c/EPA
United States
Environmental Protection
Agency
Industrial Environmental Research
Laboratory
Research Triangle Park NC 2771 1
EPA-600 9-82-OOBc
July 1982
Research and Development
Third Symposium on the
Transfer and
Utilization of Participate
Control Technology:
Volume III. Particulate
Control Devices
-------
EPA-600/9-82-005C
THIRD SYMPOSIUM ON THE
TRANSFER AND UTILIZATION OF
PARTICULATE CONTROL TECHNOLOGY
VOLUME III. PARTICULATE CONTROL DEVICES
Compiled by:
F.P. Venditti, J.A. Armstrong, and M. Durham
Denver Research Institute
P.O. Box 10127
Denver, Colorado 80208
Grant Number: R805725
Project Officer
Dale L. Harmon
Office of Environmental Engineering and Technology
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
Prepared for:
INDUSTRIAL ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NC 27711
-------
DISCLAIMER
This report has been reviewed by the Industrial Environmental
Research Laboratory-Research Triangle Park, North Carolina, Office of
Research and Development, U.S. Environmental Protection Agency, and
approved for publication. Approval does not signify that the contents
necessarily reflect the views and policies of the U.S. Environmental
Protection Agency, nor does mention of trade names or commercial products
constitute endorsement or recommendation for use.
ii
-------
ABSTRACT
The papers in these four volumes of Proceedings were presented at the
Third Symposium on the Transfer and Utilization of Particulate Control
Technology held in Orlando, Florida during 9 March through 13 March 1981,
sponsored by the Particulate Technology Branch of the Industrial Environ-
mental Research Laboratory of the Environmental Protection Agency and
coordinated by the Denver Research Institute of the University of Denver.
The purpose of the symposium was to bring together researchers,
manufacturers, users, government agencies, educators and students to
discuss new technology and to provide an effective means for the transfer
of this technology out of the laboratories and into the hands of the users.
The three major categories of control technologies — electrostatic
precipitators, scrubbers, and fabric filters — were the major concern of the
symposium. These technologies were discussed from the perspectives of
economics; new technical advancements in science and engineering; and
applications. Several papers dealt with combinations of devices and
technologies, leading to a concept of using a systems approach to partic-
ulate control rather than device control. Additional topic areas included
novel control devices, high temperature/high pressure applications, fugitive
emissions, and measurement techniques.
These proceedings are divided into four volumes, each volume contain-
ing a set of related session topics to provide easy access to a unified
technology area.
ill
-------
VOLUME III
VOLUME I. COAL FIRED BOILERS—CONTENTS x
VOLUME II. ELECTROSTATIC PRECIPITATORS—CONTENTS ... xiv
VOLUME IV. ATYPICAL APPLICATIONS—CONTENTS xviii
Section A - Scrubbers
THE CALVERT SCRUBBER 1
S. Calvert, R.G. Patterson and S. Yung
FLUX FORCE/CONDENSATION SCRUBBER SYSTEM
FOR COLLECTION OF FINE PARTICULATE EMISSIONS
FROM AN IRON MELTING CUPOLA 10
S. Calvert and D.L. Harmon
DEMONSTRATION OF HIGH-INTENSITY-IONIZER-ENHANCED
VENTURI SCRUBBER ON A MAGNESIUM RECOVERY
FURNACE FUME EMISSIONS 21
A. Prem, M.T. Kearns and D.L. Harmon
A NEW ENTRY IN THE HIGH EFFICIENCY SCRUBBER FIELD .... 33
L.C. Hardison and F. Ekman
PERFORMANCE OF PARTICULATE SCRUBBERS AS
INFLUENCED BY GAS-LIQUID CONTACTOR DESIGN
AND BY DUST FLOCCULATION 43
K.T. Semrau and R.J. Lunn
INVESTIGATION OF VENTURI SCRUBBER EFFICIENCY
AND PRESSURE DROP 51
R. Parker, T. Le and S. Calvert
SCRUBBER TECHNOLOGY AND THE INTERACTION OF
A UNIQUE STRUCTURE AS MIST ELIMINATOR 60
G.C. Pedersen
NOVEL ANNULAR VENTURI SCRUBBER DESIGN REDUCES
WASTE DISCHARGE PROBLEMS 71
H.P. Beutner
CONSIDERATION OF THE PERTINENT DESIGN AND
OPERATING CHARACTERISTICS ESSENTIAL FOR
OPTIMIZATION OF VENTURI SCRUBBER PERFORMANCE 80
H.S. Oglesby
V
-------
VOLUME III CONTENTS (cont.)
Page
APPLICATION OF SCRUBBERS FOR PARTICULATE
CONTROL OF INDUSTRIAL BOILERS 90
M. Borenstein
APPLICATION OF HIGH ENERGY VENTURI SCRUBBERS
TO SEWAGE INCINERATION 102
F.X. Reardon
AN INCINERATOR SCRUBBER THAT WORKS:
A CASE STUDY Ill
C. Menoher
EVALUATION OF ENTRAINED LIQUOR CONTRIBUTION TO
TOTAL MASS EMISSIONS DOWNSTREAM OF A WET SCRUBBER . . .119
W. David Balfour, L.O. Edwards and H.J. Williamson
Section B - Fabric Filters
A DUAL-BEAM BACKSCATTER BETA-PARTICLE GAUGE
FOR MEASURING THE DUST CAKE THICKNESS ON OPERATING
BAG FILTERS INDEPENDENT OF POSITION 128
R.P. Gardner, R.P. Donovan and L.S. Hovis
DIAGNOSING FILTER FABRIC CAPABILITIES WITH LIGHT
SCATTERING AND NUCLEI DETECTING INSTRUMENTATION . . . .140
R. Dennis, D.V. Bubenick and L.S. Hovis
ACID DEWPOINT CORROSION IN PARTICULATE
CONTROL EQUIPMENT 150
T.E. Mappes, R.D. Terns and K.E. Foster
SECOND GENERATION OF EMISSIONS CONTROL
SYSTEM FOR COKE OVENS 160
J.D. Patton
EFFECTS OF FLYASH SIZE DISTRIBUTION ON THE
PERFORMANCE OF A FIBERGLASS FILTER 171
W.F. Frazier and W.T. Davis
FUNDAMENTAL STUDY OF A FABRIC FILTER
WITH A CORONA PRECHARGER 181
K. linoya and Y. Mori
ECONOMIC EVALUATION FACTORS IN BID
EVALUATIONS--A SENSITIVITY ANALYSIS 193
J.G. Musgrove and J.E. Shellabarger
VI
-------
VOLUME III CONTENTS (cont.)
Page
FLY ASH RE-ENTRAINMENT IN A BAGHOUSE--
WHAT DOES IT COST? 201
J.G. Musgrove
WHY PERFORM MODEL STUDY OF FABRIC FILTER
COLLECTOR? 211
W.T. Langan, N.Z. Shilling, W.A. Van Kleunen and O.F. Fortune
EXPERIENCES OF A SMALL INSULATION MANUFACTURER
IN MAINTAINING COMPLIANCE WITH AIR POLLUTION
CONTROL REGULATIONS 221
R.L. Hawks
ADVANCED FABRIC FILTER TECHNOLOGY FOR
DIFFICULT PARTICULATE EMISSIONS 228
H.P. Beutner
DEVELOPMENT OF GUIDELINES FOR OPTIMUM BAGHOUSE
FLUID DYNAMIC SYSTEM DESIGN 238
D. Eskinazi, G.B. Gilbert and R.C. Carr
THEORETICAL ASPECTS OF PRESSURE DROP REDUCTION
IN A FABRIC FILTER WITH CHARGED PARTICLES 250
T. Chiang, E.A. Samuel and K.E. Wolpert
EXPERIMENTAL CORRELATION OF DUST CAKE POROSITY,
AIR-TO-CLOTH RATIO AND PARTICLE-SIZE DISTRIBUTIONS . . . .261
T. Chiang and R.L. Ostop
MODEL FOR DUST PENETRATION THROUGH A
PULSE-JET FABRIC FILTER 270
D. Leith and M.J. Ellenbecker
PERFORMANCES OF DUST LOADED AIR FILTERS 280
C. Kanaoka, H. Emi and M. Ohta
ELECTROSTATICALLY ENHANCED FABRIC
FILTRATION OF PARTICULATES 290
T. Ariman and S.T. McComas
A STAGGERED ARRAY MODEL OF A FIBROUS FILTER
WITH ELECTRICAL ENHANCEMENT 301
F. Henry and T. Ariman
vii
-------
VOLUME III CONTENTS (cont.)
Page
Section C - Granular Beds
AEROSOL FILTRATION BY A COCURRENT MOVING
GRANULAR BED: PENETRATION THEORY 311
T.W. Kalinowski and D. Leith
FUNDAMENTAL EXPERIMENTS ON A GRANULAR BED FILTER . . - .321
K. linoya and Y. Mori
DRY DUST COLLECTION OF BLAST FURNACE
EXHAUST GAS BY MOVING GRANULAR BED FILTER 332
A. Wakabayashi, T. Sugawara and S. Watanabe
Section D - Novel Devices
IRON AND STEEL AIR POLLUTION CONTROL
USING MAGNETIC SEPARATION 341
D.C. Drehmel, C.E. Ball and C.H. Gooding
TECHNICAL AND ECONOMIC EVALUATION OF TWO
NOVEL PARTICULATE CONTROL DEVICES 353
R.R. Boericke, J.T. Kuo and K.R. Murphy
TM
THE ELECTROSCRUBBER111 FILTER—APPLICATIONS
AND PARTICULATE COLLECTION PERFORMANCE 363
D. Parquet
HIGH EFFICIENCY PARTICULATE REMOVAL WITH
SINTERED METAL FILTERS 373
B.E. Kirstein, W.J. Paplawsky, D.T. Pence and T.G. Hedahl
APPLICATION OF ELECTROSTATIC TECHNIQUES TO
THE REMOVAL OF DUST AND FUME FROM THE
INDUSTRIAL ENVIRONMENT 382
S.A. Hoenig
THE DRY VENTURI 393
A.J. Teller and D.R.J. Roy
FIBER BED FILTER SYSTEM CONTROL OF
WELDING PARTICULATES 393
J.A. Bamberger and W.K. Winegardner
THE USE OF GLASS CAPILLARY FILTERS TO
CLASSIFY ACTINOLITE FIBERS 406
J.W. Gentry, T.C. Chen, S.W. Lin and P.Y. Yu
vlii
-------
VOLUME III CONTENTS (cont.)
Page
ULTRA-HIGH EFFICIENCY FILTRATION SYSTEMS
(AIR RECIRCULATION) 417
R.W. Potokar
THE WET WALL ELECTROSTATIC PRECIPITATOR 428
J. Starke, J. Kautz and K-R. Hegemann
Section E - Mechanical Collectors
TROUBLESHOOTING MULTIPLE CYCLONES ON
FUEL-OIL-FIRED BOILERS 438
F. Crowson and R.L. Gibbs
COLLECTION EFFICIENCIES OF CYCLONE SEPARATORS 449
P.W. Dietz
ELECTROSTATICALLY AUGMENTED COLLECTION
IN VORTICAL FLOWS 459
P.W. Dietz
HIGH PERFORMANCE CYCLONE DEVELOPMENT 468
W.G. Giles
AUTHOR INDEX 481
IX
-------
VOLUME I
COAL FIRED BOILERS
Section A - Fabric Filters
Page
COAL PROPERTIES AND FLY ASH FILTERABILITY 1
R. Dennis, J.A. Dirgo and L.S. Hovis
PULSE-JET FILTRATION WITH ELECTRICALLY
CHARGED FLYASH 11
R.P. Donovan, L.S. Hovis, G.H. Ramsey and J.H. Abbott
ELECTRICALLY CHARGED FLYASH EXPERIMENTS IN A
LABORATORY SHAKER BAGHOUSE 23
L.S. Hovis, J.H. Abbott, R.P. Donovan and C.A. Pareja
ELECTROSTATIC AUGMENTATION OF FABRIC FILTRATION .... 35
D.W. VanOsdell, G.P. Greiner, G.E.R. Lamb and L.S. Hovis
FABRIC WEAR STUDIES AT HARRINGTON STATION 45
R. Chambers, K. Ladd, S. Kunka and D. Harmon
SPS PILOT BAGHOUSE OPERATION 55
K. Ladd, W. Hooks, S. Kunka and D. Harmon
REVIEW OF SPS INVESTIGATION OF HARRINGTON STATION
UNIT 2 FABRIC FILTER SYSTEM 65
K. Ladd, S. Kunka
A SUMMARY OF PERFORMANCE TESTING OF THE APITRON
ELECTROSTATICALLY AUGMENTED FABRIC FILTER 75
D. Helfritch and L. Kirsten
FABRIC FILTER OPERATING EXPERIENCE FROM SEVERAL
MAJOR UTILITY UNITS 82
O.F. Fortune, R.L. Miller and E.A. Samuel
EVALUATION OF THE 25 MW KRAMER STATION BAGHOUSE:
TRACE ELEMENT EMISSION CONTROL 94
M.W. McElroy and R.C. Carr
CHARACTERIZATION OF A 10 MW FABRIC FILTER
PILOT PLANT 95
W.B. Smith, K.M. Gushing and R.C. Carr
SPECIFYING A FABRIC FILTER SYSTEM 107
R.L. Ostop and D.A. Single
X
-------
VOLUME I CONTENTS (cont.)
Page
EVALUATION OF THE 25 MW KRAMER STATION BAGHOUSE:
OPERATIONAL FACTORS IN PARTICULATE MATTER
EMISSION CONTROL 118
R.C. Carr and M.W. McElroy
PULSE-JET TYPE FABRIC FILTER EXPERIENCE AT AIR TO
CLOTH RATIOS OF 5 TO 1 ON A BOILER FIRING PULVERIZED
COAL 120
G.L. Pearson
SELECTION AND OPERATION OF BAGHOUSES AT R.D. NIXON
STATION, UNIT #1 129
R.C. Hyde, J. Arello and D.J. Huber
POTENTIAL FOR IMPROVEMENT IN BAGHOUSE DESIGN 138
R.M. Jensen
REVIEW OF OPERATING AND MAINTENANCE EXPERIENCES WITH
HIGH TEMPERATURE FILTER MEDIA ON COAL-FIRED BOILERS . . .148
L.K. Crippen
Section B - Electrostatic Precipitators
PILOT DEMONSTRATION OF THE PRECHARGER-COLLECTOR
SYSTEM 157
P. Vann Bush, Duane H. Pontius
REMEDIAL TREATMENTS FOR DETERIORATED HOT SIDE
PRECIPITATOR PERFORMANCE 165
R.E. Bickelhaupt
EVALUATION OF THE UNITED McGILL ELECTROSTATIC
PRECIPITATOR 176
D.S. Ensor, P.A. Lawless, A.S. Damle
PREDICTING THE EFFECT OF PROPRIETARY CONDITIONING
AGENTS ON FLY ASH RESISTIVITY 185
R.J. Jaworowski and J.J. Lavin
S03 CONDITIONING TO ENABLE ELECTROSTATIC
PRECIPITATORS TO MEET DESIGN EFFICIENCIES 197
J.J. Ferrigan, III
ENHANCED PRECIPITATOR COLLECTION EFFICIENCIES
THROUGH RESISTIVITY MODIFICATION 206
D.F. Mahoney
XI
-------
VOLUME I CONTENTS (cont.)
Page
DEVELOPMENT OF A NEW SULFUR TYPE ASH CONDITIONING . . . .216
R.H. Gaunt
OPERATING EXPERIENCE WITH FLUE GAS CONDITIONING
SYSTEMS AT COMMONWEALTH EDISON COMPANY 226
L.L. Weyers and R.E. Cook
THE APPLICATION OF A TUBULAR WET ELECTROSTATIC
PRECIPITATOR FOR FINE PARTICULATE CONTROL AND
DEMISTING IN AN INTEGRATED FLY ASH AND SO2 REMOVAL
SYSTEM ON COAL-FIRED BOILERS 236
E. Bakke and H.P. Willett
FIELD EVALUATIONS OF AMMONIUM SULFATE CONDITIONING
FOR IMPROVEMENT OF COLD SIDE ELECTROSTATIC PRECIPITATOR
PERFORMANCE 237
E.G. Landham, Jr., G.H. Merchant, Jr., J.P. Gooch and
R.F. Altman
EVALUATION OF PERFORMANCE ENHANCEMENT OBTAINED
WITH PULSE ENERGIZATION SYSTEMS ON A HOT SIDE
ELECTROSTATIC PRECIPITATOR 253
W. Piulle, L.E. Sparks, G.H. Marchant, Jr. and J.P. Gooch
A NEW MICROCOMPUTER AND STRATEGY FOR THE CONTROL
OF ELECTROSTATIC PRECIPITATORS 265
K.J. McLean, T.S. Ng, Z. Herceg and Z. Rana
ASSESSMENT OF THE COMMERCIAL POTENTIAL FOR THE HIGH
INTENSITY IONIZER IN THE ELECTRIC UTILITY INDUSTRY . . . .272
J.S. Lagarias, J.R. McDonald and D.V. Giovanni
APPLICATION OF ENERGY CONSERVING PULSE ENERGIZATION
FOR PRECIPITATORS-PRACTICAL AND ECONOMIC ASPECTS . . . .291
H. H. Petersen and P. Lausen
Section C - Dry SQ? Scrubbers
SO2 REMOVAL BY DRY INJECTION AND SPRAY ABSORPTION
TECHNIQUES 393
E.L. Parsons, Jr., V. Boscak, T.G. Brna and R.L. Ostop
DRY SCRUBBING SO2 AND PARTICULATE CONTROL 313
N.J. Stevens, G.B. Manavizadeh, G.W. Taylor and M.J. Widico
FIBER AND FABRIC ASPECTS FOR SO2 DRY SCRUBBING
BAGHOUSE SYSTEMS 323
L. Bergmann
xii
-------
VOLUME I CONTENTS (cont.)
Page
TWO-STAGE DRY FLUE GAS CLEANING USING CALCIUM
ALKALIS 333
D.C. Gehri, D.F. Dustin and S.J. Stachura
CONTROL OF SULFUR DIOXIDE, CHLORINE, AND TRACE
ELEMENT EMISSIONS FROM COAL-FIRED BOILERS BY FABRIC
FILTRATION 341
R.J. Demski, J.T. Yeh and J.I. Joubert
Section D - Scrubbers
FLYASH COLLECTION USING A VENTURI SCRUBBER—MINNESOTA
POWER'S COMMERCIAL OPERATING EXPERIENCE 352
C.A. Johnson
AUTHOR INDEX 361
xiii
-------
VOLUME II
ELECTROSTATIC PRECIPITATORS
Section A - Fundamentals
Page
MATHEMATICAL MODELING OF IONIC
CONDUCTION IN FLY ASH LAYERS 1
R.B. Mosley, J.R. McDonald and L.E. Sparks
MEASUREMENTS OF ELECTRICAL PROPERTIES
OF FLY ASH LAYERS 13
R.B. Mosley, P.R. Cavanaugh, J.R. McDonald and L.E. Sparks
LASER DOPPLER ANEMOMETER MEASUREMENTS OF PARTICLE
VELOCITY IN A LABORATORY PRECIPITATOR 25
P.A. Lawless, A.S. Damle, A.S. Viner, E.J. Shaughnessy and
L.E. Sparks
PROGRESS IN MODELING BACK CORONA 35
P.A. Lawless
A COMPUTER MODEL FOR ESP PERFORMANCE 44
P.A. Lawless, J.W. Dunn and L.E. Sparks
MEASUREMENT AND INTERPRETATION OF CURRENT
DENSITY DISTRIBUTION AND CHARGE/MASS DATA 54
M. Durham, G. Rinard, D. Rugg and L.E. Sparks
THE RELATIONSHIP BETWEEN GAS STREAM TURBULENCE
AND COLLECTION EFFICIENCY IN A LAB-SCALED
ELECTROSTATIC PRECIPITATOR 66
B.E. Pyle. J.R. McDonald, W.B. Smith
PARTICLE DEPOSITION PROFILES AND REENTRAINMENT
IN A WIRE-PLATE ELECTROSTATIC PRECIPITATOR 76
E. Arce-Medina and R.M. Felder
PARTICLE TRANSPORT IN THE EHD FIELD 87
T. Yamamoto
SURFACE REENTRAINMENT OF COLLECTED FLY ASH IN
ELECTROSTATIC PRECIPITATORS 97
M. Mitchner, M.J. Fisher, D.S. Gere, R.N. Leach and S.A. Self
ELECTROMECHANICS OF PRECIPITATED ASH LAYERS . IQQ
G.B. Moslehi and S.A. Self ' ' '
xiv
-------
VOLUME II CONTENTS (cont.)
Page
EXPERIMENTAL MEASUREMENTS OF THE EFFECT OF
TURBULENT DIFFUSION ON PRECIPITATOR EFFICIENCY 120
G.L. Leonard, M. Mitchner and S.A. Self
CAN REENTRAINMENT BE EXPLAINED USING A NEW
PRECIPITATOR FORMULA? 130
S. Maartmann
A LABORATORY FURNACE FOR THE PRODUCTION OF
SYNTHETIC FLY ASH FROM SMALL COAL SAMPLES 141
K.M. Sullivan
COMPUTER SIMULATION OF THE WIDE PLATE
SPACING EFFECT 149
E.A. Samuel
SIMULTANEOUS MEASUREMENTS OF AERODYNAMIC SIZE
AND ELECTRIC CHARGE OF AEROSOL PARTICLES IN REAL
TIME ON A SINGLE PARTICLE BASIS 160
M.K. Mazumder, R.G. Renninger, T.H. Chang,
R.W. Raible, W.G. Hood, R.E. Ware and R.A. Sims
APPLICATION OF LASER DOPPLER INSTRUMENTATION TO
PARTICLE TRANSPORT MEASUREMENTS IN AN ELECTROSTATIC
PRECIPITATOR 169
M.K. Mazumder, W.T. Clark III, R.E. Ware, P.C. McLeod,
W.G. Hood, J.E. Straub and S. Wanchoo
THE APPLICATION OF MEASUREMENTS OF AEROSOL
CHARGE ACQUISITION BY BIPOLAR IONS TO THE PROBLEM
OF BACK CORONA 179
R.A. Fjeld, R.O. Gauntt, G.J. Laughlin and A.R. McFarland
IDENTIFICATION OF BACK DISCHARGE SEVERITY 189
S. Masuda and Y. Nonogaki
Section B - Operations and Maintenance
MODELING OF ELECTROSTATIC PRECIPITATORS WITH RESPECT
TO RAPPING REENTRAINMENT AND OUTLET OPACITY 199
M.G. Faulkner, W.E. Farthing, J.R. McDonald and L.E. Sparks
NEW PRECIPITATOR TECHNOLOGY FOR PARTICULATE
CONTROL 208
J.R. Zarfoss
XV
-------
VOLUME II CONTENTS (cont.)
Page
AN APPLICATION SUMMARY OF HIGH ENERGY SONIC
CLEANING APPLIED TO ELECTROSTATIC PRECIPITATORS 218
M.J. Berlant
THE IMPACT OF INTELLIGENT PRECIPITATOR CONTROLS 230
N.Z. Shilling, R.O. Reese and J.A. Fackler
AN ENERGY MANAGEMENT SYSTEM FOR
ELECTROSTATIC PRECIPITATORS 242
R.R. Crynack and M.P. Downey
RELATIONSHIP BETWEEN ELECTROSTATIC PRECIPITATOR
PERFORMANCE AND RECORDKEEPING PRACTICES 252
S.P. Schliesser
AN OPERATION AND MAINTENANCE PROGRAM FOR
A PHOSPHATE ROCK ELECTROSTATIC PRECIPITATOR 262
D.B. Rimberg
Section C - Advanced Design
ELECTROSTATIC PRECIPITATOR PERFORMANCE
WITH PULSE EXCITATION 273
D. Rugg, M. Durham, G. Rinard and L.E. Sparks
DEVELOPMENT OF A CHARGING DEVICE FOR HIGH-RESISTIVITY
DUST USING HEATED AND COOLED ELECTRODES 283
G. Rinard, M. Durham, D. Rugg and L.E. Sparks
THE EVALUATION OF NOVEL ELECTROSTATIC PRECIPITATOR
SYSTEMS USING A TRANSPORTABLE PROTOTYPE 295
G. Rinard, M. Durham, D. Rugg, J. Armstrong,
L.E. Sparks and J.H. Abbott
ANALYSIS OF THE ELECTRICAL AND CHARGING
CHARACTERISTICS OF A THREE ELECTRODE PRECHARGER . . . .304
K.J. McLean
PARTICLE CHARGING IN AN ELECTROSTATIC
PRECIPITATOR BY PULSE AND DC VOLTAGES 314
L.E. Sparks, G.H. Ramsey, R.E. Valentine and J.H. Abbott
PARTICLE COLLECTION IN A TWO STAGE ELECTROSTATIC
PRECIPITATOR WITH VARIOUS COLLECTOR STAGES 326
L.E. Sparks, G.H. Ramsey, R.E. Valentine and J.H. Abbott
HIGH INTENSITY IONIZER DEVELOPMENT 334
M.H. Anderson, J.R. McDonald, J.P. Gooch and D.V. Giovanni
xv i
-------
VOLUME II CONTENTS (cont.)
DEMONSTRATION OF AIR POLLUTION SYSTEMS HIGH
INTENSITY IONIZER/ELECTROSTATIC PRECIPITATOR ON
AN OIL-FIRED BOILER 349
G.A. Raemhild, A. Prem and F. Weisz
PRIMARY AND SECONDARY IONIZATION IN AN
ELECTRON BEAM PRECIPITATOR SYSTEM 358
W.C. Finney, L.C. Thanh, J.S. Clements and R.H. Davis
INFLUENCE ON PARTICLE CHARGING OF ELECTRICAL
PARAMETERS AT DC AND PULSE VOLTAGES 370
H.J. Joergensen, J.T. Kristiansen and P. Lausen
BOXER-CHARGER MARK III AND ITS
APPLICATION IN ESP'S 380
S. Masuda, H. Nakatani and A. Mizuno
THE PERFORMANCE OF AN EXPERIMENTAL
PRECIPITATOR WITH AN ALL-PLATE ZONE 390
J. Dalmon
THE PHYSICS OF PULSE ENERGIZATION OF
ELECTROSTATIC PRECIPITATORS 404
L. Menegozzi and P.L. Feldman
ADVANCED ELECTRODE DESIGN FOR
ELECTROSTATIC PRECIPITATORS 405
S. Bernstein, K. Ushimaru and E.W. Geller
Section D - Industrial Applications
PROBLEMS IN APPLYING AN ELECTROSTATIC
PRECIPITATOR TO A SALVAGE FUEL-FIRED BOILER 415
C.R. Thompson
THE APPLICATION OF ELECTROSTATIC PRECIPITATORS
TO BOILERS FIRING MULTIPLE FUELS 425
R.L. Bump
AUTHOR INDEX 435
xvii
-------
VOLUME IV
ATYPICAL APPLICATIONS
Section A - Advanced Energy Applications
Page
HIGH TEMPERATURE PARTICLE COLLECTION WITH
A.P.T. EPxP DRY SCRUBBER 1
S. Yung, T. Lee, R.C. Patterson, S. Calvert and D.C. Drehmel
PARTICLE COLLECTION IN CYCLONES AT HIGH TEMPERATURE
AND HIGH PRESSURE 2
R. Parker, R. Jain, S. Calvert, D.C. Drehmel and J. Abbott
OPERATING RESULTS OF ELECTROSTATIC PRECIPITATORS
AT HIGH TEMPERATURE AND HIGH PRESSURES 3
P.L. Feldman and K.S. Kumar
CONTROL OF PARTICULATES IN PROCESS AREA 12, SOLVENT
REFINED COAL PROCESS 15
W.H. Wilks, P.D. Wilkinson and J.A. Schlosberg
NON-PLUGGING RETAINING STRUCTURE FOR GRANULAR
BED FILTER FOR HTHP APPLICATION 26
A.M. Presser and J.C. Alexander
PARTICULATE EMISSIONS CONTROL FROM A COAL-FIRED
OPEN-CYCLE MAGNETOHYDRODYNAMICS/STEAM POWER PLANT ... 36
H.H. Wang and T.E. Dowdy
REAL TIME COARSE PARTICLE MASS MEASUREMENTS IN
A HIGH TEMPERATURE AND PRESSURE COAL GASIFIER
PROCESS TREATMENT 46
J. Wegrzyn, J. Saunders and W. Marlow
THE DESIGN, ENGINEERING, AND STARTUP OF A VENTURI
SCRUBBER SYSTEM ON AN OIL SHALE OFF-GAS INCINERATOR ... 55
P.A. Czuchra and J.S. Sterrett
FLUIDIZED-BED COMBUSTION HOT FLUE GAS CLEANUP
PERSPECTIVE ON CYCLONES AND OTHER DEVICES ... 63
R.F. Henry and W.F. Podolski
PRESSURIZED AND NON-PRESSURIZED ACOUSTIC
AGGLOMERATORS FOR HOT-GAS CLEANUP APPLICATIONS 73
K.H. Chou and D.T. Shaw
xviii
-------
VOLUME IV CONTENTS (cont.)
Page
ALKALIS AND THEIR CONTRIBUTIONS TO CORONA CURRENT
AT HIGH TEMPERATURE AND HIGH PRESSURE 74
R.W.L. Snaddon
HOT GAS CLEANUP IN PRESSURIZED FLUIDIZED
BED COMBUSTION 83
L.N. Rubow and M.G. Klett
VENTURI SCRUBBING FOR CONTROL OF PARTICULATE
EMISSIONS FROM OIL SHALE RETORTING 95
G.M. Rinaldi and R.C. Thurnau
OVERVIEW OF THE DEPARTMENT OF ENERGY'S PRESSURIZED
FLUIDIZED-BED COMBUSTOR CLEANUP TECHNOLOGY PROGRAM . . .105
W.E. Moore
THE CYCLOCENTRIFUGE™--AN ADVANCED GAS/SOLIDS
SEPARATOR FOR COAL CONVERSION PROCESSES 116
P.R. Albrecht, J.T. McCabe and W. Fedarko
Section B - Fugitive Emissions
DEMONSTRATION OF THE USE OF CHARGED FOG IN
CONTROLLING FUGITIVE DUST FROM LARGE-SCALE
INDUSTRIAL SOURCES 125
E.T. Brookman, R.C. McCrillis and D.C. Drehmel
THE CONTROL OF FUGITIVE EMISSIONS USING WINDSCREENS . . .135
D. Carnes and D.C. Drehmel
THE INFLUENCE OF AGGREGATE PILE SHAPE AND
ORIENTATION ON PARTICULATE FUGITIVE EMISSIONS 145
D- Martin
SPRAY CHARGING AND TRAPPING SCRUBBER FOR
FUGITIVE PARTICLE EMISSION CONTROL 155
S. Yung, S. Calvert and D.C. Drehmel
IMPROVED STREET SWEEPER FOR CONTROLLING URBAN
INHALABLE PARTICULATE MATTER 156
S. Calvert, H. Brattin, S. Bhutra, R. Parker and D.C. Drehmel
A WIND TUNNEL FOR DUST ENTRAINMENT STUDIES 168
A.S. Viner, M.B. Ranade, E.J. Shaughnessy, D.C. Drehmel
and B.E. Daniels
XIX
-------
VOLUME IV CONTENTS (cont.)
