>
Figure 13(a). Ring around the discharge wire
to stabilize the corona.
Ring
3.2
-------
istics of the wire is checked at room temperature with a wire-to-cylinder
corona system shown in Figure 13(b), and it is almost negligible as shown in
Figure 13(c).
The probe is inserted in front of the middle of the three discharge
wires attached with the ring at a position that the electrode 1 faces to the
ring. Figure 12(b) shows the current-voltage characteristic of the ESP at
the gas temperature T = ca. 140 deg.C and without dust loading. Figures
14(a),(b),(c) and (d) show the probe characteristics measured at this temper-
ature under dust-loaded condition with C. =5 j^Nm at the ESP inlet. The
resistivity of the dust is about 1,$ x 10 ohm-cm. The original local
potential Vp can be easily obtained from these figures. Plotting against
the distance x from the collection plate we get Figure 15.
All the results described above are very satisfactory, providing a firm
bases for the application of the modified probe in full-scale ESP's.
5. CONCLUSIONS
The bipolar current probe to determine the severity of back corona
through separate measurement of i and i in a back corona field has been
"T
modified so as to be applied to a full-scale ESP in practical operation. The
probe can be also used for the measurement of space potential distribution in
the bi-ionized field of the ESP.
The modified probe system is fabricated with a heat-resistant alumina
ceramic, and comprizes a dust cleaning mechanism. It is powered with solar
batteries to drive the current measuring electronics at a high potential
position. The signal transmission is made through optical links and the
spark protection of the system by a sphere spark gap.
The operation under a pulsating corona voltage is possible, so far as
the probe voltage be supplied from the corona voltage source through a volt-
age divider. In addition, the space potential can be measured also with a
DC probe power supply even in this case. However, the use of the DC probe
power supply may produce error in i and i .
The stabilization of corona points is necessary if the probe is used in
a clean wire-to-plate corona system, but no such precautions are necessary
when used in a practical ESP field.
The results of performance tests made on the modified probe are very
encouraging, suggesting that it could be used in a practical ESP for diagno-
sis and monitoring of back corona. It also provides a method of laboratory
evaluation of dust precipitability.
ACKNOWLEDGMENTS
The authors wish to express their sincere gratitude to Dr. R. Altman
and Mr. R. Ritlenmeyer of EPRI, and Mr. G. W. Driggers of CE for their
26-12

-------
<
a
-600 -
-400 -
-200 -
0	-5	-10	-15
Probe Potential V (kV)
Figure 14(a). x = 40 mm, without dust.
-1000
-800
- 5	-10	-15
Probe Potential V (kV)
Figure 14(b). x  35 mm, with dust.
-800
-600
g -400
-200
-1000
-800
g -600
400
-200
-10
-15
-20
-10
-15
-5	-10	-15
Probe Potential V (kV)
Figure 14(c). x - 40 mm, with dust.
-5	-10	-15
Probe Potential V (kV)
Figure 14(d). x  45 ram, with dust.
Figure 14. Probe characteristics. VEp: -35 kV, J: -0.25 mA/m , T: 140 C.
26-13

-------
-50
-40 -
-30 ~
o
* -20
-10 -
T
 : with dust
X : without dust

0	50	100 Wire
Distance from the Plate x (mm)
Figure 15. Probe-measured potential in a laboratory ESP field.
26-14

-------
support given to this research. The authors are also indebted to Dr. J. L.
DuBard, Mr. C. Landham and Mr. C. Lindsey of Southern Research Institute for
their help and discussions. Finally, the authors wish to thank Mr. T. Noda
and Mr. H. Terai of Sumitomo Heavy Industries for their competent help and
advice.
The work described in this paper was not funded
by the U.S. Environmental Protection Agency and
therefore the contents do not necessarily reflect
the views of the Agency and no official endorsement
should be inferred.
REFERENCES
1.	Masuda, S. and Nonogaki, Y. Sensing of Back Discharge and Bipolar Ionic
Current. Journal of Electrostatics. Vol. 10. Elsevier, Amsterdam,
Netherlands, 1981. 73 pp.
2.	Masuda, S. and Nonogaki, Y. Bi-Ionized Structure of Back Discharge
Field in an Electrostatic Precipitator. Record of IEEE/IAS 1981 Annual
Meeting in Philadelphia, Pennsylvania. October 1981. 1111 pp.
NOMENCLATURE
radius of the probe.
dust concentration at the inlet of the ESP.
width of the probe electrode 2.
positive ionic current density,
negative ionic current density,
corona current.
ionic current detected by the probe electrode 1.
ionic current detected by the probe electrode 2.
ionic current detected by the probe electrode 3.
I at V = V .
X3 at V = V
corona current density.
length of the probe electrode 2.
gas temperature.
gas velocity.
probe potential.
corona voltage.
original local potential at the center of the probe,
distance from the collection plate to the probe.
26-15

-------
A METHOD FOR PREDICTING THE EFFECTIVE VOLUME
RESISTIVITY OF A SODIUM DEPLETED FLY ASH LAYER
Roy E. Bickelhaupt
Bickelhaupt Associates, Inc.
Meadowood Subdivision, Box 54N
P.O. Box 3532
Carbondale, IL 62901
Ralph F. Altman
Electric Power Research Institute
3412 Hi 11 view Avenue
Palo Alto, CA 94303
ABSTRACT
Research has been conducted to develop a technique for predicting the
effective volume resistivity of a fly ash layer that has been exposed to
circumstances leading to a hot-side precipitator problem known as sodium
depletion. Eight fly ashes, representing a major portion of U.S. coals,
were evaluated by subjecting 0.5 cm ash layers to a continuously applied
voltage gradient of 4 kV/cm for periods of time up to 35 days at a temper-
ature of 350C (662F). Resistivity was determined at 350C before the
test started and after the long period of applied voltage used to create
the sodium depleted situation. In this condition, resistivity was also
determined at 280C (536F) for an average electric field intensity of
4 kV/cm and immediately prior to breakdown. This experimental method can
be used with any ash sample to estimate resistivity in a sodium depleted
condition. But in addition, a correlation resulting from these resistivity
data has been developed. This correlation is combined with an existing
technique for predicting resistivity to produce a method for calculating
the expected fly ash layer resistivity after sodium depletion has occurred.
27-1

-------
A METHOD FOR PREDICTING THE EFFECTIVE VOLUME
RESISTIVITY OF A SODIUM DEPLETED FLY ASH LAYER
INTRODUCTION
Several years ago when electrostatic precipitators located on the hot
side of the air preheater were coming into extensive use, several serious
problems were observed including mechanical difficulties such as poor plate
alignment and other attendant effects which resulted from faulty design,
excessive temperatures, and thermal cycling. These mechanical problems not
only caused poor precipitator performance, they masked and contributed to
other problems. Precipitator fouling has also been a significant problem.
This problem is principally related to ash chemistry, fine particles and in-
advertent incursions of excessive moisture and is also influenced by poor gas
distribution, poor rapping and excessive grain loading. A third major problem,
subsequently called "sodium depletion", was also observed. While this problem
is aggravated by fouling and mechanical problems, sodium depletion can have a
devastating effect on the performance of a mechanically sound, "clean" pre-
cipitator. This article is concerned with the latter problem and gives a
method for measuring and predicting the resistivity of an ash layer in a
sodium depleted condition.
The sodium depletion problem is characterized by a time-dependent
degradation of precipitator performance. It has been observed that performance
can be substantially restored by thorough sand blasting or washing the pre-
cipitator collection plates only to have the deleterious effect occur again with
time. A most interesting observation was that after a period of operation the
precipitator electrical characteristics demonstrated back corona suggesting very
high resistivity, while in situ and laboratory measurements indicated the
resistivity was in a desirable range for ash collection. These observations
were made with both "clean" and fouled precipitators.
A generally accepted explanation (J[) of the time dependent hot-side
precipitator performance deterioration referred to as sodium depletion has
been given. Briefly, the argument states that a thin layer of fly ash becomes
permanently attached to the collection plates. This occurrence is promoted
by high inherent ash resistivity, cementitious ash, and frequent incursions
of moisture. It can also be influenced by collection plate material, surface
finish, and ineffective rapping. However, the layer can be present in a pre-
cipitator with no operational problems and cannot be completely removed by
either normal or power-off rapping.
After a long period of time, weeks to months, under electrical stress,
the effective resistivity of this thin layer increases greatly, and the
precipitator responds as if it were collecting a very high resistivity ash
even though the measured inherent ash resistivity is not high. The explanation
states that the positive charge-carrying alkali metal ions present in the
tenaciously adherent ash layer migrate away from the collection plate leaving
a glassy matrix that must depend on another conduction mechanism to transport
charge. It has been suggested that this mechanism is controlled by oxygen
ion migration and that this change in rate-controlling mechanisms leaves the
thin ash layer with a significantly higher resistivity.
27-2

-------
Remedial procedures have been discussed {2) arid several techniques have
been tried with mixed results. Thorough plate cleaning by washing and/or sand
blasting gives relief; however, the success of this procedure depends on the
length of time until it has to be repeated. A specific sodium conditioning
technique has been reported (3) as a suitable procedure. In this case, it is
important to uniformly distribute the sodium compound throughout all of the
coal to be burned and to make sure the sodium addition is compatible with the
boiler operation. The judicious selection () of coals can also solve the
sodium depletion problem. Ash from select coals can either serve as a form
of sodium conditioning or develop an adherent layer that has a resistivity
level that can be tolerated by the particular precipitator. Other forms of
conditioning with sodium compounds or proprietary compounds have also been
reported {5). Finally, the effect of pulse energization devices for control-
ling the electrical characteristics of the precipitator to combat the back
corona resulting from sodium depletion have been described (6).
For several reasons it would be desirable to be able to predict the
effective resistivity of an ash layer after sodium depletion has occurred.
The data would be useful for: a) selecting coals for compatibility with
specific existing precipitators, b) trouble shooting poorly performing
precipitators, c) sizing new hot-side precipitators, and d) estimating the
lowest tolerable operating temperature for a given coal/precipitator combina-
tion. A technique (7j is available for predicting the inherent fly ash
resistivity before sodium depletion has occurred as a function of temperature
and field strength from the composition of the fly ash or properly prepared
coal ash. This report discusses additional work conducted to develop an
experimental method for determining the resistivity of depleted ash and to
expand the resistivity predictive model to include the prediction of the
resistivity of ash layers in the sodium depleted condition.
EXPERIMENTAL APPROACH
As described in reference 7, fly ash resistivity at 350C (662F) can
be predicted using experimental results correlating resistivity with ash
composition and with electric field intensity. From these predicted data
and Equation 1 below, resistivity as a function of temperature in the hot
side range, 250C to 500C (482F to 932F) can be computed.
log pv = log pQ + log e (S/T)	(1)
where:
pv = volume resistivity of fly ash at any temperature,
p0 = complex ash parameter involving the number of mobile charge carriers,
S = slope of the high-temperature, linear portion of the log p vs (1/T)
relationship in K,
e = natural logarithm base,
T = absolute temperature, K.
27-3

-------
To expand this approach for the prediction of the resistivity of a
sodium depleted ash layer, additional data were required as follows:
	the amount of resistivity increase that will occur due to the develop-
ment of a stable, rate-controlling, sodium-depleted layer,
	the effect of electric field intensity on the resistivity of a stable,
rate-controlling, sodium-depleted layer, and
	the effect of the sodium-depleted layer on the slope of the
resistivity/temperature relationship.
A stable layer is one that has attained a quasi-equilibrium condition of
thickness and composition as a result of competing mechanisms. It exhibits
little additional attenuation of current density with time.
To be able to evaluate the required data with respect to inherent ash
resistivity and ash composition, a matrix of eight fly ashes was selected.
The chemical compositions are shown in Table 1. Ashes 304, 305, and 306
are typical of those ashes produced from bituminous coals from the eastern
U.S. The main difference in chemical composition among these ashes is the
variation in iron concentration which normally increases with an increase
in sulfur in the coal. Ashes 301, 302, 303, 307, and 308 are typical of those
ashes produced from lignites and sub-bituminous coals from the western U.S.
The principal variations among these ashes are the alkaline earth concentra-
tion (calcium plus magnesium) and the sodium concentration. These eight
ash compositions represent a large fraction of the coals mined for steam
generation. It was anticipated that this collection of ashes would allow
data interpretation either in terms of composition or magnitude of inherent
resistivity. The predicted inherent resistivity values as well as the
measured initial resistivity values at the start of the experimentation are
also given in Table 1. From these data it can be seen that a two order of
magnitude variation in resistivity was represented by the selected ashes.
To satisfy the objective of the experiment, it was necessary to evaluate
the resistivity/temperature relationship after the ash layers had demonstrated
that they had attained a stable sodium depleted condition. To achieve this
goal, it was decided to apply a constant voltage for a long period of time
until current density showed only a minor rate of degradation. The constant
voltage, 2 kV, was applied to ash layers 0.5 cm thick yielding a 4 kV/cm
electric field intensity in a parallel plate resistivity cell. Furthermore,
it was arbitrarily decided that the test would continue until the current
density was decreasing at a rate of<10% in 100 hours. The constant voltage
technique was used because it allowed four ashes to be evaluated simultaneously.
An electrometer was periodically inserted into the circuit to determine the
current passing through a given sample. This was done without interrupting
the voltage to the other samples. The apparatus used to conduct these
experiments has been previously described (8).
After the current density of the four samples simultaneously under test
has reached a degradation rate of <10%/100 hours at 350C (662F) and 4 kV/cm,
resistivity was determined, and the temperature was lowered to 280DC (536F).
When thermal equilibrium was reached, resistivity was determined again, and
resistivity as a function of electric field intensity was determined until
dielectric failure occurred. The test environment during the entire program
was air containing 4% water.
27-4

-------
Table 1.
CH0OCAL OCMPOSrna* AMD RESiariVHY VALUES FCR TfE EKM FLY ASHES EVKUSOHD
Resistivity, ohn an
Ely A^i Chemical Caipoeition, Vteight Perceit afta: Ignition at 750C Preiirtod	Start of test
tfa. Li20 tfe2 K2O MgO GaO Al^j S102 TtOz P203 S03 352C, 4 kV/cm	350C, 4 kV/an
301	0.03 0.51 1.7 1.3 4.4 5.0 25.8 59.0 1.7 0.31 0.35 7 E 9	6 E 9
302	0.01 0.29 0.71 1.8 12.6 4.1 24.6 52.9 1.0 0.13 0.24 9 E 10	8 E 10
303	0.05 0.34 0.42 6.3 19.5 4.3 24.1 41.2 1.5 0.31 0.94 1 E 11	2 E 11
304	0.04 0.19 2.7 0.85 0.56 4.1 32.2 56.4 2.3 0.15 0.18 8 E9	6E9
305	0.05 0.34 3.1 1.1 2.2 12.5 27.1 50.5 1.8 0.33 0.57 2 E9	1E9
306	0.03 0.46 2.4 0.91 3.8 21.4 20.7 46.9 1.5 0.29 1.2 1 E 9	1 E 9
307	0.01 2.8 0.62 1.1 12.8 4.1 25.6 50.4 0.8 0.19 0.41 1 E9	1E9
308	0.02 1.8 0.32 6.2 30.9 5.5 19.8 30.8 1.7 1.1 4.1 8 E9	6E9
27-5

-------
Although the eight ashes used in this work collectively represent a
very high percentage of all fly ashes produced from U.S. coals, the following
test procedure is suggested for the reader who is concerned with an atypical
ash or unusual precipitator conditions. A procedure identical to that
described above for this research is suitable with the exception of test
temperatures. For a specific ash or installation, the depleted layer should
be developed at the temperature encountered or expected under normal full-load
conditions and the lower temperature used should be the lowest expected under
reduced-load conditions.
RESULTS AND DISCUSSION
CURRENT DENSITY/TIME RELATIONSHIP
Using the apparatus described in Reference 8, current passing through
each test cell was periodically recorded. Calculated current density plotted
as a function of time is shown in Figure 1 for all eight ash samples. The
tests were done in groups of four ashes: 301, 304, 305, and 306 in the first
group and 302, 303, 307, and 308 in the second group. The test was terminated
when the current density for every ash in a given group was decreasing at a
rate of <10% per 100 hours. In each group, the current density was decreasing
at a rate lower than this for one or more of the ashes when the test was
termi nated.
This data acquisition method has been used extensively in the past with
many different types of ash. The test has been conducted with parallel-plate
resistivity cells, as in this case, and with a wire-plate corona discharge
apparatus. All results have been similar in that current density as a func-
tion of time shows a rapid attenuation in the early part of the test followed
by an additional slow decay as the test continues into hundreds of hours.
This time-dependent decrease in current density is the result of the
development of a stable sodium-depleted layer contiguous with the positive
electrode and reflects the time-dependent'performance loss by a precipitator.
For the chosen test conditions, ashes 302 and 303 are good examples of
fly ashes that will produce the sodium-depletion problem in a hot-side
precipitator. After 840 hours (35 days), the current density of ash 303 was
<2 nanoamps/cm2 under the ideal conditions of a laboratory test. Past
experience indicates that the average current density could be less for the
imperfect situation in a precipitator.
RESISTIVITY/TEMPERATURE RELATIONSHIP
Figure 2 shows the effect of the sodium-depletion phenomenon on the
resistivity/temperature relationship for two of the eight ashes evaluated.
Current density data and average electric field intensity data were used to
calculate resistivity at various stages of the test. Point A indicates the
inherent resistivity of ash 302 at 350C and E = 4 kV/cm. This is the
resistivity value before the sodium-depletion test was started. Point B is
the resistivity value for ash 302 at 350C and E = 4 kV/cm after the voltage
had been applied for 840 hours. Without removing or changing the voltage
gradient, the sample was cooled to 280C. Point C shows the resistivity
of ash 302 under these conditions. Finally at 280C as a function of
average electric field intensity, resistivity was determined for the sample
27-6

-------
FIGURE 1. Current Density versus time of applied
voltage, E = 4 kV/cm.
i
|
>
>
uj
tf
352
2S3
441
OF	4M	541	ftM	664	741	028
FIGURE 2. Typical resistivity/temper-
ature relationships after 840 hours of
applied voltage.
27-7

-------
with a sodium-depleted ash layer. Point D is the	resistivity value determined
just prior to electrical breakdown. All ashes in	the depleted condition
showed a tendency for arcing at about 8 kV/cm and	electrically failed at an
average voltage gradient of 9 to 10 kV/cm using a	0.5 cm ash layer.
Although not readily discernible, the slopes of the resistivity versus
reciprocal absolute temperature lines in Figure 2 are not identical.
Comparison of these slopes with similar data generated for Reference 7
indicates that the slopes associated with sodium-depleted ash layers are
slightly greater than the slopes calculated from inherent volume resistivity
data. Resistivity predictions, taking into account surface conduction due
to adsorbed moisture, indicate that the resistivity values for ashes 307
and 308 at 280C probably have been slightly lowered by the presence of mois-
ture in the test environment. Thus, the slopes for these two samples have
not been used in this work. For the required computations, an average slope
of 4788K was used (S-j in equation 2). This value 7% includes the slopes
for all ashes except 307 and 308.
Continuing with ash 302 as an example, this ash would probably be
successfully collected at the full load temperature of 350C with a new
precipitator sized for ash 302 with a resistivity of 8 x 10^ ohm cm. After
operating several weeks or months, a thin, tenaciously held, sodium-depleted
ash layer would coat the collection plates producing an effective resistivity
of about 5 x 10'' ohm cm. When the unit is operated at reduced load and a
much lower precipitator temperature prevails, the resistivity increases to
a value >1 x 10^2 ohm cm. This sequence of events leading to unacceptable
precipitator performance has been observed by many coal-burning power
generating stations with hot-side precipitators.
EFFECTIVE RESISTIVITY/INHERENT RESISTIVITY RELATIONSHIP
Previously, miscellaneous tests conducted with respect to the sodium-
depletion problem had indicated that one could expect about an order of
magnitude increase in resistivity as sodium depletion occurred. The present
information allows one to examine a correlation between the inherent
resistivity and the resistivity determined after a stable sodium-depleted
layer has been produced. Figure 3 shows a plot of resistivity values for
all ashes at 350C and E = 4 kV/cm in which the y axis denotes the values
determined after the test to establish the sodium-depleted layer and the
x axis denotes the values obtained before the test was started. The data
point for ash 302 in Figure 3 is simply point B of Figure 2 (y axis) plotted
versus point A of Figure 2 (x axis).
In the upper right corner of Figure 3 is the equation of the line computed
from a least squares analysis of the data. The coefficient of correlation
R = 0.97 indicates a good fit between the data and the computed line.
Using this equation, one can determine that ashes having inherent (before-
depletion) resistivities of 1 x 10^ ohm cm and 1 x 10ohm cm will increase
by factors of 12.6 and 5.8, respectively, due to sodium depletion. This
agrees reasonably well with the aforementioned rule of thumb which used a
factor of 10.
27-8