Page
TECHNIQUES AND EQUIPMENT FOR MEASURING INHALABLE
PARTICULATE FUGITIVE EMISSIONS I79
H.J. Kolnsberg
BALLOON SAMPLING TO CHARACTERIZE PARTICLE
EMISSIONS FROM FUGITIVE SOURCES 188
J.A. Armstrong and B.C. Drehmel
AN ELECTROSTATICALLY CHARGED FOG GENERATOR FOR
THE CONTROL OF INHALABLE PARTICLES 200
C.V. Mathai, L.A. Rathbun and D.C. Drehmel
RELATIVE EFFECTIVENESS OF CHEMICAL ADDITIVES
AND WIND SCREENS FOR FUGITIVE DUST CONTROL 210
D.C. Drehmel and B.E. Daniel
PARTICULATE IMPACT COMPARISON BETWEEN CONTROLLED
STACK EMISSIONS FOR A 2000 MW ELECTRICAL GENERATING
STATION 222
H.E. Hesketh and F.L. Cross
OPERATING EXPERIENCE AND THE TECHNIQUES IN THE
CONTROL OF COAL DUST EMISSIONS FROM LARGE
STORAGE PILE AT NANTICOKE TGS 232
N. Krishnamurthy, W. Whitman and Y.V. Nguyen
Section C - Opacity
MODELING SMOKE PLUME OPACITY FROM PARTICULATE
CONTROL EQUIPMENT 242
D.S. Ensor, P.A. Lawless, S.J. Cowen
TETHERED BALLOON PLUME SAMPLING OF A PORTLAND
CEMENT PLANT 252
J.A. Armstrong, P.A. Russell, M.N. Plooster
THE RELATIONSHIP OF FLY ASH LIGHT ABSORPTION TO
SMOKE PLUME OPACITY 264
S.J. Cowen, D.S. Ensor
Section D - Measurements
A SPECIAL METHOD FOR THE ANALYSIS OF
SULFURIC ACID MISTS 275
P. Urone, R.B. Mitchell, J.E. Rusnak, R.A. Lucas and
J.F. Griffiths
XX
-------
VOLUME IV CONTENTS (cent.)
Page
A MICROCOMPUTER-BASED CASCADE-IMPACTOR
DATA-REDUCTION SYSTEM 285
M. Durham, S. Tegtmeyer, K. Wasmundt and I.E. Sparks
DEVELOPMENT OF A SAMPLING TRAIN FOR STACK
MEASUREMENT OF INHALABLE PARTICULATE 297
A.D. Williamson, W.B. Smith
INHALABLE PARTICULATE MATTER SAMPLING
PROGRAM FOR IRON AND STEEL: AN OVERVIEW
PROGRESS REPORT 306
R.C. McCrillis
DEVELOPMENT OF IP EMISSION FACTORS 317
D.L. Harmon
INHALABLE PARTICULATE EMISSION FACTOR PROGRAM
PURPOSE AND DEVELOPMENT 326
F.M. Noonan and J.H. Southerland
INHALABLE PARTICULATE EMISSION FACTORS FOR BLAST
FURNACE CASTHOUSES IN THE IRON AND STEEL INDUSTRY . . . .335
P.D. Spawn, S. Piper and S. Gronberg
INHALABLE PARTICULATE EMISSIONS FROM VEHICLES
TRAVELING ON PAVED ROADS 344
R. Bohn
QUALITY ASSURANCE FOR PARTICLE-SIZING MEASUREMENTS . . .353
C.E. Tatsch
PARTICULATE EMISSIONS CHARACTERIZATION FOR
OIL-FIRED BOILERS 363
D. Mormile, S. Hersh, B.F. Piper and M. McElroy
A CONTINUOUS REAL-TIME PARTICULATE MASS MONITOR
FOR STACK EMISSION APPLICATIONS 373
J.C.F. Wang, H. Patashnick and G. Rupprecht
Section E - Mobile Sources
STUDIES OF PARTICULATE REMOVAL FROM DIESEL EXHAUSTS
WITH ELECTROSTATIC AND ELECTROSTATICALLY-
AUGMENTED TECHNIQUES 383
J.L. DuBard, M.G. Faulkner, J.R. McDonald, D.C. Drehmel
and J.H. Abbott
XXI
-------
VOLUME IV CONTENTS (cont.)
Page
STUDIES OF PARTICULATE REMOVAL FROM DIESEL EXHAUSTS
WITH MECHANICAL TECHNIQUES 395
M.G. Faulkner, J.L. DuBard, J.R. McDonald, D.C. Drehmel
and J.H. Abbott
UPDATE ON STATUS OF CONNECTICUT'S CONTROL PROGRAM
FOR TRANSPORTATION-RELATED PARTICULATE EMISSIONS . . . .406
H.L. Chamberlain and J.H. Gastler
AUTHOR INDEX 413
xxii
-------
THE CALVERT SCRUBBER
BY: S. Calvert and R. G. Patterson
Calvert Environmental Equipment Co.
5191 Santa Fe St.
San Diego, CA 92109
S. Yung
Air Pollution Technology, Inc.
4901 Morena Blvd., Bldg. 400
San Diego, CA 92117
ABSTRACT
TM
The Calvert Scrubber is a new and unique development in fine particle
scrubbing. It is an advanced gas-atomized spray scrubber which can outperform
the conventional Venturi scrubber for both particulate and gaseous pollutant
removal.
The Calvert Scrubber is a very energy efficient fine particle scrubber
and is especially well suited to utilty applications because it can meet
both particulate and sulfur dioxide removal requirements with a single control
device. Other potential applications include metallurgical industries, chem-
cal industries and any other industry where fine particle scrubbers are used.
Pilot scrubbers up to 3,000 CFM capacity have been built and tested.
Experimental data for fine particle collection are presented. Collection ef-
ficiency data are presented as a function of particle aerodynamic diameter.
The 50% cut diameter is presented as a function of pressure drop and energy
input to the scrubber. Performance is compared to conventional Venturi scrub-
bers and other fine particle scrubbers.
-------
INTRODUCTION
This paper presents the results of laboratory and pilot plant studies
for a new type of scrubber, the Calvert Scrubber , which has superior capa-
bility for particle collection and gas absorption. Particle collection effi-
ciency as a function of particle size i s presented for the Calvert Scrubber.
Comparison of the Calvert Scrubber to a Venturi scrubber using the cut power
method shows that the Calvert Scrubber can provide the same fine particle
collection capability as a Venturi scrubber but with only 50% of the power
required by a Venturi scrubber.
CALVERT SCRUBBER
The Calvert Scrubber is a gas atomized spray type scrubber whose patent-
ed design is a significant advancement over the Venturi and similar scrubbers.
Its characteristics are as follows:
1. Its design is simple and rugged, comparable to the conventional
Venturi and orifice type scrubbers.
2. It is capable of plug free operation, since it does not contain any
small orifices or nozzles.
3. It has capability for high efficiency collection of fine particles.
4. It has a lower power requirement than a conventional Venturi or
orifice type scrubber.
5. It has higher mass transfer efficiency than a conventional Venturi
scrubber.
In a gas atomized scrubber the liquid is atomized by a high velocity gas
stream. Particles are collected on the drops by impaction because of the
high relative velocity between the gas and the drops in the throat section.
Gas is also absorbed by the drops before they are separated from the gas
stream in the entrainment separator.
-------
Particle collection efficiency in the throat section increases as the
relative velocity is increased or the drop size is decreased. As the drops
accelerate in the throat section the relative velocity decreases and there-
fore the collection efficiency decreases.
The Calvert Scrubber has the following five zones:
1. Particle conditiong.
2. Primary atomization.
3. Secondary atomization, "Collision Zone".
4. Fog contact.
5. Entrainment separation.
The particle conditioning zone is designed to saturate the gas stream,
remove large particles and provide some particle growth for easier fine
particle collection. Primary atomization and particle collection take place
in two Venturi throat sections as shown in Figure 1. In the primary atomiza-
tion zone liquid is atomized by the high velocity gas stream, particles are
collected and gas is absorbed the same as it would in the Venturi scrubber.
Then the Calvert Scrubber takes a step beyond the Venturi scrubber in
the secondary atomization zone. The streams from the two throats are direct-
ed toward each other so that they collide at extremely high relative velocity.
The head-on collision of the two streams at twice the individual throat velo-
city shreds the drops into finer drops and causes very intensive scrubbing.
The Calvert Scrubber collision zone results in the following advantages over
a Venturi scrubber:
1. High relative velocity causes small drops to be shredded away from
some of the large drops.
2. Drops are also blown into half-bubbles which burst into small frag-
ments.
3. Drops from one throat which collide with drops from the other throat
shatter into small drops which are expelled outward.
4- In the secondary atomization zone, the drops are exposed to the gas
stream at even higher relative velocity than in the primary (throat)
zone. As a result, both the particle collection and gas absorption
rates are abruptly increased to high levels.
-------
5. The throats are subject to much less erosion than in a Venturi
scrubber with equivalent efficiency, because the velocity in the
throats is only half the relative velocity in the collision zone.
6. Pressure loss due to gas friction in the throats is reduced by lower
velocity operation.
The gas in the fog zone is in contact with extremely small drops for an
extended period of time so diffusional process such as gas absorption and
diffusional deposition of particles are more effective. From the fog zone
the gas passes through a high efficiency entrainment separator to separate
the drops from the gas stream.
LABORATORY STUDIES
Many pilot plant tests have been completed in the course of developing
the Calvert Scrubber and measuring its performance. Some previous data have
been reported by Calvert, et al, (1) for particle collection and S02 absorp-
tion by N32C03 solution. The data reported in this paper are for particle
collection only.
The experimental method used for determination of the particle collec-
tion efficiency of the laboratory scale Calvert Scrubber involved the use of
two prototypes. One operated at air flow rates around 2.83 m^/min (100 CFM).
The other, larger unit, had a nominal 28.3 m3/min (600-700 CFM) flowrate.
The following test aerosols were used to vary the size distribution and
physical properties of the particles:
1. Fly ash.
2. TiO£ pigment.
3. Cupola dust.
4. Hydrated lime dust.
5. Dibutylphthalate dispersed by evaporation/condensation.
6. Methylene blue dispersed by spray drying.
7. Ti02 dispersed by spraying, evaporating and oxidizing tetra-
i sopropyl-ti tanate.
8. TiO? dispersed by spraying and combusting a suspension of tetra-
isopropyl-titanate in alcohol.
-------
The first four test aerosols were dispersed by means of a dust feeder with a
screw feed. Deagglome ration was accomplished by means of a high pressure air
jet ejector followed by a cyclone separator to remove any over sized mater-
ial. The gas leaving the cyclone was then diluted with ionized air from a
Polonium 210 source to neutralize any charged particles.
Particle size distribution and concentration at the scrubber inlet and
outlet were determined by means of cascade impactors. The cascade impactors
were operated with either greased aluminum or glass fiber substrates depend-
ing on the physical properties of the test aerosol. The experimental proce-
dure used was essentially the same as that which has been used by A.P.T. in
its field evaluations, Calvert, et al , (2) and in laboratory and power plant
studies. The cascade impactors were calibrated in accordance with the "Cas-
cade Impactor Calibration Guidelines", Calvert, et al , (3), developed by
A.P.T. and adopted by EPA as their guideline method.
Air flow rates were measured by means of Venturi meters, which were
calibrated with pi tot tube traverses. Water flow rates were measured by
means of calibrated rotometers and Venturi meters.
Particle Collection Res ul ts
Particle collection efficiency as a function of particle size was com-
puted from the cascade impactor data by using three methods. All three,
visual, spline fit, and finite difference, are based on the ratio of the
slopes of the inlet and outlet cumulative mass concentrations as a function
of aerodynamic particle diameter. These methods measure the slope of the
inlet and outlet curves at the same particle diameter and then compute the
penetration as the ratio of the slope of the outlet curve to the slope of the
inlet curve. Aerodynamic particle diameter is defined as follows:
where d = aerodynamic particle diameter,
pa
d = physical particle diameter, ym
-------
p = particle density, g/cm3
C' = Cunningham slip correction factor, dimensionless
o
Typical experimental particle collection data for the 28 m /min Calvert
Scrubber prototype are shown in Figures 2 and 3. These figures show that the
particle penetration increases with decreasing particle size until a maximum
penetration is reached and then particle penetration decreases with further
decrease in particle size. This is the result of the increased particle
collection in the collision and fog zones of the Calvert Scrubber. These
figures also show similar results for two different types of test dust when
the scrubber is operating at a liquid to gas ratio of 1.33 1/m (10 gal/MCF)
and a gas phase pressure drop of 40 cm W.C. (16 in W.C.). Figure 3 shows the
improved performance when the scrubber is operated at a liquid to gas ratio
of 2 1/m3 (15 gal/MCF) and gas phase pressure drop of 71 cm W.C. (28 in W.C.).
The particle collection performance of the Calvert Scrubber can be com-
pared to a Venturi scrubber using the cut power method developed by A.P.T.,
Calvert, et al, (4). Since most runs resulted in particle penetrations less
than 50% for all particle sizes, an equivalent cut diameter was determined
from the overall penetration.
The equivalent cut diameter for the Calvert Scrubber is compared to that
of a Venturi scrubber in Figure 4. This figure shows that the Calvert Scrub-
ber requires less pressure drop for the same cut diameter as a Venturi scrub-
ber. For a cut diameter of 0.3 ymA, the Calvert Scrubber requires a pressure
drop of 76 cm W.C. (30 in. W.C.) compared to 140 cm W.C. (55 in W.C.) for a
Venturi scrubber.
CALVERT SCRUBBER PILOT PLANT
The Calvert Scrubber pilot plant was designed to be entirely self-con-
tained and requires only electrical and water connections. The scrubber
system auxiliaries (pumps, tank, instrumentation and control systems) were
installed on an 8' x 10' trailer which is easily transported from site to
site. An exhaust fan is carried on the trailer and installed along with
lightweight FRP or carbon steel ducting at the site. The Calvert scrubber
pilot plant module was designed to be free standing to maintain flexibility
-------
in locating the pilot plant at an industrial site. Typical set up times are
4-8 hours.
Gas flow rate through the pilot scrubber can be varied from 1000-2500
ACFM. Liquid circulation rates as high as 100 6PM can be achieved. The de-
sign of the scrubber allows maximum flexibility in terms of the particle con-
ditioning, primary and secondary atomization zones. The ducting system,
which can be transported on the trailer, contains sampling parts to allow
measurement of the inlet and outlet particle mass concentration as well as
particle size distributions.
The performance of the Calvert Scrubber pilot plant has been determined
at an industrial site and in laboratory tests on the following types of par-
ticles:
1. Anhydrous borax.
2. Fly ash.
3. Hydrated lime dust.
Fractional particle penetration data for the Calvert Scrubber pilot plant was
similar to that shown in Figures 2 and 3 for the laboratory scale Calvert
Scrubber.
CONCLUSIONS
The Calvert Scrubber is well suited for scrubbing fine particles. Par-
ticle collection and gas absorption are greatly enhanced in the secondary
atomization zone where the drops formed in the primary atomization zone are
shredded into smaller drops. These smaller drops provide better mass trans-
fer and additional fine particle collection because of the renewed high rela-
tive velocity. The power requirement for the Calvert Scrubber is lower than
other scrubbers, such as the Venturi, packed and plate types.
-------
REFERENCES
1. Calvert, S. and S. Yung. "Advanced Developments in Novel Particulate
Control Technology". Paper No. 79-56.1. Paper presented at the 79th
Annual Meeting of the Air Pollution Control Assoc., 1979.
2. Calvert, S., N. C. Jhaveri, and S. Yung. "Fine Particle Scrubber Per-
formance Tests". EPA-650/2-74-093, October 1974.
3. Calvert, S., C. Lake, and R. Parker. "Cascade Impactor Calibration
Guidelines". EPA-600/2-76-118, 1976.
4. Calvert, S. "Engineering Design of Fine Particle Scrubbers". J. Of Air
Pollution Control Assoc. 24: 929-934 (1974).
-------
H20
DIRTY GAS
H20
CLEAN GAS
\t
/l\
H20
DIRTY GAS
H20
CLEAN GAS
Fig. 1 CALVERT SCRUBBER COLLISION ZONE
<£
O
<
CC
H
UJ
UJ
60
40
20
10
1
0.2
AP = 41 cm W.C.
CUPOLA DUST
CALVERT
SCRUBBER
0.2 0.5 1 2 5
AERODYNAMIC PARTICLE DIAMETER.pmA
F\g. 2 CALVERT SCRUBBER PERFORMANCE
COMPARED TO A VENTURI.
z
o
cc.
\-
Ul
UJ
a
30
20
10
5
1
0.5
0.1
0.001
AP»71cm W.C. -
HYDRATED LIME
0.3 0.5 1 23456
AERODYNAMIC PARTICLE DIAMETER,(imA
Fig. 3 LABORATORY SCALE CALVERT
SCRUBBER DATA
CC
H
HI
O
1.0
0.5
0.4
0.3
0.2
VENTURI
4 6 8 10 20 30 40 50
PRESSURE DROP.in. W.C.
Fig. 4 CUT / POWER RELATIONSHIP
-------
FLUX FORCE/CONDENSATION SCR11BRFR SYSTEM FOR 001 I FCTIQN QF
FINE PARTICULAR EMISSIONS FROM AN |RQN MFITIKIR CUPOLA
By: Seymour CaI vert
Air Pollution Technology, Inc.
4901 Morena Blvd., Suite 402
San Diego, CA 92117
Dale L. Harmon
Part IcuI ate Technology Branch
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
ABSTRACT
A f I ux force/condensation (F/C) scrubbing system for the control of
partlculate emissions from an Iron and steel melting cupola demonstrated
Its Industrial feasibility during a 6-month test period.
A particle collection efficiency of 84% was required to meet the
emission IImit for a melting rate which ranged around 12,500 kg/hr. The
demonstration plant was designed so as to enable experimental determination
of operating characteristics when attaining efficiencies around the
emission limit. Particle size distributions, flow rates, pressure drops,
temperatures, and other parameters were measured and utilized for
performance analysis and comparison with theoretical predictions.
After an initial period during which equipment and operating methods
were refined, the system performed well and without significant difficulty.
An optimized F/C scrubbing system for this application would require only
about 65% of the predicted power requirement for a conventional high energy
system.
INTRODUCTION
Flux force/condensation (F/C) scrubbing has been developed by Air
Pollution Technology under EPA-sponsorship for the past several years. The
object of F/C scrubbing Is to reduce the power requirement for collection
of fine particles as compared to conventional high energy scrubbers. The
Improvement In scrubber performance due to condensation effects Is most
apparent In the submlcron size range where the mechanism of Inertlal
Impactlon is difficult to apply economically. The condensation of water
vapor from a hot saturated gas can be caused by contacting the gas with
cold IIquId.
10
-------
Three condensation effects are utilized in F/C scrubbing. The
suspended particles In the gas act as condensation nuclei, resulting in
growth of particle mass due to the condensation of water vapor. The
transfer of water vapor toward the cold liquid results in diffusiophoresis,
causing particles to collect by movement of the condensing vapor toward the
cold surface. Simultaneously, the temperature gradient established In the
condenser causes a thermal force which leads to collection by thermo-
phoresis. The three effects enhance the particle collection efficiency of
the F/C scrubber system.
Flux force effects on particles have been known for many years and the
background has been reviewed and discussed In depth by many authors. The
research and development work supported by the EPA had the objectives of
first assessing the potential of F/C scrubbing, then developing rational
means for exploiting it, and then demonstrating the technology.
This report presents the results of a demonstration project to test
further the design methods and economics of F/C scrubbing on a large scale
in an industrial environment. An F/C system for the total exhaust gas
stream from a 12,500 kg/hr Iron melting cupola was designed, built, and
tested. The operating experience Included summer, fal I, and winter in a
demanding environment and resulted In a good test of the equipment design
and materials selection.
SYSTEM DESIGN
As a replacement system In an operating plant, the F/C scrubber had to
be designed to fit in the space around the existing scrubber system. Once
the new apparatus was in place it had to be connected and the old system
disconnected with the minimum loss of productive operation of the cupola.
This and the space constraints imposed by the foundry operations had the
major Influence on F/C system layout.
Conservation of project cost required that the maximum amount of
existing auxiliary equipment be used. Thus, the old afterburner, scrubber
pump and sump, fan, and stack were used.
The F/C system was built to operate In two basic configurations,
Induced draft (I.D.) and forced draft (F.D.). The F.D. mode has the
potential advantage of reducing the fan power requirement by about 6%, with
the possible disadvantage of entrained liquid entering the fan.
The system flow diagram Is shown In Figure 1 for the I.D.
configuration. Only the ducting was modified to switch from forced to
Induced draft; the remaining components remained unchanged.
Process Description
The cupola melts scrap Iron and steel using heat generated by the
combustion of coke. Raw materials when the program started were scrap
structural steel, engine blocks, metal stampings and turnings, cast Iron
piping and fittings, and other miscellaneous metal parts. During the
11
-------
construction period the cupola was converted to water cooled operation and
the metal charge was changed to smalI steel scrap.
The scrap and coke are charged together In 1,800 kg batches with an
Iron-to-coke-mass ratio of 6. Smal I amounts of I Imestone, 20-30 kg, are
also charged to form slag. The nominal rate Is 12,500 kg/hr and the normal
tuyere air flow rate Is 3.5 DNmVs.
The exhaust gas from the bed of material in the cupola Is composed
typically of 14? CO, 13? C02, and 73? N2. Additional combustion air which
enters the charging door at a rate of 2.3 DNmVs allows for complete
conversion of CO to C02. An Ignition burner ensures that a flame source is
available for combustion of the CO to CO 2. (It was not needed or used
during Initial tests.)
Combustion Is completed In a brick I Ined afterburner. The exhaust
gases should leave the afterburner tank at a temperature ranging from 800
to 1,000°C. The exhaust gas flow rate at this point Is about 6.6 DNmVs.
After the cupola modification and the change to steel scrap charge, the
cupola gas would not burn consistently so the gas temperature was lower and
more variable than In preliminary tests.
The hot gas leaves the afterburner and Is cooled In the saturator by
water sprays and the temperature approaches the adlabatlc saturation
temperature. Typically, the saturation temperature achieved ranges from
71 °C to 77°C. The maximum water spray rate In the saturator is 4.5 l/s (75
gpm).
The condenser and cool ing tower system has a design heat rejection
rate of 9.6 x 106 J/s (33 x 10s Btu/hr). The water flow rate can be varied
up to 69 l/s (1,100 gpm). The cold gas temperature leaving the counter-
current packed-bed condenser is about 55°C at design flow rates.
The exhaust gases leave the condenser and enter either the exhaust fan
or the scrubber, depending on the ducting configuration. The primary
particle collection device In the system Is a gas-atomized scrubber with a
variable area throat, an integral water sump, and an entrainment
separator.
The fan used In the demonstration plant was re-built from that used In
the previous system and Is a straight radial blade design. The flow rate
of the fan Is approximately 14 AmVs (30,000 cfm) at -96 cm W.C.
(-38 in. W.C.).
The condenser water comes from the cooling tower which is designed to
cool 69 l/s (1,100 gpm) of water from 66°C to 38°C with ambient air. This
Is accomplished by evaporative cooling.
The settler was designed to remove particles collected by the
scrubbing liquid. A 1.3 l/s (20 gpm) stream of sump water Is clarified by
the settler. The thickened sludge (10-30? by wt) is disposed of as land
fill. The clarified water (100 ppm solids) is returned to the sump.
12
-------
PERFORMANCE TESTING
In the course of system start-up, a significant problem with the
original exhaust fan caused excessive power consumption and tripping of the
electrical overload device. This condition continued until a new fan
wheel, which A.P.T. designed, was installed. Some problems due to foaming
and entralnment carry-over had to be resolved, but system operation was
otherwise quite good during the test period. No lost production time was
experienced by the foundry due to problems associated with the operation of
the F/C scrubber system.
The major test program variables were the condenser liquid flow rate
and scrubber pressure drop. The condenser IIquld flow rate affected the
amount of cool Ing In the condenser and, hence, the condensation ratio. The
scrubber pressure drop could be varied by changing the scrubber throat flow
area and the scrubber IIquid-to-gas ratio.
Samples were usual ly taken at the Inlet to the condenser and at the
scrubber outlet. For some runs samples were also taken at the scrubber
inlet. Sampling at three locations allowed the condenser performance and
scrubber performance to be evaluated separately.
The particle size distribution and mass concentration were measured
with University of Washington cascade Impactors equipped with precutters.
Several modified EPA Method 5 total filter runs were made to determine
total mass concentration for compliance tests and to confirm impactor mass
loadings.
From measurements of cumulative mass concentration as a function of
particle size, particle penetration was computed as a function of particle
size. The data was analyzed using a computer program which calculated
impactor cut points and cumulative mass concentrations based on weight gain
per stage, impactor flow rate, and Impactor calibration data.
Performance Model
The system performance model can be used to predict the emission from
a F/C scrubber system installed on a pollutant source with a known particle
concentration and size distribution. The model allows Independent
variation of condensation ratio and scrubber pressure drop.
Without going into the details of the mathematical model, the basic
concepts and outline of the approach will be discussed. The gas leaving the
source Is hot and has a water vapor content which depends on the source
process. The first step is to saturate the gas by quenching It with water.
In the condenser there is diffuslophoretic and thermophoretic
deposition as well as some Inertlal impactlon (and, perhaps, Brownlan
diffusion). The particles In the gas leaving the condenser will have grown
In mass due to the layer of water they carry.
Subsequent scrubbing of the gas will result In more particle
13
-------
collection by Inertial Impactlon. There may be additional condensation,
depending on water and gas temperatures, and Its effects can be accounted
for.
The mathematical model for the F/C demonstration plant accounts for
the mechanisms and assumptions outlined below:
Saturator
1. Gas Is humidified and cooled to adlabatlc saturation temperature.
2. No condensation occurs.
3. Particle collection In the saturator Is negligible.
Condenser
1. Particles are collected by Impactlon In packed column.
2. Condensation occurs, causing growth of particles.
3. Col lection occurs In condenser due to dlffuslophoresls.
Scrubber
1. Grown particles are collected by Impactlon In scrubber.
2. Negligible condensation occurs.
EXPERIMENTAL RESULTS
Experimental performance measurements which were made during the last
2 months of the test period represent the best level attained after system
adjustment was nearly completed. The "36 series" of runs were In the I.D.
mode.
Runs 36/12 and following were made after the foaming problem was
resolved and the scrubber liquid piping revised to give better purging of
the Internal sump. The "37 series" of runs were made in the F.D. mode with
the exception of 37/21 and following, which were in the I.D. mode.
The hot gas temperature leaving the afterburner was for most of the
runs significantly lower than anticipated. The design-basis hot gas
temperature was 1,000°C, as measured In prelImlnary tests. The maximum
temperature measured was about 850°C. The low gas temperature resulted In
a reduction In the attainable condensation ratio and the maximum was about
0.2 g H20/g D.G. (dry gas). It had been anticipated that a condensation
ratio of 0.30 g H20/g D.G. could be achieved with the higher gas
temperature.
Particle Mass and Size
The average particle mass concentration at the Inlet to the condenser
was 2.4 g/DNm3, which Is 33/8 higher than the design basis mass
concentration of 1.8 g/DNm3.
Condenser inlet mass concentration measurements varied significantly
from run to run, which reflects the problem of having to sample during
various periods In the charging cycle. Inlet sampling times had to be much
shorter than outlets because the Inlet concentrations were so high.
Consequently, one inlet sample would not be representative of the period
14
-------
over which the outlet sample was taken. When scrubber penetrations were
computed on the basis of the Inlet and outlet samples for each run, there
was great and random variation In the results.To cope with this difficulty
an average condenser Inlet size distribution was used in determining the
grade penetration curves.
The average cumulative mass inlet concentration distribution was used
in combination with the cumulative mass outlet concentration distribution
for each run to compute the grade penetration relationship for that run.
Figures 2 through 5 present experimental and predicted grade penetration
curves for runs 36/10 through 36/13.
The grade penetration curves show that predictions and experiment
compare well In some cases but not In others. Experimental penetrations at
around 1 fimA tend to be higher than predictions, and this is true for both
F/C and non-F/C conditions.
The performance in the F.D. mode was erratic and not improved as much
by condensation as it was in the I.D. mode. Some of the condensed water
may have evaporated In the fan so that the particles entering the scrubber
were smaller than those entering the fan. It was not possible to resolve
this question during the test period.
The variation of computed penetrations appears to be caused by the
uncertainty of the inlet size distribution. The agreement between
predictions and experiment for many of the series 36 runs Is considered
good for field experimental conditions, where the equipment Is being
modified and the source process is cyclic.
ECONOMIC ANALYSIS
An optimum F/C scrubber system design for this application would not
be significantly different from the demonstration plant and the process
design would remain substantially unchanged. The major modifications
recommended for an optimum system Involve equipment redesign and
specifications to reduce capital expense and to ensure that the gas
temperature would be high.
Significant cost savings can be achieved by using a single combined
condenser/scrubber vessel. Experiments with the demonstration plant showed
that the vessel diameter and the packing depth can be smaller than used.
Because of space requirements, the cooling tower had to be elevated on
a platform above the saturator and sump. Consequently the most compact
tower was chosen, not the most economical or energy efficient. A wood-
filled cooling tower would have been the best choice If space were not at a
premium. Such a tower would require only 22 kW (30 HP), compared to 44 kW
for the tower used on the demonstration plant. The wood packing may have
some advantage over epoxy coat steel for corrosion resistance. A second
circulating water pump would be required for the system but the total cost
would be less because an expensive structural steel platform would not be
needed.
15
-------
The saturator, which was designed to fit into the available space at
the foundry, would be smal ler in an optimum system. The settler used on
the demonstration system provided acceptable performance, however the
condenser, scrubber and sump all were subject to deposition of sandy
material which required removal at frequent intervals. Sloped bottoms
should be used in all of these vessels.