-------
M
Ui
K

4
E
>
V
e
ft
4
a
>
-
>
-
a
.1.	I	J	I	I	1	L	
7 * io'	io"
Rf SISTIVtTY OHM CM 3SOC 4 l>V/cm Bf FOOf ItST
FIGURE 3. Effect of long-time applied
voltage on fly ash resistivity at 350C.
27-9

-------
Although additional data could alter the intercept and slope of the
line shown in Figure 3, it is interesting to note that the magnitude of the
effect of sodium depletion on resistivity apparently decreases as the inherent
resistivity increases. Intuitively this relationship seems reasonable. Assume
that within the sodium-depleted region conduction is controlled by the migra-
tion of unbridged and unassociated oxygen ions in the glassy ash matrix in
a manner analogous to that of a soda-lime-silica glass subjected to electrode
polarization (9j. Because the number of mobile oxygen ions and/or the
mobility of these ions are lower than the equivalent characteristics of the
alkali metal ions, the resistivity of the sodium depleted zone is greater than
the inherent ash resistivity. Alkali metal ions destroy oxygen ion bridging
in the silicon-oxygen network and leave the oxygen ion unassociated when the
alkali metal ion drifts away under the influence of the electric field.
Therefore, a large concentration of alkali metal ions produces a low inherent
resistivity and yields a large concentration of mobile oxygen ions in the
sodium or alkali depleted region. For this reason, one would not expect all
the current-density/time curves shown in Figure 1 to asymptotically approach
a single limiting value. As the alkali metal concentration decreases, the
inherent resistivity and the effective resistivity of the sodium depleted
ash both increase. It can be rationalized that the difference between these
two resistivity values should become smaller as the alkali content becomes
smaller because as the alkali content approaches zero, the ash resistivity
should become a very high value similar to that for pure vitreous silica
and show less effect due to electrode polarization.
Obviously the phenomenon is more complex than described above. Volume
resistivity is also influenced by the presence of iron and calcium in the ash
structure. In addition it is believed that the chemical diffusion of sodium
toward the anode due to the sodium concentration gradient established by the
electric field affects the ultimate current density/time relationship.
RESISTIVITY/ELECTRIC FIELD INTENSITY RELATIONSHIP
In Figure 4, resistivities measured just prior to electrical breakdown
have been plotted versus resistivity values determined for E = 4 kV/cm. The
ash layers were in the sodium depleted condition at 280C. The line represents
a linear regression least squares analysis of these data, and the included
equation defines the line. This equation is subsequently used to incorporate
the effect of electric field intensity into the prediction of resistivity for
ash layers subjected to sodium depletion.
From Figure 4 it can be seen that at high levels of resistivity, e.g.,
1 x 10^3 ohm cm, there is only an attenuation of about 20% due to an increase
in field strength from 4 kV/cm to values approaching electrical breakdown.
At a moderate resistivity level of 1 x 10'' ohm cm, the resistivity is reduced
by a factor of 3 as the field strength increases from 4 kV/cm to breakdown.
One can speculate that the glassy structure of the ashes containing the higher
concentrations of alkali metal ions are more susceptible to distortion due to
electrical stress.
27-10

-------
O 301
O 306
	302
	303
#	307
*	306

t-
u.
<
W
s
<
a
V
o
R
4
2
tog v ' t * S leg
I
X
o
1- 2 357?
>
>
a
* * 10,J	10"
RSSISTIVtrv OHM cm AT JtO C. 4liV/cm. AFTER TtST
FIGURE 4. Effect of spark voltage on fly
ash resistivity after long-time applied
voltage.
DETERMINATION OF THE VOLUME RESISTIVITY OF A SODIUM DEPLETED ASH LAYER
Volume resistivity as a function of temperature can be determined experi-
mentally as described above, or it can be determined by computation using
the expressions shown in Figures 3 and 4. The requisite information for
27-11

-------
computing the resistivity is the chemical composition of the fly ash or coal
ash. It is recommended that ignited (750C) fly ash samples should be used
and that the coal ash samples should be prepared as stated below. Coal samples
should be converted to ash using the ASTM D271 procedure followed by a second
ignition at 1050C  10C instill air for 12 to 16 hours (overnight). It is
emphasized that all the research done in this laboratory has utilized this
ashing procedure.
The ash samples are analyzed for their major elemental constituents
according to the following general outline. Three methods of attack are used
to place the ash in solution: a) a sodium carbonate fusion for sulfur deter-
mination, b) a sodium hydroxide fusion for aluminum and silicon determination,
and c) a sulfuric, hydrofluoric and nitric acid digestion for the elements
lithium, sodium, potassium, magnesium, calcium, iron, titanium, and phosphorus.
Sulfur is determined by ion chromatography. Phosphorus, silicon, and aluminum
are determined by colorimetry. The remaining elements are determined by atomic
absorption. The elemental analysis is reported in weight percent as oxides.
Having a chemical analysis from a qualified laboratory, the following
calculation is performed to obtain the atomic percentages of lithium plus
sodium, magnesium plus calcium, and iron. After normalizing the reported
weight percentages to sum 100%, each oxide percentage is divided by the
respective oxide molecular weight to obtain the mole fraction. Each mole
fraction is divided by the sum of the mole fractions and multiplied by 100
to get the molecular percentage as oxides. Each molecular percentage is
multiplied by the fraction of cations in the given oxide to obtain the atomic
percentage of the cation. The results of such a calculation are shown in Table
2. In this example, the sum of the lithium plus sodium atomic concentrations
is 0.53%, the sum of the magnesium plus calcium atomic concentrations is 15.9%,
and the iron atomic concentration is 0.7%.
Using the equations shown in Figures 3 and 4 and the Arrhenius-type
expression given in equation 1, the volume resistivity of a sodium depleted
ash layer can be computed as a function of temperature. Elsewhere (10), the
detailed step-by-step procedure for the development of equation 2 below is
given.
log pV(j = - 7.3759 + 1.0412 log Pvl*a + S]/T	^2)
where:
Pvd = volume resistivity of the sodium depleted ash layer at electrical
failure and any temperature T,
pvi-a = inherent volume resistivity of the fly ash at 625K and E = 4 kV/cm,
S] = 4788.1K = average slope of log P vs 1/T experimentally determined
from curves of the type shown in Figure 2, and
T = temperature in K.
The inherent volume resistivity as a function of the fly ash composition at
625K and E = 4 kV/cm can be obtained from equation 3. This equation was
developed from reference 7.	?7_i9

-------
Table 2.
TABULATION OF DATA FOR THE CALCULATION OF THE ATOMIC
CONCENTRATION OF CATIONS IN COAL ASH
Atomic
Determined Normalized Molecular Mole Molecular	Catioriic	Concentration
Oxide Wt % Mt t Weight Fraction Percentage Fraction	of Cation
Li20 0.02 0.02 29.88 0.00067 0.044	2/3	0.029
Na20 0.70 0.70 61.98 0.01129 0.747	2/3	0.498
K20 1.00 1.00 94.20 0.01062 0.703	2/3	0.469
MgO 5.0 5.00 40.31 0.12404 8.208	1/2	4.104
CaO 20.0 20.00 56.08 0.35663 23.598	1/2	11.799
Fe203 4.0 4.00 159.70 0.02505 1.658	2/5	0.663
A1203 23.0 23.00 101.96 0.22558 14.927	2/5	5.971
SiO 2 44.0 43.98 60.09 0.73190 48.429	1/3	16.143
Ti02 1.0 1.00 79.90 0.01252 0.828	1/3	0.276
P205 0.60 0.60 141.94 0.00423 0.280	2/7	0.080
S03 0.70 0.70 80.06 0.00874 0.578	1/4	0.144
Sum 100.02 100.00 	 1.51127 100.000	40.176
27-13

-------
TABLE 3.
COMPUTER PROGRAM, FORTRAN DATA PRINTOUT
1BE PREDICTION OF THE EFFECTIVE RESISTIVITY OK A
SODIUM DEPLETED ASH LAYER IS BASED ON WORK DONE BY
DR, ROY BICKELHAUPT OF SOUTHERN RESEARCH INSTITUTE.
THE RESEARCH MAS SPONSORED BY THE ELECTRIC POWER
RESEARCH INSTITUTE, DR. RALPH ALTMAN, PROJECT MANAGER,
FLY ASH 1
CORRECTED

ASH
ASH
ATOMIC

ANALYSIS
ANALYSIS
CONCENTRATION


		

LI20
0,02
0.02
0,030
NA20
0,70
0.70
0,498
K20
1,00
1.00
0.468
MGO
5.00
S, 00
4.103
CAO
20,00
20.00
11.79/
FEJ03
4.00
4.00
0, 663
A 1,303
23.00
23.00
5,969
SI 02
44,00
43.99
16.14b
TI02
1.00
1 ,00
0,276
P20S
0.60
0.60
0, 080
S03
0,70
0.70
0, 145
SUM
100,02
100,00
40,173
SUM OF LITHIUM AND SnDIUM ATI1MIC CONCENTRATIONS	0.53
SUM OF MAGNESIUM AND CALCIUM ATOMIC CONCENTRATIONS 15,9
IRON ATOMIC CONCENTRATION	0.7
1000/TCK)
DEG K
DEG C
DEG F
RHO(VI)
RHOCVD)

 - - -
-----
- 		


1.30
769
496
9 2 b
1.0E+09
6.HE+09
1.40
714
441
826
2.8K+09
2,1E+10
I. SO
667
394
740
7.5E+09
6,1E+10
1.60
625
352
665
2.0E+10
l,9fc>ll
1,70
586
315
599
5.6E+10
S.6E+U
1.80
$56
282
540
1.5E+11
1.7E+12
1.90
526
253
468
4.1E+11
5.1E+12
MOTE J BECAUSE THE PREDICTED RESISTIVITY VALUES ARt: VEkY
SENSITIVE TO SEVERAL ASH COMPOSITIONAL FACTORS, OilE MUST
EXERCISE GREAT CARE IN THE SELECTION AND PREPARATIUh UF
COAL AND ASH SAMPLES, FURTHERMORE, THE QUALITY OF THE
QUANTITATIVE CHEMICAL ANALYSIS WORK IS OF GREAT IMPORTANCE,
WHEN USING THIS PROGRAM, THE COAL ASH SHOULU bE PROPUCFli
USING ASTM D271 PROCEDURE FOLLOWED BY A SECOND IGNITION AT
1050 DEGREES C + OR - 10 DEGREES C IN SXILL AIR FOR AT
LEAST 10 HOURS.
27-14

-------
log p . = 8.3911 - 1.8916 log x - 0.9696 log y + 1.2370 log z
VI 3
(3)
where:
p . = defined above,
Kvi a
x = atomic percentage of lithium plus sodium,
y = atomic percentage of iron, and
z = atomic percentage of magnesium plus calcium.
Substituting the expression for Pvia> equation 3, into equation 2 and simplify-
ing, the sodium-depleted volume resistivity can be calculated from the fly ash
composition and temperature.
log p . = 1.3609 - 1.9695 log x - 1.0095 log y + 1.2880 log z	(4)
va + 4788.1/TK.
Combining and simplifying equations from reference 7, an expression similar to
equation 4 can be given to calculate the inherent volume resistivity from the
fly ash composition and temperature.
log p  = 1.2159 - 1.8916 log x - 0.9696 log y + 1.2370 log z	(5)
V1 + 4334.5/TK
where:
p . = inherent volume resistivity for a given fly ash composition,
E = 12 kV/cm, and any temperature TK.
Equations 4 and 5 can be used to manually calculate the predicted resistivity
of a fly ash layer before and after the sodium depletion phenomenon has
occurred. Computer programs in FORTRAN and BASIC are given in reference 10
to execute these computations. Using the ash composition given in Table 2,
the output of the computer program is shown in tabular and graphical form,
Table 3 and Figure 5, respectively. In the upper part of Table 3 the fly
ash compositional data are given. The atomic concentrations were calculated
as described above and illustrated in Table 2. The columns of resistivity
data labeled RHO(VI) and RHO(VD) were computed from equations 5 and 4, respec-
tively. These two columns of data are plotted as a function of temperature
in Figure 5.
The chosen example illustrates the pronounced effect of sodium depletion
and temperature on the effective resistivity of an ash layer. At the typical
hot-side precipitator temperature of 352C for full-load steam generating
conditions, the inherent fly ash resistivity is 2.0 x 1010 ohm cm, a desirable
level in the absence of the sodium depletion phenomenon. However, after
sodium depletion has occurred in the adherent ash layer and low-load steam
generating conditions are in effect, the precipitator temperature can be
315C and the effective resistivity is 5.5 x 101' ohm cm, a most undesirable
level. This situation is additionally aggravated when the boiler load is
being increased. The particulate concentration increases rapidly while the
precipitator temperature lags.
27-15

-------
PLY ASH I
lo'i.
ASH
COMPOSITION
O RHO (VI)
~ RHO(VD)
L 120
NA20
K20
MGO
CAO
FE203
AL 203
S102
T 102
P205
S03
0.02
0. 70
1 -00
5.00
20.00
4.00
23.00
43. 99
I .00
0.60
0. 70
3 I 000/T *
496 "C
925 F
TEMPERATURE
FIGURE 5. Computer plot of inherent and sodium-
depleted volume resistivities versus temperature.
27-16

-------
SUMMARY
The work reported in this paper has resulted in the development of two
new tools, one analytical and one experimental in nature. The analytical
correlation developed from data for eight fly ashes can be used to estimate
the effective (sodium depleted) resistivity of fly ash in hot-side precipita-
tors for the temperature range that is typical for the operation of these
units. This procedure makes it possible to assess the effect of a large
number of coal supplies on hot-side precipitator performance quickly and at
very little expense. The experimental method that was used to develop the
correlation can be used to determine the exact resistivity of sodium-depleted
fly ash under conditions that realistically represent the operation of a
hot-side precipitator. The use of either approach will make possible a more
logical selection of the coal supply for a plant with a hot-side precipitator.
ACKNOWLEDGMENT
This research was made possible by the financial support of the Electric
Power Research Institute, Coal Combustion Systems Division.
At the time this research was conducted, Dr. Bickelhaupt was a member of
the staff of Southern Research Institute, Birmingham, AL 35255.
27-17

-------
REFERENCES
1.	R.E. Bickelhaupt, "An Interpretation of the Deteriorative Performance of
Hot-Side Precipitators," JAPCA 30:812 (1980).
2.	R.E. Bickelhaupt and L.E. Sparks, "Remedial Treatments for Deteriorated
Hot-Side Precipitator Performance," EPA-600/9-82-005a 1:165 (1982).
3.	J.P. Gooch, et al, "Improvement of Hot-Side Precipitator Performance with
Sodium Conditioning - An Interim Report," JAPCA 31_:291 (1981).
4.	Private communications.
5.	a) P.B. Lederman, et al, "Chemical Conditioning of Fly Ash for Hot-Side
Precipitation," EPA-600/7-79-044a 1:79 (1979).
b) R.P. Bennett and A.E. Kober, "Chemical Enhancement of Electrostatic
Precipitator Efficiency," EPA-600/7-79-044a 1:113 (1979).
6.	a) W.V. Piulle, et al, "Evaluation of Performance Enhancement Obtained
with Pulse Energization Systems on a Hot-Side Electrostatic Precipitator,"
EPA-600/9-82-005a 1:253 (1982).
b) J.P. Gooch, et al, "An Evaluation of Pulse Energization on a Hot-Side
Electrostatic Precipitator," Electric Power Research Institute Final
Report CS-2634, November 1982.
7.	R.E. Bickelhaupt, "A Technique for Predicting Fly Ash Resistivity,"
EPA-600/7-79/204 (1979).
8.	R.E. Bickelhaupt, "Measurement of Fly Ash Resistivity Using Simulated
Flue Gas Environments," EPA-600/7-78-035 (1978).
9.	D.E. Carlson, et al, "Electrode Polarization in Alkali-Containing
Glasses," J. Am. Cer. S., 55^:337 (1972).
10. R.E. Bickelhaupt, "Method for Predicting the Effective Volume Resistivity
of Sodium-Depleted Fly Ash Layers in Hot-Side Electrostatic Precipitators,"
Electric Power Research Institute, CS-3421, Final Report.
27-18

-------
ANALYSIS OF AIR HEATER - FLY ASH - SULFURIC ACID VAPOR INTERACTIONS
Norman W. Frisch
N.W. FRISCH ASSOCIATES, INC.
Kingston, New Jersey 08528
ABSTRACT
Estimation of ash resistivity level for cold-side precipitator design
requires knowledge of SOj concentrations at the ESP entrance. Unfortunately,
a number of these measurements have indicated a highly variable S0g/S02
ratio. Sources and mechanisms involved in this variable factor are
discussed. In particular, this analysis focuses on the air heater as a
contributor to SGj concentration changes.
An analysis of possible interactions between ^SO^ vapor, fly ash and
regenerative air heater surfaces was made using a numerical finite-difference
method. Situations considered included a fly ash-free system and coal-firing
cases involving either a clean or dirty air heater. Equations representative
of condensation and adsorption mechanisms were integrated into the finite-
difference equations to account for the removal of acid from the flue gas.
This unique Heater Interaction Model (HIM) approach provides an
excellent problem solving tool for such complex issues as:
optimization of gas mixing downstream of air heaters for the
elimination of high resistivity ash zones
investigation of likely causes of malperformance of precipitators
and
cost-effective injection of conditioning agents into various
chambers of an ESP system.
Our basic approach is also of interest for non-SOg containing agents
which condense from the vapor state to a particulate form. Air heater
fouling problems which may occur in NHg-based N0X control processes can be
simulated by our models.
28-1

-------
ANALYSIS OF AIR HEATER-FLY ASH-SULFURIC ACID VAPOR INTERACTIONS
INTRODUCTION
Sulfur trioxide and its hydration product, sulfuric acid, exert
important effects in fossil fuel combustion systems. These include dew point
elevation (1), participation in high temperature metal attack (2), low
temperature acid condensation with resulting corrosion (3, 4, 5) possible
precipitator performance enhancement (6, 7, 8, 9) as well as plume opacity
effects (10). The state of the art fly ash resistivity prediction method
(8) relies on knowledge of SOg concentrations at the cold-side ESP inlet.
Unfortunately, measurements at a number of coal-fired installations have
shown that this value, relative to SO^ concentrations, nearly covers a five-
fold ratio range (0.0015 to 0.0070) (li).
This paper first reviews the literature on the major SO-j-genersting
mechanisms operating in fossil fuel combustion processes. The variability of
the several contributing processes is emphasized. Sinks for SOj are also
noted.
Attention is then focused on the often neglected but important
interactive effects which occur in the regenerative air heater feeding cold-
side ESP's. The paper discusses the development of a series of computer-
based approaches which describe temperature distributions (gas, air, metal)
in air heaters, local rates of acid condensation with related acid
concentrations, and especially the extent of competitive acid adsorption on
both suspended and deposited ashes within the air heater. The inefficiency
of sulfuric acid conditioning in certain situations is discussed, as well as
suitable remedial approaches. Specific ash types prone to this difficulty
are characterized.
ORIGIN OF SULFUR TRIOXIDE
SOURCES OF SULFUR
All fossil fuels contain sulfur. Natural gas in its raw state may
contain more than 50% hydrogen sulfide. In petroleum, sulfur occurs as
sulfides, mercaptans, polysulfides and thiophenes. In coal, sulfur
occurrence is of sufficient interest to warrant a classification - Forms of
Sulfur, ASTM D2492-79; included are the three predominant forms - organic,
pyritic and sulfate sulfur. In coals, organic sulfur exists mainly as
sulfides, disulfides, mercaptans and thiophenes (12). The most widely
studied group, pyritic sulfur, consists of FeS2 largely as the cubic form
pyrites, as well as the less common orthorhombic marcasite (13). Other
28-2

-------
important inorganic sulfides which occur are sphalerite (ZnS), chalcopyrite
(CuFeS2), pyrrhotite (Fe^_xS) and arsenopyrite (FeAsS) (14). Sulfate sulfur
arises in many cases from weathering of sulfides; varieties of iron or alkali
metal iron sulfates are believed to be of this origin. Important calcium-
containing sulfates found in some U.S. Western coals are gypsum (CaSO^E^O),
anhydrite (CaSO^) and bassanite (CaSO^'l/2H2O) (14). Another frequently
occurring sulfate is barite (BaSO^). Elemental sulfur occurs on occasion.
Oxidizable sulfur compounds in the fuel are converted predominantly to
sulfur dioxide during high temperature combustion. Very low levels of SO3
can be formed by direct oxidation of SO2 because the equilibrium for the
reaction
S02 + 1/2 02 ^ S03
is far to the left at flame temperatures (16). Thus, with 4% residual oxygen
in the gas, levels of SO3/SO2 on the order of 0.002 are the theoretical
limit. In practice, considerably higher levels are found; at the air heater
entrance, SO3/SO2 are typically 0.01-0.03 in situations involving coals with
non-alkaline ashes.
There are three likely sources for the observed sulfur trioxide: direct
high temperature reaction of SOj with atomic oxygen, decomposition of sulfate
minerals and catalytic production from SO* + 0 at intermediate temperature
levels.
DIRECT REACTION OF S02 WITH ATOMIC OXYGEN
At flame temperatures, molecular oxygen undergoes a slight dissociation
producing active oxygen atoms. Levels observed in experiments involving gas
flames are on the order of 50 to 500 ppm, depending upon fuel gas composition
and especially excess oxygen and flame temperature. Higher values of the
latter two variables result in increased levels of atomic oxygen. The direct
reaction
so2 + 0 ~ so3*
results in the production of an activated SO3 molecule which can radiate its
excess energy or lose it by contact to a third body, producing sulfur
trioxide (17).
s3* * S03+ hv
S03* S03 + M
Support for this theory includes the observation of high levels of SO3
produced when burning SO- - spiked carbon monoxide and also the observed
effect of adding compounds which are capable of preferentially reacting with
atomic oxygen.
28-3

-------
In the case of carbon monoxide oxidation, atomic oxygen may be produced
by	CO + 02  C02 + 0
leading to levels as high as 20% oxidation of SC>2 to SOg in dilute SO9-
containing gases. Inhibitors such as pyridine or carbon tetrachloride
present at the 3% level can reduce SOg formation by 50% (18).
Further support for this theory is shown by the absence of SOg
production from the 870C burning of high-sulfur coal. This is a condition
at which no atomic oxygen is produced.
Once SO2, SO3 and 0^ co-exist, they react to establish the equilibrium
levels defined by the equilibrium constant
P
s3
K = 	5-5-
P x P
so2 *o2
where Kp depends solely on temperature at atmospheric pressure and P desig-
nates the partial pressure of the indicated species.
At flame temperature, the kinetics of the SOg dissociation reaction is
sufficiently fast to permit some decomposition of the thermodynamically
excess SOg to form SO2 and Oj. Even in short cooling periods (0.1-0.2 s),
serious reductions in SOg concentrations are observed. The data of Hedley
(19), shown in Figure 1, support the following series reaction scheme
k ^	k2
S02 + 0 S03 S02 + 1/2 02
If k^ and k are taken to be independent of temperature, i.e.,
activation energies for the reactions are low, the series kinetics yields the
following dependence of SOg concentration on time
kl
S03 = [0] (k k ) [EXP (-kjtJ-EXP (-k2t)]
where [0] represents the initial concentration of atomic oxygen. Thus, SOg
peak concentrations are expected at t = ln (kj/k2)/(kj-k2) with subsequent
decay.
Experimentally, ^2^1 ^as keen found to be about 8, and relatively
insensitive to temperature. Similar patterns of peak SOg concentrations
followed by decay have been reported by others.
Moisieeva and Podolyak (20) report peak SOg concentrations which are
about 3% of the SO2 level in the case of 11% excess air firing; their
28-4

-------
corresponding value is 6% for 24% excess air. (These data refer to oil
firing.) These peak SOg concentrations were reported to be 4 to 5 times
higher than the long exposure levels. Ahmed and Zaczek (21) report a similar
pattern of peak SOg conversions.
A corollary of these observations is that boiler size influences the SOg
level produced because gas cooling rates are significantly dependent on
boiler size. Large utility boilers have rates on the order of 500 to 1000
K/s while industrial units may have rates of 2000 K/s or more.
The role- of excess air has been clearly demonstrated in oil-firing
situations. SOg concentrations in full-scale plants can be controlled by
operation at low excess air levels. Figure 2 shows the results of laboratory
experiments, demonstrating the role of oxygen concentrations and fuel sulfur
content (22). The effect of increased oxygen here is likely to be the sum of
two effects: (1) increased atomic oxygen content, and (2) higher catalytic
rate of conversion of SO2 (see below). These data suggest that at least 6%
of the sulfur in the gas phase was SOg.
100
d.80
cL
I 60
(V
k.
 <0
c
o
o
if)
>20
! 1 1 ; i : !
I >'!! | 1 
0 Duplicate samples, run C
-

4-


-
- X




-
/ 0




-
1 1
f
1 t
! !
1 ! 1
0
1 1
0-04
008	012
Seconds
016
Figure 1. Decay of flame
temperature-produced SOg (19).
2 6% S
20% S
lev. s
14 v. s
11% s
0 88V.S
0 06 10 15 2-0 25 30 35 40
Oxygen,*/. */v
Figure 2. Role of Excess Air on
Sulfur Trioxide Yield (22)
It should be noted that operation at extremely high levels of excess air
may result in lowered SOg levels as a consequence of much reduced flame
temperatures in spite of the presence of high oxygen concentrations.
DECOMPOSITION OF MINERAL MATTER
Sulfate minerals exposed to flame temperatures are potential sources of
SOg. Johnstone (1) discussed the direct decomposition of CaSO^. This
reaction has been extensively studied. Jakisch (23) considered the reaction
of Fe2Og with CaS04 to yield SOg. Kinjiro (24) reported the direct
decomposition in a neutral or oxidizing atmosphere at 10201050C. Wickert
(25) demonstrated the enhanced decomposition of CaSO^ at 1000C when admixed
with Si02 or NaCl + Si02. These data suggest that at high temperature, Si02
acts as a strong acid, competing with SC^ for reaction with CaO. Similarly,
kaolin enhances the decomposition pressure of CaSO*. MgSO^ is appreciably
less stable than CaSO^. Barite, of course, retains SOg tenaciously.
28-5

-------
It is interesting that potassium-containing minerals (illite or K-
decomposit ion of the lat	_	!6).
Iron (ferrous) sulfate, which occurs as a result of weathering in some
coals is readily decomposed thermally. Ostroff (27) reported a decomposition
temperature of 537C vs. 895 for MgSO^ and 1149 for CaSO^. In Argon, FeSO^
begins to decompose at 500-550C, yielding Fe^, S02 and S03 (28). In air,
the decomposition sequence is complicated by the catalysis of SC>2 oxidation
by Fe20g .
In general, it is believed that mineral decomposition provides a minor
source of SOg compared to reaction with atomic oxygen or catalysis. This is
in direct contrast to the early view (1) that mineral decomposition could
entirely explain the presence of SO^ in boiler flue gases.
CATALYSIS OF S02 OXIDATION
Sulfur dioxide oxidation over heterogeneous catalysts has been
extensively studied. Commercially, V20ij is an active material resistant to
poisons. While vanadium probably plays little role in coal firing
situations, high V oils have exhibited high [~9% SO^/iSOn + SOj)] levels in
the combustion gas, especially at high excess air levels C29).
Iron oxide, Fe20j, is a moderately active catalyst with peak activity
about 625C. It is the prevalent catalytic species involved in coal firing
situations; its moderate activity is balanced by the high concentration of
iron oxide typically present in boiler systems. Barrett (30) has
demonstrated the sharp increase (Fig. 3) in SO3 concentrations downstream of
Fe20j-coated tubular specimens. While MgO-coated specimens were relatively
inactive, bare steel samples developed appreciable activity after several
hours.
Ash deposited on convective section tube banks can substantially
accelerate the catalytic production of SCXj; distinct difference in SO3 levels
are seen at a plant with clean and dirty tube surfaces. (31)
It is generally agreed the Fejf^ can be an active catalyst when present
on fly ash surfaces. Fe^O^ is known to be inactive; presumably iron can also
be inactivated by reaction with acidic species in the ash (silica, alumina).
With moderate to high sulfur coals the potential exists for the
production of substantially higher SOg levels by catalysis than by direct
reaction with atomic oxygen at flame temperatures. Substantial effort has
been directed towards suppressing the catalytic activity of tube bank
surfaces. Emphasis has been on the use of additives (31), especially
alkaline earth oxides. It was concluded that the equilibrium between the
metal oxide and S0-j controls the extent of suppression of SOj formation. MgO
could suppress SOj beyond stoichiometric equivalence, possibly by adsorbing
S02 from the gas stream.
exchanged muscovite),
strongly accelerate the
28-6

-------
IMPORTANT S03 SINKS
CORROSION
High-Temperature Attack
At temperatures as low as 550-600C, wastage of tube metal may occur due
to liquid-phase attack by complex low melting alkali iron trisulfates (2,33)
of the form MgFeCSO^)^, where M represents K, Na or combinations thereof.
The formation of the trisulfates may proceed through the reaction of normal
alkali metal sulfates (from the fly ash), iron oxide on the tube surface and
SO3 in the gas. The iron oxide may originate in the fly ash or from the tube
surface. Substantial concentrations of SO3 have been shown (34) to be
necessary to sustain the alkali trisulfates at high temperatures.
Low-Temperature (Cold-End) Corrosion
Sulfuric acid vapor-containing gas exposed to surfaces below the dew
point undergo formation of a liquid phase rich in sulfuric acid. The
condensation rate observed (3, 5) with the metal surface near or at the dew
point is generally quite low compared to the peak rates observed with
surfaces 30 to 50C below the dewpoint (oil firing case). The data of
Rylands and Jenkinson are shown in Figure 4.
200
160
I 120
a
K>
 eo
40
0
'

1 t I t 1 1
-
a
r4 1
* cooling 


0



-

Bore steel 	
:
/o



MgO cooling


-J	

1
1 1 1 1
1
2 3 4 5 6
Tim, hr
Dewpoint
-- ,  300
Surface temperature.T
Figure 3. S0j concentration down-
stream of bare and coated steel
specimens (30).
Figure 4. Dependence of condensation
and corrosion rate on surface
temperature (3).
Condensation rates in coal-firing situations have been reported by Lee,
et al (35); their data show similar peak rate behavior with respect to
surface temperature. They also report the effect of low excess air firing in
a combustion research boiler. With 0.6% residual oxygen, acid dew point was
only about 130F compared to the normal dewpoint for 3-4% oxygen of 280F.
Coal sulfur in this case was 6.2%, ash 13.5%.
The rates of condensation are generally characterized by the rate of
conductivity buildup (RBU) using dew point meters and especially by the peak
28-7

-------
rate (PR) at the peak rate temperature (PRT). Below about a RBU of 100
Vamp/min corrosion rates are generally modest. An alternate approach for
measuring condensation rates is the acid deposition probe; these are constant
temperature probes which permit collection of condensate associated with a
specific temperature.
Corrosion of steel surfaces which occur at condensation are largely
related to the condensation rates, and relatively insensitive to acid
condensate concentration (36) (Under typical conditions, the acid condensate
is in the range of 65-85% (w) HgSO^). When conditions prevail in which quite
dilute acid is deposited, strongly accelerated corrosion rates are observed.
The fact that peak condensation rates occur at surface temperature
30-50G below the dew point has been explained by Land (37) as follows: with
relatively high surface temperatures, condensation rate is dependent upon the
difference between the partial pressure of SOg in the gas phase and the
equilibrium value corresponding to the surface temperature; however, when the
surface temperature is sufficiently low, fog formation occurs and the rate of
removal from the gas suffers. The equations reported by Land are used in
the analysis of cold-end condensation in the air heater section of this
paper.
Dew Point Relationships
Data, both experimental as well as thermodynamic, have been developed to
predict the effect of SO-j vapor concentrations and moisture content of a gas
on the dew point and resulting acid condensate concentration (3 8,3 9,40,41).
Banchero and Verhoff (42) have analyzed several sets of data and produced an
equation which relates acid dew point, SO3 and H2O content of the gas.
The use of the data of Abel which gives the partial pressure of H2SO4
over known acid solutions as a function of temperature in conjunction with
that of Greenewalt (41) provide a convenient methodology for our problem.
Greenewalt provides equations for the partial pressure of water vapor over
sulfuric acid solutions as a function of temperatures. Figure 5 is our
representation of the combined data over the acid concentration range of 65-
85%, for water vapor content of the gas from 5 to 20%.
In our use of these data we use a cubic spline fit to retrieve an
equilibrium SO3 partial pressure value, and the corresponding acid condensate
composition for an input of dew point temperature and gas phase moisture
content. When equilibrium values were needed outside the range of the Abel-
Greenewalt data, we fitted spline-derived values via regression analysis to
an equation of the form
P* = EXP (a + &/T 4 ^m(T) + 6T)
where P* is the equilibrium partial pressure of H2SO4 at temperature T K,
The coefficients are functions of gas phase moisture content.
28-8

-------
ADSORPTION ON FLY ASH
The beneficial effect of high coal sulfur content on ash resistivity of
many U.S. coal has been known for many years (43). The conditioning benefits
of H2SO^ were recognized as early as 1912 (44). The effect of adsorbed S03
on fly ash surfaces of various character was first quantified by Bickelhaupt.
The use of supplemental S03 or HjSO^ for high resistivity fly ash
situations has been practiced for forty years in the U.S. Vaporized sulfuric
acid and SO^ produced by high temperature catalytic oxidation over ^2^5 are
the standard approaches used by conditioning system manufacturers. The
claims of some manufacturers notwithstanding, it is recognized that minimal
benefits have been achieved with some ashes under some conditions. With some
Australian ashes, benefit of sulfuric acid conditioning has been both
variable and too limited to warrant continued plant use. Alternate agents
are preferred. (45)
The work of Bickelhaupt and the author's exposure to several performance
situations prompted his interest in the areas to be discussed in the analysis
section. We mainly concern ourselves with the phenomena which occur in a
regenerative air heater with gases containing sulfuric acid vapor.
THE ANALYSIS
We analyze the operation of a series of rotary air heaters with respect
to average outlet gas temperature and also effectiveness. We then estimate
the condensation rates of sulfuric acid in an ash-free environment. The
behavior of sulfuric acid vapor in an ash-containing gaseous environment as
it passes through the air heater is also presented. The dependence of the
extent of acid adsorption is shown to depend on a number of variables. The
behavior of a dirty air heater is also analyzed with respect to competing
adsorption of acid by the stagnant ash layer on the rotor heat exchange
surface.
THE ROTARY REGENERATIVE AIR HEATER
Air heaters serve to transfer sensible heat from combustion gas to
combustion air. The rotary regenerative (Ljungstrom) type accomplishes this
by periodic contact of the rotor surface with the hot gas, and then releasing
the heat to the combustion air. Figure 6 shows the heat exchange surface
arranged in baskets.
Typical inlet gas temperatures are in the range of 600-800F; average
gas outlet temperature for utility installations is about 250 to 375F. The
outlet temperature levels are generally based on such considerations as the
temperature required to dry the fuel, the cost of coal energy, the cost of
heat recovery equipment and the acid dew point range.
28-9

-------
h2o vapor conc., molh
Figure 5. Equilibrium data for the	Figure 6, Rotary Regenerative
system I^SO^ - HjO - Air.	Air Heater
(Total pressure = 1 atm.)
AIR HEATER TEMPERATURES
We require gas and metal temperatures at points within the air heater on
the gas side since the extent of the pertinent phenomena (condensation,
adsorption, etc.) is influenced by these levels.
There have been a number of approaches (46, 47, 48, 49, 50, 51) to this
problem. These include numerical graphical methods, algebraic solution (high
rotative speed case), characteristic solutions and finite-difference methods.
The latter is convenient and especially useful for the subsequent situations
to be considered.
For the case of the periodic-flow heat exchanger, without leakage or
longitudinal conduction, four dimensionless parameters serve to define the
solution (Figure 7),
(MASS FLOW RATE x THERMAL CAPACITY)air
UMNCMX =
CRCMN =
SYMM
NTUO =
(MASS FLOW RATE x THERMAL CAPACITY)gaS
ROTOR MASS x RPH x ROTOR THERMAL CAPACITY
(MASS FLOW RATE x THERMAL CAPACITY)air
Figure 7. Four
(HEAT TRANS. COEFF x HEAT TRANS. AREA)air	Basic Air Heater
(HEAT TRANS. COEFF x HEAT TRANS. AREA)gaS	Parameters
CMN E [1/(HEAT TRANS. COEFF x HEAT TRANS. AREA)i]
I = AIR, GAS
Figure 8 shows the elementary cells into which both the gas and air
streams are subdivided. Heat transfer occurs in cross-flow (Figure 9) as the
cooled surface moves (from left to right) contacting the combustion gas
(flowing from top to bottom). The heat exchange equations which define the
outlet gas and surface temperature for a cell based on the inlet temperatures
are linear for a reasonable number of subdivisions in the gas and rotor
surface streams.
28-10

-------
OAS FLOW
ROTOR FLOW >
ROTOR RECYCLE
CONVERGENCES)
AIR FLOW
ROTOR FLOW
TRX(I-1,J)
GAS FLOW
TGX(I,J-1)
I,JTH
CELL
TRX(I.J)
y
TQX(I.J)
TQX{I,J)-TQX(Iil-1) -K1[TaX
-------
The gas temperature spread at the air heater outlet ranged from 59F
down to about 16F at the lowest gas outlet temperature of 237F. Even
higher spreads are observed in some installations.
ACID CONDENSATION RATES
A clean hot gas containing sulfuric acid and water vapor when cooled by
passage through an air heater will lose I^SO^ from the vapor if the local air
heater surface is below the acid dew point temperature.
The rate of condensation is often expressed as being proportional to the
difference of the partial pressure of sulfuric acid in the gas, and the
equilibrium partial pressure. Thus, the rate, for dilute gases at 1
atmosphere, may be expressed as kg moles of acid/(h-m ) and is given by
R = ^ (P.-P*)	Mechanism I
\j	8
2/3
where a is the a constant (P[\/p) (Da/D(J)
PA is density of sulfuric acid vapor
P is density of gas
is the diffusion coefficient for sulfuric acid vapor
D(J is the thermal diffusivity of the gas
H is the convective heat transfer coefficient
C is the thermal capacity of the gas
(Note	is the Lewis number.)
This condensation rate is applicable provided the gas temperature in any
region does not fall below the dew point. If the latter occurs and fog
formation results, then the rate of condensation falls (37). In that case,
it is given by
r = (yDu)1/3 WQ
dT
(Tg-T$) Mechanism II
where Tg is the temperature of the gas, and
Tg is the surface temperature.
Calculations were made using the condensation program for the six air
heater designs. Figure 11 shows the results expressed as the fraction of
acid condensed versus gas outlet temperature. The dotted line represents the
prediction based on Mechanism I (no fog formation); the curve for Mechanism
II shows the extent of condensation when fogging occurs. In practice,
intermediate rates may occur.
The local rate of condensation in the air heater is shown versus rotor
surface temperature in Figure 12 (average gas outlet temperature is 269F).
Mechanism I over-predicts the condensation rate; Mechanism II predicts a
condensation rate equal to that of Mechanism II at the inlet to the
condensation zone; subsequently, the rate drops markedly as fog formation
occurs. (Peak rate corresponds to 22 yg SOj/h/cm*). The program also
predicts acid condensate composition.
28-12