It Is also recommended that a moving belt filter be used Instead of
the gravity settler to remove this sandy material. These can be purchased
at a lower cost than the settler which was used on the demonstration plant.
In order to evaluate the potential benefits of an F/C scrubber system
as compared to a conventional one, the costs for a venturi scrubber system
were estimated. The process conditions and emission criteria were the same
for both systems.
Capital Cost Estimates
The capital cost of an optimum F/C system would consist of direct
costs for delivered major equipment and other items such as Installation,
piping, and electrical, along with indirect costs such as engineering,
construction overhead, and contingencies. Table 1 shows the estimated
direct and indirect costs for the F/C system and the high energy scrubber.
All costs were adjusted to December 1979 by means of the Marshall and
Stevens cost index.
TABLE 1. DIRECT AND INDIRECT COSTS
F/C SYSTEM
DI rect
Equipment
InstalI at ion
Instruments
Piping and Ducting
Electrical
Site Preparation
Total Direct Costs
Indirect
Engineering
Construction Overhead
Contractor's Fee
Contingency
Total Indirect Costs
Fixed Capital Investment
Ratio
1.00
0.40
0.10
0.40
0.05
0.05
2.00
0.40
0.45
0.10
0.40
1 .35
3.40
Cost, $
$102,380
40,952
10,238
40,952
10,238
5f119
$209,879
40,952
46,071
10,238
40f952
$138,213
$348>092
CONVENTIONAL
SCRUBBER
Ratio
1.00
0.40
0.10
0.40
0.10
0.05
2.05
0.40
0.45
0.10
0.40
1.35
3.40
Costf $
$ 73,765
29,506
7,377
29,506
7,377
3,687
$151,218
29,506
33,193
7,377
29f506
$99,582
$250,800
16
-------
OPERATING COSTS
The operating costs for the air pollutlon control system consist of
the annual cost of the utilities (power and water), raw material, and
maintenance. Table 2 shows the power requirement for both the F/C scrubber
system and a conventional scrubber. The F/C scrubber system would require
only 65% of the power of the conventional scrubber. The exhaust fan for
the conventional scrubber would require more than twice the power of the
F/C system.
TABLE 2. SUMMARY OF POWER REQUIREMENTS
Power, kW (HP)
F/C Scrubber Conventional Scrubber
Exhaust Fan 156 (210) 332 (445)
Saturator Pump 10 (13) 10 (13)
Scrubber Pump 2 (2) 3 (4)
Cooling Water Pumps 33 (44)
Cooling Tower Fan 22 (30)
Total Power Required 223 (299) 345 (462)
The estimated total operating costs are summarized In Table 3. The
total operating cost for the F/C scrubber system was found to be $131,230
compared to $150,300 for the conventional scrubber. The annual operating
cost of the F/C scrubber system would be about $19,100 less than the cost
for a conventional scrubber for the conditions stated. If the operating
time were reduced below 4,000 hr/yr the F/C system would be less favorable,
but longer operating time would make it more favorable. Power cost will
also have an obvious effect on the economics.
TABLE 3. SUMMARY OF ESTIMATED ANNUAL OPERATING COSTS
F/C Scrubber Conventional
Item Unit Cost System Scrubber
Capital Cost § 10% of F.C.I. $ 34,810 $ 25,080
Maintenance Materials 6 J>% of 10,440 7,520
F.C.I.
Labor § 8 hr/wk shift, 7,200 7,200
$9/hr
Operating § 12 hr/wk shift, 10,800 10,800
Labor $9/hr
Power Cost* $0.065/kWhr 57,980 89,700
Water Use $0.035/1,000 liters 2,000 2,000
Raw Soda ash $0.265/kg 8,000 8,000
Materials flocculant $1.15/1 iter
TOTAL OPERATING COST $131,230 $150,300
*Power cost estimated for 4,000 hr/yr.
17
-------
CONCLUSIONS
The following conclusions can be drawn from the results of this study:
1. The performance data were In general accord with predictions based
on design methods developed by A.P.T. on previous EPA contracts.
2. The F/C scrubber system was capable of complying with all
applicable air pollution regulations for the site.
3. The F/C scrubber system needs only about 65% of the predicted
power requirement of a conventional high energy scrubber to
achieve the same performance.
4. The particle mass concentration and size distribution emitted
during the cupola charging cycle varied greatly. Continuous or
semi-continuous cupola charging with a conveyor system would
probably reduce the variation In emissions and gas temperature.
5. The solid/liquid separation presented some maintenance problems.
Additional attention should be given to equipment design, both to
prevent settling In the scrubber equipment and to Improve solids
separation In the water treatment system.
6. Corrosion of the cooling tower packing was severe. Acids must be
neutralized In order to prevent significant deterioration of steel
components In the system. Maintenance of the proper circulating
water pH Is also Important If stainless steel materials are used
because of the chloride content present.
7. Operation over a 6-month period showed that the F/C system Is
capable of performing well and without significant difficulty In
the demanding conditions of a ferrous foundry. During the entire
test period there were no production stoppages due to the F/C
scrubbing system.
8. F/C scrubbing becomes more economically advantageous than
conventional scrubbing as either power cost or plant operating
time Increases.
RECOMMENDATIONS
Experience with the demonstration plant has led to several
recommendations about equipment changes; most of them are discussed above
under "Economic Analysis." The essential alms of the recommendations are:
1. Minimize the costs of the equipment required to obtain F/C
effects.
2. Maximize the cupola flue gas temperature.
3. Provide adequate solids removal capacity for the scrubbing liquid
recycle stream.
18
-------
VO
CUPOLA
WATER
TREATMENT
Figure 1. F/C demonstration plant flow diagram, induced draft.
-------
1C
^
s^
: ;
K
q1 * 0.08
•---PREDICTED
MEASURED
0.3 0.4 0.5
1.0
2.0 3.0 4,0 5.0
10.0
50
40
30
20
10
0.3 Q.4 0
f
q' * o.i
. =._-—PREDICTED
•MEASURED
0 4.0 5.0
10.0
Figure 2. Particle penetration, Run 36/10
Figure 3. Particle penetration, Run 36/11
50
40
30
20
Z 10
KgCT
"ilfi*'
TF
-;SJ;:: v
— -'-K:
•iti
T-T-N-
w
11
0,1 -
:;HH~:{S; ny
m^m~
-
--------- PREDICTED
"" - MEASURED
IE
0.3 0.4 0,5
1.0
2.0 3.0 ^.0 5.0
10.0
Fioure 4. Particle penetration, Run 36/12
10
LU _
D. 5
i -'-1
w
^k_
m
0.16 ~
,.:. PREDICTED
"" MEASURED
0,3 0.4 0.5
1.0
2.0 3.0 4.0 5.0
10.0
dpfl,
Figure 5. Particle penetration, Run 36/13
-------
DEMONSTRATION OF HIGH-INTENSITY-IONIZER-ENHANCED
VENTURI SCRUBBER ON A MAGNESIUM RECOVERY
FURNACE FUME EMISSIONS
By: Anil Prem, M. T. Kearns
Air Pollution Systems, Inc.
18642 - 68th Avenue South
Kent, Washington 98031
Dale L. Harmon
Environmental Protection Agency
Industrial Environmental Research Laboratory
Research Triangle Park, N. C. 27711
ABSTRACT
3
A 33,960 Am /hr (20,000 acfm) permanent demonstration system,
consisting of the Air Pollution Systems' High Intensity Ionizer (HII)
and a variable throat venturi scrubber, has been installed on a magnesium
recovery furnace at Teledyne Wah Chang Albany, Albany, Oregon. The
particulate emission from this source is predominantly submicron in
nature. Fairly high stable electric fields of 10-12 kV/cm have been
successfully maintained in the HII. Comparing the measured charge/mass,
and the calculated charge/mass based on the measured particulate size
distribution at the outlet of the scrubber, the HII appears to be imparting
fairly high charge to the incoming particulate. Preliminary performance
test results indicate higher performance of the venturi scrubber when
the HII is energized for all the venturi pressure drops tested. The
penetration reduction due to HII increased as the venturi pressure drop
was increased, and as the particle size decreased.
INTRODUCTION
Particulate emissions from industrial stack gases are usually
collected by one or more types of three conventional devices—electrostatic
precipitators, scrubbers, and fabric filters—the design of which has
virtually remained unchanged since the turn of the century. However,
with the increased need for fine particulate control, the cost of these
particulate collection devices is also substantially increasing. This
has resulted in a flurry of research and development activity to produce
more cost effective pollution control systems.
The objective of the novel device program of EPA/IERL-RTP's Particu-
late Technology Branch was to identify and evaluate new technology or
new combinations of well studied mechanisms that are cost effective in
controlling fine particle emissions. Air Pollution Systems's (APS) High
Intensity Ionizer (HII) falls in this category. Due to its unique
geometry, HII can maintain very high electric fields, resulting in a
significant improvement in the particle charging capability compared to
a conventional particle charger.
21
-------
Using an HII, APS has developed an electrostatically enhanced
venturi scrubber (Scrub-E) that is capable of high fine particle collection
at a reduced pressure drop compared to a conventional venturi scrubber.
Based on results of an EPA sponsored test of a laboratory scale unit
(1), APS won a contract from EPA to demonstrate the Scrub-E on an industrial
application. A magnesium recovery furnace which predominantly emits
fine particulate was selected as the demonstration site. A 33,960
Am /hr (20,000 acfm) demonstration Scrub-E system was designed, fabricated,
and installed. This paper describes the APS Scrub-E, the system setup
at the demonstration site, and the preliminary test results.
APS Scrub-E
Conventional scrubbers usually have low particle efficiencies in
the 0.01 to 2.0 pm diameter range. This low particle collection efficiency
in the fine particle size range is caused by conventional wet scrubbers'
primary use of inertial impaction. In order to increase the fine particle
collection efficiency of conventional scrubbers, it is necessary to
increase the energy input substantially. For high efficiency collection
of fine particles at a moderate energy consumption, one must apply some
particle collection force that is not dependent on increasing the relative
velocity between the drops and gas stream. Electrostatics is one such
promising particle collection force.
APS Scrub-E utilizes electrostatics to improve the performance of a
conventional venturi scrubber. The APS HII is utilized to highly charge
the particulate in the flue gas. The negatively charged particle sees
an induced positive charge on the neutral droplet (charged particle
image force) in the venturi scrubber and is attracted to it. This
results in increased collection efficiency. The higher the particle
charge, the higher the improvement in collection efficiency.
The APS HII, due to its unique electrode geometry, is successful in
maintaining a highly stable intense corona discharge. The cathode is a
solid metal disc supported by a structurally reliable tube section
centered in a cylindrical anode arrangement. The electrode configuration
produces a substantially uniform three dimensional field which is a
principal factor in the greater electrical stability. Typical operating
field strengths of 10-13 kV/cm have been obtained compared to 3.5 kV/cm
for wire electrode geometry in an industrial flue gas.
In addition to the very high electric fields, the HII electrode
geometry produces a concentrated field with ion densities of 10 - 10
ions/cm , many times that obtained in the wire electrode geometry. As a
result of higher fields and ion densities, both the level of charge
acquired by the particles and the rate of charging are significantly
higher. For example, even with the significantly lower residence time
in an HII compared to an electrostatic precipitator, the level of charge
obtained by the particles exiting an HII are two to three times higher
compared to the particles exiting an ESP. This means that particles can
be charged to a high level in a very small volume resulting in a
22
-------
very compact system that can be retrofitted to existing scrubbers.
Demonstration Project
The site selected for the demonstration project was a magnesium
recovery furnace at Teledyne Wah Chang Albany (TWCA), Albany, Oregon.
TWCA is a large production facility for zirconium and hafnium metals,
and columbium and tantalum alloys.
The particulate gas stream to the demonstration unit originates in
four emission sources. Two of these are the magnesium recovery furnaces
where the spent magnesium, as MgCl2, is reduced to magnesium metal for
recycle. MgCl2, along with the flux, is loaded into crucibles which are
placed in a gas fired furnace. Once the reduction is completed, first
the magnesium metal and then the slag are manually ladled from the
furnace. The third emission source is the MgCl2 furnace which is essentially
a concentrator operation to recover whatever magnesium is left in the
crucibles. The fourth emission source is the smokehouse, where the
MgCl2 crucibles are heated in a burnout operation.
All the furnaces primarily emit MgO fume. The majority of the MgO
fume emission is attributed to the magnesium recovery furnaces during
the ladling process. Significant emissions occur during the slag ladling
from the magnesium recovery furnace. The magnesium reduction process is
a batch operation. Each furnace is tapped four or five times a day.
Prior to designing the demonstration unit, the magnesium recovery
furnace emissions were sampled to characterize the particulate. Scanning
electron microscope (SEM) photographs indicated that the greatest pro-
portion of the particulate were less than 1.0 ym in diameter. The size
distribution of the particulate was also measured using a University of
Washington Mark III cascade impactor. The mass mean diameter of the
particulate was found to be 1.48 pm and the average mass concentration
0.19 g/DNm . The above measurements indicated that the emissions are
predominantly ultrafine particulate. Brief resistivity measurements of
the garticulategin the APS laboratory indicated fairly low resistance (3
x 10 to 1 x 10 ft-cm).
Figure 1 is a flow diagram of the EPA demonstration system at
Teledyne Wah Chang Albany, Albany, Oregon. Flue gas from the four
emission sources enters the demonstration system at the top of the HII
module. The HII module consists of seven anode cylinders each with a
concentric probe with three discharge electrodes. The flue gas stream
is distributed amongst these cylinders to charge the particulate prior
to entering the venturi scrubber.
The venturi scrubber is a variable throat design utilizing a Bombay
type damper to vary the throat opening. A 90° elbow is directly below
the venturi. The gas stream leaving the venturi enters a conventional
separator and exits to the atmosphere through two in-line induced draft
fans, operated individually or in series.
23
-------
TO
EXISTING
SCRUBBER
j
re*
-,—DAMPERS
n
r
INLET
TEST
STATION
-HMTRSET
IONIZER
MODULE
VARIABLE
THROAT
VENTURI
SCRUBBER
SMOKEHOUSE
Mg RECOVERY
FURNACE
MgCI2 FURNACE
Mg RECOVERY
FURNACE
RECYCLE PUMP
TO DRAIN
DEMISTER PAD
SPRAY NOZZLES
Figure 1. EPA demonstration system flow diagram.
-------
The pressure drop across the venturi is controlled by the Bombay
damper and the gas flow rate through the system can be controlled by
both the inlet and outlet dampers. The liquid flow rate to the venturi
can be controlled by the valves in the liquid flow lines. Liquid to the
venturi is supplied from the recycle tank. Kimre demister pads, retrofitted
into the separator to eliminate droplet carryover, are wetted with 100
percent fresh water.
Inlet sampling ports to the system are upstream of the Scrub-E
inlet damper. Outlet sampling ports are approximately four duct diameters
upstream of the stack outlet.
Results
The first preliminary performance tests on the APS Scrub-E at TWCA
used EPA Method 5 and University of WashingtonJtfark III impactors. The
main gas flow rate was maintained at 25,500 Am /hr and the liquid flow rate
at 52 m /hr. The venturi pressure drop (AP) was maintained at 51 cm
HO. Both HII "on" and "off" cases were evaluated.
As the tests progressed, a significant droplet carryover from the
separator became apparent. This droplet carryover made it almost
impossible to accurately determine system performance. The liquid flow
rate was arbitrarily reduced to 45 m /hr and, in the final two tests, 22
m /hr. Neither reduction alleviated the droplet carryover problem.
In another effort to alleviate the droplet carryover problem, APS
installed demister pads in the outlet section of the separator. A few
spray nozzles were installed upstream and downstream of the demister pad
to irrigate them. This alleviated the droplet carryover problem.
System performance was next evaluated, as a function of venturi AP
for both HII "on" and "off" cases. The main gas flow was maintained at
25,500 Am /hr and the liquid flow rate 30 m /hr. University of Washington
Mark III cascade impactors were used for the performance evaluation.
For the HII "on" tests, the HII was left in the automatic mode.
Inlet mass concentration during the above tests varied from 0.025
to 0.16 g/DNm . Outlet mass concentration varied from 0.005 to 0.053
g/DNm . Inlet mass mean diameter ranged from 0.52 to 0.79 ym. Outlet
mass mean diameter ranged from 0.36 to 0.47 ym. Figure 2 illustrates
overall collection efficiency as a function of venturi AP for both the
HII "on" and "off" cases. Overall, the collection efficiency for the
HII "on" cases was higher compared to the HII "off" cases. For both the
HII "on" and "off" cases, collection efficiency increased as venturi AP
was increased.
Figures 3 and 4 illustrate collection efficiency as a function of
AP across the venturi scrubber for particles ^3 m and ^1 ym diameter,
respectively. As with overall collection efficiency, the collection
efficiency was higher for the HII "on" case than for the HII "off."
25
-------
85
o
o
8
65
55
45
• HI I "ON"
O HI I "OFF"
50 75 100 125 150
PRESSURE DROP ACROSS VENTURI, cm W.C.
Figure 2. Overall collection efficiency as a function
of pressure drop across the venturi scrubber.
85
u
I 75
U
8
55
45
50 75 100 125 150
PRESSURE DROP ACROSS VENTURI, cm W.C.
Figure 3. Collection efficiency of particle; -' 3 pm
as a function of pressure drop across the venturi
scrubber.
26
-------
u.
Ul
75
65
55
HI) "ON"
O HII "OFF"
50 75 100 125 150
PRESSURE DROP ACROSS VENTURI, cm W.C.
Figure 4. Collection efficiency of particles < 1 urn as a
function of pressure drop across the venturi scrubber.
1.0
e
0?
Ul
§
E
Ul
X » - 025 (NON-WETTABLE)
O f - 05 (WETTABLE)
• SCRUB-E PERFORMANCE AT
TWCA (HII "OFF")
0.1
I I I I 1 ! I I
5 10 20 50 iOO 300
PRESSURE DROP ACROSS VENTURI, cm W.C.
Figure 5. Representative cut diameters as a function of pressure drop for
venturi scrubbers (3).
27
-------
Also, for both the HII "on" and "off" cases, collection efficiency
increased as venturi AP was increased.
Though the preliminary test results were encouraging, based on the
past data, APS felt that the electrostatic charge on the particulate was
not being fully utilized. At this point, APS reviewed all the data to
date and attempted to arrive at a brief test program that would help
optimize system operation.
An effort was made to determine how well the venturi scrubber per-
formance at TWCA agreed with the general correlation between cut diameter
and pressure drop. The aerodynamic cut diameter was calculated using
simultaneous inlet and outlet particle size distribution for the Scrub-E
with the HII "off." Figure 5 illustrates the aerodynamic cut diameters
for various AP's across the venturi scrubber at TWCA along with the
representative cut diameters as a function of AP for a venturi scrubber
(2). The data shows that, with the HII "off," the venturi scrubber
performance at TWCA agrees fairly well with the general correlation of
cut diameter with AP for venturi scrubbers used to collect wettable
particulate.
During the optimization tests, it was noted that the sparkover
voltage in the HII was lowered significantly during the slag tap. Once
the tap was completed, the HII voltage would go back up to 100 kV. This
lower sparkover voltage was probably related to significant current
suppression created due to high concentration of small particles and the
presence of three discharge electrodes in the HII.
In an attempt to decrease the sparking, the gas flow through the
system was increased to 37,400 Am /hr. No effect was noted on the
sparkover voltage; however, the water from the separator started re-
entraining. This water droplet carryover was alleviated by reducing the
gas flow through the system to approximately 34,000 Am /hr.
Due to the fairly low HII sparkover voltage during the slag tap,
the unit was turned "off" for visual inspection. The HII and the scrubber
appeared to be in fairly normal condition and the sparking appeared to
be in the anode section. The seven discharge electrode probes were
dismantled and two out of three electrodes on each probe were removed to
decrease sparking. The probes were put back into the unit with only the
last electrode. On restarting the unit, for the no-tap case, the HII
was operating at 110 kV - 15 mA (higher voltage and current/electrode) .
During the slag tap, the HII was sparking between 80 and 100 kV, compared
to around 60 kV earlier, indicating significant improvement.
Using EPA Method 17, brief mass tests were conducted to determine
system performance. The outlet was sampled using an APS charge/mass
probe which gave both the mass loading and charge/mass at the outlet of
the Scrub-E. To accurately determine the performance of the Scrub-E for
the two distinct taps (magnesium and slag), tests were conducted so as
to bracket each phase individually. The main gas flow rate was maintained
at 34,000 Am /hr, the liquid flow rate to the venturi at 42 m /hr,
28
-------
venturi AP at 51 cm HO, and the water flow rate to the demister at 4.5
m /Tax.
Inlet mass concentration during the magnesium tap ranged from
0.0087 to 0.044 g/DNm (average 0.021 g/DNm ); for-the slag tap, inlet
mass concentration ranged from 0.061 to 0.16 g/DNm (average 0.090
g/DNm ). The charge/mass measured at the Scrub-E outlet ranged from 85
to 655 yC/g. On the average, the values were higher for the magnesium
tap than for the slag tap. The Scrub-E performance improved when the
HII was "on." The collection efficiency on the average was of the same
magnitude (72.5 percent) for both the magnesium and slag taps when the
HII was "on." For the HII "off" case, the collection efficiency for the
slag tap was higher than the magnesium tap (61 percent vs. 42 percent).
Table 1 summarizes preliminary test results to date. Bearing in
mind that the quantitity of data obtained is very limited, and that no
definite conclusions can be drawn from it, a few general trends can be
seen. The Scrub-E collection efficiency is increased when the HII is
The penetration reduction with the HII "on" appears to increase
both as the particle size decreases and the venturi AP increases. Since
the inertial forces decrease as a function of particle size, the relative
effect of electrostatics should increase as the particle size decreases
(3). Increase of venturi AP should increase the atomization of the
liquid resulting in smaller droplet size. The relative effect of
electrostatics increases with the decrease in the droplet size (3).
The baseline efficiency, for a given venturi AP and approximately
the same liquid/gas flow ratio, decreased as the gas flow rate was
increased, possibly due to the lower residence time in the scrubber.
However, the penetration reduction with the HII "on" increased, probably
due to higher sparkover voltage in the HII during the tap.
Throughout the preliminary tests, the number of furnace taps during
a day was around three, compared to nine taps when the site was selected.
This is related to the lower demand of the zirconium metal. Due to the
manual nature of the process, the emissions are very erratic and no two
taps are alike. This is reflected in the scatter of the mass emission
and charge/mass data. The reduced number of taps a day appears to have
changed the particle characteristics. Both the grain loading and the
mass mean diameter of the particulate appear to have decreased.
Conclusions
3
A 33,960 Am /hr (20,000 acfm) demonstration system, consisting of
an Air Pollution Systems' High Intensity Ionizer and a variable throat
venturi scrubber, has been installed on a magnesium recovery furnace at
Teledyne Wah Change Albany, Albany, Oregon. Preliminary tests were
conducted by APS to evaluate the performance of the system. Tests were
conducted to study the effect of pressure drop across the venturi
29
-------
00
o
TABLE 1
SUMMARY OF PRELIMINARY TEST RESULTS
CASHFLOW
(Am /hr)
25,500
it
"
it
ii
ii
ii
ii
ii
34,000
LIQUID FLOW
(in /hr)
30
ii
it
ii
ii
ii
it
ii
ii
42
VENTURI AP
(cm H20)
64
95
127
64
95
127
64
95
127
74
TYPE OF
MEASUREMENT
C.I. -Overall n
ii
n
C.I. -Part -3 ym
n
n
C.I. -Part -1 ym
it
n
EPA Method 17
BASE EFFICIENCY
(%)
57
67.5
76
52.5
62.5
71
47
57.5
67
51.5
PENETRATION
REDUCTION
(%)
21
29
33
23
30
34
26
40
41
43
-------
scrubber for both the HII "on" and "off" cases. Problems were encountered
during the testing which resulted in some modifications to the system.
Conclusions drawn from the preliminary test results to date are:
1. The APS Scrub-E installation has been fully functional for all
aspects of operation.
2. Most of the particulate emissions occur during the magnesium
recovery furnace tap. There are two distinct phases within a single
tap: ladling of magnesium and slag.
3. For the whole tap, the average mass mean diameter at the
scrubber inlet was 0.64 ym, and 0.41 ym at the outlet.
4. Emissions are generated over a very short time span and are
very sporadic, which makes testing difficult.
5. The retrofitted Kimre packing in the separator eliminated the
earlier carryover problems.
6. The initial HII configuration with three discharge electrodes
gave a lower sparkover voltage and current during the tap. Increasing
the main gas flow rate from 25,500 Am /hr to 37,400 Am /hr did not
improve the HII electrical characteristics.
7. Removing two of the three discharge electrodes from the probe
in the HII significantly improved the electrical characteristics.
8. High charge/mass values were obtained at the outlet of the
scrubber for both the magnesium and slag taps.
9. For the HII "off" tests, the venturi scrubber performance data
is in fairly good agreement with the general correlation of cut diameter
with pressure drop for venturi scrubbers.
10. The collection efficiency for both the HII "on" and "off"
cases increased as the pressure drop across the venturi increased.
11. The collection efficiency with the HII "on" was higher than
with the HII "off."
12. The penetration reduction due to HII increased as the venturi
pressure drop was increased, and as the particle size decreased.
ENDNOTES
Acknowledgements
The demonstration projected has been supported by Industrial Environmental
Research Laboratory, U. S. Environmental Protection Agency, RTF, N. C.,
31
-------
under Contract No. 68-02-2666, Dale L. Harmon, Project Officer. Special
appreciation is noted to Gerald Sing of Teledyne Wah Chang Albany, for
his support and cooperation on this project.
References
1. Calvert, S., et al. (1976), "APS Electrostatic Scrubber
Evaluation," EPA-600/2-76-154a (NT1S No. PB 256335).
2. Calvert, S., (1976), "Engineering Design of Fine Particle
Scrubbers," JAPCA, 24, No. 10, p. 929.
3. Prem, A. and M. J. Pilat (1978), "Calculated Particle
Collection Efficiencies by Single Droplets Considering
Inertial Impaction, Brownian Diffusion and Electrostatics."
Atmospheric Environment, Vol. 12, pp. 1981-1990.
32
-------
A NEW ENTRY IN THE HIGH EFFICIENCY SCRUBBER FIELD
By: L.C. Hardison, President
Air Resources, Inc.
600 N. First Bank Drive
Palatine, Illinois 60067
Frank Ekman, Ph.D., P.E.
US Environmental Protection Agency
(Current Affiliation)
R.R. 5
Barrington, Illinois 60010
ABSTRACT
The use of Venturi, flooded-disk and various orifice-type scrubbers for
wet collection of fine particles at high energy levels and high levels of
particulate collection efficiency is well known. In general, these scrubbers
may be characterized as roughly equivalent in performance level and various
models are competitive only with respect to price and mechanical considera-
tions and ease of maintenance or operation.
The MVS (Modular Venturi Scrubber) invented by Frank Ekman and being
marketed by Air Resources, Inc. is the latest entry into this competitive
field, and has several advantages in size, cost and simplicity over more
conventional designs. This paper describes the development, design and oper-
ation of. the MVS units.
INTRODUCTION
The ARI MVS(TM)" Modular Venturi Scrubber represents a significant im-
provement in the art of wet particulate collection. While it involves no
essentially new principles of collection of particulates by impaction, it
does represent the embodiment of proven techniques into a modular system
which has distinct performance advantages, which can be manufactured at lower
cost and which can be adapted to larger gas volumes than existing techniques.
This paper reviews the evolution of Venturi scrubbers and the derivation
of the MVS(TM) design. It describes a unique application concept made possi-
ble by the modular construction, and develops application criteria for accom-
modation of the modular design to industrial processes.
VENTURI SCRUBBER DEVELOPMENT
The development of the ARI MVS(TM)(1) does not represent a scientific
breakthrough, in which new or startling different physical principles are
invoked to accomplish particulate collection at previously unobtainable low
levels of energy consumption. Rather, it is the culmination of many years
of gradual development of the Venturi scrubbing principle originally embodied
in the Pease-Anthony scrubber.
33
-------
During World War II, a method was needed to disperse the newly discov-
ered DDT over the Pacific islands, by air, to kill mosquitos. Dr. E.D.
Comings and others at the University of Illinois adopted the dry venturi
meter to this use. A few years later, Dr. H.F. Johnstone of the University
of Illinois, was consulting on a Pease-Anthony spray tower that was not per-
forming properly. Johnstone, remembering the success of the venturi in
creating droplets in an airplane slip-stream, inserted a venturi with nozzles
into the duct ahead of the spray tower. The purpose of the venturi was
strictly to supply fine droplets so the spray tower would work more effi-
ciently. The combination worked well and everyone was delighted. Later,
someone turned off the nozzles in the spray tower and scrubbing efficiently
was not diminished at all. The Venturi Scrubber was born!
The typical configuration of the first scrubber, as shown in Figure 1,
was not changed appreciably for many years. Converging and diverging angles
were held at 25° and 7°, just as they were on the existing dry Venturi. At
first, throats were round as in a conventional Venturi, and nozzles were
placed upstream of the throat in elaborate patterns to distribute the water
evenly across the throat. Unfortunately, particulate built up on the nozzles
and their support structure so that the design was modified to remove the
spray nozzles from the gas stream. This was done by making the Venturi
throat rectangular, up to 14 inches in width by whatever length was required
for the capacity. Nozzles were positioned on the long side of the Venturi,
flush with the gas stream. However, the traditional 25° and 7° angles were
maintained, with the result that 20 feet of length, or more, were needed for
the Venturi to contract from duct diameter down to throat width and back
again for large ducts. The mist eliminator continued for many years to be
designed to spray-tower dimensions.
The result was a new design of scrubber that was much more efficient
than anything that had gone before, but one that took up a great deal of
room, and one that had to be individually designed for each installation. In
later years, after the first patents expired, competition entered the field
and a variety of changes were made in the design. However, the changes were
mostly minor and cosmetic, and not much progress was made to ward evolution
of a truly functional design.
With this historical background, the development of the MVS Modular
Venturi Scrubber design and application concept follow logically, and are
described in the subsequent sections of this paper.
DEVELOPMENT CONCEPT
A principal disadvantage of Venturi scurbber is that the size of the
system had a marked influence on the design of the, enturi approach section,
throat and diffuser. Circular Venturis have long suffered from the size
limitation imposed by the inability to wet the gas stream uniformly when
wetter wall designs such as that incorporated into the Pease-Anthony scrubber
are used. In very large sizes, the throat length required for uniformly
mixing the water droplets from the outer wall into the center of the throat
section is sufficiently long that all of the droplet acceleration has taken
34
-------
place before all of the gas has been wetted. This imposed a nearly immutable
upper limit on the size of simple Venturis. Various devices have been used
to minimize this problem, such as the elongated rectangular slot Venturi and
the flooded disk. These approaches are illustrated in Figures 2 and 3.
Unfortunately, the problem can be mitigated, but not eliminated com-
pletely, for large installation. Therefore, the possibility exists that in
any given application, a large unit will not function as well as small proto-
types or test units.
The MVS(TM) design, as illustrated in Figure 4, seeks to bring a new
approach to the field. First of all, the throat of the Venturi, including
both converging and diverging sections, is now the space between two round
parallel bars spaced about 1 inch apart. If 3/4 inch I.P.S. pipe is used for
the bars, as it usually is, the length of the Venturi in the direction of
flow is just over 1 inch. Now a series of bars placed side by side having a
depth of only 1 inch in the direction of flow, displace a whole Venturi that
can easily be more than 20 feet long.
The MVS(TM) modular design concept assures that no upper size limit is
imposed on the scrubber design. Each element of gas flow is treated identi-
cally, and there is substantially no variation in performance from the smal-
lest units to those of extremely large size. None of the elements incorpor-
ated into the design have inherently decreasing efficiency levels with in-
creasing size, nor to any of them have features which are inherently more
costly for scrubbers of extremely large or extremely small size. Rather,
all of the elements can be treated as identical "building blocks" and assem-
bled to match the requirements of any particular scrubbing job. The next
step was to miniaturize the spray-tower collection, which was done by using
Z-vanes. These steps were accomplished by an earlier patent(1) issued to
one of the authors.
The new patent, assigned to ARI, carried simplification one step further
by modularizing construction. Further, problems in the design of high energy
scrubbers relate to the difficulty in achieving de-entrainment of the scrub-
bing liquid from the gas stream after the throat of the Venturi. The
Pease-Anthony scrubber illustrated in Figure 1 often incorporates a cyclonic
separator in which the efficiency of de-entrainment varies inversely with
the diameter of the cyclonic section. Whereas small units incorporating this
design are highly effective, larger units are both cumbersome to design and
inherently less efficient in the separation of scrubbing liquid. Compound
entralnment separator designs which use the cyclonic action only for a
"rough cut" followed by a chevron or mesh mist eliminator have been used
successfully, but inherently require compromise between the optimum design of
the cyclonic section and the design of the chevron or mesh containment
vessel. Also, when such separators are located in a vertical gas stream, the
mechanism for draining collected water back into the scrubber becomes cumber-
some and, at very large sizes, it is nearly impossible to provide sufficient
slope to the streams to assure adequate drainage of liquid to the side walls.
This means that the entrainment separators must be operated at a velocity low
enough that the liquid droplets can fall downward against the rising gas
velocity without being re-entrained.
35
-------
With modular organization of the Venturi, all elements of gas in the
scrubber are treated exactly the same for the first time, so that all por-
tions of gas can be treated to maximum efficiency. It also means that scrub-
bers can be prefabricated to reduce cost. Prefabrication means that custo-
mers can be given almost instant delivery instead of the several months re-
quired in many instances. For very large units that are difficult to ship,
shipping of prefabricated elements is indicated. These can be quickly and
easily assembled on site. Lastly, a considerable number of scrubbers have
to be situated in extremely crowded plants. With standard scrubbers such
expensive tactics as tearing out walls on roofs are required, and, even then,
the scrubber sometimes has to be constructed on site. With the MVS (TM)
design, flat prefabricated sections can be brought in through existing pas-
sageways and can then be assembled in the field. The small size of the MVS
reduces the space and money required for installation substantially.
THEORY
When a sphere of diameter D^ (in Venturi scrubbers the sphere is usually
a water droplet) sweeps through a gas, the flow lines part to let it through,
as shown in Figure 5.
Small particulates suspended in the gas do not follow the flow lines
because of inertia. As shown, some impinge upon the droplet while others
that are further away from the center line of the sphere escape. If the
distance between symmetrical flow lines that carry particles that just barely
impinge is "X", an efficiency of collection can be as the ratio of areas
perpendicular to the gas stream, or X^/D^. Other properties of the gas
stream and of the particle must be taken into consideration, and these are
all assembled together in the correlation shown in Figure 6. Here target
efficiency, r\, is correlated against a Separation Number which can be ex-
pressed as:
18y Dd
Where: Ns = Separation Number, Dimensionless
d = Particle Diameter
Dd = Diameter of Droplet
Pp = Particle Density
Vo - Relative Velocity Between Droplet
and the Gas Stream
U = Gas Viscosity
Several things can be seen be examining the Separation Number. First,
collection efficiency varies as the square of the particle diameter, directly
as the relative velocity, and inversely as the diameter of the water droplet.
Particle density and gas viscosity cannot be changed, and in industrial
gases, particles vary in diameter over an enormous range. Nevertheless, as
efficiency varies as the square of particle diameter, it is very obvious that
efficiency falls off sharply for the smaller particles, and they are much
more difficult to collect.
36
-------
Design can influence the two remaining properties, V0 and Dd- In a gas
stream, a small particle reaches the velocity of its carrier gas almost in-
stantly. A water droplet, typically having a diameter of from 100 to 200P,
takes a little longer, but very quickly travels at substantially the same
velocity as its carrier gas. For V0, the relative velocity, to have an ap-
preciable value then, there must be acceleration in deceleration of the gas.
Droplet size can also be changed within limits. Nukijama and Tanasawa(2)
have developed a dimensional equation which can be used to estimate the
mass-mean diameter created in a enturi throat. For average conditions it
can be written:
Dd = 16,_5QO + 1.5 L1'5
V0
Where: Dd = Droplet Diameter, Microns
V0 = Differential Velocity, Ft/Sec
L = Liquid-Gas Ratio, Gallons per 1000 ACF of Gas
Even though the equation does not estimate Dd precisely, it does show
that Dd varies inversely as V0. A high value of V0 then raises the Separa-
tion Number directly by inversely affecting Dd which is the denominator.
As V0 is the relative velocity, the function of a Venturi scrubber is to
provide acceleration and the deceleration of the gas stream to obtain this
relative velocity. In conventional Venturis this occurs over a length of
many feet. In the MVS(TM), however, acceleration and the deceleration occurs
over the length of the throat which is now only about 1 inch. Consequently,
both acceleration and deceleration are much greater than before and effi-
ciency improves.
MVS(TM) SYSTEM DESCRIPTION
Implicit in the modular design concept is the ability to apply the
MVS(TM) scrubber to scrubbing jobs with substantially no special design con-
siderations relating to the size of the flowing gas stream or the physical
configuration needed to accommodate the scrubber.
For many years, it has been the practice in the scrubber industry to
find a limited number of scrubber sizes which can be applied on the basis
of past experience, and to custom design intermediate sizes, to tailor the
physical design of the scrubber to the required geometry of the process, or,
more frequently, to tailor the physical design of the process to match the
required geometry of the scrubber. With the MVS (TM) Modular Venturi Scrubber
approach, it is possible for the system designer to select the Venturi scrub-
bing components by a simple two-step procedure:
1. Select the pressure drop requirement to provide sufficient collec-
tion efficiency for the application in question.
2. Calculate the gas flow rate on a saturated basis and select the
37
-------
number of modules required.
The design engineer can, at this point, elect to purchase a scrubbing
unit tailored to hold the desired number of modules, or he can design an
enclosure to purchase throat and the entrainment modules to use within it.
A few very simple principles apply to the orientation of the elements.
The venturi throat element can be operated in any orientation whatever.
Gas flow is ordinarily horizontal, but can be vertical, upwards or downwards
or at an angle.
The mist eliminator consists of two sections as shown in Figure 7.
These sections are designed for operation with horizontal gas flow, and
with collected liquid draining downward at a slight angle tc vertical to the
bottom of the scrubber. If these elements are piled one upon another in
configuration more than two elements high, it is necessary to provide a
"gutter" at the bottom of each two-section layer to drain the collected water
to a side wall, thereby preventing re-entrainment in the high velocity gas
leaving the separator elements. Elements designed for vertical gas flow with
an angle of approximately 30° to the horizontal plane can be obtained for
configurations which accommodate vertical gas flow.
There is no requirement that the Venturi section and mist eliminator
sections operate with gas flow in the same general direction. For example,
a vertical down flow of gases through the Venturi scrubber can be matched
with horizontal gas flow through the entrainment separators as shown in
Figure 7 if this mechanical arrangement fits better into the process scheme.
Because it is economical to select standard module sizes for mass pro-
duction, the gas flow required may not match precisely the combination of a
convenient combination of elements. For example, 5000 saturated ACFM fits
conveniently in the smallest standard module. A flow rate of 50,000 ACFM
would, consequently, require ten standard modules. This fits conveniently
in an arrangement five modules wide by two high, but it may not be convenient
to accommodate a scrubber with this width to height ratio.
How can the size be modified so as to provide a nearly square cross-
section? Two ways of handling this problem are suggested. One approach is
to select nine modules, and to operate with a slightly higher pressure
drop than the category specified, or a combination of slightly higher pres-
sure drop at-slightly lower liquid flow rate. An alternative approach is to
use a configuration four modules wide by three high, wich would normally
have a capacity of 60,000 saturated ACFM. The oversized arrangement can be
modified by the insertion of a few clip-on covers between the rods, so as
to reduce the cross-section area available for gas flow in the desired ratio
of five to six. This is accomplished by simply blanking off every sixth
opening-
Problems of turndown are handled similarily. Venturi scrubbers are
basically constant flow devices. For application where substantial varia-
tion in gas flow rate is expected, the system should be designed to allow the
induction of ambient air or recycle of clean gas to the scrubber inlet to
38
-------
keep the gas flow volume approximately constant.
However, some circumstances will be encountered under which variation
gas flow rate must be accommodated. These can be handled on a continuous
basis by varying the liquid flow rate above and below the design flow rate
to accommodate variations of gas flow below and above than for which the
system was designed. In order to obtain substantial turndown, it is neces-
sary to resort to banks of spray nozzles which can be turned on as the flow
decreases so as to provide a uniform spray pattern.
The design criteria for the MVS(TK) Modular Venturi Scrubber system are
summarized in the table presented as Figure 8, and in the performance char-
acteristics in Figures 9 and 10.
CONCLUSION AND SUMMARY
The MVS(TM) Modular Venturi Scrubber design represents a substantial
improvement in the mechanics of application of the mature Venturi scrubbing
principle to particulate collection problems. The modular Venturi scrubber
approach provides for scrubbers varying in size from very small to very large
without any variation in performance due to geometry of the system.
The modular scrubber permits the design engineer to fashion a venturi
scrubber from standardized low-cost elements and arrange them to suit the
geometry of his application. Conversely, he may select from among standard
housing designs prefabricated units covering the size range to 500,000 ACFM.
ENDNOTES
1. U.S. Patent Number 3,488,039.
2. Nukiyama, S. and Tanasawa, Y. Transactions of the Society of Mechanical
Engineers (Japan) _4, No. 14, 86 (1938)'.
39
-------
Ct-KANED HAS
QUTLKT
CLEAN GAS
OUT
CYCLONIC
SEPARATION
WATER
OUTLET
FIGURE. 1
TYPICAL
PEASE-ANTHONY SCRUBBER
CONFIGURATION
FIGURE 2
RECTANGULAR THROAT
FIGURE 3
FLOODED DISK SCRUBBER
40
-------
setwwmu * _
FIGURE B
PARTICULATE MATTER IMPACT/ON
AT WATER DROPLET
FI6URE 6
EFFICIEHCY AS A FUMCT/OU OF
SEPARATION HUMBER
ENTKAINMEH7
SEPARATORS
FI6UKE &
TYPICAL D/Mfr/S/Of/1: OF
FIGURE 7
LINE DWAWMS OF A MVS™ SCRUBBER WITH VERTICAL
THROAT AND HORIZONTAL ENTRAPMENT SEPARATORS
SCFWI
sooo
to.ooo
2QOOO
3QOOO
45,000
&O.OOO
iAKtfe
A
li'-H"
li'-lf
f^'-S"
/s'-a'
15' &
is: s"
x *n
a
3-0"
3:0-
6-0'
£'-o-
4'-0-
f.O'
£4 O
C
3- Z"
4:&-
s-.io-
T-4"
B-O-
•
H
7-t,"
T-l,'
T-i,'
T-f
T-l,"
7-6"
/
/-'-'•
/-'3
£•;?
i'-f
i'-v'
? z"
j
Z'-O"
•?:o-
Z'-o
7'-0~
?O
?' (7
41
-------
I
-
40
K
Vi 1O
^c
if
sst//te
WtHSii '
0&PW//O0O
2
size.
PERCENT Of X*rtO
GAS A/ OU
PRESSURE
F/GUXS fO
42
-------
PERFORMANCE OF PARTICULATE SCRUBBERS AS INFLUENCED "BY
GAS-LIQUID CONTACTOR DESIGN AND BY DUST FLOCCULATION
By: Konrad T. Semrau
SRI International
Menlo Park, CA 94025
Robert J. Lunn
Donaldson Company, Inc.
Minneapolis, MN 55440
ABSTRACT
Several types of gas-liquid contactors were compared experimentally,
using both flocculated and deflocculated test dusts. Performances were
compared on the basis of the effective friction loss across the contactors.
The various devices generally gave very similar or identical performances
except over some limited ranges of operating variables in which inferior
performance was encountered. A fiber-bed contactor gave superior performance,
apparently by superimposing filtration upon scrubbing. Dust flocculation
afforded increased collection efficiencies, but with some collectors partial
deflocculation occurred before contact of the dust with liquid was effective.
INTRODUCTION
A recent paper (1) summarized the conclusions from various investigations
of the relationship of particulate scrubber efficiency to the energy expended
in gas-liquid contacting (the contacting power). The present paper presents
a summary of the results of tests made under a common set of conditions with
a variety of gas-liquid contactors. The tests were made under laboratory
conditions using standardized test dust in both dispersed and flocculated
conditions.
EXPERIMENTAL METHODS AND EQUIPMENT
The primary investigation was made with the DP scrubber, which had a
maximum air flow capacity of about 180 ftVmin and is shown schematically in
Figure 1. The DP scrubber was actually a test assembly in which the gas-
liquid contactor and the entrainment separator could be changed as desired.
Atmospheric air carrying dust entered the scrubber through the sparger tube,
the lower end of which was immersed in water in the contactor pot. The air
and entrained water then passed through the contactor (orifice, venturi, or
fiber bed) located at the gas exit of the contactor pot and flowed upward to
the entrainment separator. Two different entrainment separators were used.
The first was a simple cyclone. The second was a combination of an inertial
separator followed by a wire-mesh separator, shown schematically in Figure 1.
The inertial separator embodied a simple 90 change of direction for the air
stream. The wire-mesh separator consisted of the upper part of a Donaldson
43
-------
A-8073 engine-intake air cleaner.
The DP scrubber effectively incorporated multiple stages of gas-liquid
contacting, including the sparger, the designated contactor, the duct to the
entrainment separator, and the entrainment separator itself. In one arrange-
ment, termed the impinger, a nozzle was fitted to the immersed lower end of
the sparger tube.
The contactor pot also served as the reservoir for the scrubbing water,
which was neither added nor withdrawn during a test run. There was no pro-
vision for measuring the circulation rate of the water. The gas pressure
drop through the scrubber could be changed by varying the air flow rate, the
water depth in the contactor pot, or the contactor.
Tests were also made with a Donaldson A-8073 engine-intake air cleaner,
using both water and lubricating oil as scrubbing liquids.
Additional tests were made with a venturi scrubber (Figure 2) made from
glass and provided with several alternative water-fed arrangements. A
metered flow of water was fed once through to the scrubber. The cyclone was
used as the entrainment separator.
The general test procedure used was that prescribed in the SAE Air
Cleaner Test Code (2). Weighed charges of dust were fed to the scrubber, and
the entire exit air stream was filtered to determine the penetration. The
principal test dust used was the standardized Air Cleaner Fine (ACF) test
dust, 39% of which is in the size range 0 to 5 t*m. Some tests were also made
with the 0 to 5 ^m fraction of dust classified from the ACF dust.
The dust charges were fed from a vertical elutriator, which delivered
the dust in a flocculated state. To obtain a dispersion of deflocculated
dust, the stream from the elutriator was drawn through and discharged from a
compressed-air ejector.
RESULTS AND DISCUSSION
The initial tests, which provided a baseline for the remainder of the
investigation, were made with the DP scrubber, using the sparger alone and in
combination with the orifice and venturi contactors (Figure 3). Performance
curves were established for both flocculated and deflocculated ACF dust, and
these curves from Figure 3 are repeated for reference in the following
Figures 4-7. All three contactor arrangements gave essentially the same
results for given levels of the effective friction loss. The sparger/orifice
combination alone was used to establish a performance curve for deflocculated
0-5 Mm dust (Figure 3).
In the initial tests, the cyclone was used as the entrainment separator.
Later tests were made with the combination inertial and wire-mesh separator.
When compared, the two sets of results agreed if the gas pressure drop across
the wire-mesh separator unit was subtracted from the total measured pressure
drop. Evidently, insufficient water reached the mesh section to produce
44
-------
significant gas-liquid contacting there. Hence, the energy loss across the mesh
section did not contribute to scrubbing efficiency.
The A-8073 air cleaner had a contacting scheme somewhat similar to that
of the DP scrubber. The air entered the cylindrical air cleaner through a
concentric sparger tube that dipped into an oil bath in the bottom of the
cleaner body, then flowed upward through a thick pack of screens that formed
the principal gas-liquid contactor. The scrubber air then flowed upward
through the wire-mesh separator for removal of the residual droplets of
entrained oil. Apparently, the quantity of oil reaching the wire-mesh sep-
arator was small. The only method for changing the pressure drop across the
A-8073 unit was by varying the air flow. At high air flow rates, the
velocities of air entering the oil bath were probably high enough to sweep
the inner cup in the oil bath free of oil at the surface where the air stream
turned upward.
The A-8073 unit was tested with both flocculated and deflocculated ACF
dust. No difference could be distinguished between the results obtained with
oil and with water as the scrubbing liquid. As the other studies had sug-
gested, the pressure drop across the wire-mesh separator section was sub-
tracted from the total pressure drop to give the effective friction loss,
which reached only 9 inches of water at the maximum practical air flow rate.
The efficiencies obtained significantly exceeded those obtained in the base-
line studies with the sparger/orifice contactor in the same range of effec-
tive friction loss (Figure 4). To simulate the same scrubbing action, the DP
scurbber was fitted with a fiber-bed contactor composed of steel wool, and
with this unit it was feasible to operate in a higher range of effective
friction loss without using inconveniently high air flow rates. The data
points obtained with the DP/fiber-bed scrubber on deflocculated ACF dust
clearly fell along the same performance curve as did the data points for the
A-8073 (Figure 4). With the flocculated ACF dust, the efficiency of the
A-8073 unit actually decreased at effective friction losses exceeding about
4 inches of water, which correspond to increasing velocities of the entering
air. Comparison of these results with those for deflocculated dust clearly
indicate that deflocculation of the initially flocculated dust was taking
place in the A-8073 before effective contacting with liquid was achieved.
Further confirmation was obtained by tests with the DP/fiber-bed arrangement
at higher levels of effective friction loss; the resulting data points fell
along a rising performance curve consistent with that for the A-8073 when
operated in the range of effective friction loss under 4 inches of water
(Figure 4).
The use of a fiber-bed as a contactor gave definitely superior perfor-
mance, suggesting that filtration (particle deposition on fibers) was being
added to scrubbing (particle deposition on droplets). However, extrapolation
of the performance curve for fiber-bed scrubbing of deflocculated ACF dust
suggests that the superiority of performance might disappear in the effective
friction loss range of 30-35 inches of water. In that range, the residual
dust would be in the lower submicrometer range at which the coarse steel wool
fibers would give only very low collection efficiencies. Extrapolation of the
performance curve for flocculated dust indicates relatively superior perfor-
mance continuing into a much higher range of effective friction loss, which
45
-------
would be consistent with the larger effective particle size of the
flocculated dust.
With the impinger contactor, the energy for gas-liquid contacting was
derived from the jet of air entering the contactor pot. The performance
curves with both flocculated and deflocculated ACD dust showed flatter slopes
than the corresponding baseline curves obtained with the DP/orifice config-
uration, with a trend to relatively poorer performance with increasing effec-
tive friction loss derived from increasing air jet velocities (Figure 5). A
study of the jet velocity-pressure drop relationship indicated that the higher
jet velocities may have coincided with increasing channeling of the air jet
through the liquid in the contactor pot, which could account for the rela-
tively poor performance.
The first tests with the venturi scrubber (Figure 6) were made with
water injection at the four throat taps. The venturi was placed in a hor-
izontal orientation. Visual inspection indicated that most of the con-
tacting and energy dissipation took place in the diffuser. A marked transi-
tion in the flow pattern took place at a liquid-to-gas ratio of about 11.5
gal/1000 ft^. At ratios below this value much of the water appeared to be
held in a thick boundary layer, or eddy, in the diffuser. Some channeling
of dust-laden air through the contacting zone may have taken place under
that condition. At liquid-to-gas ratios above the critical value, the water
broke up into a dense spray that appeared to fill the diffuser uniformly.
In tests made with liquid-to-gas ratios under the critical value for the
flow transition, both deflocculated and flocculated dusts were collected with
efficiencies far below those obtained in the corresponding baseline tests
(Figure 6). Nevertheless, with liquid-to-gas ratios above the critical
value, the data points for collection of deflocculated ACF dust are in excel-
lent agreement with the baseline performance curve. In corresponding tests
with flocculated ACF dust, the three data points for tests with venturi
throat velocities of 235 to 260 ft/sec also gave excellent agreement with the
baseline performance curve. On the other hand, the three data points for
tests at throat velocities of 280 to 320 ft/sec fall substantially below the
baseline curve, probably indicating partial deflocculation of the dust in the
throat upstream of the water injection points.
A few additional tests were made with water injection at various other
locations (Figure 7). With feed at the four forward throat taps, virtually
the entire length of the venturi throat was used in gas-liquid contacting.
With all the water injection configurations, the data points were generally
in reasonably good agreement with the baseline performance curve. Upstream
water feed with a liquid-to-gas ratio of 51 gal/1000 ft^ and a throat velo-
city of 155 ft/sec gave a data point falling along the established perfor-
mance curve. On the other hand, the same configuration with a liquid-to-gas
ratio of only 1.3 gal/1000 ftj and a throat velocity of 400 ft/sec gave an
efficiency far below the baseline value, despite the visual appearance of a
good distribution of spray entering the venturi throat.
46
-------
CONCLUSIONS
With the exception of the A-8073 air cleaner and the DP/fiber bed
scrubber, all the scrubber configurations gave essentially the same perfor-
mance when operating in their best modes. Nevertheless, at least three of
the contactors displayed performances that were less than optimum over some
part of their ranges of possible operating conditions. With the DP/impinger
scrubber and the venturi scrubber with throat feed taps, the inferior per-
formance may have been associated with channeling of the air flow. On the
other hand, the experiments with the venturi scrubber with upstream water
feed appeared to indicate that low water rates as such tended to give infer-
ior performance, even in the absence of a flow pattern that could properly
be termed "channeling." In general, there were at least indications that low
or very low liquid-to-gas ratios may coincide with inferior performance.
Well-established performance curves such as those of Figure 3 appear to
represent the maximum performance that can be obtained from a scrubber on a
given dust or aerosol. Deviations such as those observed in this investiga-
tion have been in the direction of inferior performance and appear to be
characteristic of some range of operating conditions rather than of the par-
ticular device. The fiber-bed contactor was the only device that showed a
positive deviation from the performance curve, and this was apparently asso-
ciated with the introduction of a particle deposition mechanism not normally
operative in conventional scrubbers.
ACKNOWLEDGMENT
This investigation on which this paper is based was supported by the
Donaldson Company, Inc., Minneapolis, Minnesota.
ENDNOTES
1. Semrau, K. T. Practical Process Design of Particulate Scrubbers.
Chem. Eng. 84: 87-91, September 26, 1977.
2. Society of Automotive Engineers. Air Cleaner Test Code, SAE J726b.
SAE Handbook, Part 2. Warrendale, PA, Soc. Auto. Engrs., 1979.
47
-------
Sparger
Wire-Mesh
Entrainment.
Separator
I i
L J
•^
I
y
**r-
/Contactor
L Contactor
Pot
JA-363522-1
FIGURE 1 DP SCRUBBER WITH WIRE-MESH
ENTRAPMENT SEPARATOR
Throat
Feed
Forward
Throat-
Feed Taps
Upstream
Feed Tap
Single
Throat
Feed Tap
Alternative
Venturi
Convergence
JA-363522-2
FIGURE 2 VENTURI GAS-LIQUID CONTACTOR
C/D
H
z
DC
LU
LL.
CO
CC
h-
DC
LU
CO
Contactor
Sparger
Sparger/Orifice
Sparger/Venturi
Defloc. Floe. Defloc.
ACF ACF 0-5
•
A
D
I
1 10 70
EFFECTIVE FRICTION LOSS — inches of water
JA-363522-3
FIGURE 3 PERFORMANCE CURVES FOR DP SCRUBBER
48
-------
CO
DC
HI
LI-
CO
o
DC
LU
GO
S
D
A-8073
Sparger/Fiber Bed
1 —
1 10 70
EFFECTIVE FRICTION LOSS — inches of water
JA-363522^1
FIGURE 4 PERFORMANCE CURVES FOR A-8073
AND DP/FIBER BED SCRUBBERS
8
tr
LU
LL.
CO
DC
DC
LU
00
Defloc.
ACF
o
Floe.
ACF
1 10 70
EFFECTIVE FRICTION LOSS — inches of water
JA-363522-5
FIGURE 5 PERFORMANCE CURVES FOR DP/IMPINGER
SCRUBBER
49
-------
C/3
DC
LU
LL
C/5
z
<
QC
t-
U-
O
oc
LU
CO
Liquid-to-Gas
Ratio
Below Flow
Transition
Above Flow
Transition
Defloc.
ACF
1
Floe.
ACF
1 10 70
EFFECTIVE FRICTION LOSS — inches of water
JA-363522-6
FIGURE 6 PERFORMANCE CURVES FOR VENTURI SCRUBBER
WITH THROAT WATER-FEED TAPS
oc
LU
LL.
LO
QC
LU
CO
Water Feed Points
Forward Throat Taps
Single Throat Tap (1/4-in.)
Single Throat Tap (1/16-in.)
Single Upstream Top
High Water Rate
Low Water Rate
I
D
Defloc.
ACF
o
v
A
0
D
Floe.
ACF
1 10 70
EFFECTIVE FRICTION LOSS — inches of water
JA-363522-7
FIGURE 7 PERFORMANCE OF VENTURI SCRUBBER
WITH VARIOUS WATER FEED POINTS
50
-------
INVESTIGATION OF VENTURI SCRUBBER EFFICIENCY AND PRESSURE DROP
By: R. Parker, T. Le, and S. Calvert
Air Pollution Technology, Inc.
4901 Morena Blvd., Suite 402
San Diego, CA 92117
ABSTRACT
The Venturi scrubber performance model (EPA-600/2-77-172) gives good re-
sults when used to predict the performance of industrial scrubbers. However,
insufficient data have been available to make a detailed parametric evaluation
of this model. A parametric study is needed in order to properly predict Ven-
turi scrubber performance in nonconventional environments such as high pres-
sure gas streams.
Pilot plant data on a 7.6cm (3 in.) throat diameter Venturi scrubber are
presented. Flow rates from 10 to 20m3/min (350-700 CFM) and liquid-to-gas
ratios from 1 to 3£/m3 (7.5 - 22.4 gal/MCF) were used. Throat length and dif-
fuser length were also varied.
These parameters resulted in throat velocities from 30 to 70m/s and pres-
sure drops from 10 to 50 cm W.C. Experimental data showed higher penetrations
than predicted, especially for particles larger than 1 or 2 ymA aerodynamic
diameter.
INTRODUCTION
The Venturi scrubber performance model (1) gives generally good results
when used to predict the performance of industrial scrubbers. However, field
tests do not allow sufficient control over operating parameters, and field
test data are not sufficiently accurate to enable an adequate evaluation of
the model. On the other hand, available laboratory data are scarce and are
generally for too small a scale to be representative of industrial scrubbers.
Experimental performance data for a pilot scale Venturi scrubber are needed
in order to evaluate the available mathematical models and to provide a more
adequate basis for further refinement of models.
The purpose of this project was to evaluate the Venturi scrubber model
on a pilot scale and obtain scrubber performance data which could be used as
a baseline for evaluating Venturi scrubbers operating at high gas pressure.
High pressure Venturi scrubbers are important components in many advanced coal
conversion processes currently being developed. This work has been funded un-
der a contract with the U. S. Department of Energy.
THEORY
Yung, et al (1) developed a revised model for Venturi scrubber perform-
ance from the basic equations governing particle collection and pressure drop
derived by Calvert (2, 3) and Boll (4). Several assumptions were made:
1. The flow is one-dimensional, incompressible and isothermal.
51
-------
2. Liquid drops are uniformly spread across the duct and the drop
diameter is invariant with axial distance.
3. Drops are of uniform diameter.
4. There is no wall loss of liquid drops.
5. At any cross section of the scrubber, liquid fraction is small.
Therefore, at any location gas velocity can be calculated by
the following equation:
G PGA
Pressure Drop
Yung found that pressure drop predictions by a modified Calvert's equa-
tion and by Boll's equation agree with experimental data. The modified
Calvert's equation has the following form:
AP = 1 x 10"3i F UG£
where:
AP = pressure, cm W.C.
Up = gas velocity in the throat, cm/s
F = fraction of the gas velocity which is attained by
the liquid drops, dimensionless
Particle collection by inertial impaction depends on the relative velocity
and the drop holdup, and both of these decrease as the drops approach the gas
velocity (F-KL) .
Yung defines a dimensionless throat length, L, by the following:
3 £ C p
L = t Do G (3)
2ddPL
£ = throat length or distance between liquid injection
point and the exit of throat, cm
C^ = drag coefficient at the liquid injection point
pp = gas density, g/cm3
d, = Sauter mean drop diameter given by the empirical
correlation of Nukiyama and Tanasawa (5), cm
p, = liquid density, g/cm3
Yung shows that L = 2 or 3 is the optimum throat length and a value of F =
0-82 accurately predicts pressure drop.