-------
MECHANISM II
220 240 260 200 300 320 340
AVERAGE OAS OUTLET TEMPERATURE. 
-------
imply a significant extent of adsorption at low degrees of saturation.
F igure 13 .
Examples of BET
Isotherms
0 -2 .4 .6 ,8 1.0
XA
Ashes with strongly basic surfaces are characterized by large values;
C^, according to the BET theory, is given by
CA = EXP (AH/RT)
where AH is the enthalpy difference between the heat of adsorption between
the solute and the ash surface and the heat of condensation of
the solute,
R is the universal gas constant and
T is the absolute temperature of sorption.
Thus, for 02000, a monolayer forms on the ash with relative saturation
of only 0.02. This corresponds to an f^SO^ concentration of only 0.2 ppm, at
275F with 8% (v) H/>0 in the gas With C-2, the monolayer requirement would
correspond to an acid concentration of nearly 5 ppm.
Characterization of surface acidity can be accomplished by a variety of
approaches, including titration with bases (n-butylaxnine, pyridine.) with
suitable indicators, use of a series of Hammett indicators, and adsorption
isotherms. Surface acidity depends largely upon detailed surface composition
which may differ substantially from bulk composition. Thermal exposure can
be an important influencing parameter.
Bickelhaupt (8) has proposed the use of the following criteria to
distinguish "normal" ashes from those which are less responsive to SOj
conditioning namely, that they contain less than 1 atomic % iron and less
than 5 atomic % (Ca + Mg).
In the U.S. low sulfur East Kentucky (and some W. Virginia) coals
produce such ashes, A typical bulk ash composition of the former might be
55% Si02, 33% A1203 , <4% FejO, , l%(CaO + MgO)and < 0.25% NajO. Such an ash
will be quite refractory witn fusion temperatures in excess or 1600 C. When
derived from strip-mined coal, contamination with aluminum-containing clays
can aggravate SO3 responsiveness considerably.
28-14

-------
A large number of Australian coals yield quite acidic ashes which is
probably one of the reasons that some Australian investigators decry the role
of sulfur trioxide in influencing ESP performance. One notorious example is
derived from Tarong coal (Nanango district of Queensland); its ash
composition is 61% SiC^, 3 4% AI2O3 , 3.4% FeoOj , 0.2% (CaO +MgO) and 0.1%
NanO. Further, the coal sulfur's content is aDout 0.3% with 28% ash. Many
Indian coals also yield acidic ashes.
The adsorption model, in the clean rotor situation, postulates
adsorption equilibrium at the outlet of every air heater cell. Obviously,
the extent of acid adsorption in the air heater based on this model does not
require investigation of interior points; the conditions at the outlet will
define the extent of H^SO^ removal, as well as the number of monolayers of
acid adsorbed on the ash particles. Our module was developed to give
information concerning adsorption in the interior region of the air heater,
as the module served as a stepping stone to more complex modules.
The application of the finite-difference approach to the adsorption
situations (on fly ash, or on both fly ash and ash deposits) is considerably
more complicated mathematically than the simple temperature situation of
figure 9. In the adsorption case we require the highly accurate solution,
for each cell, of 3 or 5 simultaneous equations, some of which are non-
linear .
Using the six cases previous noted, we calculated the average number of
layers of sulfuric acid adsorbed on the fly ash. Figure 14 shows the extent
of SO3 removal (inlet gas-8 ppm H2SO4) with acidic and basic ashes. Below
about 250F, the acidic and basic ashes behave similarly with respect to the
average uptake of acid. At higher gas outlet temperatures we see a large
difference in behavior of the two types of ashes; there is strongly reduced
uptake of acid, especially for the acidic ash. Figure 14 refers to a coal
with a 10% ash level.
Ash loading and ash size distribution also exert important effects with
respect to acid uptake.
Calculations reveal that with outlet temperature spreads as small as
40F, resistivities of acidic ashes can range from 2.5 (10) ohm cm in the
cold part of the air heater duct to more than 4 (10) ohm cm in the hot part
due to the difference in adsorption of acid.
Induced Conditioning
With respect to conditioning, the predicted effect of added SOg is shown
for a gas outlet temperature of 289F with acidic and basic ashes. From
Figure 15 one might deduce that there is only about 5 ppm SO3 additional
requirement for achieving useful acid uptake levels with the acidic ash
compared to the basic ash. This is not generally so for several reasons.
The basic ash may have a lower resistivity level when compared at equal
average levels of acid uptake. This could follow from the higher
conductivity of the basic ash constituents without conditioning as well as
from the better uniformity of conditioning achievable with the basic ash.
28-15

-------
Also, as expected, some segments of the acidic ash are predicted to undergo
little uptake of acid. If the air beater flow stratification persists into
the precipitator, some precipitators cells may operate under poor electrical
conditions and yield low collection efficiencies.
tz
QA8 TEMPERATURE 289'F
HaSO* FEED CONC., PPM
s
Figure 14. and Values For	Figure 15. Predicted
Three Fly Ash Samples	Conditioning Effect For Two
Fly Ash Samples
The use of the adsorption model for such situations can provide insight
into the cause of malperformance as well as guidance for obtaining optimal
conditioning levels with a minimal expenditure of acid. The model also
enables one to anticipate leakage of unadsorbed acid into the atmosphere
where operation is not conducted at proper conditions.
COMPETITIVE ADSORPTION - FLY ASH VS. ASH DEPOSITS
This specific analysis was prompted by a situation in which H2SO4
concentrations in the inlet gas to various chambers of an ESP system were
found to vary nearly fivefold (0.6 to 2.9 ppm; at a sister unit the value
was 4 ppm). Air heater inlet concentrations averaged about 6 ppm.
Measurement of water-soluble sulfate on the ashes did not indicate that the
differential quantity of HjSO^ lost from the gas was adsorbed on the
corresponding fly ash. Further, the 0.6 ppm ^SO^ zone produced the highest
resistivity ash (~4(10) ohm cm) and exhibited the poorest electrical
operating parameters.
It was postulated that the losses of acid were occurring by adsorption
on stagnant ash deposits on the air heater surface or by condensation on the
cold metal surface. The lower temperature of the deposits compared to that
of the fly ash would favor adsorption on the former.
In our analysis we considered the competitive adsorption of ^SO^ on the
two categories of ash in each air heater cell with some subsequent stripping
of ^SO^ from the ash on the air side. It was necessary to involve the
boiler system since now some SO3 was being recycled to the boiler in the
combustion air. Figure 16 shows the boiler system with its SO3 sources and
sinks. In the system we considered the SO^-atomic oxygen reaction, the
28-16

-------
prompt partial reversion of excess SOg to SO2 and the low temperature
heterogeneous catalysis of SO2 to SO3.
Figure 17 shows that in the acidic ash situation the stagnant ash
deposits compete (in this equilibrium analysis) much more successfully for
adsorption than does the fly ash. With basic ashes, the competition is
less one-sided. The low sulfur content of many coals with acidic ashes
further aggravates the competitive adsorption situation.
		 100	
Q
X
80
CO
o
i
>
o
z
UJ
a
COMBUSTION
O*
<0,
X
 \ I AW HEATCft
UTS	X 1
<
O
A|R	RELATIVE ROTOR SURFACE DUST LOADING
Figure 16. Boiler System S0g Sources	Figure 17. Loss of H^SO^ Due To
and Sinks	Competitive Adsorption on Ash
Deposits
There are several more complex aspects which were not introduced into
this limited analysis. First, fly ash surfaces are notoriously heterogeneous
and a simple BET isotherm with CA=constant=f(T) only should be replaced with
one which better describes this heterogeniety by expressing a dependence of
on X^, the degree of saturation. While we neglected kinetics aspects of
the adsorption process, it is imperative to note that adsorption kinetics
with basic ashes are likely to be appreciable faster than with acidic ashes.
The latter appears to be a significant factor involved in induced
conditioning of various fly ashes.
CONCLUSIONS
Complex phenomena occur in air heaters treating gas streams containing
H2SO4 vapor. Cold end corrosion results as acid condenses on cold metal
surfaces; when the gas temperature falls below the acid dew point, fog may
form.
With respect to ESP performance, adsorption of H2SO4 on the ash surface
can serve to supply a sufficiency of conducting ions to facilitate current
flow across a deposited ash layer. A simple equilibrium model (BET) provides
guidance as to the adsorptive behavior of ash surfaces of various acidities
with H2SO4 vapor in the range of air heater temperatures. Under some
conditions fly ash exiting the air heater can exhibit a wide distribution of
acid loadings, largely dependent upon local temperature levels, and the
availability of H2SO4. In these cases insufficient gas mixing downstream of
the air heater can lead to marginal performance in some zones of the ESP when
28-17

-------
other conductive ash species are lacking.
Competitive adsorption of acid between fly ash and air heater ash
deposits tends to favor the ash deposits since they are at a lower
temperature. In situations in which acidic ashes are involved, the adsorbed
H2SO4 on the stagnant ash layer is more readily stripped on the air side
(than in the case of more basic ashes). It was shown that the competition for
H2S04 adsorption between fly ash and rotor ash deposits is less favorable
from the standpoint of the fly ash in acidic ash cases. Thus normally less
attractive situations involving acidic ashes can be appreciably penalized
when air heater ash deposits occur. Modeling of this situation involved, in
addition to the air heater interactions, boiler system generation of SOg by
atomic oxygen attack of SO2, as well as catalytic oxidation of SC^.
These unique HIM computer models have proved themselves useful in a
number of situations. Among them are the ability to
anticipate poor local (natural) conditioning of fly ash and suggest
more suitable operating conditions
provide guidance as to an optional gas mixing approach to eliminate
high resistivity ash zones
define details of a cost-effective SO3 conditioning scheme which
provides proper local SO3 injection concentrations.
Our basic approach is also of interest for non-SO^ containing agents
which condense from the vapor state to a particulate form. Air	heater
fouling problems which may occur in NH^-based N0X control processes	can be
simulated by our models.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the views of the Agency and no official
endorsement should be inferred.
28-18

-------
REFERENCES
1.	Johnstone, H.F. An electrical method for the determination of the dew
point of flue gases. Univ. Illinois Eng. Exp. Sta. Circ. No. 20, 1929.
2.	Borio, R.W., Hensel, R.P., Wilson, E.B., and J.W. Leonard. The control
of high-temperature fire-side corrosion in utility coal-fired boilers.
Office of Coal Research, Dept. of the Interior, Research and Development
Report No. 41, Apr. 1969, Washington.
3.	Rylands, J.R. and J.R. Jenkinson. The acid dew point. J. Inst. Fuel 27:
299-318, 1954.
4.	Harlow, W.F. Causes of flue gas deposits and corrosion in modern boiler
plants. Proc. Inst. Mech. Engr. (London) 160: 359-368, 1949.
5.	Mittelbach, G. Kondensation von SchwelfelsKure aus Rauchgasen und der
MePvorgang auf electrischen Taupunktfuhlem. VGB Kraftuerkstechnik 53:
675-682, 1973.
6.	Bickelhaupt, R.E. Measurement of Fly Ash Resistivity Using Simulated
Flue Gas Environments, EPA-600/7-78-635 U.S. Environmental Protection
Agency, Research Triangle Park, NC, 1978, 23 pp.
7.	Bickelhaupt, R.E. The effect of sulfuric acid vapor on fly ash
resistivity, presented at Fourth Symposium on the Transfer and
Utilization of Particulate Control Technology, Oct. 11-15, 1982,
Houston, TX.
8.	Bickelhaupt, R.E. A Study to Improve a Technique for Predicting Fly Ash
Resistivity with Emphasis on the Effect of Sulfur Trioxide (Draft Report
to U.S. Environmental Protection Agency, Mar. 1983).
9.	Potter, E.C. and C.A.J. Paulson. Does sulphur dominate fly ash
collection in electrostatic precipitators? In. EPA-600/9-80-639a. Second
Symposium on the Transfer and Utilization of Particulate Control
Technology, Vol.I, 184, 1980.
10.	Pilat, M.J. and J. M. Wilder. Opacity of monodisperse sulfuric acid
aerosols, Atmospheric Environment 17: 1825-1835, 1983.
11.	Southern Research Institute measurements supplemented by data from the
author.
12.	Attar, A. and W.H. Corcoran. Sulfur compounds in coal. Ind. Eng. Chem.
Prod. Res. Dev. 16: 168-170, 1977.
13.	O'Gorman, J.V. and P.L. Walker, Jr. Mineral Matter and Trace Elements
in U.S. Coals. Research and Development Report No. 61 Prepared for
Office of Coal Research, U.S. Department of the Interior, 1972.
14.	Montano, P.A., Granoff, B. and T.D. Padrick. Role of impurities In.:
B.R. Cooper and W.A. Ellingson (Eds.), The Science and Technology of
Coal and Coal Utilization. Plenum Press, NY, 1984.
15.	Mitchell, R.S. and H.J. Gluskoter. Mineralogy of ash of some American
coals: Variations with temperature and source. Fuel 55: 90-96, 1976.
28-19

-------
16.	Bodenstein, M. and W. Pohl. Gleichgewichtsmessungen an der
Kontaktschwefelsaure Z. Electrochem. 11: 373, 1905.
17.	Dooley, A. and G. Whittingham. The oxidation of sulphur dioxide in gas
flames. Trans. Faraday Soc. 42: 354-362, 1946.
18.	Whittingham, G. The oxidation of sulphur dioxide in slow cmbustion
processes. Trans. Faraday Soc. 44: 141-151, 1948.
19.	Hedley, A.B. A kinetic study of sulphur trioxide formation in a pilot
scale furnace In H.R. Johnson and D.J. Littler (Eds) The Mechanism of
Corrosion by Fuel Impurities. Butterworths, London 1963, p 204-215.
20.	Moisieeva, T.A. and V.E. Podolyak. The effect of the rate of cooling of
combustion products in the furnace on the formation of sulphur trioxide.
Teploenergetika 20: 14-17, 1973.
21.	Ahmed, S.H. and B.J. Zaczek. A kinetic study of factors influencing SOo
formation in a premixed laminar flame. J. Inst. Fuel. 50: 107-110, 197/.
22.	Laxton, J.W. Influences of excess oxygen and metallic additives, In H.R.
Johnson and D. J. Littler (Eds) The Mechanism of Corrosion by Fuel
Impurities, Butterworths 1963, p 23 2-237.
23.	Jakisch, H. Clinkering in the combustion chamber below the softening
temperature of the ash. Arch Warmewirt 23 : 211-214, 1942.
24.	Kinjiro, F. Decomposition of calcium sulphate at high temperature. Sekko
to sekkai No. 7 , 2-5, 1952.
25.	Wickert, K. Fouling of heating surfaces in steam generator boilers. Part
I Composition, physical behaviour and sintering of combustion chamber
dust. Brennstoffe, Warme, Kraft 10: 1, 1958. Part II Chemical reactions
in the combustion chamber and measures to counteract the resulting
damage, Brennstoffe, Warme, Kraft 10: 101, 1958.
26.	Baker, D.C. and A. Attar. Sulfur pollution from coal combustion. Effect
of the mineral components of coal on the thermal stabilities of sulfated
ash and calcium sulfate. Environ. Sci. Technol. 15: 288-293, 1981.
27.	Ostroff, A.G. and R.T. Sanderson. Thermal stability of some metal
sulfates, J. Inorg. & Nuclear Chem. 9: 45, 1958.
28.	Pechkovskii, V.V. Formation of metal sulfides and oxides in the thermal
decomposition of manganese and iron sulfates. Zhur. Priklad Khim 32:
2613-2618, 1959.
29.	Homolya, J.B., Barnes, H.M. and C.R. Fortune. A characterization of the
gaseous sulfur emissions from coal and oil-fired boilers, In Proc. 4th
National Conference on Energy and Environment, Cincinnati, OH, 1976, p
490-494.
30.	Barrett, R.E. High temperature corrosion studies in an oil-fired
laboratory combustor, Trans. ASME, J,Eng. Power, Ser. A 89: 288-296,
1967.
28-20

-------
31.	Beidick, H. and R. Reifenhauser. Catalytic SO* Formation as Function of
Boiler Fouling Combusion 52: 17-21, 1980.
32.	Zetlmelsl, M.J., McCarthy, K.J. and D.F, Laurence. Investigation of some
factors affecting nonflame generation and suppression of SO*. Ind. Eng.
Chem. Prod. Res. Dev. 22: 710-716, 1983.
33.	Levy, A. and E.L. Merryman, Interactions in sulphur oxide-iron oxide
systems, Trans. ASME J.Eng. Power, Ser. A, 89: 297-303 , 1967.
34.	Anderson, C.H., and C.W. Goddard, Equilibrium SOj pressures of inner
layers of fireside deposits from coal boilers. J. Inst. Fuel. 41: 3 57 
363 , 1968.
35.	Lee, G.K., Friedrich, F.D. and E.R. Mitchell. Paper 4. Effect of fuel
characteristics and excess combustion air on sulphuric acid formation in
a pulverized-coal-fired boiler. Dept. of Energy, Mines and Resources,
Mines Branch (Canada) 1967. 9 pp.
36.	Fontana, M.G. and N.D. Greene. Corrosion Engineering (2nd ed.). McGraw-
Hill Book Co., NY, 1978, p 225.
37.	Land, T. The theory of acid deposition and its application to the dew-
point meter. J. Inst. 50: 68-75, 1977.
38.	Miiller, P. Study of the influence of sulfuric acid on the dew point
temperature of the flue gas. Chemie-Ing.-Tech. 31: 345-351, 1959.
39.	Abel, E. The vapor phase above the system sulfuric acid-water. J. Phys.
Chem. 50: 260-283 , 1946.
40.	Gmitro, J.I. and Vermuelen, T. Vapor-liquid equilibria for aqueous
sulfuric acid. AIChE Journal 10: 740-746, 1964.
41.	Greenewalt, C.H. Partial pressure of water out of aqueous solutions of
sulfuric acid. Ind. and Eng. Chem. 17: 522-523, 1925.
42.	Banchero, J.T. and F.H. Verhoff. Evaluation and interpretation of the
vapour pressure data for sulfuric acid aqueous solutions with
application to flue gas dewpoints. J. Inst. Fuel 50: 76-80, 1975.
43.	White, H.J. Electrical resistivity of fly ash. Air Repair 3: 79-87,
1953.
44.	White, H.J. Fifty years of electrostatic precipitation. J. Air Poll.
Control Assoc. 7: 167-177, 1957.
45.	Tassicker, O.J., Some aspects of electrostatic precipitation research in
Australia. J. Air Pollution Control Assoc. 25: 122-127, 1975.
46.	Hahnemann, H.W. Approximate calculation of thermal ratios in heat
exchangers including heat conduction in direction of flow, National Gas
Turbine Establishment Memorandum No. M36, 1948.
47.	B. H. Schultz. Regenerators with longitudinal heat conduction, IME-ASME
General Discussion on Heat Transfer, London, England, 1951, p 440-443.
48.	Coppage, J.E. and A.L. London. The periodic flow regenerator, Trans
28-21