Particle Penetration
Most industrial Venturi scrubbers have dimensionless throat lengths
52
-------
between 2 and 3, and the infinite throat length model may be used to
their penetration. The Venturi scrubber penetration for an infinite
length is giVen by:
/ 0.7 \
4K + 4.2 - 5.02 K0'5 I K /tan
. B P° PO \ po /
predict
throat
Pt
= exp
0 • 5
K
po
(4)
where:
^ = penetration for particles with diameter d , fraction
B = dimensionless liquid-to-gas ratio parameter
B = -^ —
w
Pa Gt
10
Do
QL = liquid flow rate, m3/s
QG = gas flow rate, m3/s
K = inertial parameter =
d = particle aerodynamic diameter, ymA
Up = velocity of the gas in the throat, cm
U = gas viscosity, g/cm-s
(5)
(6)
EXPERIMENT
Approach
The experimental setup is shown in Figure 1. The Venturi throat was 7.8
cm in diameter and 12, 27, 43, or 58 cm in length. The diffuser section was
50 cm long with an exclusive expansion angle of 7°. The entrainment separ-
ator consisted of four rows of chevron baffles arranged in a zigzag pattern.
Fly ash was redispersed, then injected directly into the blower inlet to
promote good mixing. The flow rate was varied from 10 to 20 m3/min (350 to
700 CFM) and the liquid-to-gas ratio was set at 1, 2, or 3 £/m3 (7.5, 15,
22.5 gal/MCF). The throat velocity varied from about 30 to 70 m/s with most
runs falling between 40 and 50 m/s. The scrubber pressure drop (Venturi only)
varied from 10 to 50 cm W.C.
Cascade impactors were used at the scrubber inlet and outlet to measure
particle mass concentration and size distribution. These data were used to
compute grade penetration curves, which were used to determine the scrubber
cut diameters.
53
-------
Because our data are most consistent between about 0.8 and 3.0 umA, we
used an extrapolation technique to obtain cut diameters below 1 umA. This
technique assumes the general relationship:
Ptd = exp (-A dp| ) (7)
where "A" is a constant. We used the measured penetration at 1 ymA to predict
the effective cut diameter. Scrubber performance was evaluated in terms of
the cut diameter as a function of pressure drop.
Results
The pressure drop data are plotted against predictions in Figure 2. The
agreement between theory and data is excellent even for the short throat
lengths.
Scrubber cut diameters are plotted as a function of pressure drop in Fig-
ures 3, 4 and 5 for liquid-to-gas ratios of 1.2, 2.2, and 3.0 £/m3 respective-
ly. The solid lines are theoretical predictions from Yung's model.
Discussion and Conclusions
For all conditions studied, the experimental cut diameters are larger
than predicted by theory. This implies that the theory over-predicts the
Venturi scrubber collection efficiency. For a liquid-to-gas ratio of 1 to 2
Vro3 (the most common range for industrial scrubbers) the predicted cut dia-
meters are approximately 20% smaller than those measured. The pressure drop
theory does a good job of predicting the experimental data.
ENDNOTES
1. Yung, S.C0 et al., "Venturi Scrubber Performance Model", EPA 600/2-77-172,
August 1977. Also J.A.P.C.A., 27, 4, 1977 and E.S.&T., 12, 456, 1978.
2. Calvert, S.,. "Scrubbing," Chapt. 6 in "Air Pollution," A. Stern, Vol. IV,
1977.
3. Calvert, S. et al., "Scrubber Performance for Particle Collection, " AlChe
Symposium Series, Volume 70, No. 137, 1970.
4. Boll, R0H., "Particle Collection and Pressure Drop in Venturi Scribber, "
Ind. Eng. Chem. Fund 12_,_ 40, 1973.
5. Nukiyama, S., and Tanasawa, Y. Transactions of the Society of Mechanical
Engineers (Japan). Vol. 4, 86, 1938.
54
-------
SUMP
DUST
GENERATOR
OUTLET SAMPLE
JL- I
ENTRAPMENT
SEPARATOR
BLOWER
INLET
VENTURI SAMPLE
FIGURE 1. EXPERIMENTAL SETUP
55
-------
E
o
o
UJ
DC
CO
<
LLJ
60
50
40
30
20
10
I I I I I I I I 1 I I I I I I I I I I I I
•
: THROAT LENGTH
•
1 O 12cm
; A 27
: O 43
: V 58
: O
O
o -
i i i i i i i i i i i i i i i i i i i i i i i i i i i r
0 10 20 30 40 50
PREDICTED AP, cm W.C..
60
FIGURE 2. PRESSURE DROP DATA
56
-------
<
E
cc
LU
H
LU
Q
H
Z>
o
3.0
2.0
1.0
0.5
0.4
0.3
0.2
T T
^
THEORY
THROAT LENGTH
O 12cm
V 27
A 43
0.1 I I I I I
j L
5 10 20 40 60
PRESSURE DROP, cmW.C.
FIGURE 3. CUT-POWER PLOT FOR L/G=1l/m3
57
-------
cc
LU
H
LU
H
D
O
3.0
2.0
1.0
0
0.5
0.4
0.3
0.2
0.1
THROAT LENGTH
O 12cm
V 27
A 58
O 43
5 10 20 40 60
PRESSURE DROP, cm W.C.
FIGURE 4. CUT-POWER PLOT FOR L/G=2l/m3
58
-------
CC
LU
I-
III
o
I-
D
O
3.0
2.0
1.0
0.5
0.4
0.3
0.2
0.1
i i i i
I I T
THEORY
O
THROAT LENGTH
O 12cm
A 43cm
i i i i
5 10 20 40 60
PRESSURE DROP, cm W.C.
FIGURE 5. CUT-POWER PLOT FOR L/G=3l/m3
59
-------
SCRUBBER TECHNOLOGY AND THE INTERACTION
OF A UNIQUE STRUCTURE AS HIST ELIMINATOR
BY: GEORGE C. PEDERSEN, P.E.
KIMRE, INCORPORATED
P.O. BOX 570846
PERRINE, FL 33157 USA
ABSTRACT
Mist eliminators should not be an afterthought addition to a scrubbing
system. As the last line of defense, they should be an integral part of
the control technology. The role of Kimre's unique patented structure as
a mist eliminator for use with other scrubbing devices is considered.
The wider range of physical properties now available yields new capabilities
in systems design. This media is being used successfully on four continents
and is a standard for many United States scrubber manufacturers.
INTRODUCTION
We show how the scrubber situation and the scrubber technology used interact
on the selection of a mist eliminator. Emphasis will be placed on the
B-GON^M patented mist eliminator structure. In order to explore this
interaction, first of all we will consider just what this structure is,
familiarize the reader with the material; and then also familiarize him
with the use of this material. Basic considerations in the selection of
mist eliminator structure, as it relates to the scrubbing technology will be
emphasized by case history analysis. Conclusions regarding these con-
siderations are made.
There is a considerable overlap between scrubbing technology and mist
eliminator technology. We are attempting to consider only the mist
eliminator aspects, but it is impossible to completely differentiate
functions. The case histories given are examples drawn from Kimre's
extensive experience as consultant and supplier to the industry over the last
six and a half years.
BACKGROUND
In the beginning -- which in this case was quite some time ago -- there was
only an idea: If it were possible to produce a structure made of
monofilaments, wherein all the monofilaments were perpendicular to the flow
of the gas, then this should be an efficient mist eliminator. The technical
and commercial development of this concept led to the evolution of a wide
range of structures. These structures can be made to vary in void fraction
up to 97^; they are available in fiber diameters from 31 microns (just
slightly more than 1 mi 1) up to 62 mi 1s -- a range of 60 to 1 in fiber
diameter (Figure 1).
60
-------
TM
KIMRE B-GON FIBERS
©
2 MIL 4 M|L
16 MIL 32 MIL
37 MIL
62 MIL
1 MIL 11 MIL
BRINKS H.P. KNITTED MESH
TI-32
KIMRE INC.
PERRINE FLA.
DUMPED PACKING
FIBER SIZE
COMPARISON
FIGURE 1
-------
The commercial evolution of the structure's use led to the formation of
Kimre, Incorporated, with its own manufacturing facilities South of Miami
Florida.
Technical Development
The structure used in the B-GON Mist Eliminator and the KON-TANE patented
Tower Packing products are based on a unique means of interlocking fibers
so that they maintain a specific geometric orientation. The simplest form
of this structure is diagrammed in Figure 2. The methodology of producing
these structures from a wide range of fiber diameters was developed, but at
that time there wasn't any proof that the structure actually functioned.
Commercial tests were run in the Spring of 197^ by Calspan Corporation, and
the performance parameters were verified.
It was demonstrated that this structure was amenable to analysis, based on
single-fiber studies by others for collection of mist and particulates.
There is a curious exception when predictions are compared to actual results.
For relatively "small" drops for which the effectiveness is expected to drop
off in a pronounced fashion, the decrease in performance is nowhere near as
rapid as is anticipated. We have theories for why this occurs, but no bona
fide answers. The specific surface areas and fiber diameters range over far
too wide a scale to describe velocity units by the conventional
"Souder-Brown" Equation. A method was developed to predict pressure drop,
flooding, and efficiency for the different configurations for almost any
range of physical conditions.
It was also found that the structure could be made at high void fractions
with smaller fiber diameters than were available with other manufacturing
methods. The manufacturing methods developed also led to the ability to
make the product in large pieces, which has technical advantages during
installation, and to prevent problems of bypassing around the edge of
sectioned pads. Other technical advantages are its strength and stability.
The breaking strength of some of the coarser materials exceeds 1,000 pounds
per inch of width. In some cases, the materials can take enormous loads,
and this led to a supplementary business in the support of beds of other
materials.
Commercial Development
Kimre, Incorporated was founded to commercialize these structures starting
with the basic structure and personal financing (with no government
assistance), manufacturing equipment was developed, a manufacturing plant
was put into operation, and an international sales force was developed.
Through our contacts with almost every part of the air pollution control
and chemical processing industries, Kimre's technology and products are
well-proven. Today, the products are specified extensively by many scrubber
manufacturers,engineering/constructor firms, consultants, and a variety of
other buying influences.
62
-------
ON
DIAGRAM OF FLOW THROUGH
B-GON STRUCTURE
FIGURE2
-------
Kimre is not a scrubber manufacturer. We supply control concepts and
internals applicable to air pollution control and chemical processing
operations. The control concepts utilizing our products and technology are
wider-ranging than other basic approaches. Literally, the products have
been used from the extreme of keeping out birds to removing half-micron-
size particulate or mist, to the removal of "rain", or the removal of re-
acting glue; from one millimeter of mercury to 2500 psi; from 100 degrees
below zero to 300 degrees above zero; from less than 1 cfm to over half a
million cfm. Chemical resistance requirements have varied from removal of
pure water to removal of aqua regia, including radioactive operations.
Physically, the B-GON™ Mist Eliminators span a wider range of physical con-
figurations and usability than is available from any other single source.
In cases where we didn't already produce something that would satisfy a
need, custom-designed and custom-made products have been used.
«... DISCUSSION
For the purposes of this paper, we define a mist eliminator to be a device
primarily intended for the removal of relatively non-viscous (less than
10,000 cp) mist from a gas stream. Such mist might include entrainment,
fine mist intentionally allowed to pass through other parts of the control
device, or condensation mist. While it is recognized that mist may include
particulates collected in the rest of the scrubber, the mist eliminator's
function rs not to remove otherwise uncaptured material from the gas -- i.e.,
it's not an absorber or particulate scrubber in its own right. It should be
recognized of course, that in many cases it's hard to differentiate
between the mist eliminator and the scrubbing section.
Based on this interpretation of a mist eliminator, and considering situations
where the particulates are being collected, then:
It must be anticipated that the mist eliminator should never be a more
efficient scrubbing device than the scrubber.
To ignore this rule invites a disastrous installation. We have noted
many times in conversations that otherwise knowledgeable people will install
a mist eliminator that's going to be so efficient in collecting remaining
particulates that it will plug in short order.
As secondary considerations we offer:
1. Consider everything.
2. Don't overdo efficiency.
There are a lot of considerations that go into the selection of a mist
eliminator; all of these must be considered, even though the answers to the
consideration in many cases is routine. However, in others the decision
becomes surprisingly complex.
64
-------
1 . Flooding.
Easy to say, but what's it mean?
2. Pluggage resistance.
3. Access! bi 1 i ty .
4. Spatial and temporal distribution of the gas flow.
5. Corrosion.
6. Temperature.
7. Bypassing.
8. Supports.
A great deal of "black art" has materialized over the selection of mist
eliminators. It usually isn't necessary, and should be avoided wherever
possible. In almost all cases, it's possible to at least make a narrow
selection of possibilities. When based on thorough analysis, it's surprising
how often the final decision bears no resemblance to the original idea for
the mist eliminator.
Perhaps the best way to consider these interactions is to look at specific
exampl es :
A. Ferti 1 izer Plant
In 1975, Kimre bid on a mist eliminator for a 22-foot-d iameter scrubbing
tower and lost the job for less than $200. The mist eliminator was
specified only as "6- inch-thick polypropylene".
In early 1977, the unit started up; and in about 7 weeks the mist eliminator
plugged up so solidly that the pressure drop caused the support beams to bend
upward, and it tore the mist eliminator sections loose, causing significant
damage to the inside of the vessel and duct. This was on top of 20 feet of
packing. The original knitted mesh mist eliminators could not be cleaned
and had to be discarded.
Those familiar with fertilizer plant operations recognize a classic case of
difficult service. The incoming gases contain soluble fluorides which
react in the liquid to give a silica-containing deposit. Incoming gases
also contain dust, and the scrubbing liquid itself contains the same
compound, and others besides, which cause substantial problems. However,
one does not expect to find tar-like materials.
At the request of the owner, Kimre personnel visited the site and made a
number of recommendations:
1. A man-supporting bottom grid be installed, and that a B-GON Mist
Eliminator in very large sections, each approximately 6 feet wide, be used.
The use of the bottom grid allows the mist eliminator to be installed in
large sections, and merely rolled out piece by piece. It also provides
structural integrity for the entire assembly. The large pieces of B-GON
Mist Eliminator can be removed through the manway very readily. The
ultimate objective of all this is to minimize the impact of an upset
situation, and so the changeover of mist eliminators could be made as
rapidly as possible.
65
-------
2. They obtain 2 mist eliminators, with one as a standby, each comprised
of k layers of style 37/97 and k layers of style 16/97-
3. A lightweight top hold-down grid, to hold the entire assembly in place.
i». Pressure taps be added to provide a means to monitor the pressure
drop increase.
The mist eliminator was obviously a better scrubber that the scrubber
(reaction products were detected in the mist eliminator). By putting the
37/97 on the downstream -- i.e., bottom — side of the mist eliminator, it
was anticipated this would remove the great majority of the water drops
prior to collection on the finer 16/97 material. Also, since the 37/97
material drains liquid at a very high rate, the total amount of water
held in the pad at any time would be greatly reduced, decreasing the rate of
the deposition from liquid phase reactions. The combination provided the
ability to make a changeover at a very rapid rate, and would allow the mist
eliminator sections to be cleaned of almost anything that could get on them.
Immediately after start-up, the mist eliminators worked fine, but after 8
weeks' service, excessive pressure drops occurred across the mist
eliminator, and the change-out was made. The changeover to the spare mist
eliminator went very smoothly, but when we examined the mist eliminator
there was a great surprise.
The bottom material, the 37/97, contained a small amount of a light red
scale removable by flexing the pads and washing them. However, the 16/97
on the top side of the mist eliminator was completely plugged up with a
tarry black material. The user developed a way to clean these by soaking
them in a commercial cleaning solution, so that this tarry material was
completely dispersed. The pads were exchanged on an approximate 8-week
cycle.
After about a year the customer decided to purchase additional 37/97 and
discontinue the use of 16/97 material. Our customer continued along very
happily with the installation. I was notified in 1980 that they had found
the source of the original pluggage problem. They were using an oil-based
de-foamer in the recycled scrubbing liquor. It worked well as a de-foamer,
but reaction products plugged the 16/97 part of the mist eliminator. A
change of de-foamer eliminated the original problem.
It is interesting that the tarry material did not collect on the coarser
material -- it only occurred on the finer material, style 16/97, thus
emphasizing in a negative way the selectivity due to fiber diameter.
We think it's obvious that the original specification calling for only 6
inches of a polypropylene mist eliminator was inappropriate. Considering
the technology that was available, or at least well known to most people, it
was, however, understandable.
B. Condensation Mist
66
-------
It frequently occurs that relatively hot, moisture-laden gas is cooled
in a scrubber. I wonder how often anyone ever thinks about the condensation
mist that can occur in such situations. One can get condensation mist where
cooling a gas with liquid, when heating a cool gas with hot liquid, or just
by mixing 2 saturated gas streams. Such condensation situations can by
mystifying.
One of our customers came to us with an emission problem which they had
been attempting to solve for over 2 years. They had added more packing to
a crossflow scrubber and had provided additional cooling, to no avail.
Kimre personnel found this perplexing, and suggested a series of tests to
derive the cause of the actual problem. After three false starts, it was
finally determined that mist was going into the mist eliminator at a drop
size too small for the mist eliminator to catch. This mist was then going
into the blower, where it was warmed up slightly, and water evaporated. The
contaminants ended up in the stack as a combination of gaseous and particu-
late components.
The solution to that problem was very simple: We merely added a mist
eliminator section efficient enough to catch that size droplet. The cost
was completely insignificant, and the results very dramatic — a factor of
5 reduction in emissions.
Consider Everything
In this case, many different alternatives had been considered in the
original analysis, but one crucial alternative was missed. The original
mist eliminator would never have removed the sub k-$ microns which actually
caused the emissions.
In the example given above, the first material we installed was too fine
and plugged up quickly. We had to back off to a coarser material adequate
to give the required performance. The most common error that we see is
overdoing the efficiency by specifying too fine a material.
Analagous problems are commonplace on hydrochloric acid scrubbers, which
are usually complicated by the presence of large amounts of entrainment.
Recently, one of our accounts in West Germany, ordered -- without specifi-
cation for use — a mist eliminator comprised of many layers of our style
k/36. After testing It, he complained bitterly that it didn't work, and
then provided us with the process information. We had to report back to
them that it would have been impossible for the unit he ordered to ever
have removed any mist. Given the conditions and the vessel size, this very
fine mist eliminator was always flooded. It would take the mist that was
there and make it from very small drops into large drops, but would not
remove anything. We suggested that he throw that one away, and replace it
with a composition of our design which included allowance for flooding.
The curious part of all that was that the unit which would work cost
less than one-fifth as much as the original design specified by the user.
67
-------
C. Flux-Force Condensation
This is one of those operations in which it is not obvious whether the mist
eliminator is a scrubber or a wist eliminator. What do you call it when
you take aerosols, use thermodynamic driving forces to cause those aerosols
to"grow" into much larger aerosols, and then collect them in a mist
eliminator? Is the mist eliminator a scrubber, or is the section where you
grow the particles the scrubber?
For the purposes of this paper, we're not concerned about how the growth can
be accomplished; it can. The drops containing the contaminants have been
increased in size, but they still aren't very large, and they still have to
be collected. In these situations the coarser materials have been removed
prior to beginning the flux-force condensation techniques. Only fine
particulate and soluble materials are present, which helps relieve the
pluggage problem, but it doesn't necessarily prevent plugging since the
materials may be reactive. It takes considerable judgment and complete
information to arrive at a suitable decision.
A customer of ours has what we believe to be the world's largest installation
at around 100,000 cfm. In this case there was an existing vessel, so we
were constrained to relatively high velocities of well over 10 feet per
second. It took us several changes to reach a suitable balance between
efficiency, pressure drop, and pluggage resistance. We eventually ended up
with a composite involving k different styles: Our 4/96, 8/96, 16/96, and
37/9^ materials.
The savings to the user were in the seven-figure range. Other technologies
that might have been used were: Venturi, followed by a routine mist
elimination device; a candle-type mist eliminator; or an electrostatic
preci pi tator.
In this particular case, the primary consideration was very high efficiency
(at about 1-1/2 microns), flooding resistance, and pluggage resistance.
This is a commercial contract incinerator owned by a well-known company.
D. Flooding: Localized and Otherwise
One of our customers, located in Missouri, operates an incinerator for the
destruction of chlorinated hydrocarbons. This facility had been on-stream
for some time, and operated marginally satisfactorily at design conditions
of around 13 feet per second in the mist eliminator. They did have one
problem, and that was in the winter there was condensation occurring in the
stack, which was blowing out the top of the stack and causing local
problems.
When the throughput of the unit was increased a very small amount --
feasible from the incinerator and collection standpoint -- the mist
eliminator overloaded and substantial emission problems resulted. A
substantial problem developed because their total demand exceeded their
capacity. Our customer related to us that he did a continent-wide search
68
-------
for a mist eliminator that would handle his very high flow rate and very high
liguid-load conditions.
Although we considered the application routine, there were no other responses
to his inquiries. An order was placed with Kinnre for one of our very high-
flow-capacity units, comprised of style 37/97- This order was placed on an
emergency basis. The unit was manufactured, shipped, installed and in
service in 5 days. The result: elimination of the problem -- and a good
many happier employees. Even the receptionist commented on what an improve-
ment this change had made.
In our opinion, the original design velocity was higher tha it should have
been. Of course, that does keep down the capital cost. The alternative
to B-GONTM would have been expensive. The customer elected to add a second
entire eliminator at the top of the stack. This B-GON loads directly
down on the top of a stack enlargement.
In a similar situation, another of our customers operates a rock-dryer kiln
fired with oil. This produces some sulfuric acid, and of course a heavy
load of particulate, as well as some decomposition products from the rock.
The particulate is intended to be collected in a circular enturi of lurgi-
type design. The gas discharges in a radial direction inside of the vessel,
the gas load turns upward, passes through a mist eliminator, and through an
overhead stack located directly over the mist eliminator.
How does 100 gallons a minute of entrainment from a 90,000 cfm system seem
to you? We diagnosed the problem as one of too high a velocity, particularly
in view of the highly irregular flow distribution inherent in that type of
desi gn.
The solution: Drop a high-capacity B-GONTM Mist Eliminator right on top
of the existing chevrons, add a few hold-down grids on top, tie it all in
place -- and the problem disappeared. Total cost: Less than $5,000.
Alternative cost -- well, nobody ever added all the figures up, but it
certainly would have been over $150,000. Subsequently, we've converted 2 or
3 similar units for the same reason.
The combination of relatively high velocities and flow non-uniformity with a
high liquid load is a common problem. Most of the situations that we've
run into can be handled by a simple B-GON™ Mist Eliminator of high-capacity
design. However, in some cases it has been necessary to go to a more
radical solution: A variable-resistance mist eliminator. In this situation,
the mist eliminator is put together so the flow resistance is higher in the
areas where the flow is the greatest, causing a correction of the flow
distribution, and therefore lower peak velocities.
One last example of this type of problem How would you like to have a
scrubber for which the emissions are higher than the inlet? One of the
clients of a consultant knowledgeable about our products has that problem.
They have a Venturi system with chevron mist eliminators, using a
69
-------
contaminated scrubbing liquid. The mist carryover from the mist eliminator
is contributing higher emissions than the incoming gas. Re-design of this
system to accommodate a B-GON™ Mist Eliminator of appropriate design is
currently under way; installation is anticipated within a few months.
E. Electrostatically-Enhanced Scrubbers
There is a great deal of interest in electrostatically-enhanced scrubbing
for the collection of fine particulates. The best scrubbing appears to be
with the finest droplets, which requires an efficient mist eliminator.
This situation is essentially analagous to that of flux-force condensation,
and Kimre has been working with other parties -- one of whom, Dr. Stuart A.
Hoenig, will be reporting at this Symposium on his work.
SUMMARY
There's more to the selection of a mist eliminator than plugging some
numbers into the Souder-Brown equation. Tremendous strides have been
made in the science of predicting mist eliminator performance, but
there's still some art.
Faced with a mist elimination situation, consider all the aspects. Consider
the cost; consider the downside risk due to uncertainty in the process.
Consider scenarios of what might happen under particular conditions -- you
could save somebody a lot of trouble.
Don't be like the fellow who ordered a mist eliminator that cannot
conceivably work at 5 times the cost of one that will. And most
assuredly, don't be the other fellow, who in an attempt to save 2%
on the price of a mist eliminator, selects one that is marginal.
70
-------
NOVEL ANNULAR VENTURI SCRUBBER
DESIGN REDUCES WASTE DISCHARGE PROBLEMS
By: Heinz P. Beutner, Ph.D., President
Interel Corporation
7108 South Alton Way, Bldg. M
Englewood, Colorado 80112
ABSTRACT
Characteristics of a new type annular venturi scrubber system, developed
by Leisegang Umwelttechnik of West Germany, are presented.
The scrubber has internal circulation of the scrubbing liquid and
allows solids buildup in the liquid to 20% or more. The liquid spray is
generated without use of nozzles by contact of the gas with the liquid sur-
face. Solids are discharged either as concentrated slurry or as sludge by
means of a scraper.
The annular venturi scrubber achieves highest energy efficiency at any
pressure drop from 1 to 80 inches WG. It can be operated as variable flow
venturi by regulating the water level. Fractional particle removal
efficiencies as a function of pressure drop are presented. Examples of
installations on product dryers, incinerators, and in foundry and steel mill
operations are described.
INTRODUCTION
The venturi effect in particulate scrubbing is based on the rapid growth
of water droplets by condensation and coalescense during the high velocity
flow of the saturated gas through the venturi zone. In the process, dust
particles are collected by collision with water droplets (Fig. 1).
The basic principle has been applied in the Leisegang Annular Venturi
Scrubber combined with internal generation of the water spray. The system
offers the advantage that liquid with a high solids content is recirculated
internally without need for external piping, pump, settling tank, and spray
nozzles.
The system achieves the highest possible energy efficiency for any
pressure drop from 1 to 80 inches WG. Because of the very low amount of
waste water generated, the low scrubbing energy required and the low main-
tenance needs (no recirculation pump or sensitive spray nozzles), this
advanced design offers the user substantial operating cost savings compared
with other types of scrubber systems.
The Leisegang Annular Venturi Scrubber is protected under U.S. patent
No. 3,876,398 issued April 8, 1975, and patents in all major countries.
71
-------
Venturi Throat
CLEAN GAS
Venturi Annulus
FIGURE 1.
PRINCIPLE OF ANNULAR VENTURI SCRUBBER
-------
Principle of Design
The basic scrubber design is shown in Fig. 2. The venturi annulus is
formed by a cone positioned in a vertical tube inside the scrubber vessel.
The gas to be cleaned enters the scrubber vessel and impacts on the liquid
surface before entering the narrow annulus between the tube and the cone. The
strong acceleration of the gas to high velocity in the annulus and change of
direction breaks droplets from the liquid surface and carries them into the
gas stream.
The opening of the annulus can be adjusted mechanically to change the
pressure drop through the scrubber. The position of the liquid level can be
regulated either by a simple mechanical overflow weir or by control of the
liquid inflow and outflow based on the set point of a liquid level sensor or
a pressure differential transmitter. The latter approach maintains a constant
pressure drop across the scrubber with variable gasflow conditions.
The gas leaving the venturi orifice decelerates and regains static
pressure as the annulus increases in size. A deflector plate knocks down
large droplets before the gas enters a demister section. The demister
typically consists of S-shaped, parallel arranged lamella plates of polypro-
pylene or metal. Liquid drains off on collection edges while the gas passes
through the spaces between the plates at a minimum of pressure loss. Solids
in the liquid are easily flushed down with the liquid.
The liquid collected in the upper housing section returns by gravity
flow to the tank section of the scrubber. The liquid tank can be designed
for turbulence to maintain solids in suspension for removal by a slurry pump
or as a settling tank to promote sedimentation of solids for removal by a
built-in sludge scraper. Standard scrubber sizes from 600 to 60,000 cfm are
available. The larger sizes include multiple venturi tubes.
When compared with other scrubber systems, the Leisegang venturi
achieves maximum possible energy utilization for particulate removal, i.e.,
for any required removal efficiency for a given 'dust problem, the overall
power requirements for the Leisegang annular venturi are always equal to or
better than those of other scrubber designs (Ref. 1, 2). This statement
applies over the complete range of pressure drops used in practice, from very
low energy scrubbing to very high energy scrubbing. It should be noted that
the Leisegang Annular Venturi system produces wa.ter spray from the energy
supplied to the system fan for gas movement. Other venturi systems require a
separate pump to produce high pressure atomized water. In this case, the
pump power must be added to the fan power requirements when comparing power
requirements.
The Leisegang Annular Venturi Scrubber can be operated at any selected
pressure drop from as low as 1 inch WG (25mm WG, or 250 Pascals) to 80 inches
WG (2,000mm WG, or 20,000 Pascals).
The efficiency of particulate removal in a venturi scrubber is a
function of the pressure drop and particle size. The fractional removal
efficiency for quartz dust (specific density 2.6g/cm3 or 162.2 Ibs/ft3),
shown in Fig. 3, is based on actual test results.
73
-------
FIGURE 2
LEISEGANG ANNULAR VENTURI SCRUBBER
INCLUDING SLUDGE SCRAPER FOR SOLIDS DISCHARGE
74
-------
Ul
48 inches
W/A» 20.2 N
32 inches /
W/A > n.O
Water to Air ratio * 7.0 gal/1000 ft
Fractional efficiency for quartz shown,
(density 2.6 a/cm3 or 162.2 Ibs/ft3)
Correction for other types of dust:
d (micron) *d quartz V/ d*"ity
*
0.2 0.3
0.5 1.0 2.0
Particle Size (micron)
5.0
10.0 20.0
FIGURE 3
FRACTIONAL PARTICULATE REMOVAL EFFICIENCY
FOR LEISEGANG VENTURI SCRUBBER
-------
Variable Venturi Operation
A simple approach to variable venturi operation is a continuous control
of the water level in the scrubber relative to the annulus opening. This is
best accomplished by a differential pressure controller that automatically
turns on and off the flow of fresh water at a constant rate of slurry dis-
charge from the scrubber. Alternatively, the slurry discharge rate can be
varied at constant water feed rate. A typical schematic for variable
venturi operation is shown in Fig. 4.
The variable venturi operation by control of the liquid level allows the
turn-down of the gas flow rate to 30% of design without loss in scrubbing
efficiency.
Installation Experience
Typical uses of the Leisegang Annular Venturi Scrubber to date are in
the foundry industry, on product dryers of various types, and on incinerator
emissions.
A large installation of five identical scrubber units, each for 36,000
cfm gasflow, controls all emission sources from the Autocast Foundry near
Johannesburg, S. Africa (Fig. 5). The foundry makes parts for Toyota cars.
Except for a small amount of overflow liquid, there is no requirement for
liquid waste processing, since all solids are discharged as a sludge. The
scrubber meets the required outlet dust level of 50 mg/Nm3 (0.023 gr/scf) at
a pressure differential of 14 inches WG.
A typical installation of a high energy type scrubber (60 inches
pressure differential) is a small installation processing only 280 cfm from
a dissolver for nuclear reactor elements. Because of the radioactivity of
the collected solids and liquid aerosols, the Leisegang scrubber was selected
to avoid any external pumping loop. The stainless steel scrubber includes a
pressure differential sensor and controller to maintain the required liquid
overflow (Fig. 6).
An installation in the U.S. in FRP construction on incinerator exhaust
includes a packed bed section above the venturi section for absorption of
hydrochloric acid. The venturi scrubber removes gases only partially
because of the very short contact time between gas and liquid. The packed
bed contactor, using caustic reagent, allows gas absorption to very low
outlet levels.
A major installation on a foundry cupola was started up in 1979 at
Klockner Werke in Manstaed, W. Germany. The venturi scrubber, operating at a
pressure differential of 40 inches WG, follows a spray dryer installation
with cyclone. The spray dryer system was required to remove sulfur dioxide.
The sludge from the scrubber is recirculated into the spray dryer and the
only waste discharge is solid product from the cyclone.
ENDNOTES: 1. Wicke, M. Fortschrittsberichte der VDI Zeitschriften R 3
Vr 33, VDI Verlag, 1970 ' ' '
2. Leisegang, L. Swiss Chem 1, Nr. 12, 1979
76
-------
Constant draft (variable gas volume)
Constant water feed
Variable speed
slurry discharge
FIGURE 4
TYPICAL CONTROL DIAGRAM FOR VARIABLE VENTURI OPERATIONS AT CONSTAT DELTA P
-------
FIGURE 5
VENTURI SCRUBBER INSTALLATION ON FOUNDRY OPERATIONS
AT AUTOCAST PTY NEAR JOHANNESBURG, S. AFRICA
78
-------
RAW GAS
CLEAN GAS
a)
b)
c)
d)
e)
Gas Inlet
Gas Outlet
Demister
Scrubbing Zone
Liquid Return Pipe
FIGURE 6
HIGH ENERGY SCRUBBER
FOR RADIOACTIVE AEROSOLS
(DELTA P = 60 INCHES WG)
79
-------
CONSIDERATION OF THE PERTINENT DESIGN AND OPERATING
CHARACTERISTICS ESSENTIAL FOR OPTIMIZATION OF
VENTURI SCRUBBER PERFORMANCE
By: H. Scott Oglesby
National Council of the Paper Industry
for Air and Stream Improvement, Inc.
Southern Regional Center
P.O. Box 14483
Gainesville, Florida 32604
ABSTRACT
The use of medium to high energy venturi scrubbers in the pulp and paper
industry has increased substantially over the past eight to ten years. Ac-
companying this increased use is a wider spectrum of applications, and a demand
for a sustained higher level of performance to meet the current and emerging
regulatory requirements. Accordingly, there is a need to review the critical
factors affecting performance relative to optimization, process compatibility
and energy requirement minimization. In this respect, it is essential that
considerations be given to the pertinent design and operation parameters in
assessing performance. This paper addresses these issues and develops a
methodology for troubleshooting scrubber performance.
INTRODUCTION
For many years the application of venturi scrubbers was almost exclusively
confined to controlling submicron particulate emissions from the lime kiln, but
more recently these devices are being applied to combination fuel-fired boilers,
recovery furnaces, smelt dissolving tank vents, and on causticizer vents, for
the same purpose. Also, with the emerging requirements for flue gas desul-
furization on new coal-fired power boilers, it is anticipated that the appli-
cation of wet scrubbers will grow at an accelerated pace.
DESIGN FUNDAMENTALS
Venturi scrubbers have been used in a variety of applications for many
years. Experience has been vast, and yet a condensation of this experience in
terms of a reliable design equation has been lacking. Perhaps the earliest
published work relating design and performance parameters empirically was that
of Stairmand (1), which appeared in 1965. Also, a very useful relationship was
developed by Hesketh (2) in 1974, which related the pertinent scrubber operating
variables. More recently, results of a number of highly comprehensive studies
have been published (3) relative to optimizing the design and performance of
venturi scrubbers.
Since the performance of a venturi scrubbing system is achieved at the
expense of a high pressure drop in the gas stream, accompanied by high fan
operating cost, designs of system which for a given gas flow rate and partic-
ulate removal efficiency, have minimum pressure drop, may be considered optimal.
80
-------
Gas-Liquid Relationship
It has been shown to be generally true that scrubber efficiency increases
with increased energy consumption, whether this is achieved by employing
high gas velocities in the throat, or using a high water/gas ratio. It has
also been shown (1) , however, that at any particular throat velocity there
is an optimum pressure drop at which the scrubber should be operated. This
is shown clearly in Figure 1, in which the interrelation between efficiency,
water/gas ratio, and pressure drop is given. The points of maximum efficiency
for a given pressure drop are joined to produce the economic design curve
shown, which indicates that there is a particular combination of operating
parameters which gives maximum efficiency for minimum pressure drop.
It is known that the target efficiency of a droplet increases by de-
creasing the diameter and increasing the relative velocity between the gas
and the droplet. Figure 2 was developed from published data (1) to show
that the optimum droplet size and the induced droplet size are equal for
only a particular throat velocity for any given dust size. Thus, there is
an optimum throat velocity for any given set of conditions.
Throat Geometry
Fundamental to scrubber design is the throat area which is specified
based upon the saturated gas volume and pressure drop requirements. A more
recent analysis (4) of performance suggest that other aspects of throat
geometry may be important in optimizing performance. For instance, it has
been shown (5) that increasing venturi length will increase the collection
efficiency of particles greater than approximately 0.5 ym in size without a
commensurate increase in pressure drop. Further it has been demonstrated
(4) that the pressure recovery is strongly influenced by the throat length.
Other important aspects of throat geometry are those which relate to
throat coverage. Regardless of the concepts which are utilized in designing
a scrubber, it is imperative that these provide for maximum throat coverage
over the full operating range.
Elbow and Entrainment Separator
The diffuser and regain section of the venturi scrubber discharges
into the elbow which turns the gases and liquid into the entrainment separator.
A "flooded" elbow, which provides for a pad of liquid in the horizontal
section, has become standard design in recent years. This liquid pad
provides a surface for turning the gases, thus reducing wear on the elbow,
and serves as an additional impaction zone for collecting particulates
remaining in the gases.
The design criteria for venturi scrubber entrainment separators, which
is straightforward, involves cyclonic separation design principles for
collecting liquid droplets as small as 15 ym in diameter. From the foregoing,
it is obvious that inlet volumes to the separator must be limited to fluctu-
ations of ±15 percent, if high separating efficiencies are to be maintained.
81
-------
If the separator is not properly designed and/or operated, carryover or
reentrainment of liquid droplets will occur resulting in increased dust lost
to the atmosphere. Also, a poorly designed separator can cause excessive
pressure drop across the system, thus reducing overall capability.
System Ancillaries
Equally important to the venturi scrubber and its major components,
relative to optimizing performance, are the pertinent ancillaries which
comprise the total system. These items, and the vital role they play, are
addressed in the sections which follow.
OPERATING PRINCIPLES
Once the scrubbing system is placed in operation, it is essential that
the actual conditions be compatible with that on which the design was based,
if optimal performance is to be achieved. Because the venturi scrubber is
extremely flexible, it is relatively easy to operate the system beyond the
range intended, at a sacrifice in performance and/or energy consumption.
Relationship Between Operating Design Parameters
Those responsible for operation and maintenance of venturi scrubber
systems should be provided with the appropriate operating and performance
curves such as the one shown in Figure 3. If the curves or the data for
plotting such are not available, then the relationship developed by Hesketh
(2) as summarized in Table 1, may be utilized for this purpose. Performance
curves such as these are essential in employment of diagnostic routines for
optimization.
TABLE 1. VENTURI SCRUBBER PRESSURE DROP RELATIONSHIP (2)
2 A0.133_0.78
AP = Vt P£ A L
1270
Where: AP = Venturi Pressure Drop, inches W.G.
L = Liquid to Gas Ratio, Gal/1000 ACF
pe = Gas Density Downstream from Venturi, Ib/ft
Vt = Throat Velocity of Gas, ft/sec
A = Throat Cross-section Area, ft
Control Options
Continuous measurement of throat pressure differential is the most
direct way to monitor scrubber performance. However, in order to optimize
performance a comprehensive instrumentation package is required as indicated
in Figure 4. Control options include a liquid control valve, variable speed
pump drive, an adjustable throat, or various combinations of these devices
coupled with feed-back control. More complex control systems may include
variable speed I.D. fan drive, or fan damper controls as well.
82
-------
Separator Limitations
Since the separator is so important to the overall performance of the
scrubber system, and its operation range is more limited than the scrubber,
set-points should be utilized to activate an alarm (and record) when the
scrubber is operating outside the capability range of the separator. Although
this is a passive control approach, at least the operator would be aware that
the system was not operating under optimal conditions.
CONSIDERATIONS FOR OPTIMIZING PERFORMANCE
AND MINIMIZING ENERGY REQUIREMENTS
This section will briefly address those innovations which may be utilized
to improve scrubber performance or reduce energy requirements at the same
level of performance. The choice for optimization, of course, will depend
upon the specific application and the degree of system modification which can
be justified.
Preconditioning of Gases
It has been stated (6) and demonstrated that the key to low power scrub-
bing involves the judicious use of sufficient gas retention after quenching
to effect conditioning, thus enabling agglomeration and absorbtion of partic-
ulate fume and gases. The impact on particle size distribution due to particle
growth resulting from water vapor condensation is shown in Figure 5. A
recent, modern version of this concept (7), called flux force/condensation
scrubbing, claims to use only 70 percent as much power as a conventional
scrubber system would. While the theory surrounding this concept is somewhat
complex, applications using simple spray nozzels to quench gases prior to the
scrubber and resulting in improved performance have been reported (8).
Multiple Throat Arrangement
Recent experimental work (5) has demonstrated that two venturi throats
in series can improve scrubber performance by as much as 30 percent over a
single throat at the same pressure drop. It may perhaps be too early to
judge whether this same level of performance can be achieved by the more
simple preconditioning approach, and if in fact the multiple throat approach
does provide an additional advantage.
Separator and Duct Configuration
System energy consumption to a greater extent and performance to a
lesser degree can be affected by inlet duct configuration and separator
design. Good ductwork design procedures are well documented (9) and should
be reviewed if this part of the system is suspect. As previously mentioned,
the importance of the separator cannot be over emphasized, hence good sepa-
rator design, operation and maintenance are essential for optimizing
performance and energy consumption.
I.D. Fan Location and Design
Since performance of a venturi scrubber depends oh development of the
83
-------
required static pressure, the induced draft (I.D.) fan is a vital component^
of the system. A well designed fan will be an asset in the quest for optimizing
performance and energy consumption (10,11).
The power required to drive wet scrubber I.D. fan can vary in a given
situation depending on whether the fan is before (scrubber under pressure) or
after (scrubber under vacuum) the scrubber. When the I.D. fan is operated in
the "cold" (after) position, the brake horsepower (BMP) reduction can be 20-
35 percent compared to a fan located in the "hot" (before) position. This
reduction is achievable due to the lower saturated gas volume and lower
equivalent static pressure (ESP) prevailing at saturated temperature, there-
fore reducing the fan capacity and power requirements. In lieu of the energy
savings, the system static capability (and efficiency) may be increased by
placing an existing I.D. fan after the scrubber.
TROUBLESHOOTING VENTURI SCRUBBERS
In the interest of maintaining continuity of high performance it is
essential to develop a troubleshooting procedure with routine followup. A
lot more can be said about this aspect than time will allow, but a brief
summary if provided. The most important tool in diagnosing poor performance
is the use of the operating and performance curves, to determine how close
actual operating conditions relate to design. Usually, this will allow one
to zero in on the area of suspect.
Static Pressure
Development of a system static pressure profile is another useful tool
in troubleshooting a system. This is not overly time consuming and will
allow an immediate assessment of the pressure drop distribution relative to
design.
Separator Velocity
Perhaps the most common reason for poor scrubber performance is poor
entrainment separation. Reentrainment may be described by a number of
mechanisms, but is primarily due to either too high or too low of a vertical
velocity in the separator, as depicted in Figure 6. Knowing the separator
dimensions and gas volume being processed, it is a simple matter to determine
this velocity. The separator may also be the source of excessive pressure
drop.
Throat Area
For fixed throats the throat area selected for a given application
should be in the mid-range, so that the liquid flow rate required to maintain
the pressure differential is neither too high nor too low. High liquid
flowrates, in addition to consuming excessive energy, tend to aggravate
reentrainment conditions in the separator. On the other hand, low liquid
rates usually result in reduced efficiencies and poor throat coverage.
Recently, we have seen increased use of the adjustable (variable) throat,
84
-------
and this provides an excellent mechanism for optimizing performance. The
intent of this concept is to maintain the throat velocity which achives
highest efficiency at lowest pressure drop. Unfortunately, the flexibility
afforded by this concept also allows operation to drift outside the optimal
range.
Pump and Scrubbing Fluid Control
The scrubbing fluid pump, piping, valves, sump tank and controls must be
designed and operated to respond to the scrubber demands. Quite often the
scrubber operating range will exceed the pump capacity, which results in
reduced efficiency. Also, where systems operate with high fluid recirculation
rates it is essential that adequate attention be given to the sump tank,
piping and liquid level control arrangements.
Fan Characteristics
Generally, the system's static pressure requirements are stipulated, and
it is readily obvious as to whether the I.D. fan will meet this demand.
However, there may develop, due to a combination of physical conditions, a
situation where a fan "falls off its curve" and will not perform as specified
(11). Therefore, it is essential to have the fan performance curves and
"test block" specifications available during performance analysis. This
information is also useful in determining if the fan may be upgraded to
develop more static or volume capacity.
CONCLUSIONS
The pertinent aspects of optimizing venturi scrubber performance relative
to maximizing efficiency and minimizing energy consumption have been reviewed.
Reliable design equations and relationships have been developed, and recent
studies have extended their use toward optimizing performance. Also, today
there are available excellent relationships for developing operating and
performance curves to assist in achieving and maintaining continuity of high
performance.
Mechanisms are now being advanced which offer the opportunity to boost
venturi scrubber system efficiencies, or reduce energy requirements for the
same level of performance. Some of these concepts are new in principal while
others have been utilized in the past, but were not fully quantified until
recently.
The importance of the cyclone entrainment separator, and the part it
plays in overall system optimization cannot be overemphasized. The entrain-
ment separator is much more susceptible to volume swings than the scrubbers
and the failure to recognize this aspect is one of the major reasons for poor
performance. Optimal performance may indeed be difficult to achieve without
devoting proper attention to the major ancillaries which comprise the complete
system.
With the development of appropriate operating and performance curves for
a specific system, a troubleshooting routine may readily be performed in the
85
-------
interest of optimization. In troubleshooting a system, the major components
to be considered in the analysis are the I.D. fan, throat, entrainment sepa-
rator, and liquid flow control equipment.
REFERENCES
1. Stairmand, C. J., Removal of Grit, Dust, and Fume from Exhaust Gases
from Chemical Engineering Processes, The (British) Chemical Engineer,
Dec. 1965, pp. CE310-326.
2. Hesketh, Howard E., Fine Particle Collection Efficiency Related to
Pressure Drop, Scrubbant and Particle Properties, and Contact Mechanism,
JAPCA, Vol. 24, No. 10, Oct. 1974, pp 939-42.
3. Goel, K. C. and K. G. T. Hollands, Optimum Design of Venturi Scrubbers,
Atmospheric Environment, Vol. 11, pp. 837-845, Pergamon Press 1977.
4. Overcamp, Thomas J. and Steven R. Bowen, Effect of Throat Length on
Pressure Loss in a Venturi Scrubber, Presented at the 73rd Annual Meeting
of the APCA, Montreal, Quebec, June 22-27, 1980.
5. Muir, D. M. and Y. Miheisi, Comparison of the Performance of a Single
and Two-Stage Variable-Throat Venturi Scrubber, Atmospheric Environment,
Vol. 13, pp. 1187-1196, Pergamon Press 1979.
6. Taeler, David H., Gas Scrubbing Pays Double Dividend, Minerals Process-
ing, Jan. 1964, Jan. 1964, pp. 24-27.
7. Anon., Science/Technology Concentrates, Chemical and Engineering News,
Dec. 3, 1979.
8. Oglesby, H. S. and A. V. Barron, Economic Considerations for Selection
of Pollution Control Systems, Presented at TAPPI 22nd Engineering Confer-
ence, Sept. 19-21, 1967, Atlanta, Ga.
9- Jorgensen, Robert, Editor, Fan Engineering, Sixth Ed., Chap. 3, Buffalo
Forge Co., Buffalo, N.Y.
10. Perry, Robert E., Power Conservation through Proper Fan Operation,
Southern Pulp and Paper Manufacture, June 1977, pp. 29-36.
11. Barron, Alvah V., Fan Application for High Energy Scrubbers, Presented
at the TAPPI 26th Engineering Conference, November 18, 1971, pp. 609-
622.
12. Oglesby, H. Scott, Design and Operating Characteristics of Venturi
Scrubbing Systems which contribute to Excessive Reentrainment, Proceed-
ings of the 1977 NCASI Southern Regional Meeting, Special Report No. 77-
07, Oct. 1977, pp. 109-121.
13. Calvert, Seymour, Get Better Performance from Particulate Scrubbers,
Chemical Engineering, Oct. 24, 1977, pp. 133-140.
86
-------
IOC I
"96
H
H
TMROAT V£l_. JCOft/MC
-------
*
B"
-
o
5"
VI
I690F
S»T'0 DENSITY >O.05i**/FT
SAT'OCFM «o.7Ti XACFU
73.SOOACFM
68,530 ACrS-
63,500 ACFUIOESiGN
98,500 AC!
33.5OO ACFM
«0 GPMIEESISN)
•
1
t
I
I
!
8 10 12 14 l« 18 20 22 !« IS 28 *J
VEMTUR! SCRUBBIWi LIOUIO RATE (iiU-CNS/IOOOCFM)
FIGURE 3. SCRUBBER OPERATING VARIABLES (12)
BLEED- OF1LTC PROCESS
CIRCULATING POMP ~
FIGURE 4. VCHTURI SCRUBBING SYSTEM COMPONEWS JiND INSTRUMENTWION
88
-------
.
§04
0.1 I—J
I 2
I I I
iO
BO
93
JO «0 50 60 70
MASS (% UNCCRSIZE
FIGURE S. PAKTICLE uROWTH KESULTINC FiiCM WATER-VAPOR CONDENSATION
SEDCCES POWER nEHANJ! (13)
DOWNWARD FLOW
W»TER FU>W WT£ (CC/ KWUIE)
•mvELoctrr m/srr]
6. VERTICAL UP AKD DOUN FU K Cfi.lPHBISON OF ENTRAINED FRACTION IN
UPWRO AND DOKNWARO COCUBREttT ANNULAS rLOW IN 1.25 cm I.D. TU2E
89
-------
APPLICATION OF SCRUBBERS FOR PARTICULATE
CONTROL OF INDUSTRIAL BOILERS
By: Murray Borenstein
Neptune AlrPol Inc.
Sub. of Wheelabrator-Frye Inc,
Englewood, NJ 07631
ABSTRACT
Cyclone dust collectors were formerly used as the means of dust control
on bark boilers, but are no longer adequate. More stringent requirements
call for higher efficiency collection, and scrubbers have been successful in
meeting the codes. This paper will explore the operating parameters, oper-
ating experience, and test results of scrubbers on wood and coal fired
boilers, particularly in the pulp and paper industry. It will also review
the power requirements, materials of construction, scrubber design and
mechanical collectors.
Background
Prior to the 1970's mechanical dust collectors were almost the exclu-
sive method of pollution control on industrial wood and coal fired boilers.
These were principally the multiple cyclone type with cast iron tubes ranging
in diameter from 6" to 12". Gas enters the cyclones axially, being given its
rotation spin by vanes. The smaller the diameter, the greater the centrif-
ugal force, and the higher the efficiency. Fly ash is highly abrasive, and
the cyclones must be constructed of iron of relatively high Brinell hardness.
Efficiency falls off rapidly as particle size falls below 10 microns.
Typical efficiency for a multiple cyclone collector handling flue gas con-
taining fly ash from a pulverized coal fired boiler is about 80%. Thus, with
an inlet loading of 3 gr/scf, it is obvious that an outlet loading of .6 gr/
scf would not suffice for today's standards.
For wood fired boilers, the bark char is particularly light, specific
gravity of .2 to .5 so that the efficiency would be even less. It has been
demonstrated that a second stage mechanical type collector in series with
the first may increase overall efficiency by 2 - 5%. This percentage varies
depending on the ability to design a proper system for primary hopper evacu-
ation or out-tube shave off which contains the finer material. Even with
the most efficient two-stage type units which are reported capable of pro-
viding a total of 90-95% efficiency, they will still not meet federal and
state standards of about . l#/m BTU input.
It should be noted that particle size distribution of dust at outlet of
primary dust collectors have over 50% of the material less than 5m. The
particles smaller than 5m have a very low terminal velocity and are relative-
ly unaffected by inertial or gravitational force so that the probability of
collection of these size particles is poor for any second stage mechanical
cyclone device.
90
-------
Further, particles of less than 5m range also are the major contributors to
stack opacity and substantial removal is necessary in order to meet low
Ringleman member requirements.
In some installations where dust emission is very marginal, a shave off
arrangement in which a portion of the dust rich gas of about (10-50%) of the
volume can be scrubbed and then recombined with main gas stream. Reports on
few installations of this type have not been successful in meeting current
stringent codes.
With the advent of the more stringent state codes of the 1970's and the
Federal Clean Air Act of 1977, it was apparent that secondary air cleaning
for industrial boilers would be required. In 1979, the EPA promalgated the
NEW SOURCE PERFORMANCE STANDARDS (NSPS) for boilers generating electricity
greater than 73 MW (250 MM BTU/hr) which limits the emission to .03 Ibs/MM
BTU. Boilers less than 73 MW are handled by state codes which generally re-
quire emissions ranging from .1 to .8 Ibs/MM BTU. Some of the larger indust-
rial boilers exceed this capacity and would come under the federal standards.
The standards also call for maximum emissions of 1.2 Ibs. SO-/MM BTU plus a
70-90% reduction of S02, depending upon the sulfur content of the fuel. In
almost all cases of coal burning boilers, the user is faced with SO removal
as well as particulate.
The user has a choice between 4 alternate methods of secondary control:
1. Wet Scrubber
2. Fabric Filter
3. Electrostatic Precipitator
4. Dry Scrubber
Each method has its pros and cons, but this paper will deal with Wet
Scrubbers only.
Advantages of Wet Scrubbers
1. Have the capability to collect both particulate and
absorb gases.
2. Function in wet corrosive and/or explosive gas mixtures.
3. Not sensitive to high temperatures.
4. Occupy less space than fabric filter or electrostatic
precipitator.
Disadvantages of Wet Scrubbers
1. Higher power consumption.
2. Water and sludge disposal problems.
3. Highly visible steam plume.
4. Possible corrosion problems if material of
construction not properly selected.
91
-------
Types of Scrubbers
All scrubbers operate on the principal of inertial impaction between a
droplet of liquid and a solid particle, or impingement of a high velocity
particle on a wetted surface. To achieve the desired impaction, water is
finely divided into droplets and accelerated to achieve its impact velocity.
This may be accomplished by injecting jets of water into a high velocity gas
stream (such as is done in the Venturi Scrubber) or generating a fine spray
in high pressure spray nozzles. The types of scrubbers that have been used
on industrial boilers are: . TT
1. Venturi
2. Flooded Disc
3. Impingement Tray
4. Dynamic Scrubber
This paper will deal primarily with Venturi Scrubbers since they are
the type most commonly used for particulate collection, although they also
have the capability for gas absorption.
See Figs. 1, 2, 3, 4
Fuels
In determining the design parameters for scrubbers for industrial
boilers, the design engineer must review several factors, such as:
1. What type of fuel?
a. Coal
b. Bark and Wood
c. Oil
d. Combinations of the Above
2. In coal fired boilers, what is the method of firing?
a. Pulverized
b. Spreader Stoker
c. Chain Grate Stoker
d. Under Feed Stoker
3. What is the excess air rate?
Coal fired boilers have heavier dust loadings and finer particulate sizes
than wood fired; and, therefore, require higher pressure drops in the range
of 12-18" w.g. water gauge to achieve emissions of 0.1 Ib/MM BTU. Bark or
wood, when efficiently burned, has relatively light loadings and requires
pressure drop in the range of 8-10" w.g. Combinations of wood and coal
would need pressure drop somewhere in between.
See Fig. 5
92
-------
Participate Removal 99.9%
High S02 Removal
MECHANICAL
DUST COLLECTOR
RECIRCULATION
TANK
Fig. #1
AUTOMATIC
PIA.
SMffiS
PlA.
c^-O^=-D^
E.L&OW
7
&L&VATIOM
FIGURE 2. AIRPOL BARK BOILER SCRUBBER SYSTEM
93
-------
1^==! ]'| T/O =»LT~. gU_
H
T/O =LT~. EL.
\ \ri.or; .^
7~1
\ /fl'ii
s^ssW^—--^—x«>
f^TT \ J' LRt
§, i ijz
-if- !l
uPtacarr Bu.
I !
FIGURE 3. AIRPOL VENTURI SCRUBBER AND SEPARATOR
GENERAL ARRANGEMENT
94
-------
FIGURE 4. AIRPOL S02 ABSORPTION TOWER
-------
O
CO
5-1
tn
tn
C
fO
O
O
0.5
0.2
O.I
0, OS
0.06
O.O4
o.oz
CXO!
10 15 20 25 30 35
Pressure Drop (in. w.g.)
SCRUBBER PERFORMANCE ON COAL FIRED BOILERS
Fig. #5
96
-------
Oil fired boilers, although having very light loadings, have extremely
fine, submicron particles which would require a relatively high pressure
drop for efficient removal. Scrubbers are not usually designed for particu-
late alone on oil fired boilers, but only when SO collection is also
required.
Studies done by the Mid-West Research Institute have shown that there
is a substantial difference in the particulate size of the fly ash in the
flue gas, depending upon which firing method is used. For example: The
mass median diameter of the particles for these four methods is as follows:
Method of Firing Median Particle Diameter
(microns)
Pulverized 20
Spreader Stoker 48
Chain Grate Stoker 75
Underfeed Stoker 16
Oil 2
Since the particle size determines the pressure drop required, the method
of coal firing must be taken into consideration by the design engineer.
Our experience has shown that the rate of excess air to the boiler is
an important factor in the particle size formation. When the excess air
rate starts to exceed 40%, we have noticed that efficiency of fly ash
collection begins to decrease.
Venturi Scrubber Design
Since boilers operate at various loads, gas volume to the scrubber will
vary as much as 30 to 100% of capacity. At the reduced gas volume, the pres-
sure drop will tend to decrease because of lower velocity in the Venturi
throat, resulting in lower efficiency.
It is necessary to compensate for this condition by restricting the
throat area, maintaining the design throat velocity which in turn will main-
tain the efficiency. This is done by means of an adjustable venturi throat
which may consist of either side dampers, movable plug type cone, or flooded
disc. For automatic adjustable throat operation, controls can be installed
to maintain constant pressure drop or constant furnace draft. Today, most
Venturi Scrubber manufacturers inject water at the top of the Venturi, main-
taining a falling film of water down the convergent section to prevent any
dust build-up at the junction line of dry gas and water. This prevents the
formation of wet/dry line deposits. In addition, our company also directs
jets of water directly into the throat to insure good water coverage at the
throat.
After the gas is scrubbed in the Venturi, the liquid must be separated.
Separators most commonly used today are either centrifugal or impingement
type.
97
-------
With the centrifugal separator, gas enters tangentially and spirals upward
through a cylindrical tank throwing the water against the walls by centrifu-
gal force which then drains down by gravity. However, as the gas volume gets
large; and, consequently, the diameter of the separator gets very large, the
centrifugal force decreases inversely proportionately to the diameter, de-
creasing the efficiency of droplet separation. It then becomes necessary to
use an impingement type separator, such as a chevron mist eliminator. Our
company uses a diameter of about 14 ft. as the border between a cyclonic and
an impingement type separator. With chevron mist eliminators, it is neces-
sary to have spray headers for an intermittent wash.
Materials of Construction
For boilers burning 100% bark or wood, carbon steel is a satisfactory
material of construction. However, most boilers that burn primarily wood
also burn one of the fossil fuels in combination, or at least have it on
stand-by. Therefore, practically all scrubbers today are designed for cor-
rosion resistance. Type 316L stainless steel is the preferred material of
construction. When sulfur bearing fuels are burned and scrubbing is done
with water which is recirculated, the water will quickly build up to a sat-
urated soltuion of sulfurous acid, with a pH down to about 2. This is low
enough to attack the 316L stainless steel so that the scrubbing water must
be raised to a pH of at least 4 by means of a caustic liquor.
An alternate material for corrosion resistance is carbon steel lined
with polyester fiberglass. While this is satisfactory for ductwork and
stacks, it should not be used in the high abrasion areas of the scrubber,
such as the Venturi, or the tangential inlet of the separator. Experience
has shown that these coatings do not stand up over the long run in these
areas. Also, if organic coatings are used, provision must be made against
high temperature attack in case of loss of water. One of the advantages of
stainless steel is that temperature is not a factor.
Occasionally, there may be a high chloride content of the scrubbing
medium which ultimately would attack the 316L stainless steel. In such
cases, the stainless steel with a higher molybdenum content, such as 317L
stainless steel has been used. Alternately, polyester fiberglass may also
be used if proper temperature controls are installed.
Outlook for Industrial Boilers
Present regulations of the DOE require that new industrial boilers of
greater than 100 MM BTU must use coal rather than oil or gas, unless this
proves to be impractical. The DOE also has the power to require existing
boilers over 100 MM BTU to switch from oil or gas to coal.
It is fortunate that the United States has an abundance of coal but,
ironically, this is the dirtiest of the fuels. This means that air pollu-
tion equipment manufacturers should be kept busy over the next several years.
98
-------
Market projections by Mcllvaine show the projected sales of various
air pollution devices over the next several years.
See Figs. 6, 7
Acknowledgments
1. "Technology Assessment Report lor Industrial Boiler Applications:
Particulate Control"
Industrial Environmental Research Lab.
U. S. Environmental Protection Agency
Research Triangle Park, NC 27711
2. "Air Pollution Control Market for Industrial Boilers"
The Mcllvaine Co.