-------
ASME 75: 779-787, 1953.
49.	Lambertson, T.L. Performance factors of a periodic flow heat exchanger,
TranB ASME 80: 586-592, 1958.
50.	Bahnke, G.D. Effect of longitudinal heat conduction on rotary
regenerators, MS thesis, U.S. Naval Postgraduate School, Monterey, CA.
1962.
51.	Mondt, J.R. Heat exchanger solid and fluid temperature distributions, J.
Engr. for Power, Trans ASME, Series A, 86: 121-126, 1964.
52.	Brunauer, S. and P.H. Emmett. The use of low temperature van de Waals
adsorption isotherms in determining the surface areas of various
adsorbents, J. Amer. Chem. Soc. 59: 2682-2689, 1937.
53 . Brunauer, S., Emmett, P.H. and E. Teller. Adsorption of gases in multi-
molecular layers, J. Amer. Chem. Soc. 60: 309-319, 1938.
54.	Brunauer, S., Deming, L.S., Deming, W.E, and E, Teller. On a theory of
the van der Waals adsorption of gases, J. Amer. Chem. Soc, 62: 1723-
1732,1940.
55.	Brunauer, S., Skalny, J and E.E. Bodor. J. Colloid Interface Sci. 30:
546-552, 1969.
56.	Ditl, P. and R.W. Coughlin. Sorption and diffusion interactions with
fly ash of SO, in air, SO, in air, S09 + H,0 in air and S0,+H,0 in air,
Environ. Sci Techno 1. 11: 701-706, 1977.	J 1
28-22

-------
Session 11: ESP: FUNDAMENTALS II
Philip A. Lawless, Chairman
Research Triangle Institute
Research Triangle Park, NC

-------
EXPERIMENTAL STUDIES OF SPACE CHARGE EFFECTS IN AN ESP
D.H. Pontius and P.V. Bush
Southern Research Institute
Birmingham, Alabama 35255-5305
ABSTRACT
As particles become charged in an electrostatic precipitator they inter-
fere with the normal corona discharge by establishing relatively immobile,
local electric field effects. These effects, often referred to as space
charge, were demonstrated in a wire-pipe ESP by injecting resuspended fly ash
into the system. A simple mathematical model was derived for this ESP con-
figuration to verify the behavior of the system in the presence of substan-
tial space charge effects.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
29-1

-------
I. INTRODUCTION
The fundamental concept of electrostatic precipitation is elegantly sim-
ple. Particles suspended in a gas may be captured by charging them with ions
in an electrical corona discharge and forcing them to impact on a solid sur-
face by the effect of an electric field. Given this basic principle, along
with the knowledge of what is required to produce a corona discharge, the
development of specific configurations of electrodes to perform the operation
is straightforward. This information is not, however, enough to define how
large the device should be for a specific application.
Among the early attempts to develop a quantitative description of elec-
trostatic precipitation was that of W. Deutsch (1). He derived an expression
for the rate of precipitation, based upon certain broad assumptions and
approximations. Since that expression is independent of the concentration of
suspended dust in the gas to be cleaned, it integrates very neatly into an
exponential model for a precipitator of arbitrary size. The Deutsch model
has been modified and refined in a number of different ways to suit specific
needs, but the fundamental ideas remain useful and effective (2,3).
A model is a representation of an object, a system, or a concept. The
utility of a model depends upon the nature of the correspondence, or isomor-
phism, between the subject and the model; the fidelity is determined by both
the accuracy and the consistency of the details. Within the general defini-
tion of a model, the operative correspondence can take on any of a great
variety of forms, including those of a simple physical transformation of
scale, a qualitative description, or a set of abstract mathematical state-
ments. The validity of a model depends entirely upon its intended use.
The present discussion concerns an experimental investigation and the devel-
opment of a mathematical model designed to evaluate the effect of particulate
space charge on the behavior of an electrostatic precipitator.
II. PRELIMINARY EXPERIMENTS
The effects of the ionic space charge can be demonstrated by a pair of
simple experiments as described in the following part of this discussion and
illustrated in Figure 1. First a pair of identical, isolated, parallel wires
are energized at equal values of voltage but opposite polarity. If the volt-
age is high enough, a bipolar corona discharge will occur between the two
wires, and if the experiment is done in air at room temperature, the ionic
currents due to the positive and negative ions will be nearly the same. The
symmetry of the apparatus results in almost identical distributions of posi-
tive and negative charge, so that on the macroscopic scale the net space
charge in the system is essentially zero.
The plane of symmetry midway between the two wires is an equipotential
surface at ground potential. If a grounded metal plate is now inserted on
the location of that surface, the electrostatic description remains practi-
cally unchanged, but the distribution of space charge is radically altered.
The positive and negative ions remain separated until they effectively recom-
29-2

-------

NEGATIVE
POTENTIAL
-V
POSITIVE
IMAGE POTENTIAL
+V
REAL SPACE
IMAGE SPACE
A. WIRE-PLANE APPARATUS
POSITIVE
POTENTIAL
+V
NEGATIVE
POTENTIAL
-V
i
'"^EQUIPOTENTIAL PLANE
V = 0
B. WIRE-WIRE APPARATUS
Figure 1. Bipolar corona, space charge neutralized (a). Twin unipolar corona (b).
29-3

-------
bine on the grounded plane, Thus, in the space on either side of the plate
there is a monopolar distribution of ionsa space chargesimilar to that
found in an electrostatic precipitator.
If the ionic space charge were, in general, ineffective at suppressing
corona discharge, the following results should be observed. At any given
value of the applied potential, the sum of the currents measured for each of
the two wires with the grounded plate present (i.e., the total ionic current
intercepted by the plate) would be the same as the current measured at either
wire in the other configuration. But from the experimental results shown in
Figure 2 it is clear that the neutralization of the space charge in the mixed
bipolar experiment permits a much more vigorous corona discharge than occurs
in the other experiment.
'I '
WIRE-WIRE
w 140
 100

WIRE-PLATE
38 42 46 50 54 58
APPLIED VOLTAGE, kV
Figure 2. Comparison of conduction characteristics of bipolar and mirror-twin unipolar
corona discharges.
The point of the previous discussion is to demonstrate the extent to
which a corona discharge is dominated by the distribution of space charge in
the system. In the experiment described, only an ionic space charge was pre-
sent, but additional suppression of the corona discharge results from a par-
ticulate space charge. It is that phenomenon which will be addressed in the
present discussion.
III. SPACE CHARGE EXPERIMENTS IN THE LABORATORY
Elementary experiments involving space charge associated with particu-
late loading in a precipitator were carried out in a cylindrical device with
a segmented collecting electrode. This system is equipped with a 32-channel
29-4

-------
electrometer so that accurate measurements of current density as a function
of position can be made rapidly. The data are accumulated in computer stor-
age, and statistical analyses are carried out. Typically, 400 measurements
are acquired in a period of about one minute. The internal diameter of the
cylinder is 24.8 cm, and the active length is 40.6 cm. It is mounted in a
wind tunnel which can provide a gas flow of controllable velocity, tempera-
ture and dust loading.
Experiments were carried out using both positive and negative corona
polarity. Fly ash was injected into the gas by sandblaster. In each experi-
ment, the principal measurements included determination of V-I characteris-
tics, particle size distribution, and electrical charge on the particles. It
is notoriously difficult to resuspend the finer fraction of fly ash particles
by mechanical means; however, the results were adequate to provide a useful
demonstration of the effects of space charge.
Figure 3 shows voltage-current characteristics for the device with posi-
tive energization. Clearly, there is a nearly linear shift of the curve to
the right when ash is injected into the system. This is just the behavior
expected when particulate space charge exists in a precipitator. The other
data tend to support this conclusion. In Figure 4 a reasonably uniform dis-
tribution of current is indicated in the presence of clean air, and Figure 5,
corresponding to approximately the same applied voltage but in the presence
of dust in the gas stream, shows a similar distribution, slightly reduced in
magnitude. It is significant, however, that the current density is reduced
DIRTY, NO ASH
INJECTION ,
/ ASH INJECTION
40
VOLTAGE, kV
Figure 3. Conduction characteristics of the cylindrical system with positive energization.
29-5

-------
4.0
V =46.0kV
I = 85.0 mA
3.0
<
a.
H
Z
UJ
cc 2.0
oc
3
O
1.0
0 0 . n n i til n
4 6 8 10 12
DISTANCE FROM INLET, in.
14
16
4.0
Figure 4. Current distribution for positive corona, dean air.
V =46.3 kV
I =75.0 mA
3.0
<
a.
| 2.0
e
D
O
1.0
0.0UI
4 6 8 10 12
DISTANCE FROM INLET, in.
fl
16
Figure 5. Current distribution for positive corona, dirty air.
29-6

-------
even at the closest point to the inlet. This phenomenon implies that essen-
tially complete charging had occurred in an interval of time too small to be
resolved by the apparatus, or that the particles were already charged to a
relatively high level before they were introduced into the precipitator.
Triboelectric charging in the ash dispersion system can account for such an
effect. The latter explanation was confirmed by a series of measurements
with a Millikan apparatus. Figure 6 shows the distribution of charge as a
function of particle diameter at the inlet of the device, and Figure 7 gives
the same information at the outlet. They are practically identical. Since
no significant change in the status of particulate charge occurred in the
precipitator, a uniform effect of space charge is to be expected.
10
10
CE
X
u
-J
CE

ftlO
Ld
10
10
10u
DIAMETER.MICRONS
Figure 6. Particle charge as a function of diameter, measured at the inlet of the ESP.
Energizing the system with negative polarity led to somewhat more inter-
esting results in terms of the evolution of the particulate space charge with
time. Since the entrant space charge in this instance is opposite in polar-
29-7

-------
UJ
3
0
o
2
0
LU
I	1
QC
LU
0
UJ
1 0
1 0
DIAMETER,MICRONS
Figure 7. Particle charge as a function of diameter, measured at the outlet of the ESP;
positive corona.
ity to the ionic space charge, the current density is increased at the inlet
end of the precipitator, as can be observed by comparing Figures 8 and 9.
The charge is quickly neutralized and negative charge begins to accumulate on
the particles, resulting in progressively greater suppression of corona cur-
rent toward the outlet end of the device. The measurements of charge distri-
bution by particle charge, shown in Figure 10, indicated that the magnitude
of the charge at the outlet was approximately the same as was achieved in the
experiment done with positive polarity. In the negative experiment, however,
the charge on the particles underwent a complete reversal of polarity, which
was graphically traced by the variation of current density with distance down
the length of the precipitator.
29-8

-------
8.0
V = -48.0 kV
I =136 mA
6.0
<
a.
2 4.0
cc
IT
D
O
2.0
0.0
ill
6	8 10 12
DISTANCE FROM INLET, in.
14
16
Figure 8. Current distribution for negative corona, dean air.
8.0
V = -42.3 kV
I = 120 mA
6.0
<
a.
5 4.0
0C
CC
D
O
2.0
0.0
uJ
4 6 8 10
DISTANCE FROM INLET, in.
12
14
Hi
16
Figure 9. Current distribution for negative corona, dirty air.
29-9

-------
4	t --i _o_.
-L	I-	-i-	L. .. J , I fc
1 0
DIAMETER,MICRONS
Figure 10. Particle charge as a function of diameter, measured at the outlet of the ESP;
negative corona.
IV. THEORETICAL MODEL
The model to be discussed in the following paragraphs shares many of its
basic assumptions with the Deutsch model. The most significant area of
departure is that in the present model it cannot be assumed that the rate of
precipitation is independent of the concentration of dust. This is a delib-
erate complication, brought about by the inclusion of the charge on the par-
ticles in the overall charge distribution. As a result, the rate of precipi-
tation becomes dependent upon the concentration of particles and the charging
conditions to which they have been subjected. The simple exponential behav-
29-10

-------
ior no longer applies, and a single algebraic expression for the overall
collection efficiency of a precipitator cannot be derived. Assumptions and
approximations retained from the Deutsch model include the following: (1)
Turbulent mixing of the aerosol maintains the distribution of suspended par-
ticles uniform over any section of the precipitator normal to the direction
of gas flow. (2) Effects of electrostatic forces and turbulence produce no
net transfer of particles in a direction parallel to the general flow of the
gas. (3) The electric field will be assumed to be a function of position,
applied potential and space charge; no discontinuities, instabilities or
incongruities will be admitted.
If the constraints upon the model seem overly stringent and perhaps
unrealistic, it should be borne in mind that the specific parameter under
investigation is the distribution of the particulate space charge. The goal
is to provide a quantitative assessment of the extent to which the operation
of an electrostatic precipitator is modified by that space charge, given a
description of the particle concentration and distribution by size at the
inlet of the control device. Provisions will also be made for the possibil-
ity that the entrant particles already possess some specified level of
charge, due to triboelectric or other effects, before arriving at the inlet
of the precipitator.
The set of assumptions upon which this model is based are chosen to
minimize the number and length of computational procedures while dealing
quantitatively with the presence of particulate space charge in an electro-
static precipitator. This approach results in a smoothed-out model in which
the finer details are treated statistically. In the limit, as the particu-
late space charge is reduced toward zero, the result converges to that of the
Deutsch model.
In most instances electrical coronas are controlled, or limited, by the
distribution of charge in the space between the electrodes. The effect of
ionic space charge is treated in the fundamental theory of corona discharges.
In an electrostatic precipitator, however, there is an additional component
to the space charge resulting from the capture of ions by dust particles
suspended in the gas. The electrical mobility of the charged particles is
much smaller than that of the ions, so the resulting particulate space charge
tends to build up until a significant redistribution of the electric field
occurs. The field near the corona discharge electrode becomes reduced, caus-
ing a reduction in the ionic current injected into the system. But at the
passive electrode the strength of the electric field increases. The result
of the former effect is to diminish the charging rate, and the latter pro-
duces an increase in the migration velocity of the particles. The presence
of a strong particulate space charge may therefore be either beneficial or
detrimental, depending upon which of the two competing effects dominates.
The general concept of space charge as outlined above is well known, but
it is a difficult mechanism to build into a mathematical model. The rate of
particle charging depends upon the ion current density and the local strength
of the electric field, among other things. The current density and field
strength, in turn, are strongly affected by the charge distribution in the
system, including both the ions and the charge on the particles. (For com-
29-11

-------
pleteness, the contribution of free electrons to the charge distribution can
be included, but this effect will be considered negligible for the purposes
of this discussion.)
The working geometry for developing the fundamental solutions will be
that of coaxial cylinders. The solutions will be, as implied in our intro-
ductory remarks, essentially two-dimensional; the axial components of all
quantities will be considered negligible. The configuration is advantageous
from both theoretical and experimental grounds.
Because of the relationships between the mechanisms of particulate
charging and the effect of the resulting space charge on the charging rate,
an iterative method has been developed. The calculation begins with a value
of zero for the particulate space charge. Given parameters for the distribu-
tion of particles by size, the procedure computes the charge accumulated by
the particles during a small increment of time. The calculation is then
repeated with the inclusion of the small increase in space charge. Depletion
of particles from the system, and the concomitant removal of charge are com-
puted in terms of the values of electrical mobility calculated for the parti-
cles and the electric field strength at the collecting surface, as in the
model of Deutsch.
V. MATHEMATICAL DEVEL0R4ENT
If end effects are neglected in a cylindrical system containing only
equipotential surfaces that coincide with coaxial cylinders, then the elec-
tric field is entirely radial, and from basic electrostatic theory, the mag-
nitude of the electric field E(r) can be expressed in the form
A(r)
E
-------
assuming that we choose the surface r=b as the reference for zero potential.
Application of this method thus consists of specifying or calculating a dis-
bution of space charge, carrying out the integration in Equations (3) and
(4), and then evaluating A in terms of the boundary conditions V(a)=V, the
applied potential, and V(b;=0.
UNIFORM DISTRIBUTION OF SPACE CHARGE
The simplest nontrivial problem in the geometry in question is that in
which a uniform distribution of space charge exists. Cooperman (4) has
explained that a uniform distribution of ionic space charge is a very good
approximation if the current density is small. Because of the normally
turbulent behavior of the gas in a precipitator, an assumption of uniform
particulate space charge is also a good first-order approximation.
Therefore, taking p(r)=p, a constant, Equation (3) becomes
Xa p(r 2 - a2)
E(r) - 	 + 	 ,	(5)
2irer	2er
and the potential function calculated from Equation (4) is
^a	p b2 - r2
V(r) = 	 n(b/r) +  [	a2&n(b/r)l .	(6)
2 ire	2e 2
Now, setting V(a)=V and solving for Ag yields
x = 1_ |V	1 [b2 _ a 2 (1+2 n(b/a) ) ] } .	(7)
a n(b/a)	4e
The first term on the right-hand side of Equation (7) represents the
electrostatic contribution, and the second term indicates the reduction in
linear charge density on the corona wire required to maintain the imposed
potential on the wire in the presence of the space charge.
Inserting the value of Afl from Equation (7) into Equations (5) and (6)
provides the following expressions for E(r) and V(r):
V	o	b2 - a2
E(r) = 	 + 	 [r2	] ,	(8)
rn(b/a) 2er	2n(b/a)
p(b2 - d2) n(b/r) p
V(r) - [V - 			]	*  (b2 - r2) .	(9)
4e	i.n(b/a) 4e
CORRESPONDENCE WITH TOWNSEND'S THEORY
Up to this point in the discussion, no distinction has been made between
ionic and particulate contributions to the space charge. The results
expressed in Equations (7) through (9) are static.
29-13

-------
Townsend's theory expresses corona current as a quadratic function of
the applied potential. In Cooperman's notation (but in MKS units)
8tt .k
1 = -ZtVt v(y-v) ,	(10)
b2n(b/a)	
where I is the current per unit length of corona wire, and k is the ion mobi-
lity. The quantity VQ is the corona inception voltage, defined by the rela-
t ionship
VQ = aEQn(b/a)	(11)
E0 is the critical field strength, or the minimum value of the field required
to sustain ionization. Then the corona inception voltage is the applied
potential required to produce an electric field strength EQ at the surface of
the discharge electrode.
Now, as evidenced by Equation (8), the presence of a space charge will
decrease the electric field strength at the surface of the corona wire for a
given value of applied voltage. In any event, no corona discharge can occur
until the value of E(a) reaches EQ. So, from Equation (8), expressing this
condition where VQ is the inception voltage in the presence of the space
charge,
Vn	p	- a2 n
E = 	C	 + 	 a2 - 	]	(12)
0 an(b/a) 2ea	2n(b/a)
which yields
or
Vq = aEQn(b/a) + [b2-a2(l+2n(b/a))] ,
v0 - VQ +  [b2-a2(l+2Jtn(b/a))] .	(13)
The corona inception voltage therefore shifts	upward by an amount pro-
portional to the space charge. Clearly, the value of p in Equations (12) and
(13) includes only particulate space charge, since	there is no ionic current
at the point of corona inception.
Because the general effect of the particulate space charge, under any
conditions, is to reduce the linear charge density on the corona wire by an
amount proportional to the density of the space charge, the presence of a
given distribution of particulate space charge causes a shift in the entire
voltage-current characteristic curve. The magnitude of the displacement must
be the same as the change in the inception voltage derived above. Conse-
29-14