2970 Maria Ave.
Northbrook, 111. 60062
3. "Controlling Bark Boiler Fuel Gas Emissions"
Louis A. Mason
Neptune AirPol Inc.
Sub. of Wheelabrator-Frye Inc.
Englewood, NJ 07631
99
-------
C;VU Of :-.:L'lV!i; tD.'LER Ai> ;;-aUT!t"< fi'M^OL FCJ:»-!NT ^.vrNTS
J Ell)ions
3001
loci
SROwTH OF TOTAL INDUSTRIAL BOILER POLLUTION CONTROL EOUj
1S79 1930
KecKanical (rig. 6)
1B33 'SS4 1985
100
-------
Sreafc Southern Paper - Power Boiler Adjustable Throat
Power Boiler Scrubber
GREAT SOUTHERN PAPER
101
-------
APPLICATION OF HIGH ENERGY VENTURI SCRUBBERS
TO SEWAGE INCINERATION
By: Francis X. Reardon
Metcalf & Eddy, Inc.
Boston, Massachusetts
ABSTRACT
Wet scrubbers were installed on the flash drying sludge incinerator furnaces at the Allegheny
County Sanitary Authority plant as part of a program to upgrade the furnaces, thereby removing
mechanical dust collectors which were not capable of complying with the emission code under for-
mulation in 1969 by the Allegheny County Health Department.
Features of the engineering application of the flooded disc scrubber to the furnaces, the anti-
cipated performance of the scrubbing system, the actual test results and operation of the units
which continue to perform successfully today are explained and discussed. It is our purpose to
adhere as closely as possible to the engineering aspects of the intended and continuing furnace
rehabilitation program.
INTRODUCTION
The Allegheny County Sanitary Authority has been engaged in a program of incinerator
improvements for many years. In 1969, incinerator modifications included a major program in
which pulverized coal facilities and all ancillary equipment for coal-firing in the incinerator fur-
naces were demolished and removed. Natural gas burners were installed to fulfill the auxiliary fuel
requirements for sludge burning and have resulted in greatly improved environmental and opera-
tional conditions within the plant. Venturi scrubbers were installed. Modifications to vapor recyc-
ling ductwork and to the air and gas systems for each furnace to reduce and limit air infiltration
and auxiliary fuel costs were completed. The furnaces continue to require a consistently high level
of operator attention and maintenance. :
Basic engineering problems requiring a solution prior to the scrubber selection are noted as
follows:
102
-------
1. The selection of the required capacity or flow rating of the scrubbers consistent with the
expected excess air levels achievable in the flash drying furnaces. It was apparent that
there were many potential sources of infiltration of ambient or furnace room air into the
furnace systems. An intensive maintenance program was imminent because of forthcom-
ing repairs and rehabilitation to the heat exchanger sections of the furnaces, new auxili-
ary fuel burners, vapor return ductwork, cage mill repairs, and dryer cyclones. The pro-
posed new sludge conveyor system providing vacuum filter cake to sludge mixers would
require modification to effect sealing against room air.
2. An assessment of venturi scrubbers operating at a medium pressure drop, say in the 15-
to 25-inch range, to meet the proposed EPA New Source Performance Standards for
sludge incinerators. A proposed emission of 0.031 gr. per dscf (grains per dry, standard
cubic foot) eventually became 1.30 pounds of particulates per ton of dry sludge input.
The Allegheny County Health Department, Rules and Regulations, Article XVII, Smoke
and Air, was being discussed at hearings and was to be adopted while contracts for scrub-
bers and other equipment were out for bids.
3. A reliable estimate of the expected dust load from the flash drying furnace, considering
the use of natural gas burners for auxiliary fuel in lieu of pulverized coal burners.
4. The physical and chemical characteristics of the anticipated fly ash.
5. The selection of an alloy steel for the scrubber which would withstand the corrosive and
erosive attack of particulate-laden flue gases.
DESIGN CONSIDERATIONS
Factors which were to affect furnace operation and subsequently scrubber operation were the
result of both furnace modifications and the installation of a new vacuum filter facility. It is proper
and truthful to say that not all of the effects of the furnace modification and rehabilitation were or
could have been foreseen in planning the scrubber installation.
At this point in time, the furnaces were to operate with filter cake, and during the scrubber
emission testing the vacuum filters and the newly installed sludge conveyors were put into initial
operation. Prior to this time, a liquid sludge had been pumped to the furnaces. The run-in period
for the new sludge conveyors and the feed to mixers and cage mills required much attention before
a steady sludge feed, necessary for the emission testing, was realized.
Prior to the writing of specifications for the scrubber system, it was noted that excess air
levels of gases from the furnaces were on the order of 180 to 190 percent. These measurements
103
-------
were made with exit gases at 600 deg F to 700 deg F during operation with mechanical collectors.
Combustion Engineering Company heat balances are customarily based on 40 percent excess air for
sludge burning in flash drying furnaces. Thus it is seen that the data available to the designers
spanned a broad range of conditions.
It was decided that the scrubber exhaust gases at saturation could be approximately 200,000
pounds of wet gas per hour and that the excess air level resulting from excessive infiltration might
reach 200 percent before system adjustments and furnace modifications could achieve a reduction.
The maximum test run conducted on Furnace No. 4 indicated a gas flow of 189,200 pounds of gas
per hour at 160 deg F or 49,456 acfm.
The specified scrubber duty required each unit to handle 85,000 acfm of gases leaving the
furnace preheater at approximately 1,000 deg F. The dust loading entering the scrubber was
specified as four (4) grains per scfm. The emission rate was specified not to exceed 0.05 pounds
particulates per 1,000 pounds of flue gas. Testing was to be based on 60 to 100 percent of max-
imum rated furnace capacity.
The physical constraints involving the location of scrubbers, induced draft fans, breechings,
and integration with a new sludge conveyor system, and the existing flash drying furnaces
required a specific arrangement of venturi scrubber and separator vessels. Figures 1 and 2 are pro-
vided which give plan views and sectional elevations of the installed equipment in the incinerator
building. Scrubbers were specified to have a flanged, vertical gas inlet to the venturi section, a cyc-
lonic separator with a tangential inlet and a bottom, vertical outlet, a hopper bottom with a flanged
outlet, water connections and structural support brackets. The flooded disc scrubber selected suited
the structural and geometric requirements.
The scrubbers and the downstream breeching were constructed of Incoloy Alloy 825. The
inlet duct from the air preheater hood of the furnace to the scrubber is constructed of annealed car-
bon steel plate conforming to ASTM Designation A-283-67. Expansion joints in the Incoloy
breechings were constructed of Incoloy Alloy 825. All of the breechings from each scrubber were
constructed as individual breechings from the fan to the stack.
Scrubber water is provided from the plant secondary treatment effluent discharge. The
adjustable disc is perpendicular to the gas flow and forms an open annulus of varying size as it tra-
vels in the tapered throat section of the scrubber. Gases flowing through the opening have a high
degree of disturbance and at the same time, water is introduced through the large diameter disc
support pipe and is ejected across the face of the disc. The principle, as described by the manufac-
turer, is based on high velocity gas striking the disc face, shearing water particles at the edge of the
disc, and transforming the water into a fine mist, which is thoroughly mixed with the gas. The
specified pressure drop for the system was not to exceed 20 inches w.g. The recommended pressure
drop for the scrubbers conforms to the specifications, that is, 20 inches w.g.
104
-------
An ash slurry system consists of a slurry tank designed to receive a continuous ash slurry dis-
charge from each pair of scrubbers. Fly ash pumps convey the slurry by force main to the plant ash
pits.
SUMMARY OF TEST RESULTS
Data from the stack emission test report prepared by The Rust Engineering Company and
issued in February 1973 is presented which indicates that all test runs resulted in values below the
allowable emission rate of 0.05 pounds per 1,000 pounds of flue gas in accordance with the contract
documents. Tests were made in the incinerator breechings downstream of the scrubbers and the
induced draft fans, before the breechings leave the incinerator building to enter the stack. Four (4)
test runs were made on unit No. 2 and three (3) runs were made on unit No. 4. Highlights of the
test results are tabulated below:
Run Flue gas flow Paniculate emission % excess
No. ACFM Ib/hr Ib/hr lb/1000 Ib gas air
2SO-1A 43,733 158,200 6.53 .0413 157.23
2SO-2 47,121 169,200 3,18 .0188 157.23
2SO-3 34,885 133,600 1.00 .0075 174.72
2SO-4 43,362 157,500 6.08 .0386 174.72
4SO-1 44,027 158,500 2.85 .0180 165.64
4SO-2 50,643 175,200 6.66 .0380 203.11
4SO-3 49,456 189,200 8.32 .0440 136.42
The first digit of the Run No. designates the incinerator unit, SO designates "scrubber outlet"
and the last digit designates the sequence of the run on that incinerator. Test 2SO-1 was not com-
pleted, so its successive test was designated 2SO-1 A. During all of these runs, a pressure drop of 20
inches was maintained at the scrubber.
OPERATION
The mechanical functioning and the response of the flooded disc scrubbers to changing fur-
nace conditions resulting from sludge feed variations, sludge moisture variations, and furnace off-
105
-------
gases for any reason was immediately perceived to be excellent. The rugged construction of the disc
actuator, the hydraulic system and the basic venturi section has assured reliable service. It is
apparent that the equipment has been designed to serve well in many applications demanding
durability such as foundry and blast furnace service.
A review of the history of the scrubber systems starting with scrubber No. 3 in August 1971,
No. 1 in November 1971, No. 2 in February 1972, and No. 4 in May 1972 revealed a severe wear-
ing or "thinning out" in the tapered side wall, just below each disc. Each worn area spanned the
center line of the long arc of the inlet duct from the furnace to the scrubber. The area of "thinning
out" or reduction in metal thickness was estimated to be approximately 24 inches long by 6 to 8
inches high. None of the remainder of the scrubber shell or the disc itself exhibited any wear. One
scrubber had exhibited wear in 352 hours of operation. Others had required patching by welding
carbon steel plate to the outside of the shell in 1,500 to 2,000 hours.
A laboratory analysis of the fly ash deposited in the disc area revealed that the silica content of
the ash was over 55 percent on a dry basis. Subsequently, the contractor and the scrubber manufac-
turer proposed and installed a refractory-lined throat area. A 3/16-inch thick, carbon steel plate
section was used to replace the Incoloy 825 material in the sloped throat of the scrubber. Applied to
the steel plate are 4 inches of fiberfrax insulation, a bed of Pennwalt HB mortar and 2 inches of
Norton Company CN163 silicon carbide shapes. This type of lining has been applied in blast fur-
nace applications. From present reports, it is still resisting wear well and is expected to have long
life. Subsequent to the rapid erosion of scrubber throats experienced at the Alcosan plant, our
specifications have required that venturi throat sections have an allowance for refractory linings. It
has not been necessary to line any of the scrubbers used on sludge furnaces for municipal sewage
treatment that we have designed since Alcosan.
From an analysis of the volumes of flue gas flow through the scrubbers during the test runs
conducted during January 1973, particularly noting the maximum scrubber outlet gas flow of
189,200 pounds per hour during run 4SO-3, the scrubbers have performed at the intended design
point. During an analysis of furnace operations in 1974 and 1975, heat balances were prepared
utilizing the Combustion Engineering Company computer program. Data from these illustrate the
volume, temperature and moisture of the furnace off gases to the scrubber.
YR 1974 Feed 21,000 Ibs
Dry solids 4,670 Ibs
Furnace - off gas temperature, deg F 800
Furnace - off gas volume, cfm 75,921
Moisture in off gas, gr/lb d.a. 1,409
Excess air, percent 172
Gross heat input, Btuh 53,647,485
106
-------
YR 1975 Feed 20,000 Ibs
Dry solids 4,000 Ibs
Furnace - off gas temperature, deg F 760
Furnace - off gas volume, cfm 76,989
Moisture in off gas, gr/lb d.a. 1,255
Excess air, percent 200
Gross heat input, Btuh 52,024,090
Feed 20,000 Ibs
Dry solids 4,000 Ibs
Furnace - off gas temperature, deg F 760
Furnace - off gas volume, cfm 69,985
Moisture in off-gas, gr/lb d.a. 1,394
Excess air, percent 180
Gross heat input, Btuh 49,408,301
Feed 20,000 Ibs
Dry solids 4,000 Ibs
Furnace - off gas temperature, deg F 760
Furnace - off gas volume, cfm 66,741
Moisture in off gas, gr/lb d.a. 1,471
Excess air, percent 170
Gross heat input, Btuh 48,196,630
Noting the specified scrubber duty requiring it to handle 85,000 acfm at approximately 1,000
degrees F. and comparing the furnace off-gas volumes displayed in the heat balance calculations,
one observes that the specified volume exceeds the computed volumes, even at such high excess air
levels. It is also significant that the dry solids loading to the furnaces during the emission testing
exceeded the feed values utilized in the computer program examples.
SUMMARY AND CONCLUSIONS
The discussion of the suitability of the flooded-disc scrubber for the intended service and its
performance at Alcosan cannot terminate without a well considered attempt to describe, in simple
terms, the sensitive furnace system in which the unit functions. As we have noted, efforts to
reduce infiltration air and to conserve auxiliary fuel have continued since the onset of the furnace
modifications and the construction of the sludge preparation and filter facilities.
Improvements have been ongoing in the roof mounted sludge burners and in the gas burners.
Gas burner flames have been shaped to avoid impingement and the resultant slagging on the walls
and roof arches of the furnaces. A furnace cooling air ductwork system was devised utilizing sludge
107
-------
combustion air which flows through the furnace preheater to cool specific areas of the roof arches
and walls. A parallel benefit of restoring the preheater air flow circuit would be to depress the gas
outlet temperature to the scrubber, thereby improving furnace efficiency and reducing the input
of auxiliary fuel.
The reduction of slagging had been a plant priority for many years under operation with the
mechanical collectors and the slag burden continued as the rehabilitated furnaces continued to
operate utilizing wet scrubbers. The slag problem had caused the destruction of furnace target
walls at times and had become so severe that, even if target walls were not burdened sufficiently to
cause a furnace shutdown, other slag-burdened areas would necessitate a halt to furnace operation.
The need to air cool specific areas of the furnace to prevent slagging was in direct contradic-
tion to the requirement to greatly reduce excess air to an amount only necessary for completing
combustion of the sludge solids.
This discussion, noted in the opening remarks as intending to limit itself to the engineering
application and performance of the flooded disc scrubber for this plant, might expand into many
undertakings which were of major concern to the plant modifications and operation. Nevertheless,
we trust that the system in which the scrubber was applied is more readily understood. The writer
gained a sincere respect for the Authority and it's people due to the willingness they applied to the
solution of many difficult operational problems.
108
-------
©
o
VO
THIRD FLOOR PLAN ELEV. 772.54
SCALE 1/8- = I'-O"
FIG. 1 THIRD FLOOR PLAN
-------
AIR DUCT CONTROL SCHEMATIC
FIG. 2 SECTION
-------
AN INCINERATOR SCRUBBER THAT WORKS: A CASE STUDY
By: Charles Menoher, Product Manager
Zurn Industries, Inc.
Air Systems Div.
245 North Center Street
Birmingham, AL 35204
ABSTRACT
A midwestern chemical company has recently started up a medium energy
scrubber on an industrial trash burning incinerator. Compliance testing
verified an outlet particulate loading of less than the guaranteed 0.15
#/l,000 pounds of dry gas. This level of performance was achieved with
30" W.C. of venturi pressure drop, with final subcooling to remove vola-
tiles and condensibles. This system was purchased because of the horse-
power savings when compared to the "convential" scrubbing approach that
requires 60" W.C. or more of pressure drop. The savings will pay for
the scrubber system in four years.
"TEXT"
In January of 1980, Zurn Air Systems was contacted by a midwest
chemical company to provide a quotation for a high energy scrubber to
clean the off-gases from one of their several in-plant incinerators. The
subject incinerator had an existing low energy scrubber on it, following
the customer's quenching chamber. In this chamber, the 1,600°F off-gas
temperature is quenched to adiabatic saturation at 170°F-
The type of trash being incinerated included all of the normal range
of refuse one would expect to find being fed into an incinerator of this
service but did not, however, include organics, solvents, or heavy metals.
The Bartlett Snow incinerator operation generates an average particulate
loading of approximately 1.0 Gr/SCFD, with periodic excursions up to 2.3
Gr/SCFD, for which the dynamic type scrubber was able to achieve an outlet
average loading of only 0.3 Gr/SCFD. The Bartlett Snow unit, Model 12,
uses a batch feed, and operates 24 hours per day, seven days a week. The
incinerator is rated for an input of 3,000 pounds per hour.
Since the customer's other incinerator scrubber systems were of a con-
ventional design using a high pressure drop venturi and cyclonic separator,
the request for quotation called for a similar concept. After all, those
scrubbers were keeping their incinerators in compliance. Thus, their ex-
pectation was that a venturi pressure drop of 60" W.C. or more would also
be required for this incinerator. With the exception of Zurn, the other
bidders did not disagree with the specification.
Zurn took the position that with nearly three-quarters of the dust
specified as being less than 0.5 microns by weight, high pressure drop
alone was not going to do the job in getting down to the state requirement
111
-------
of 0.15 pounds/1,000 pounds dry gas (nearly equal to 0.08 Gr/SCFD). The
creation of this much sub-micron dust is caused by the intermittent burn-
ing of pharmaceutical powders in the incinerator.
We felt the solution to the problem lay in using a medium energy ven-
turi followed by a direct contact gas cooling tower. The rationale was
that at 30" W.C. of venturi drop, we would do 80 to 85% efficiency, and
that the remaining ten to 15% could be predictably removed only by conden-
sation scrubbing. A secondary, but very important, benefit was the sav-
ings in fan horsepower.
Credit for the concept of condensation scrubbing belongs, not to Zurn,
but to Dr. Seymour Calvert of Air Pollution Technology, from whose articles
in Chemical Engineering Magazine the concept was taken. Dr. Calvert gives
the phenomenum the esoteric name of diffusiophoresis; more frequently
referred to as condensation or mass transfer scrubbing. Associated with
mass transfer scrubbing is "particle growth", which is caused by the con-
densation of water vapor around a particle, thus enlarging the particle
mass, and making it more susceptible to collection by impaction in the
packed bed.
Diffusiophoresis occurs during condensation of water vapor into the
cooling water mass. This condensation exerts a force on the particles
that causes them also to deposit on the cooling water surface. The per-
centage of particles removed in this manner is approximately equal to
the percentage of the gas stream condensed out.
The- easiest way to visualize this phenomenum is to imagine that the
gas stream is pure steam, or water vapor. Obviously, if one condensed the
entire gas stream to water, there would be no air pollution left, because
there would be no gas stream left
Now some specifics. The gas volume from the quencher to the Zurn
venturi was designed for 23,000 ACFM at 170°F and -12" W.C. static pres-
sure. After subcooling, the volume to the I.D. fan was 13,200 ACFM at
100°F and -47" W.C. static pressure.
The cooling water supply rate of 650 GPM at 80°F condensed nearly 43
GPM of water vapor from the gas stream, which amounted to 7,500 SCFM.
This reduced the gas volume by more than 40%, and by all accounts, lowered
the particulate loading from the venturi by about 60%. Of this.60%, it is
assumed that 40% was due directly to diffusiophoretic or mass transfer
scrubbing, with the remaining 20% being removed by impaction after parti-
cle growth. As stated, this can only be assumed since there is no way it
can be readily verified.
The 150 GPM scrubbing water used at the venturi is taken from the
customer's quencher sump, and thus is already contaminated. Given the
configuration of the scrubber/cooling tower, there is no suitable inter-
mediate location between the venturi and packed bed to take samples.
112
-------
For the astute among you who might observe that an approximation could
be made by measuring the percent solids in both the supply and return lines
of both the scrubbing water and the cooling water, this is perhaps true.
However, it would pre-suppose that the cyclonic separator portion of the
vessel is nearly 100% efficient in removing the scrubbing water entrainment
before the gas stream gets to the cooling section. Since no reliable sep-
arator is that efficient, lets look at the reality of how difficult it
would be to verify how effective the cooling section is as a particulate
removing device.
Based on the performance requirement for the scrubbing system, the
allowable amount of dust to the stack is approximately 0.12 pounds/minute.
If only one GPM of the 150 GPM scrubbing water gets to the cooling section
and that gallon per minute contains only one-half percent solids, this
converts to a loading of 0.04 pounds/minute of particulate. This represents
fully one-third of the allowable dust out! It is therefore reasonable to
argue that the cooling section and subsequent mesh pad demister perform a
necessary service in lowering the entrainment to the final stack. Even so,
it is a certainty that even without entrainment to worry about, it would
take a significantly higher venturi pressure drop to meet the guarantee if
no subcooling was used.
When measured last week according to the procedures in the EPA Title
40, Chapter 1, Subchapter C, Part 60, the outlet dust loading was 0.12
pounds/1,000 pounds dry gas (against the requirement of 0.15 pounds/1,000
pounds) and the stack opacity was 10%.
Since we were hopeful that even our specified 30" W.C. of venturi pres-
sure drop was going to be more than adequate because of the subcooling that
followed, the throat was intentionally sized large enough to permit a pres-
sure drop as low as 20" W.C. Since the test results were well under the
code, there is no reason why the customer should not back off from the 30"
pressure drop design number and operate at a lower horsepower.
As alluded to previously, the much lower power consumption of the Zurn
system was extremely important to the customer. They are very happy with
the horsepower savings the system is providing compared to what they had
originally expected.
If, for the sake of discussion, they had bought a scrubbing system that
called for a 60" W.C. drop on the venturi and with no subcooling, the I.D.
fan would had to have been sized for approximately 28,000 ACFM at 75" W.C.
static pressure. Assuming a brake horsepower efficiency of 72%, the operat-
ing horsepower would be approximately 472. The fan actually used on this
installation operated at a brake horsepower of about 145. The savings on
the fan power, then, is 327 horsepower. From this must be subtracted the
horsepower of the cooling water supply pump. Assuming a 55% efficient pump
rated at 650 GPM and 40 PSIG TDH, the operating horsepower would be 27. The
total net savings in horsepower, therefore, of the installed system versus
the conventional approach, is approximately 300 HP. At $250/HP/Year, the
annual savings at that rate is $75,000. The power savings will pay for
entire installation in less than four years.
113
-------
The materials of construction used in the system are what one would
expect for this application, given the concern for chlorides. The duct
work from the quencher to the venturi if FRP; the venturi is rubberlined
carbon steel with a Hastelloy C adjustable throat; the cooling tower, duct
work to the I.D. fan, and final stack are FRP, and the I.D. fan has a rub-
berlined carbon steel housing and Hastelloy G. wheel.
The hook-up to the customer's existing quench chamber was made by re-
moving the dynamic scrubber and installing a "plenum box" in its place. The
duct work was then run from the top of the box over to the venturi inlet.
The venturi design is conventional. Given several types of venturi to
offer, we chose the one with open pipe water feed. Since the scrubbing
water is the re-used quenching water from the customer's existing sump, it
was felt that the use of spray nozzles ran the risk of pluggage. At the
very least, spray nozzles would require periodic replacement simply due to
erosion. With open pipes at low pressure, it is hoped the venturi selected
will give the customer many years of maintenance free service.
The separator/cooling tower is also of a conventional design. The
lower cyclonic separator section is isolated from the cooling section with
a ring and cone arrangement. This, of course, is to keep the dirtier,
lower pH scrubbing water from mixing with the cooling water.
The necessary cooling is achieved throug a 9' deep bed of polypropylene
packing supported on an FRP grid. The cooling water is applied to the bed
through a PVC water distributor, using no nozzles, at low pressure. Here
again, the intent was to provide the customer with a design that would re-
quire no spare parts, and thus minimize maintenance.
Finally, the cleaned and cooled gas stream passes through a polypropy-
lene mesh pad to strip out all entrainment before going to the I.D fan and
stack.
The pH of the scrubbing water is controlled by a continuous small ad-
dition of caustic from the customer's effuent stream from one of these
processes. He monitors the pH continuously, and sets the caustic flow
rate to keep the pH of the scrubbing water above four.
Other than some minor problems with instrument calibration, the start-
up went smoothly. The system has been on line for about three weeks now,
with no indication of any operational problems. The customer is pleased
with the system.
In an era of high energy costs, the scrubber industry can no longer
continue to advocate the same scrubbing techniques that were popular, or
at least acceptable, less than ten years ago. Frankly, there is little in
the Zurn scrubber system on this job that is terribly innovative. We like
to feel that the system is simply the solid application of proven scrubber
techniques.
114
-------
Endnotes
1. Chemical Engineering; August 29, 1977 and October 24, 1977.
Author Background
Charles Menoher is Product Manager of Wet Scrubber Products at Zurn
Air Systems, having been an Application Engineer in wet scrubbers at Ameri-
can Air Filter, Koch Engineering, Sheldons Engineering (Canad) and Chemico.
He has authored articles for J33 Metal Producing, and Iron and Steel Engineer.
115
-------
-------
RACETRACK VENTURI
117
-------
SEPARATOR • COOLING TOWER
118
-------
EVALUATION OF ENTRAINED LIQUOR CONTRIBUTION TO TOTAL
MASS EMISSIONS DOWNSTREAM OF A WET SCRUBBER
By: W. David Balfour, Larry 0. Edwards, Hugh J. Williamson
Radian Corporation
8501 Mo-Pac Boulevard
Austin, Texas 78766
ABSTRACT
As flue gas passes through a wet scrubber, both S02 and particulate
matter can be removed. The scrubber can, however, contribute mass through
entrainment of scrubber liquor high in suspended and/or dissolved solids.
The contribution of mass by a variety of scrubbing systems has been calcu-
lated using the concept of a chemical element balance, as:
J j j
where:
41. E mass flow rate of species j out of scrubber
P E penetration fraction of fly ash through scrubber
. = mass flow rate of species j into scrubber
C. = concentration of species j in scrubber liquor
V = volume rate of entrained scrubber liquor
Particulate compositions into and out of the scrubber were obtained from
chemical analyses (ion chromatography and inductively coupled argon plasma
emission spectroscopy) of total particulate catches. Scrubber liquor compo-
sition was obtained by similar analyses of liquor samples collected during
the particulate sampling. A weighted regression analysis was performed on
the set of equations (for each chemical species).
Results for a variety of combustion and scrubber processes show that
entrained liquor can contribute from 1 to 80 percent of the total mass
emissions from the wet scrubber.
The Clean Air Act and subsequent New Source Performance Standards (NSPS)
require that sulfur dioxide (SOa) emissions be controlled. The flue gas
desulfurization (FGD) technology most commonly used in controlling SOa emis-
sion is the wet scrubber. A variety of reagents can be used as a scrubbing
liquor including soda ash, caustic, lime or limestone. Similarly, a variety
of contactors are used to provide the necessary liquor surface area required
for the gas-to-liquid transfer. The dissolved solids content of the scrubber
liquor can be quite high due to recycling of the liquor, with addition of
reagents and make-up water as needed (Figure 1). The solids content is kept
at an acceptable level by blowing down the liquor as necessary. Any entrained
119
-------
SLOWDOWN -*-
©
FLUE
GAS OUT
MIST ELIMINATOR
WASH
CONTACTOR
FLUE
GAS IN
SCRUBBER LIQUOR
RECYCLE TANK
o
MAKEUP
WATER
-? FEED
SAMPLING POINTS
1. FLUE GAS INLET
2, FLUE GAS OUTLET
3. SCRUBBER LIQUOR
Figure 1. Schematic of a Wet Scrubber
120
-------
liquor droplets which penetrate the mist eliminator can contribute to the
overall particulate mass emission as a result of these dissolved solids in
the liquor droplets. This mechanism is commonly used to generate test aero-
sols in the laboratory.
This paper describes an approach for determining the contribution of
entrained liquor to the total mass emission downstream of a wet scrubber.
The information can be used to evaluate mist eliminator performance and
define actual versus apparent particulate penetration through the control
device. With increased concern over fine particle emissions, particularly
sulfate particles, such an evaluation can be helpful in determining the
source of such emissions.
Approach
The contribution of mass by entrained liquor to the total mass pene-
trating the scrubber can be calculated using the concept of a chemical
element balance. If one assumes that the mass flow of species j out of the
scrubber is made up of contributions from the fly ash and entrained liquor,
then the mass balance equation may be written:
Dn + c. v [i]
J J
where:
. is the mass flow rate (g/min) of species j out of the scrubber,
P is the penetration fraction of the fly ash particles,
. is the mass flow rate (g/min) of species j into the scrubber,
C. is the concentration (g/m£) of species j in the scrubber liquor,
J and
V is the volume rate (m£/min) of entrained scrubber liquor.
The mass flows into and out of the scrubber can be measured directly.
Similarly, the solids content of the scrubber liquor can be measured. The
penetration fraction and volume rate of entrained liquor can then be calcu-
lated indirectly. By solving the chemical element balance for multiple
species, greater confidence can be gained in the solution, and an indication
of the error for the solution obtained.
Several assumptions are necessary, however, to use this approach. First,
the fly ash penetration fraction must be constant for all species. This is
to say that the species selected for the chemical element balance must not
exhibit appreciable particle size dependence in the fly ash. Previous
studies (1,2) have shown that Ca, Mg, Ti, K, Mn, Na and Fe all show minimal
size dependence in fly ash. Elements such as Ba, Zn, V, Cr, Co, As, Br, Se,
Hg and Ni can show large particle size dependence, and as such may not be
suited to this chemical element balance approach. Similarly, highly
soluble species such as sulfate, chloride and fluoride may not be suited, as
discussed later in this paper.
121
-------
The final assumption requires that the scrubber liquor be homogeneous.
This assumption should be good, since most scrubbers are well mixed systems.
Notice that the volume rate calculated is for entrained liquor only and does
not include evaporated liquor. While the scrubber liquor can contain both
suspended and dissolved solids, the suspended solids are assumed not to be a
significant source of mass. Justification of this assumption is provided
later in the paper.
A variety of statistical methods may be used in the solution of this
chemical element balance. The equation used in the statistical model
development is
cj)0ut = p ^n + c. V + I + e. [2]
3 J J J
where an intercept term I has been added, ej represents a random error term,
and the other terms are as defined above (see Equation [1]). Weighted
regression analysis can then be used to determine the values of P, V, and I.
In weighted regression, the fact that the different measured values have
different error variances is taken into account; that is, P, V, and I are
chosen so as to minimize the following sum of squares:
/s? [3]
3=1 3 3
where:
. is the j predicted value of mass flow rate out of the scrubber,
sf is the error variance of the corresponding observed value
J (i>9ut, and
n is the number of species included.