-------
quently, the perturbed expression relating the current to the applied poten-
tial takes the form
X = 8^nk	 V {v-V -  [b2 - a2(l + 2n(b/a)) 1} .	(14)
b 2n(b/a)	0 4e L	H
VI. RESULTS AND CONCLUSIONS
Equation (14) can be used as it is to describe the effects of space
charge in a system where there is a uniform and constant distribution of
charge, as in the example presented in connection with Figures 4 and 5. The
predicted shift in the V-I characteristic agrees with that indicated in
Figure 3. At the time of this writing the modeling of the dynamic behavior,
as exemplified by the reversal of the space charge illustrated in Figures 8
and 9, is still in progress. The model includes a charging routine based
upon the sum of field and diffusion charging (5); it also accounts for the
collection of particles and the concomitant removal of their contribution to
the space charge by the collection process developed in the Deutsch-Anderson
model.
The experiments described in this paper have shown the extent to which a
corona discharge is dominated by the effects of space charge. When the
normal ionic space charge in a corona system is augmented by a particulate
contribution, the electric field becomes redistributed, resulting in a change
in the local field in the zone of the active corona and thus a modification
of the overall conduction characteristics of the system. Since the particu-
late charge can be determined by application of established charging theo-
ries, a straightforward linear superposition of all charges results in a
complete expression for the behavior of a system where the distribution of
charge agrees with the geometrical symmetry of the system.
REFERENCES
1.	Deutsch, W. Bewegung und ladung der elektrizitatstrager im zylinderkon-
densator, Ann. Phys. 68, 1922. 335 pp.
2.	Cooperman, P. A new theory of precipitator efficiency. Atmos. Envir.
5:541, 1971.
3.	Gooch, J.P. and Francis, N.L. A theoretically based mathematical model
for calculation of electrostatic precipitator performance. J^. APCA
25:2:108, 1975.
4.	Cooperman, P. A theory for space-charge-limited currents with appli-
cation to electrical precipitation. AIEE Trans., March 1960, p. 47.
5.	White, H.J. Particle charging for electrostatic precipitation. AIEE
Trans. 70: 1, 1951).
29-15

-------
AN ELECTROSTATIC PRECIPITATOR FACILITY FOR TURBULENCE RESEARCH
J. H. Davidson
Department of Mechanical and
Aerospace Engineering
University of Delaware
Newark, Delaware 19716
E. J. Shaughnessy
Department of Mechanical Engineering
and Materials Science
Duke University
Durham, North Carolina 27706
ABSTRACT
Turbulence is generally accepted to be detrimental to precipitator
particle collection efficiency, yet the magnitude and character of electric-
ally induced secondary flow and turbulence in full scale precipitators is
not thoroughly understood. Prior laboratory scale experimental studies
indicate that negative corona discharge generates turbulence but are re-
stricted to precipitator flow aspect ratios (height to width) much less than
those of interest in industrial design. This paper describes a large scale
laboratory precipitator aerodynamic research facility. The electrical char-
acteristics of negative corona operation are presented along with hot-film
anemometer measurements of the electrically induced turbulence. Measure-
ments include streamwise mean velocities, turbulence intensities, and one-
dimensional energy spectra for gas speeds from 0.5 to 2.0 m/sec and electric
current densities as high as 0.84 mA/m2 of plate area. It was found that
turbulence was generated throughout the precipitator. The level of turbu-
lence production increases with increasing current but is not sensitive to
increases in current density above approximately 0.3 mA/m2. At a primary
gas speed of 1.0 m/sec, a current density of 0.3 mA/m2 causes a threefold
increase in turbulence intensity over that of the background aerodynamic
turbulence (from 5 to 15 percent). The implications of these results are
discussed in terms of particle transport and precipitator collection
efficiency.
30-1

-------
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
INTRODUCTION
During the past decade, expanding research efforts aimed at understand-
ing the fundamental processes of the wire-plate electrostatic precipitator
have been directed toward increasing particle collection efficiency without
corresponding increases in specific collector area and capital cost. While
much of the research has focused on electrical and mechanical character-
istics, the impact of fluid forces on the particle collection process must
also be considered. Of all the phenomena affecting precipitator perform-
ance, the effect of the electric body force associated with a negative
corona discharge on secondary flows and turbulence levels, and thus particle
motion, is one of the least understood and perhaps one of the most signifi-
cant. The dispersion and mixing of particles resulting from any elec-
trically induced turbulence and secondary flows may be a dominant effect and
is certainly not conducive to the precipitation process in which there is a
preferred direction of particle motion. In addition to an initial reduction
in particle collection efficiency resulting from the disruption of particle
motion toward the collector plates, any turbulent scouring action at the
plates causes reentrainment of collected particles into the gas stream and
thus further reduces the collection efficiency.
While it is well known that in the case of a uniform discharge along
the wire electrodes, the electric body force on the gas produces a secondary
flow, referred to as the corona wind, the questions surrounding secondary
flow and turbulence generation by the electric field in a negative corona
discharge have not been adequately addressed. Research on this subject in
the past several years has generated much discussion on the deleterious
effects of electrically induced turbulence on precipitator performance, but
the fundamental mechanism responsible for turbulence creation is not fully
understood. Nor is there any quantitative information on this phenomenon in
full scale precipitators.
A theoretical analysis of the governing electrohydrodynamic equations
(1,2) points out that the extent to which precipitator gas flows are modi-
fied by the corona discharge is determined by the magnitude and distribution
of current density J-j. The current density in turn is determined primarily
by the electrode geometry and also the discharge polarity. The situation in
the case of the wire-plate precipitator is dominated by the structure of the
corona discharge along the wires. The majority of what is known about the
current density distribution and electrically induced flow modifications in
the wire-plate precipitator is restricted to a positive uniform discharge in
a laminar flow. The pioneering studies of uniform positive corona by
Ramadan and Soo (3), Robinson (4), Yabe et al_ (5), and Yamamoto and Velkoff
(6) indicate that electrically induced secondary flows of a circulatory
nature are possible. But because of the inherent differences in negative
and positive corona discharge, it is reasonable to assume that the character
of the corona induced flow modifications is polarity dependent. Thus these
30-2

-------
prior studies are of limited use in understanding the flows induced by a
negative discharge. Recent laboratory small scale experimental investiga-
tions of negative corona discharge by Leonard et al_ (7,8), Ushimaru et al
(9), and Kumaran (10) confirm that turbubulence is generated by a negative
discharge. It is not clear that these results are applicable to actual pre-
cipitators which have large flow aspect ratios (height to width ratio) and
essentially two dimensional flows.
To address these inadequacies, a large scale laboratory electrostatic
precipitator was built to study the aerodynamic effects of corona discharge
in a facility more representative of commercial practice. In this paper,
the facility is described and selected results of an extensive experimental
study of flow modifications caused by a negative corona discharge are dis-
cussed. Hot film anemometry measurements of mean streamwise velocities,
turbulence intensities, and energy spectra are presented for nominal gas
speeds of 0.5 to 2.0 m/sec and negative corona discharge currents as high as
0.84 mA/m^ of plate area. Experiments were conducted at room temperature
with no particle loading of the precipitator gas stream. No conclusive evi-
dence of an organized secondary flow was found, but substantial increases in
turbulence intensity were seen even at relatively low current levels. The
spectral analysis indicates that the electrically induced turbulence is
highly diffusive. The implications of these results to the particle collec-
tion process and precipitator design are significant.
EXPERIMENTAL FACILITY
THE LABORATORY PRECIPITATOR
The experimental precipitator used in this study was specifically de-
signed to study corona wind and turbulence as they occur in full scale pre-
cipitators and is unique among experimental facilities in its size and range
of operating conditions. A schematic diagram of the precipitator is shown
in Figure 1. The flow channel is 10 m long, 4.88 m high, and 24 cm wide
providing an aspect ratio of 20. Room air is drawn through the tunnel past
metal inlet scrolls, a foam air filter to remove dust and large particles,
and a corrugated paper flow straightener. The inlet scrolls provide a
smooth turn into the rectangular duct thus preventing flow separation and a
large pressure drop. The straightener eliminates any disruptive flow
patterns generated at the entrance by forcing the air stream through 8 mm
corrugations 30 cm long. Flow rate is controlled by a variable speed tube-
axial blower capable of mean flow speeds up to 2.5 m/sec.
The active precipitator test section, located 5.2 m from the entrance,
is 1.83 m in the streamwise direction. Plate to plate spacing is 24 cm.
Six 3.2 imi diameter corona discharge wires on 30.5 cm spacings are located
midway between the grounded flat metal walls. The steel wires are mounted
at the floor and ceiling of the test section to aluminum bus bars elec-
trically isolated from the test section and exterior of the tunnel by 1.3 cm
thick plexiglass. Instrumentation ports are located 30.5 cm and 46 cm up-
stream and downstream of the test section along the entire 4.88 m height. A
Del Electronics low ripple power supply rated at 75 kV and 30 mA provides
30-3

-------
Test
Section
-*"1.83m*
Flow Straightener
4.88 m
Air
Instrumentation
Ports
Exhaust
vDuct
Flow
Discharge
Wi res
Fan
Air
High
Voltage
Power
Supply
Variable
Speed
Motor
ter
Figure 1. The Laboratory Precipitator
(Scale: 1 cm = 0.8 m)
30-4

-------
the capability of negative and positive corona discharge. Only negative
corona is considered in this paper. Wire diameter, wire to wire spacing,
and duct width are common to industrial precipitators. The tunnel height,
though less than some industrial precipitators, provides a very large exper-
imental aspect ratio and the flow is two dimensional over more than 90% of
the precipitator height.
Downstream of the test section a 25 degree contraction reduces the duct
height to 2.44 m. The airstream exits the tunnel through a right angle
semicircular turn. The blower, mounted at the base of the exit turn, is
attached to a removable nylon duct leading outdoors.
INSTRUMENTATION AND PROCEDURES
The difficulties presented in measuring gas velocities inside the
active section of a precipitator are immense. Because of these diffi-
culties, the approach taken was to use hot-film anemometry to measure the
flow 46 cm downstream of the discharge wires. The basic concept is to view
the active precipitator as a black box which transforms the input flow into
an output flow. Measurements taken in the absence of a corona discharge are
compared to those obtained for a range of electrical current levels. A
great deal of effort was taken to insure against the possibility of anomal-
ous velocity measurements attributable to current flow to the hot-film probe
itself. Kumaran et al_ (11) studied this phenomenon in a small model precip-
itator. A similar study of this problem was carried out in the present
facility at the corona current levels and gas speeds of interest (1,12).
The results indicate that placing the probe 46 cm from the corona discharge
wires prevents measurement errors.
Flow measurements were made with a TSI 1050 linearized anemometer
system using 1210-20 film sensors. The hot films were calibrated with the
known steady flow provided by a TSI 1125 calibrator. Streamwise mean
velocity (u) and turbulence intensity (urm^/0") transverse profiles were
obtained by traversing the hot-film probe in 20 cm increments across the
flow channel at five vertical locations: the centerline, and 1.2 m and 2.1
m above and below the center. The linearized voltage output was signal
conditioned to remove most of the DC component and then low pass filtered at
100 Hz with a Kron-Hite 3202R filter. The signal conditioner DC component
was retained to restore the mean value to its correct level during data
analysis. The output of the low pass filter was connected to a Honeywell
SAI-48 correlator linked with a Radio Shack TRS-80 microcomputer. The cor-
relator constructed the probability distribution of velocity using 32,768
samples at 5 millisecond intervals. The 400 point probability density func-
tion obtained in this manner was stored and analyzed with the microcomputer
to obtain the mean and centered root mean square (rms) voltages which were
then converted to velocity using the linearizer gain.
The one dimensional energy spectrum is the cosine transform of the
autocorrelation function. To record the autocorrelation function, the hot-
film signal is bandpass filtered at 0.1 and 100 Hz and then fed into the
correlator which constructs the autocorrelation function. Incremental
sampling time is 5 milliseconds and the total number of samples is 131,072.
30-5

-------
The total observation time is thus 11 minutes. The autocorrelation data is
transformed from the time domain to the frequency domain by means of a
cosine transformation algorithm of Bendat and Piersol (13) on the Triangle
University Computation Center's mainframe IBM 370.
RESULTS
ELECTRICAL CHARACTERISTICS
The negative corona current-voltage characteristics are shown in Figure
2. The corona onset voltage is approximately 25 kV and sparkover occurs at
voltage and current levels of 60 kV and 20 mA respectively. Visual observa-
tion of the discharge in the large scale laboratory precipitator confirms
the nonuniformity of the current distribution on the discharge wires. At
the onset of the discharge, a few discrete corona discharge tufts or spots
(on the order of 2 to 5 mm in length) appear along the wires. These spots
occur at random positions about 15 cm apart along the length of the wires
but tend to predominate on the "sides" of the wires closest to the collector
plates. Once established the discharge tufts do not move about on the wires
since imperfections on the oxidized wires tend to stabilize the tuft loca-
tions. As the voltage is increased, more corona spots are produced and the
distribution of the spots along the wires is more uniform. In other words,
the spacing between adjacent tufts is relatively constant along the total
wire length. The number of discharge soots tends to reach a maximum at
current levels of about 5 mA (0.28 mA/m2) which is twenty five percent of
the sparkover current. The minimum tuft spacing is on the order of 3 cm.
The tufts appear along the entire length of the six discharge wires.
FLOW MEASUREMENTS
The precipitator flow is uniform over 90 percent of the duct height as
shown by the upstream and downstream zero current mean velocity profiles of
Figures 3 and 4. The number noted on each curve is the vertical distance
from the duct centerline. The profiles indicate a fully developed flow with
a slight dip in the downstream profile due to the wakes of the unenergized
wires. The overall shape of these curves is representative of those
obtained at gas speeds of 0.5 and 2.0 m/sec.
The two-dimensional nature of the flow is maintained with corona dis-
charge and thus to avoid redundancy only the centerline profiles are dis-
cussed. Figures 5, 6 and 7 are mean velocity profiles, for nominal gas
speeds of 0.5, 1.0 and 2.0 m/sec respectively, at current density levels of
0.0, 0.28, 0.56, and 0.84 mA/m2 (corresponding to currents of 0, 5, 10, and
15 mA). Clearly there is no significant current effect on the mean flow at
gas speeds of 1.0 and 2.0 m/sec. A slight flattening of the profiles
suggests increased turbulent mixing. The scatter in the mean velocity data
of Figure 5 for the low gas speed of 0.5 m/sec is due to the character of
the flow itself. The current off flow is laminar but unsteady whereas with
a corona discharge, the flow is intermittently turbulent. Thus repeatable
measurements are difficult to obtain using reasonable averaging times.
30-6

-------
1
p
c
CU
s-
S-
3
o
Figure 2.
20.	30	40		
Voltage (kV)
Voltage-Current Characteristics
for a Negative Corona Discharge
o
0)
to
I Z3
o
o
'qj
>
c
(0

-------
CJ
a>
v>
1=5
r
U
O
a>
c
10

-------
1.5
o
qj
en
1 3
>>
o
o
%
c
	1	-
0.00 mA/m2
I

Wall	t
Wi re
Figure 6. Centerline Mean Velocity Profiles,
Gas Speed =1.0 m/sec.
Wall
0.28-*
0.00 mA/m
Figure 7. Centerline Mean Velocity Profiles,
Gas Speed - 2.0 m/sec.
30-9

-------
The turbulence intensity profiles of Figures 8, 9, and 10 indicate that
turbulence intensity increases significantly due to the corona discharge at
gas speeds of interest for industrial precipitators. The irregularity of
the 0.5 m/sec current on profiles is once again a reflection of the inter-
mittent nature of the turbulence. This problem aside, the increase in
intensity caused by the discharge is striking. The intermittent behavior of
the low speed flow does not persist at gas speeds of 1.0 and 2.0 m/sec. At
a flow speed of 1.0 m/sec the intensity increases from 5 to 14% at the duct
midpoint. Nearer the walls, the increase in intensity is of less conse-
quence. This can be explained by intense turbulence production in the
vicinity of the discharge tufts on the wires. The effect is less dramatic
at a gas speed of 2.0 m/sec as shown in Figure 10. Here intensity values
are doubled by the corona discharge. Noteworthy in these results is the
insensitivity of the data to increasing the current density above 0.28 mA/nr
of plate area (5 mA).
Some of the more interesting results of this study appear in Figure 11.
Centerline turbulence intensities are plotted as a function of current den-
sity. This figure highlights the roles of gas speed and current density in
determining the magnitude of the electric body force effect on turbulence
levels. The data indicate that the effect of the corona discharge becomes
more significant as the primary gas speed is reduced and the current density
is increased up to a certain point. A remarkable feature of the data is the
insensitivity of the turbulence intensity to increasing the current density
above 0.3 mA/nr. The rate of increase of intensity with current level de-
creases dramatically at this point. This is particularly evident at gas
speeds of 0.5 and 1.0 m/sec. This leveling off of intensity occurs at ap-
proximately the current density for which the number of discharge tufts is
the maximum and the distribution of tufts is most uniform. As the corona
current is increased and the discharge tufts become more numerous, the cur-
rent density distribution becomes more uniform. It is speculated that once
the corona discharge becomes more coherent along the wires, any additional
electrical energy input results in a more organized secondary flow in the
region between the wires and the plates much like that predicted by Yamamoto
and Velkoff (6) for a positive uniform discharge. The inability to operate
the hot-film probe in the active section of the precipitator is a severe
limitation to characterizing the exact nature of any organized secondary
flows.
The energy spectra of Figure 12 are those obtained at the tunnel
centerline for a nominal gas speed of 1.0 m/sec. The energy spectra indi-
cate the distributions of the turbulent energy among the various frequencies
and provide a means of identifying any preferred frequency of the corona
generated turbulence. The spectra provide dramatic evidence of the elec-
trical turbulence generation. Like the turbulence intensity results, the
spectra do not show any marked differences due to increasing the current
density above 0.28 mA/mS Compared to the zero current spectrum, the
current-on spectra do not indicate a favored frequency but contain addi-
tional amounts of energy at all frequencies. Most of the additional energy
input is in the lower frequencies corresponding to the large scale turbulent
motions. Since turbulent diffusion is a process dominated by large eddies,
the spectra suggest that the corona discharge is responsible for increases
30-10

-------
0.84
30-
o
o
0.56
22-
0.28
14-
o
10-
0.00 mA/m
Wall
Wall
Wire
Figure 8. Center!ine Turbulence Intensity
Profiles, Gas Speed = 0.5 m/sec.
34,-
o
to
.0.84
0.54
0.28
3
-Q
6-
2
Wall
Wall
Wi re
Figure 9. Center!ine Turbulence Intensity
Profiles, Gas Speed = 1.0 m/sec.

-------
o
o
c
ai
+j
c
' 14j
QJ
u
c
QJ
0.84
0.56
0.28
0.00 mA/m
Wall
Wall
Figure 10. Centerline Turbulence Intensity Profiles
Gas Speed =2.0 m/sec.
22
O
o
p
X
18
in

3
14-
'r-
LO
c

C

<1)
U
C

-------
N
E
o
10
10
10'
10
+J
d>
o
>>
a-'
s-
o
E
10
-i
10"
10"
3 a a tfri,-1!
Current (inA/mZ-)
*	0.00
9 0.28
	0.56
<* 0.84
4-4-
10
Figure 12.

4-4-
requency
y (Hz1)0
i IIP.
Centerline Energy Spectra,
Gas Speed =1.0 m/sec.
30-13

-------
in turbulent diffusivity. Eulerian length scales (L), which are an estimate
of the average turbulent eddy size, were computed from the spectral
analysis. With no corona discharge, the length scale is 10.3 cm and with a
current density of 0.28 mA/m2 the length scale is 8.0 cm. An estimate of
the increase in particle diffusion can be made from a dimensional estimate
of the eddy diffusivity, <, as urms L where urms is the root mean square
velocity and L is the integral length scale determined from the spectral
analysis. The purely aerodynamic diffusivity is 53 cm2/sec. At a corona
current density of 0.28 mA/m2, the diffusivity is increased to 79 cm2/sec.
SUMMARY AND CONCLUSIONS
Measurements of the turbulent flow field in a large scale laboratory
precipitator confirm that the inhomogeneous negative corona discharge is re-
sponsible for marked increases in turbulence levels over that of the back-
ground aerodynamic turbulence. It was found that turbulence was generated
throughout the precipitator. The results confirm the experimental results
of corona induced turbulence obtained in bench scale model precipitators by
Ushimaru et al (9), Leonard et al (7,8), and Kumaran (10) but extend the
experimental data to a more realistic geometry and a wider range of operat-
ing conditions. The laboratory precipitator has a flow aspect ratio
(height/width) of 20. Test conditions included gas speeds from 0.5 to 2.0
m/sec and electrical corona discharge current densities as high as 0.84
mA/m2 of plate area. The level of turbulence production increases with de-
creasing gas speed and increasing current density but is not sensitive to
increases in current density above 0.28 mA/m2. The latter result is at
first glance quite surprising; however, when viewed in terms of the physical
appearance of the discharge is very reasonable. At low current densities
the discharge tufts are widely spaced in an irregular pattern on the wires.
As the current density is increased, the number of discharge tufts grows and
the spacing between tufts decreases and finally stabilizes at a current den-
sity of approximately 0.3 mA/m2. At this point, the electric body force is
more uniformly distributed along the length of the wire. In this case the
corona discharge may cause a circulating motion on the scale of the wire to
wire spacing in addition to increasing turbulence levels.
The implications of these results to the particle collection process in
industrial precipitators are significant. At typical operating conditions,
a gas speed of 1.0 m/sec and current density of 0.3 to 0.84 mA/m2, the
corona discharge is responsible for a threefold increase in turbulence
intensity (from 5 to 15 percent). In addition to the resulting increased
turbulent diffusion, any ordered vortical secondary flow which exists at
high current densities can significantly disrupt the particle collection
process.
Since the turbulent flow in a precipitator arises from both aerodynamic
and electric forces, the relative importance of the electrically induced
secondary flows and turbulence in commercial precipitators will depend upon
the magnitude of other non-ideal effects including poor aerodynamic design.
Wire-plate precipitators are used in many applications and are operated over
a wide range of conditions. Thus it is nearly impossible to project in a
30-14

-------
general way the significance of corona induced secondary flows and turbu-
lence. But it is important to realize that the turbulence created by the
electric body force is in addition to whatever aerodynamic turbulence is
present and will increase turbulent diffusion. Recognizing that turbulent
diffusion and secondary vortical motions will always oppose the migration of
charged particles to the collection plates, a reduction in these effects
will increase precipitator efficiency.
REFERENCES
1.	Davidson, J. H., Secondary Flows and Turbulence in Electrostatic Pre-
cipitators. Ph.D. Dissertation, Duke University, Durham, N.C., 1984.
2.	Shaughnessy, E. J., Davidson, J. H., and Hay, J. C., The Fluid Dynamics
of Electrostatic Precipitators: Effects of Electrode Geometry. Fifth
Symposium on the Transfer and Utilization of Particulate Control Tech-
nology, Kansas City, Missouri. August 27-30, 1984.
3.	Ramadan, 0. E., and Soo, S. L., Electrohydrodynamic Secondary Flow.
Phys. Fluids 12: 1943, 1969.
4.	Robinson, M., Effects of the Corona Discharge on Electric-Wind Convec-
tion and Eddy Diffusion in an Electrostatic Precipitator. ERDA Health
and Safety Laboratory Report, HASL-301, 1976.
5.	Yabe, A., Mori, Y., and Nijikate, K., EHD Study of the Corona Wind Be-
tween Wire and Plate Electrodes. AIAA Journal 16: 340, 1978.
6.	Yamamoto, T., and Velkoff, H. R., Electrohydrodynamics in an Electro-
static Precipitator. J. Fluid Mech. 108: 1, 1981.
7.	Leonard, G. L., Mitchner, M., and Self, S. A., Precipitation from Tur-
bulent Flows. Proc. of the Intl. Conf. on Elec. Precip., Monterey,
California, 208, 1981.
8.	Leonard, G. L., Mitchner, M., and Self, S. A., An Experimental Study of
the Electrohydrodynamic Flow in Electrostatic Precipitators. J. Fluid
Mech. 127: 123, 1983.
9.	Ushimaru, K., Butler, G. W., and Milov soroff, P., An Experimental and
Theoretical Investigation of Electrostatic Precipitators. Flow Re-
search Report #219, Flow Industries, Kent, Washington, 1982.
10.	Kumaran, A. R., Aspects of the Fluid Mechanics of Electrostatic Precip-
itators. HTGL Report No. 15-83-TR, 1983.
11.	Kumaran, A. R., Leach, R., and Self, S. A., Anomalous Readings from
Hot-Wire Anemometers in Strong Electric Fields, TSI Quarterly, July-
Sept., 9, 1983.
30-15

-------
Davidson, J. H., and Shaughnessy, E. J., Mean Velocity and Turbulence
Intensity Profiles in a Large Scale Laboratory Precipitator. ASME
Joint Power Generation Conference, Toronto, Canada, September, 1984.
Bendat, J. S., and Piersol, A. G., Random Data: Analysis and Measure-
ment Procedures, (New York: Wiley-Interscience, 1971), p. 317.
30-16

-------
ON THE STATIC FIELD STRENGTH IN WIRE-PLATE
ELECTROSTATIC PRECIPITATORS WITH PROFILED COLLECTING ELECTRODES
BY AN EXPERIMENTAL METHOD
C. E. Akerlund
Flakt AB
Vaxjo, Sweden
ABSTRACT
To introduce the highest possible energy input in an electrostatic
precipitator, the field distribution at the collecting plate must be as even
as possible. It must be realized, however, that there are peaks in the field
distribution at such exposed points as stiffening elements. Sparkover is
most probable to occur from these points.
In this report the static field at profiled collecting electrodes has
been analyzed using a combined conductive paper - computer aided design
technique. The presented results verify that the field strength rises
significantly in local areas.
Therefore investigations of the field distribution at collecting
electrodes of new design are of great importance, if sparking at too low
a voltage is to be avoided.
INTRODUCTION
In an electrostatic precipitator the transformer/rectifier sets are
normally current-regulated. To obtain maximum charging and precipitation of
the dust, the current is gradually increased up to a level where sparkover
starts. However, it is the field strength at the sparkover point - not the
current level - which is of importance for the initiation of the sparkover.
The critical field strength is highly dependent on the properties of
the gas and the dust - for example, the gas composition, gas temperature,
gas velocity and the resistivity of the dust. But if the mechanical design
of the collecting electrodes and the discharge electrodes with their frame
is unfavorable, sparkovers occur in exposed points even at low charging
current and during dust-free operations. Uniform field distribution over the
collecting electrode must be secured to avoid sparking at too low a voltage.
31-1

-------
The calculation of current-voltage characteristics and the electric
field in wire-plate precipitators has been dealt with by many authors (1),
(2), (3), (4). These methods are presented under the presumption that the
collecting plates are plane.
A light and inexpensive design of the discharge and collecting system
always includes stiffening elements in the form of beads or other projecting
parts. A stiffener with an inexpedient shape can cause local magnification
of the electric field well above the sparkover limit.
A numerical calculation of the field in the region close to earthed
profiled collecting electrodes where the potential is sufficiently high for
the initiation of sparkovers is consuming considerable time. A practical
method will more easily provide a solution with sufficiently reliable
accuracy.
Vitkovitch (5) has presented different numerical, graphical and
experimental methods in an extensive survey. Among numerical techniques, the
method of finite elements is the most important. It requires, however,
comprehensive computer programming. Graphical field plotting requires a
minimum of equipment but the work can be very laborious if the object has a
complex geometrical shape and if high accuracy is required. The electrolytic
tank can be used for two-and three-dimensional cases of Laplacian as well as
Poissonian fields but requires extensive laboratory resources. In addition,
the method is not very flexible, if changes in the object's configuration
are desired.
The combined conductive sheet - computer aided design technique
proposed in this paper is very attractive, due to its simplicity, although
it is limited to two-dimensional cases. It can be used for both static and
space-charge fields. This investigation covers fields without space charge.
EQUIPMENT
CONDUCTIVE daPER
i--r + + + +r + ++ r + "t-+*--^
Figure 1
Equipment
31-2

-------
Figure 1 shows the equipment used. The conductive paper has a graphite
coating. The resistivity of the paper is about 1250 ohms per square. A
variation of 3 % has been found from paper to paper. The paper is slightly
anisotropic but the anisotropy is so slight - about 1 % - that it can be
overlooked.
The model is painted on the paper with a conducting paint containing a
high percentage of finely divided silver. After 24 hours curing in ambient
temperature, the resistivity is 0.11 ohms per square. Several contacts were
attached to the collecting electrode. No further action is required to
ensure that the full length of the collecting plate has the same potential.
TECHNIQUE
Equipotentials can be plotted in two different ways. One method is to
find a number of points having the same potential and to draw a curve
through these points. In areas where the shape of the equipotential is more
difficult to predict, many points must be tested with the probe until the
equipotential point is determined. The searching takes much time. It was
also discovered that the field was disturbed by a large number of closely
placed holes made with the probe.
These negative experiences led to the use of another procedure. A grid
of points was drawn on the paper. The spacing of the grid was adapted to the
complexity of the equipotential's shape. The potential was then measured at
every point of the grid. Assuming that the potential varies from a measured
value in direct proportion to the distance from the measuring point, the
coordinates of the equipotential line were calculated. These coordinates
were fed into a computer-aided design system and the lines were drawn using
a spline function. The function is a parametric spline fitting through all
given points that maintains continuity up to the second derivative. Each
segment is described as a cubic function.
The field strength increases rapidly in the vicinity of corners and
sharp edges. Consequently, the measuring grid must be very fine in this area
and the measurements have to be made in several steps.
As shown in figure 1, a complete model was first painted in such a
scale that the silver-painted electrode was given a minimum width of 2.5 mm
to ensure an equal potential along the silver string. In this model
equipotential lines were drawn; one of these, 4 kV, was drawn using a very
fine grid in order to obtain a high degree of accuracy. (See figure 2.)
A second model was painted in much larger scale. (See figure 3.) A
large scale was used not only because the grid must be fine at certain
areas. It was also discovered that it is roost difficult to accurately
measure the very low voltage close to the silver string. Consequently, no
measurements were made closer to the silver string than about 4 mm in actual
scale.
31-3

-------
125.00
150.00
>25.00
5 KV
52
KV
Figure 2.	Emitting and collecting electrodes with equipotential
lines.
+	t-	+	+	K V
+	+	+	+	+ +	+	+	+ ~~~~~~ 			
+	+-	+	+	-H+-	+	+	+	+ +  +
+	+	+	+	+ +	+	4-	+	+ t+ -r f + + 4 4- 4-4 4
4
+
+
+
+
+
+
+
f
+
Figure 3.	Enlarged section of figure 2.
A voltage difference of about 28 V was used between emitting and
collecting electrodes in the models. Since the feed point of current was a
silver paint with a diameter of only about 5 mm, the voltage had to be
limited to about 30 V to prevent burn in the paper.
31-4

-------
The use of a computer-aided design system for the drawing work is a
prerequisite for a successful outcome. The technique to make the
measurements in several steps in gradually increased scale requires an exact
reproduction of the model and the equipotentials, the latter being defined
by a great number of coordinates in the spline function.
An error calculation shows that the true value of the field strength
with 95 % probability lies within a range of + 3 % from the mean value.
APPLICATION
Two profiled collecting electrodes were analyzed at different voltage
distances, i.e. the distances between emitting and collecting electrodes.
The field strengths refer to the corona starting voltages shown in table 1.
TABLE 1. CORONA STARTING VOLTAGES IN KV AT DIFFERENT DISTANCES BETWEEN
EMITTING AND COLLECTING ELECTRODES
Voltage
Corona
distance
starting
mm
voltage kV
150
43.0
200
52.5
250
61.0
Let us first revert to figure 2. In the region from the emitting
electrodes, and down to about 20 kV, the equipotentials are scarcely
affected by the geometrical form of the plate's end profile. Close to the
plate the equipotentials follow the profile and "sneak" around its edge.
This is further illustrated in figure 4.
0. 56 MM
750 V
POO V
375 V
VOLTAGE DISTANCE - 0.25 M
Figure 4.	Electric field at profiled collecting electrode.
31-5

-------
The field strength at the center of the profiled plate midway between
two emitting electrodes (figure 2) is about 137 kV/m for the three voltage
distances of 150 mm, 200 mm and 250 mm. Values calculated using both
references (1) and (3) agree with this value.
Totally different conditions will be found at the end profile. Table 2
shows the measured fifeld strength along two field lines, one at the top and
one at the edge of the profile. (See figure 4.) The field strengths are
calculated using reference (6) and the voltages and distances shown in
figure 4.
TABLE 2. FIELD STRENGTH KV/M AT TOP AND EDGE OF END PROFILE
Voltage
distance
mm
Field strength kV/m
at top of profile,
calculated as average
between
at edge of profile,
calculated as average
between
0-200 V 0-375 V 0-750 V 0-200 V 0-375 V 0-750 V
150
250
Magnifi-
cation
with
increased
voltage
distance
320
410
1.3
290
380
290
370
1.3
1.3
400
530
220
280
140
180
1.3
1.3
1.3
As shown in table 2, the field strength at the top of the profile
increases very moderately, about 10 %, if calculated as an average up to the
200 V potential line, compared with a calculation up to the 750 V line.
Totally different conditions will be found at the edge. Since the profile
has an open part, the 750 V line will not follow the plate profile but
directs itself towards the plane part of the plate and moves away from the
edge.
Contrary to the 750 V line, the 200 V line "sneaks" around the edge and
the field calculated up to the 200 V line is three times higher than the
field calculated up to the 750 V line. This illustrates the necessity to base
the field strength calculation on potential lines very close to the critical
point.
31-6

-------
Higher field-strength values would undoubtedly result if an
equipotential line had been drawn still closer to the plate than the 200 V
line. However, it should be noted that, according to Paschen's law, the
minimum voltage is 300 V for discharge in air between two plane electrodes.
Thus, in this case as well, large fields for voltages less than 200 V may
not initiate breakdown.
If the voltage distance is increased to 250 mm, the field at the top
and edge is 30 % higher than the field at the 150 mm voltage distance. (See
table 2 and figure 5.) As mentioned above, no corresponding increase is
found at the center of the plate.
	750 V
^	r- 375 V
		aoo v
wiri)rii>ini>i/>iin
VOLTAGE DISTANCE 0.15 M
DASHED LINE
UNBROKEN LINE - VOLTAGE DISTANCE 0.25 M
Figure 5.	Electric field at 150 mm and 250 mm voltage distance.
Table 3 shows the relation between the field strength at the edge of
the profile, point A, figure 4, and the center of the plate. The field
strength at the edge of the profile is almost four times the field at the
center of the plate for a voltage distance of 250 mm.
TABLE 3. RELATION BETWEEN FIELD STRENGTH AT EDGE OF PROFILE AND CENTER OF
PLATE
Voltage Relation between field
distance strength at edge of profile
mm	and center of the plate, calculated as
average between
0-200 V 0-375 V 0-750 V
150	3.0	1.6	1.0
250	3.9	2.0	1.3
31-7

-------
Another collecting electrode design is shown in figure 6. The trend in
precipitator design development is towards wider collecting plates and this
design is available in large widths. A fairly wide plate needs one or more
stiffeners when a small plate thickness is used. Such stiffeners can be
inexpensive and made easily in the roll-forming operation. A typical
stiffener design is shown in the middle section of the plate.
750.00
225.00
150.00
4 KV
A KV
150.00
225.00
150.00
Figure 6.	Emitting and collecting electrode with equipotential
lines.
As figure 7 shows, equipotentials close to the end profile are affected
by the profile but they do not go around the edge of the profile as shown in
figure 5.
The field strength along the four different lines A, B, C and D of
the collecting electrode have been calculated using reference (6).
(See figure 7 and 8.) Line A is on the open side of the end profile, while
lines B and C are on the closed side. Line D is at the stiffening bend.
31-8

-------
e
l&QQ V
7&Q...V
375 V
200 V

'50 V
VOLTAGE DISTANCE - 0.25 M
Figure 7.	Enlargement of the end section in figure 6.
-4000
500
750
3 75
20 0

VOLTAGE DISTANCE - 0.25 M
Figure 8.	Enlargement of the middle section in figure 6.
TABLE 4. FIELD STRENGTH AT POINTS A, B, C AND D IN FIGURE 7 AND 8
Field strength kV/m,	Relation between field
Voltage calculated as average strength at different
distance	between 0-200 V at	points and center of plate
mm	the lines
ABCDA	B	C	D
150	220
250	300 260 320
Magnification
with increased
voltage
distance	1.4
1.8
250 2.4 2.1 2.5 1.9
31-9

-------
As seen in table 4, the field strength is higher at point C than at
point A. A close look at the equipotential lines verifies this observation.
At the closed side of the profile, the lines follow the plate profile and
come closest to the plate at the bend. On the open side of the profile, the
lines leave the upper corner and cross obliquely down to the plane part of
the plate, while the plate at the upper corner is bent 90 degrees.
The stiffening bend represents a very moderate change in the collecting
electrode's topography and may appear to be fairly harmless from the point
of view of field magnification. It should be noted, however, that the field
strength at the bend is nearly twice as high as that at the center of the
plate.
SUMMARY
The proposed method of investigation has been applied to collecting
electrodes which have been used for many years in electrostatic
precipitators with high collecting efficiency. The results confirm, however,
that a not negligible rise in the field distribution exists in local areas
of the collecting electrode. Thus, thorough investigations of the field
distribution along collecting electrodes of new design are of great
importance if sparking at too low a voltage is to be avoided. The combined
conductive paper - computer aided design technique used in this
investigation is recommended because of its simplicity.
The work described in this paper was not funded by the U.S. Environ-
mental Protection Agency and therefore the contents do not necessarily
reflect the view of the Agency and no official endorsement should be
inferred.
ACKNOWLEDGEMENTS
The author wants to thank Professor Stig Lundquist, head of the
Institute of High Voltage Research, Uppsala, Sweden, for his most valuable
discussions and guidance. I am greatly indebted to many colleagues in the
Electrostatic Precipitator and Data Processing Departments of Flakt Industri
AB who have contributed their experience.
31-10

-------
REFERENCES
1.	J. R. Mc Donald et. al. A mathematical model for calculating electrical
conditions in wire-duct electrostatic precipitation devices. Journal of
Applied Physics, Vol 48, No. 6, June, 1977, pp. 2231 - 2243.
2.	P. Cooperman. A theory for space-charge-limited currents with
application to electrical precipitation. AIEE, (1960) pp. 47 - 50.
3.	S. Sekar et. al. On the prediction of current-voltage characteristics
for wire-plate precipitators. Journal of Electrostatics, 10, (1981).
pp. 35 - 43.
4.	J. L. Davis and J. F. Hoburg. Wire-duct precipitator field and charge
computation using finite element and characteristics methods. Journal
of Electrostatics, 14, 1983, pp 187 - 199.
5.	D. Vitkovitch. Field Analysis. D. Van Nostrand Company Ltd., London,
1966.
6.	H. Prinz. Kochspannungsfelder. R. Oldenbourg, Miinchen, 1969.
31-11

-------
THE FLUID DYNAMICS OF ELECTROSTATIC PRECIPITATORS:
EFFECTS OF ELECTRODE GEOMETRY
E. J. Shaughnessy, J. H. Davidson,t and J. C. Hay
Department of Mechanical Engineering and
Materials Science
Duke University
Durham, North Carolina 27706
ABSTRACT
The role of the electric body force in the generation of secondary
flows and turbulence in precipitators is poorly understood at present,
despite the fact that these disturbances are thought to be detrimental to
precipitator performance. This paper undertakes a fundamental analysis of
the problem using the Navier-Stokes and reduced Maxwell equations. It is
shown that the rotational component of the current density causes vorticity
production in the gas and is therefore the source of secondary flows and
turbulence. When the current density is irrotational, as in a cylindrical
precipitator, the electric body force causes a change in the pressure dis-
tribution but has no effect on the velocity field. Since the current
density distribution is determined by the electrode geometry alone, the
electrodes should be designed to minimize the rotational component of the
current density and thus reduce gas flow disturbances.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
I. INTRODUCTION
There is increasing evidence that secondary flows and turbulence gener-
ated by corona discharge influence particle transport and lower the ef-
ficiency of precipitators (Williams and Jackson (1); Robinson (2,3); Cooper-
man (4); Soo and Rogers (5); Feldman et a]_ (6); Pyle et al_ (7,8); and
Leonard et_ al^ (9,10,11)J. While secondary flows have been described by
Ramadan and Soo (12), Yamamoto et al_ (13), Robinson (14), and Ushimaru et al
tCurrent address: Mechanical and Aerospace Engineering, University of
Delaware.
32-1

-------
(15) for the wire-plate precipitator with positive corona, it appears that
negative discharge produces turbulence. The recent experimental studies by
Ushimaru et al_ (15), Leonard et al_ (10,16), Kumaran (17) and Davidson (18)
show that the discharge increases the turbulence intensity substantially at
gas speeds of 1 m/sec. Perhaps more intriguing is the suggestion (18) that
higher current levels generate both turbulence and secondary flows.
Since there is little doubt that both of these processes limit the max-
imum collection efficiency, identifying the source of these disturbances is
the first step towards their elimination. The goal of this paper is to pro-
vide a more complete understanding of the electrohydrodynamic interactions
in particle free precipitator gas flows. By examining the governing elec-
trohydrodynamic equations, it is shown that the rotational component of the
current density causes the vorticity production which generates secondary
flows and turbulence. Flippen's (19) study of irrotational electrode geome-
tries and Hay's (20) perturbation solution for the wire-plate precipitator
are particularly helpful in revealing the role of electrode geometry in
electrical vorticity production.
In turbulent flow the mean and fluctuating rotational components of
current density are both capable of generating turbulence. Current density
fluctuations, which are probably insignificant with DC energization, may be
a very significant source of turbulence when the discharge is pulsed. These
results suggest that a uniform planar discharge electrode would cause less
flow disturbance than wires.
II. ELECTROHYDRODYNAMIC THEORY
The equations describing the electrohydrodynamics of a precipitator gas
flow in the absence of particles are Ohm's law, a set of reduced Maxwell's
equations, and the classical fluid mechanics equations. The applicable
electrodynamic equations are:
Jl = 6|Pc|Ei	(2.1)
3Ei
e - = P c
3x. **
1
(2.2)
E, - -	(2.3)
' 3x.
1
3J-;
 = 0	(2.4)
where J-j is the current density, B is the ion mobility, pc is the ionic
space charge density, e is a constant gas permittivity, E-j is the electric
32-2

-------
field, and <|> is the electric potential. The basic equations for the
incompressible flow field are continuity:
9u-i
 =0	(2.5)
and the Navier-Stokes equation with an electric body force term:
3ui	3u-j	aD	32Ui
p[ + Uj ] = -  + u - + pcE^	(2.6)
j 8xJ 8*i
where p is the fluid density, u^ is the velocity, p is the pressure, and v
is the absolute viscosity. The electric body force per unit volume on gases
exposed to corona discharge may be represented as J-j/3 for positive corona
discharge and as -J-j/P for the case of negative corona discharge. In the
theoretical development that follows, the discharge is assumed to be
negative.
An important feature of this theoretical formulation of the problem is
that the electrical equations may be solved independently of the equations
governing the flow. The reverse is not true. As a consequence of the
partial uncoupling of the equations, an expression for the electric body
force can be obtained without regard to the flow conditions. Once the
electric body force is known, equations (2.5) and (2.6) may then be solved
for the velocity u-j and pressure p.
It is useful to express the body force -J-j/3 in terms of its scalar and
vector potentials i|> and R-j respectively by
1 , 3i|>	^
-Ji/B = - - ( + eiik ).	(2.7)
1	3 3x. 1JK 3x.
^	J
Substituting this expression into the Navier-Stokes equation (2.6) yields
du-i	3u-i	ap 3^U-j 1	3R|<
p[1 + u< 1] - -  + p - - - 
force field. The irrotational component of the body force,  , no
3 3xi
longer appears in the equation of motion except as a modified pressure.
Thus the irrotational component of the electric body force has no effect on
the velocity field but merely affects the pressure distribution in the gas.
32-3

-------
Very little is known of the exact nature of the current density in most
geometries and in particular for a negative corona discharge. Numerical
solutions for uniform discharge in the wire-plate geometry (Leutert and
Bohlen (21); McDonald et_ _al_ (22); Lawless and Sparks (23); Yamamoto and
Velkoff (24); and Hoburg and Davis (25)) provide no information as to how
the current density is divided among its rotational and irrotational com-
ponents. However, the recent work of Hay (20) on the single wire with posi-
tive corona discharge does provide some insight into the composition of the
electric body force in the wire-plate geometry. His perturbation solution
of the governing electrodynamic equations confirms that J-j is always rota-
tional. The solution shows that at very low current levels the rotational
component of J-j is small compared to the irrotational component. At corona
current levels comparable to those of electrostatic precipitators, however,
the rotational component of the current density dominates the solution.
An examination of the vorticity equations provides further insight into
the role of the rotational current density in affecting the gas flow. The
vorticity equations are obtained by taking the curl of the continuity and
Navier-Stokes equations (2.5 and 2.6). The equations are
dlD-j
- - 0	(2-9)
1
and
3u),-	3(1)^	32^.	9U. ,	9j,
+ UJ = u o + ui	eiik 	(2.10)
3t J 3x. 3X.2 J	3x. 0 3x.
J	J	J	J
3U|^
where ^ = eijk - is the vorticity. This	is the classical vorticity
j
equation with an added vorticity production term involving the current
density. Since the current density term is the curl of J-j, an electric
field with a rotational current density component produces vorticity in the
interior of the fluid. This vorticity production is the source of the
secondary flows described by Yamamoto and Velkoff (13) and Ramadan and Soo
(12) for a uniform discharge.
The electric vorticity production term may be decomposed by taking the
curl of equation (2.1). The result is
1 3Jk	^Pc	3E|<
 ieijk "Eijk Ek + Pc 5^'	(2'n)
J	J	J
The second term on the right hand side of this expression is zero by
equation (2.3). Thus the curl of the electric body force may be written as
1	3^k	3pc
" 3 eijk = eijk  Ek-	(2-12)
J	0
32-4

-------
This expression is zero in two circumstances. The first is whenever the
space charge density is uniform, which is physically unrealistic. The
second situation for which there is no electrical vorticity production is
when the electric field lines are everywhere normal to contours of constant
space charge density. Flippen (19) has shown that this is possible for
three geometries provided the corona discharge is uniform on one electrode.
These geometries are concentric spheres, concentric cylinders, and parallel
plates. For all other geometries, and these three if the discharge is
nonuniform, the current density has a rotational component. Thus the
electrode geometry and uniformity of the discharge completely define the
degree to which the electric body force perturbs the gas flow.
Before discussing the mechanisms by which the electric body force
produces turbulence in precipitators, it is useful to nondimensionalize the
Navier-Stokes equation using the scales I, A, U, and L for current, plate
area, velocity, and plate-to-plate spacing. The result is
3ui	3^- 3p j 32Ui i
 + u-i - = -  +  - - J-j	(2.13)
3t J 3x. 3x. Re 3x.2 f2
J	'	J
rpU20Al1 /9
where Re = pUL/y is the Reynolds number and F = [is the e1ectric
Froude number. The parameter 1/F2 is a measure of the ratio of the electric
body force to the inertia! force. Using values representative of fly ash
precipitators (U = 1 m/sec, p = 0.55 kg/m3, y/p = 0.58 cm2/sec, M 2.1 x
10^ rrr/volt-sec, I/A = 4 x 10~4 Amp/nr of plate area, L = 0.25 m) leads to
estimates for the Reynolds and electric Froude numbers of Re = 4310, F =
1.07 for a precipitator operating at 375C: at 27C, Re = 15,000 and F =
1.58.
These estimates allow us to draw several conclusions regarding the
fluid dynamics of wire-plate precipitators. Since Hay (20) has shown that
the body force is predominantly rotational in the wire-plate geometry, a
ratio of rotational electric body force to inertia! force of order one
indicates that there is a significant perturbation of the gas flow by the
electric body force. It is worth noting that parallel plate flow at these
Reynolds numbers exhibits relatively low values of eddy viscosity and
diffusivity in the absence of corona discharge, (26). Moreover, at these
speeds the flow is just above the Reynolds number at which transition occurs
from laminar to turbulent flow (18). Two conclusions may be drawn from this.
First, mechanical devices in the precipitator inlet are relatively
ineffective in smoothing out nonuniform inlet velocity profiles. And
second, particle transport by electrically generated secondary flows and
turbulence may be more significant than the transport due to aerodynamic
sources of turbulence.
32-5

-------
III. TURBULENCE
The interaction of the electric body force with turbulent flow is
described by the pertinent turbulence equations. The first step in the
analysis is to decompose the instantaneous flow variables into their mean
and fluctuating components:
u-j = u-j + u'	(2.14)
i
p = p + p'
and
where mean and fluctuating values are denoted by an overbar and a prime re-
spectively. Since EHD theory assumes that the instantaneous electric body
force and current density are completely independent of the fluid flow, the
source of a fluctuating current density must be electric in origin. With DC
energization there are a number of possible causes of current density fluc-
tuations. Among the most obvious of these are Trichel pulses and the insta-
bilities often noted with negative corona discharge. Although positive
corona discharge on a wire may produce a uniform discharge, negative corona
is usually characterized by a series of localized discharge tufts which in
some circumstances may move about the wire. Sparkover is another source of
current fluctuations as are vibrations and galloping motions of the dis-
charge wires themselves. With pulsed energization, there is obviously a
substantial current density fluctuation which cannot be neglected. The tur-
bulence equations to follow are developed assuming that fluctuations in the
current density exist.
Following the normal procedure, the decomposed quantities of equation
(2.14) are substituted into the continuity (2.5) and Navier-Stokes equations
(2.6) and time averaged to yield the classical Reynolds equations for the
mean motion:
3Uj
3x]
and
dp
aui . Bu,
p[ + U-j J = -
at J 8x. 3x.
J	i
= 0
a^u*
nj-s	a	1
+ u 	 -pi (im - -X.
SXj2	1 j 8
(2.15)
(2.16)
From the previous discussion, it can be seen that a mean current density
with a non-zero rotational component will perturb the mean velocity field.
The earlier discussion on the significance of geometry in determining
whether or not is rotational also apply to the mean current density J-j.
32-6

-------
An electrical modification of the mean flow is normally referred to as
a corona wind. The magnitude of the corona wind will depend on the
magnitude of a mean electric Froude number appropriately defined. There is
also a more subtle effect on the mean flow due to changes in the Reynolds
stress u'u1. The equation governing the Reynolds stress contains the
1 J 1
additional term - - [u|J' + u'J'] which shows that current fluctuations
3 i J J i
which are correlated with velocity fluctuations contribute to the Reynolds
stresses acting on the gas. Thus turbulence models which assume the
Reynolds stress to be only aerodynamic in origin may not be applicable,
particularly with a pulsed discharge.
A final insight into the role of the electric body force in generating
turbulence may be gained by an examination of the energy balance equation
for the mean and fluctuating motion. The mean kinetic energy per unit
volume E = 1/2 pu^ is obtained by multiplying equation (2.16) by u,-.
m	i
result in divergence form is
3Em 3 - 		 - -	9ui	8uj
 +  (Em ui + p u'u' Ui + p Ui - u Ui	( +	))
3t 3x. m J	i j 1 K J	1 dx.	3x/J
J	\J	'
y ,3ui	3uj\p -ir 3Ui	1	i-
= -  (; + -r + p u u' -	- -	u-j 37.
2 8Xj	ax. i j 3xj	0 1 1
(2.17)
Each term in this equation represents a physical mechanism which affects the
balance of mean flow kinetic energy. The divergence terms on the left hand
side reflect the usual fluxes of energy. The terms on the right hand side
 3^i
are the viscous decay term, the transfer term p u'u1  which represents
i j 3x.
the energy transfer from the mean motion to the turbulence, and a term
describing the interaction between the mean velocity and mean current
density. The transfer term is almost always negative indicating that mean
flow kinetic energy is continually being transformed into turbulent kinetic
energy. The interaction term - ITJ- represents the rate at which the mean
electric body force contributes energy to the mean velocity field. Thus the
energy provided by the mean electric field to the mean flow is ultimately
made available to the turbulence via the transfer term.
The equation governing the fluctuating kinetic energy per unit volume
E^ = 1/2 p uj2 is obtained by subtracting the Reynolds equation (2.16) from
the instantaneous form of the Navier-Stokes equation (2.8) and then
multiplying the resultant expression by u^' and averaging. The turbulent
32-7

-------
energy balance is then
at
a . - 1 , . .
+ r (Etui + ; p u-u -u  +
3x. 0 ' 2 i j j
v
2
(2.18)
Once again each term with the exception of the electric body force
1 , ,
contribution - u J may be given its usual interpretation. The divergence
3 i i
terms on the left are energy fluxes. The right hand side terms are the
fluctuating viscous dissipation, the transfer term describing the exchange
of energy between the mean and fluctuating motion, and the interaction term
1 , ,
- u J which represents the mean rate at which the fluctuating electric
3 i i
body force contributes energy to the turbulence. Since the current density
vector is directed primarily towards the collection plates, it can be shown
(18) that the energy transfer is to the transverse velocity fluctuations
which transport particles away from the collection plates.
It has been shown that the degree to which a precipitator gas flow is
modified by corona discharge is completely determined by the current
density. If the current density has a rotational component there will be
electrical vorticity production in the gas resulting in secondary flows and
turbulence. In turbulent flow, the mean and fluctuating components of cur-
rent density cause indirect and direct energy transfer to the turbulence
from the electric field. With DC energization it does not appear that cur-
rent density fluctuations are significant except possibly during sparkover.
In pulsed energization, however, this mechanism is likely to be a more sig-
nificant source of turbulence.
Since turbulence and secondary flows in precipitators cause particle
transport away from the collection plates, these effects are detrimental to
the particle collection process and must be avoided. The theory demon-
strates that the electrode geometry and discharge structure are the critical
factors controlling flow perturbations of all types. In three geometries it
is possible to avoid these disturbances entirely with a uniform discharge.
In the wire-plate geometry the current density always perturbs the flow, but
there is presumably an electrode design which delivers a desired total
current while causing minimal flow disturbances. An exciting prospect is a
planar electrode design to replace the wires. Obtaining a uniform discharge
with this geometry is an area of current research.
IV. SUMMARY
32-8

-------
REFERENCES
Williams, J. C., and R. Jackson. The Motion of Solid Particles in an
Electrostatic Precipitator, Third Congress of the European Federation
of Chemical Engineering, Symposium on the Interaction Between Fluids
and Particles, 282-288 (1962).
Robinson, M. A Modified Deutsch Efficiency Equation for Electrostatic
Precipitation, Atmos. Environment 1, 193-204 (1967).
Robinson, M. Turbulent Gas Flow and Electrostatic Precipitation, J. of
Air Pollut. Control Assn. 18(4), 235-239 (1968).
Cooperman, P. A New Theory of Precipitator Efficiency, Atmos. Environ-
ment 5, 541-551 (1971).
Soo, S. L., and L. W. Rodgers. Further Studies of the Electro-
aerodynamic Precipitator, Powder Technology 5, 43-49 (1971).
Feldman, P. L., K. S. Kumar, and G. D. Cooperman. Turbulent Diffusion
in Electrostatic Precipitators, AIChE Symposium Series 165, 73, 129-130
Pyle, B. E., D. H. Pontius, T. R. Snyder, and L. E. Sparks. Particle
Trajectory Studies in a Scale Model Electrostatic Precipitator, 73rd
Annual Meeting of the Air Pollution Control Assn., 80-49.2 (1980).
Pyle, B. E., J. R. McDonald, W. B. Smith, and L. E. Sparks. The Rela-
tionship Between Gas Steam Turbulence and Collection Efficiency in a
Lab-Scaled Electrostatic Precipitator, Third Symposium on the Transfer
and Utilization of Particulate Control Technology, Orlando, Fla.
Leonard, G. L., M. Mitchner, and S. A. Self. Particle Transport in
Electrostatic Precipitators, Atmos. Environment 14, 1289-1299 (1980).
Leonard, G. L., M. Mitchner, and S. A. Self. Precipitation from
Turbulent Flows, Proc. Intl. Conf. on Electrostatic Precipitation,
Monterey, California, 208-256 (1981).
Leonard, G. L., M. Mitchner, and S. A. Self. Experimental Study of the
Effect of Turbulent Diffusion on Precipitator Efficiency, J. of Aerosol
Science 13(4), 271-284 (1982).
Ramadan, 0. E., and S. L. Soo. Electrohydrodynamic Secondary Flow.
Phys. Fluids 12, 1943 (1969).
Yamamoto, T., S. Nakamura, and H. R. Velkoff. Numerical Study of
Secondary Flow Interaction in an Electrostatic Precipitator, in
Innovative Numerical Analysis for the Engineering Science. University
Press of Virginia, 3-12 (1980).
32-9

-------
14.	Robinson, M. Effects of the Corona Discharge on Electric-Wind Convec-
tion and Eddy Diffusion in an Electrostatic Precipitator, ERDA-HASL-301
(1976).
15.	Ushimaru, K., G. W. Butler, and P. Milovsoroff. An Experimental and
Theoretical Investigation of Electrostatic Precipitators, Flow Research
Report 219, Flow Industries, Kent, Washington (1982).
16.	Leonard, G. L., M. Mitchner, and S. A. Self. An Experimental Study of
the Electrohydrodynamic Flow in Electrostatic Precipitators, J. of
Fluid Mech. 127, 123-140 (1983).
17.	Kumaran, A. R. Aspects of the Fluid Mechanics of Electrostatic Precip-
itators, HTGL Report No. 15-83-TR (1983).
18.	Davidson, J. H. Secondary Flows and Turbulence in Electrostatic Pre-
cipitators, Ph.D. Dissertation, Duke University, Durham, N.C. (1984).
19.	Flippen, L. D. Electrohydrodynamics, Ph.D. Dissertation, Duke
University, Durham, N.C. (1982).
20.	Hay, J. C. The Electric Field and Vorticity Production in a Wire-Plate
Precipitator with Uniform Discharge, M.S. Thesis, Duke University,
Durham, N.C. (1984).
21.	Leutert, G., and B. Bohlen. The Spatial Trend of Electric Field
Strength and Space Charge Density in Plate-Type Electrostatic Precipi-
tators, Staub-Reinholt. Luft 32(7), 27-33 (1972).
22.	McDonald, J. R., W. B. Smith, and H. W. Spencer III. A Mathematical
Model for Calculating Electrical Conditions in Wire-Duct Electrostatic
Precipitation Devices, J. of Applied Physics 48(6), 2231-2243 (1977).
23.	Lawless, P. A., and L. E. Sparks. A Mathematical Model for Calculating
Effects of Back Corona in Wire-Duct Electrostatic Precipitators, J. of
Applied Physics 51(1), 242-256 (1980).
24.	Yamamoto, T., and H. R. Velkoff. Electrohydrodynamics in an Electro-
static Precipitator, J. of Fluid Mech. 108, 1-8 (1981).
25.	Hoburg, J. F., and J. L. Davis. Wire-Duct Precipitator Field and
Charge Computation Using Finite Element and Characteristics Methods, J.
of Electrostatics 14(2), 187-199 (1983).
26.	Page, F., W. G. Schlinger, D. K. Breaux, and B. H. Sage. Point Values
of Eddy Conductivity and Viscosity in Uniform Flow Between Parallel
Plates, Indus, and Engineering Chem. 44(2), 424-430 (1952).
32-10

-------