•j
In standard regression analysis, the term s. would be left out, giving all
observations the same weight; in the above, the data points with the smallest
error variances are weighted more heavily. This weighted regression pro-
cedure also prevents one or two points far away from a cluster from exerting
a dominant influence in determining the slope.
The intercept term I in Equation [2] represents the mass flow rate out
of the scrubber for a hypothetical species with 4>^n = 0 (zero mass flow rate
into the scrubber) and Cj = 0 (zero concentration in the scrubber liquor).
Clearly, such a species should have zero mass flow rate out of the scrubber,
within random error, unless there is a source which is not accounted for by
the model. Thus, the intercept term together with its standard error serves
to provide a check on the closure of the chemical element balance.
One additional statistical point should be mentioned briefly. Near
linear dependencies (multicollinearities) among the predictor variables,
122
-------
(4>j and Cj) is present, can cause spurious results in regression analyses.
This problem and its effects have been discussed widely in the statistical
literature (3) . An investigation of selected scrubber cases with larger-
than-average standard errors in the regression coefficients has revealed
that^the results presented here have not been affected significantly by multi-
collinearity. This conclusion is based on several statistical measures
(variance inflation factors in weighted ridge regression, etc.).
Sampling and Analysis
Particulate mass and chemical composition into and out of the scrubbers
were obtained from analysis of EPA Method 5 (4) and/or 17 (5) particulate
catches. Scrubber liquor composition was obtained from similar analyses of
scrubber liquor samples collected during the particulate sampling.
Collection of particulate samples into the scrubber were usually ob-
tained using an EPA Method 17 type sampling train. The instack filtration
system eliminates the possible losses in a probe during extraction of the
sample out of the stack. However, collection of particulate samples out of
the scrubber typically must be obtained using the extractive EPA Method 5
type sampling train. This requirement is necessary due to the high moisture
content of the flue gas and frequent presence of entrained droplets in the
flue gas (in the case of no reheat). The heated probe and filter holder
elevates the gas temperature sufficiently to evaporate any droplets and
prevent condensation of moisture on the filter which may result in blinding.
In both the EPA Method 5 and 17 sampling trains, filters with low chem-
ical backgrounds must be used. Gelman Micro Quartz filters have been used
with success due to the compatibility with the required digestion procedures
for the chemical analysis. Sampling times must be chosen to ensure adequate
sample for the species detection limits.
The sampling point for the scrubber liquor sample must be chosen to
ensure that the sample is representative of the liquor being exposed to the
flue gas. The liquor sample should be filtered immediately as it is being
obtained. A high capacity Millipore filtration system is suitable. The
filtration is necessary to prevent any equilibrium interaction between the
solid and liquid phases. Temperature and pH measurements can be obtained
to further characterize the scrubber process (and serve to help check the
chemical analyses results). The dissolved solids in the filtrate and the
filtered solids are both determined gravimetrically and held for chemical
analysis.
Prior to analysis, samples were digested using a mixed acid technique
(6). Elemental analysis was performed on the digested particulate and
scrubber solids samples using inductively coupled argon plasma emission
spectroscopy (ICAPES) (7). The emitted spectral bands specific for each
element are detected by separate photomultiplier tubes. The system (Applied
Research Laboratories 34000) is computer controlled which allows for simul-
taneous multi-element determination. Filter background and matrix inter-
ferences are automatically adjusted.
123
-------
Soluble species analysis (SOi*, Cl , F ) was obtained using ion chroma-
tography (1C) (8). Separate aliquots of sample were leached using a
carbonate solution. This leachate was then injected directly into the 1C
(Dionex Model 10) for qualitative and quantitative analysis.
The expected errors in these analyses are less than 15 percent for the
particulate samples and less than 10 percent for the scrubber liquor samples.
This information is used in the chemical element balance model, as previously
described.
Results and Discussion
The chemical element balance approach to determining the contribution
of entrained liquor to the total mass emissions from three individual wet
scrubber systems was investigated. Table 1 summarizes the combustion and
scrubber processes for each of these systems. All scrubbers included mist
eliminators of either a chevron or wire mesh design. Contactors included
a variety of configurations. Scrubbing reagents included limestone and
caustic.
TABLE 1. SUMMARY OF COMBUSTION AND SCRUBBER PROCESSES
System
Number
Fuel
Type
Scrubber Type
Boiler
Load
Special
Conditions
coal limestone - venturi
+ spray
oil radial flow venturi
for particulate control
oil radial flow venturi
for particulate control
550 Mw
650 Mw
hot air
injection
reheat
no reheat
650 Mw caustic
addition,
no reheat
Table 2 presents the results of the weighted least squares analysis for
each of the systems investigated. The number of species available for
inclusion in the chemical element balance are listed for each system. The
square of the multiple correlation coefficient (r2) for the analyses range
from 0.81 to 0.99, indicating valid relationships in all cases. The inter-
cept values (I) represent less than one percent of the mass rate out of the
scrubber, indicating good closure of the chemical element balance. Compari-
son of the penetration fraction (P) and the volume rate of entrained liquor
(V) with their respective standard errors indicate that the calculated values
of P and V are usually statistically significant.
124
-------
TABLE 2. RESULTS OF CHEMICAL ELEMENT BALANCE - WEIGHTED LEAST SQUARES
ro
en
Sys tern
Numb er
1
2
3
Number
of
Species
5
7
7
Penetration
Fraction
(P)
0.001
0.76
0.76
Standard
Error
in P
0.0004
0.12
0.16
Volume
Entrained
(V) U/min)
6.9
10.
5.5
Standard
Error in V
U/min)
5.
13.
12.
Intercept
(I)
(g/min)
0.033
-4.44
-12.1
Standard
Error in I
(g/min)
0.029
15.4
15.8
Correlation
Squared
(r2)
0.87
0.93
0.89
-------
Table 3 presents these results of the chemical element balance for total
mass emissions. Results are expressed as a percentage of the total mass
emission due to either 1) penetration of incident fly ash or 2) entrainment
of scrubber liquor. From these results, it can be seen that entrained
scrubber liquor can contribute from 1 to 40 percent of the total mass emis-
sions measured downstream of a scrubber. The value for system number 1 was
not calculated directly, because the total dissolved solid concentration
was not available. However, a value of 30 percent was calculated indirectly
based upon the total predicted concentrations of species entrained from
the scrubber.
TABLE 3. CALCULATED CONTRIBUTION OF SCRUBBER LIQUOR TO TOTAL MASS EMISSIONS
Percent Total Mass Due to
System
Number
1
2
3
Penetration
63
98
99
Entrainment
*
1
1
*TDS value not available.
The solutions to this chemical element balance model consistently under-
estimate the measured value of sulfate emissions. Several explanations are
offered. Sulfate does show particle size dependence, with a large fraction
of the total sulfate concentration found in fine particles. As a result,
the calculated penetration fraction may not be applicable for sulfate due
to the poorer removal of fine particles by a wet scrubber. Additionally,
gaseous SOa may be transferred directly to entrained droplets. Once in
solution, the S02 may be readily oxidized to sulfite and/or sulfate.
Finally, sulfuric acid mist may contribute to the sulfate emissions from the
scrubber. However, the sulfuric acid mist concentrations measured at these
systems cannot by themselves account for the underprediction of sulfate.
Undoubtedly, the difference between the predicted and observed values for
sulfate are due to a combination of mechanisms.
Similar discrepancies have been found for chloride and fluoride. Like
sulfate, sources of chloride and fluoride can be found in the gas phase, and
are highly soluble in the scrubber liquor. This fact may explain the
observed low predicted values for these species.
Conclusions
A chemical element balance approach has been described which determines
the contribution of entrained scrubber liquor to the total mass emissions
from a wet scrubber. A weighted regression analysis of the data is used.
126
-------
Results from a variety of combustion and scrubber processes show that the
approach provides statistically significant results with good closure of the
mass balance. Contributions from entrained liquor to the total mass emis-
sions were observed to range from 1 to 80 percent of the total mass. As
such, wet scrubbers can be a significant source of mass emissions.
ENDNOTES
Acknowledgments
The authors thank Ms. Judy Tobias for her help in the statistical analysis
of the data.
References
1. Meserole, F. B. et al. Trace Element Emissions from Coal-Fired Power
Plants. Journal of Engineering for Power, 101:620-624, 1979.
2. Ondov, J. M. et al. Emissions and Particle-Size Distributions of Minor
and Trace Elements at Two Western Coal-Fired Power'plants Equipped with
Cold-Side Electrostatic Precipitators. Environ. Sci. Technol.,
13:946-961, 1979.
3. Hoere, A. E. and R. W. Kennard. Ridge Regression: Biased Estimation
for Monothogonal Problems. Technometrics, Vol. 12, 1979.
4. EPA Method 5: Determination of Particulate Emissions from Stationary
Sources. Environmental Reporter, October 21, 1977.
5. EPA Method 17: Determination of Particulate Emissions from Stationary
Sources (In-Stack Filtration Method). Federal Register, Vol. 41,
No. 187. September 24, 1976.
6. McQuaker, N. R. et al. Digestion of Environmental Materials for Analysis
by Inductively Coupled Plasma Atomic Emission Spectroscopy. Anal. Chem.,
51:1082-1084, 1979.
7. McQuaker, N. R. et al. Calibration of an Inductively Coupled Plasma
Atomic Emission Spectrometer for the Analysis of Environmental Materials.
Anal. Chem., 51:888-895, 1979.
8. Mulik, J. D. and E. Sarviki. Ion Chromatographic Analysis of
Environmental Pollutants, Vol. 2, Ann Arbor Sci., Ann Arbor, MI, 1979.
127
-------
A DUAL-BEAM BACKSCATTER BETA-PARTICLE GAUGE FOR MEASURING THE DUST
CAKE THICKNESS ON OPERATING BAG FILTERS INDEPENDENT OF POSITION
By: Robin P. Gardner
Center for Engineering Applications of Radioisotopes
North Carolina State University
Box 5636
Raleigh, NC 27650
R. P. Donovan
Process and Chemical Engineering Division
Research Triangle Institute
Post Office Box 12194
Research Triangle Park, NC 27709
L. S. Hovis
Utilities and Industrial Processes Division
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
ABSTRACT
A prototype, dual-beam, backscatter beta-particle gauge for measuring
the dust cake thickness on operating bag filters has been designed, con-
structed, and tested. The device uses the two beta-particle sources Kr-85
and Sr-90 with an end-window GM counter so that the measurement of dust cake
thickness is essentially independent of the position of the bag in relation
to the gauge. A plastic rotating-shutter shield operated by a small electric
motor is used to provide signals alternately from each source. The shutter
position is sensed by an infrared detector and the response to each source is
appropriately routed to separate counters. The low-energy Kr-85 source pro-
vides a backscatter response that depends only on filter position while the
high-energy Sr-90 source provides a backscatter response that depends on both
filter position and dust cake thickness. The real-time simultaneous solution
of appropriate models for the two responses with a microcomputer gives the
dust cake thickness, an estimate of the standard error of dust cake thickness
due to statistical counting rate fluctuations, and the filter position for
each set of responses. Laboratory and pilot plant results indicate that the
gauge is accurate and will be useful for research purposes.
INTRODUCTION
The usual operation of bag filters for removing particulates from gas
streams is cyclic. The cycle consists of a normal filtration period fol-
lowed by a period in which the dust cake is removed, usually by shaking
or reversing the gas flow. The efficient operation of bag filters
involves the consideration of a number of operating variables for a given
system of particulates in a gas. One of the most important of these is
the dust cake thickness. While this thickness can be inferred from pressure
128
-------
drop measurements across the bag filter or by other indirect methods, it would
be very advantageous to have a direct, continuous, in situ measurement of it.
Such a measurement would be important both in research applications in the
laboratory or pilot plant and in the normal control of an industrial facility.
This paper describes the development and preliminary testing of a dual-
beam backscatter beta-particle gauge for making the desired measurement.
Others (1-3) have investigated the use of transmission beta-particle gauges
for this application. While the transmission approach has merit, it suffers
from the disadvantages that: (1) the in-situ response depends on sample
position in an unknown way, (2) access is required to both sides of the bag
filter, and (3) the measurement obtained represents the average of both sides
of the bag filter at a particular position. The development of the back-
scatter approach described here was undertaken to avoid these disadvantages.
Although the response of a backscatter type beta-particle gauge also
depends on sample position, the effect is easier to specify than that for a
transmission type gauge. Therefore, in principle, the effect of variable
sample position can be eliminated by using a dual-beam approach with two
sources of different beta-particle energies. It is assumed in this approach:
that the backscatter responses for two beta-particle sources of different
characteristic energies will depend on sample position and thickness differ-
ently, that the two responses can be modeled, and that the two models can be
solved simultaneously to give a measurement of thickness independent of
position. The parameter that is directly measured by such a device is the
combined density thickness or area! density of the cloth filter and the dust
cake layer. Therefore, to derive either the dust cake porosity or thickness
from this measurement requires an independent knowledge of the other.
BETA-PARTICLE BACKSCATTER MODELS
The final beta-particle backscatter response models are modifications of
previous models (4). The basic beta-particle backscatter response is taken
to be an exponential function that yields saturation values. The models are:
RA = {RsiA[1-e>
-------
When all the model parameters and the model responses in Eqs. 1 and 2
are known, sample density thickness x can be solved for independently of
sample distance d by using a simple trial -and-error method. The authors used
the iterative Newton-Raphson method in which a root of the function f(x) is
found from an initial estimate x0 from the relationship:
Xi = Xi_i - f(xi_i)/f'(x1.i) (3)
where x-j is the ith estimate of the root and f'(x) is the derivative of the
function f(x). In this case, f(x) is obtained by solving Eqs. 1 and 2 for
G exp(-kQd) and setting the results equal to each other. The resulting f(x)
is given by:
f(x) = (RA-RB1A-RB2A-RBA)/^ RsiACl-exp(-klX)] + RS2A[>exp(-k2x)]} -
(RB-RBlB-RB2B-RBB)/{ RslB[l-exp(-kix)] + Rs2B[l-exp(-k2x)]} = 0 (4)
After the value of x has been obtained, Eq. 1 or 2 can be solved
directly for d. An analytical estimate of the standard deviation of x can be
obtained by using the standard approach of expanding the explicit x function
in a Taylor Series and using only the first-order terms. The resulting
relationship is:
s(x) = [A2RA/t + B2RB/t]l/2 (5)
where t is the counting time for each channel in seconds and the A and B are
given by
A = l>lRsiA exp(-kix) + k2Rs2A exp(-k2x)]G exp(-kGd) (6)
exp(-kix) + k2Rs2B exp(-k2x)]G exp(-kQd) (7)
This relationship for s(x) assumes that the standard deviations of the dual-
beam responses are independent and equal to that due to counting rate fluc-
tuations only. These assumptions are not strictly valid and a more exact
method has also been employed to evaluate the s(x). This method consists of
randomly choosing sets of R/\ and RB values that have Gaussian distributions
about their true values with arbitrarily chosen standard deviations. The
resulting measured values of x and d are calculated for each set of RA and
RB and are then used with the known values of x and d in the general esti-
mator for standard deviations to obtain accurate estimates of the standard
deviations.
SOURCE OPTIMIZATION STUDY
A study was performed to determine the optimum two beta-particle sources
for the dual -beam backscatter gauge. This was accomplished by using modifi-
cations of the models given in the previous section to evaluate a design
objective function consisting of the sum of the squares of the standard
deviations in measured filter cake thickness due to: (1) statistical counting
rate fluctuations and (2) variations in beta-particle attenuation in the
130
-------
backscatter path. This objective function was minimized by an appropriate
nonlinear search program called OPTCON (5) to obtain the optimum values of
source backscatter coefficients kj and k£. The objective function was
taken as:
sf(x) = s|(x) + s2(x) (8)
where s|(x) is the objective function or the total variance in the measured
value of sample density thickness, s2(x) is the variance of x due to statis-
tical counting rate fluctuations, and s?(x) is the variance due to variable
attenuation of the beta particles in thgir backscatter path. The variance
of x due to statistical counting rate fluctuations is equal to the sum of the
variances due to each of the two dual -beam responses:
s§(x) - s2$(x) + s§s(x) (9)
To account for detection system stability as well as statistical counting
rate fluctuations, the standard deviations of the two dual -beam responses were
assumed to be 1 percent of the response values. Then standard deviations
SAS(X) ancl SBS(X) are determined by separately solving the modified versions
of Eqs. 1 and 2 simultaneously for values of fy\ and Rg equal to 1.01 times
their actual values, respectively. The differences between the values of x
obtained for 1.01 RA and RA and 1.01 Rg and RB are taken to be the standard
deviations SAS(X) and SBS(X), respectively.
The measurement variance due to variations in beta-particle attenuation
is introduced by first including additional attenuation factors in Eqs. 1 and
2. Since the values of R$2A and R$IB are small compared to the values of
RSIA and R$2B» the latter two values are neglected in the present analysis.
For the optimization study, the two dual -beam responses previously given in
Eqs. 1 and 2 become:
RA = RsiAU-expC-kix^G exp(-kGd)exp(-2uidrG)+RBlA+RB2A+RBA (10)
RB = Rs2Bl>exp(-l<2x)]G exp(-kGd)exp(-2u2drG)+RBlB+RB2B+RBB
where U] and u2 are the attenuation coefficients in cm2/g for sources 1 and
2, respectively, and rG is the atmospheric density along the backscatter path
in g/cm3. Various values of SA(X) are obtained by solving Eqs. 10 and 11
simultaneously for x for specified variations in the atmospheric density rG
and subtracting the values obtained for x with the base value of rg.
Although the standard deviation s/\(x) is actually based on variations in
atmospheric density, it can be thought of as representing a class of attenu-
ation type errors; e.g., the presence of dust particles in the gas or dust
layers on the source or detector.
The fixed parameters in the present problem include: (1) the dust cake
thickness and density, (2) the filter thickness and density, (3) the atmo-
spheric density rG, (4) the gauge background responses RBA and RBB, (5) the
measurement time interval, and (6) the gauge-to-sample distance d. The first
two of these parameters are given by the sum of their density thicknesses:
each is taken as 0.1 g/cm2 to give a sum of 0.2 g/cm^ for the design value of
131
-------
x. Atmospheric density rg is taken as 1.3 x 10"3 g/cm3. The gauge background
responses Rg/\ and Rgg are both taken to be 20 counts per second. The measure-
ment time interval is fixed by the desired response time and is taken as
10 seconds. The gauge-to-sample distance should be as small as practical, so
a value of d of 2 cm is taken. The remaining fixed parameters required in
Eqs. 10 and 11 are related to source intensities. The values of R$IA and
RS2B are taken to be 1.0 x 103 counts per second, the value of G is taken to
be 1.5 x 103, the value of kg is taken to be 0.171 cm-1, and the values of
RB1A> RB2A> RB1B» and RB2B are taken as zero.
The remaining design parameters to be optimized are uj, U2, kj, and k£,
which are the attenuation and backscatter coefficients for the two sources
chosen. According to Knoll (6) there is a power law dependence of charac-
teristic maximum beta-particle energy and the attenuation coefficient. That
relationship is:
Emax = (u/14.23)-0.722 (MeV) (12)
where u is in cm^/g units. One can also identify a relationship between the
attenuation and backscatter coefficients (4). It is taken to be approxi-
mately linear and is given by:
k = 3.27u (13)
This relationship can be used to eliminate two of the four design parameters
to be optimized.
The results of the design optimization for various levels of the attenu-
ation error are given in Table 1. The results indicate that the maximum
TABLE 1. SOURCE OPTIMIZATION RESULTS
Attenuation
Error, A TG
(g/cm3)
2.6xlO-5
1.3x10-4
2.6xlO-4
5.2x10-4
Standard Errors (x 103'
SA(X)
(9/cm2)
1.99
2.20
3.14
5.22
sS(x)
(q/cn.2)
7.22
9.51
9.96
10.37
) Source Maximum Energies
Source 1, Emax
(MeV)
3.500*
3.500*
3.500*
3.500*
Source 2, Emax
(MeV)
0.245
1.086
1.351
1.503
*The maximum value allowed.
energy for one of the sources is always the maximum allowed. The maximum
energy for the second source varies with the amount of attenuation error
chosen. For low values of the attenuation error, small second-source maximum
energies are obtained.
The commonly available beta-particle radioisotope sources are given in
Table 2 with their pertinent characteristics. The radioisotope Ru-106 emits
useful high-energy beta particles but has an impractical half-life of only
132
-------
1.0 year. Consideration of the half-lives and beta-particle energies leads
to the selection of Sr-90(Y-90) and Kr-85 as the most likely combination of
sources.
TABLE 2. CHARACTERISTICS OF COMMON BETA-PARTICLE RADIOISOTOPE SOURCES
Source
Ru-106
Sr-90(Y-90)
Tl -204
Kr-85
Pm-147
C-14
Maximum
Energy, Emax
(MeV)
3.53
2.26
0.764
0.672
0.225
0.156
Half-Life
(years)
1.0
28
3.6
10.3
2.6
5730
Attenuation
Coefficient
(cm2/g)
2.5
4.6
20.7
24.7
112.
187.
PROTOTYPE DESIGN
A prototype gauge was designed and constructed. It consisted of a plas-
tic base plate 0.63 cm thick, 10 cm long, and 10 cm wide. A thin end-
window GM tube with a 2.5 cm diameter was mounted in a hole in the center of
the plastic base. Holes were drilled on opposite sides of the GM tube for
mounting beta-particle sources. A circular 180-degree sector of plastic was
mounted on an axle so as to rotate and alternately shield each source. A
small electric motor was attached to this rotating shield with a pulley
arrangement to drive it. Finally an infrared source and detector were
mounted in line with one of the sources to detect the position of the rotat-
ing shield. This infrared detector was used in an electronic circuit to
switch the GM tube signal from one counter (A) to another (B). The resulting
device records the response of source 1 in counter A and source 2 in counter
B. A schematic diagram of the device is shown in Fig. 1.
Three different beta-particle sources have been purchased for use in the
prototype device. These include 1 millicurie (mCi) sources of Ru-106 and
Sr-90(Y-90) encapsulated in steel, a 0.1 mCi source of Sr-90(Y-90) encap-
sulated in aluminum, and two 10 mCi sources of Kr-85 encapsulated in alumi-
num. All sources are encapsulated in right circular cylinders 2.22 cm long.
The first three sources have diameters of 0.32 cm; the last two, 0.64 cm.
The windows of all sources were originally 0.025 cm thick, but the windows
on the two Kr-85 sources were subsequently milled down to about 0.008 cm or
less.
The ancillary electronics used with the device are a standard high volt-
age supply (ORTEC Model 456), a Timer Counter (ORTEC Model 773), a Dual
Counter/Timer (ORTEC Model 715), a Print-Out Control (ORTEC Model 432A), and
a Printer-Paper Tape Punch (Teletype). The rotational speed of the shield
was adjusted so that a complete revolution is obtained in exactly 3.00 sec-
onds. Therefore, the timer can be adjusted to any integer multiple of 3
seconds (3, 6, 9, 30, 60, etc.) to ensure an equal amount of time for the
counts accumulated in each channel. In addition to this, a standard pulse
133
-------
inverter was used to invert the GM tube pulses and an electronic circuit to
route the pulses according to the condition of the infrared detector was
designed and built. In the future an Apple II Microcomputer will be con-
nected to the system so that real time calculations can be made to obtain the
filter cake thickness directly from the present dual-beam responses.
ALTERNATOR SHIELD
BETA SOURCE
r////////< SHIELD
END-WINDOW
GM COUNTER
K/////////ISHIELD
BETA SOURCE 2
TO COUNTERS
TO SIGNAL SWITCHER
f-ig. i. bcnematic diagram of the prototype, dual-beam, backscatter beta-
particle gauge.
DETERMINATION OF CALIBRATION MODEL PARAMETERS
When initial responses were taken with the dual-beam gauge it was found
that counting rates were high enough with the existing sources to cause sig-
nificant dead time losses and corresponding low observed counting rates in
the GM tube. Rather than decrease the counting rates by several possible
methods, it was decided to correct for the dead time losses mathematically.
This approach has the advantage that the device will have the fastest pos-
sible response times for a particular desired accuracy level.
For GM tubes, the observed counting rate RQ is related to the true
counting rate Rj and the dead time r by:
RQ - Rj exp(-RTr)
(14)
where RQ and Rj are the observed and true counting rates in counts per second,
respectively, and r is the dead time in seconds. This is the relationship
for parallyzable (4) detectors where the total dead time in the detector is
a function of the true counting rate. The dead time r can be measured when
134
-------
the true and observed counting rates are known. One method for determining
the true counting rate is the decaying source method. In this method a
radioisotope of known half-life and sufficient initial intensity to cause
significant counting losses is placed next to the detector and the observed
counting rate is recorded as a function of time. At sufficiently long decay
times, negligible counting losses occur and the exponential decay curve with
the decay constant appropriate to the half-life of the radioisotope used can
be established. This permits extrapolating the true counting rate back to
the short decay times and determining dead time r as a function of true
counting rate.
This experiment was performed with the radioisotope Si -31 with a half-
life of 2.62 hours. The dead time r was found to fit the following rela-
tionship with true counting rate:
r = DI exp(-D2Rf) - 03 expC^Rj) seconds (15)
where Di, D?, 03, and 04 are constants with the values 1.20 x 10~4s,
2.69 x IO-5S, 2.69 x lQ-4s, and 9.71 x 10'4s. Substituting Eq. 15 for
r into Eq. 14 gives a transcendental relationship that must be solved for
the true counting rate in terms of the observed counting rate.
The counting rate as a function of sample distance from the gauge was
found to be exponential for a wide range of distances. The resulting rela-
tionship for backscatter response as a function of sample distance for all
sources is:
Rd = R06 exp(-kGd) d _> 2 cm (16)
where G is a constant, d is the distance to the sample in cm, kg is a con-
stant (found to be 0.397), and R^ and R0 are the counting rates at dis-
tances d and zero, respectively.
The remaining model parameters in Eqs. 1 and 2 were found by first
obtaining data on paper and pasteboard samples of known density thickness
at known distances from the dual -beam gauge for each radioisotope source
separately. Then a nonlinear least-squares method was used to evaluate the
parameters for the data obtained. The model parameters so obtained for each
radioisotope source are given in Table 3.
TABLE 3. MODEL PARAMETERS OBTAINED FOR VARIOUS SOURCES AT A
SAMPLE DISTANCE OF 3.9 cm
Source
Ru-106
Sr-90(steel)
Sr-90(Al)
Kr-85(No. 1)
Kr-85(No. 2)
RSIA
(c/s)
39.08
1006.
244.5
16.62
20.88
RB1A
(c/s)
139.6
1945
796.6
93.48
105.7
RSIB
(c/s)
1.729
4.250
0.0
0.0
RBIB
(c/s)
48.69
35.59
52.58
50.15
kl
(cm2/g)
10.36
11.42
11.42
46.25
52.33
135
-------
A typical set of data for obtaining the model parameters is shown plotted
versus the model values in Fig. 2.
SAMPLE DENSITY THICKNESS (g/cm2)
Fig. 2. Experimental data and calibration model for the Sr-90(Al) source.
A MICROCOMPUTER PROGRAM FOR OBTAINING REAL TIME MEASUREMENTS
A computer program has been written and tested to process the response
data from the dual-beam gauge to give a measurement of the filter and dust
cake density thickness, the distance to the filter, and the estimated
standard deviation of the density thickness. The program was written in
BASIC for the Apple II Microcomputer. The program includes a decay correc-
tion for each of the two radioisotope sources used, a correction for dead
time losses, and a trial-and-error solution of the two models for the den-
sity thickness, the estimated standard deviation of the density thickness,
and the distance to the sample.
RESULTS
Prior to calibration of the prototype device, an opportunity arose to
obtain data using the in-house shaker baghouse at EPA. The Ru-106 and Sr-90
sources were available at that time. The filter bag used was a silicone-
graphite finished fiberglass one and the test dust was Southwestern Public
136
-------
Service fly ash as described by Hovis et al. (7). This baghouse is a top-fed
inside collector. The dual-beam backscatter beta-particle gauge was mounted
on the outside of the bag about 4 cm from the bag surface. A typical
result is shown in Fig. 3 for one complete filtration and cleaning cycle.
Note that although most of the dust cake removal occurs in less than 20 sec-
onds, the shaking cycle lasts 120 seconds.
•a
c
o
u
a>
=3
O
O
LU
ID
Z
P
O
O
0
2000
8000
10000
HOOO 6000
TIME (seconds)
Fig. 3. Response data for Ru-106 source signal channel for one complete
cycle of filtration. Time increments per point are six seconds.
Some laboratory results have also been taken at the Center for
Engineering Applications of Radioisotopes, using the dual-beam gauge and
known density thicknesses of paper at known distances from the gauge. The
experimental and model responses are shown in Fig. 4. These results were
obtained for the Kr-85 (No. 2) source and the Sr-90 (Al) source combination.
By using model parameters approximately the same as those listed in Table 3
in the computer program previously described for obtaining sample density
thickness, measured values for sample density thickness and sample distance
were obtained for all experimental responses. The standard deviations of the
measured from the actual values were found to be 0.068 g/cm2 and 0.54 cm,
respectively.
DISCUSSION AND CONCLUSIONS
The accuracy of the laboratory results was somewhat disappointing.
This was primarily due to the high background counting from both
137
-------
o
o
0)
to
to o
+J LO
o
o
LU
a:
o
o
C
O
o
O)
CO
+J
O
o
LU
CC
CD
O
O
o
PO
§ L
I
O
"""o"
-O-
DiSTANCE = 4cm
O
4.5cm
5cm
CHANNEL A
{Kr-85 SOURCE)
DISTANCE = 4cm
CHANNEL B
(Sr-90 SOURCE)
0 0.05 0 iO 0 15 Q 20
SAMPLE DENSITY THICKNESS (g/cm2)
Fig. 4. Experimental and model responses for laboratory measurements of
various paper density thicknesses at various distances.
0 25
138
-------
sources and the much higher response of the Sr-90 source (about a factor of
seven). Subsequently the prototype has been modified by placing lead
cylindrical shields (20 mm diameter) around both sources. The background
response to the Sr-90 source was decreased by a factor of about six in this
manner. The resulting measurement standard deviation of the gauge was
reduced to half the original value.
Additional improvements, such as increasing the Kr-85 source intensity,
decreasing the amount of material in the superstructure for the rotating
shield, decreasing the variations in the rotation shield speed, and increasing
the accuracy of placing samples in the calibration procedure, would probably
decrease the measurement standard deviation by another factor of one-half.
The resulting dual-beam gauge should prove to be sufficiently accurate for
both research and control purposes. However, the present device can only be
used in environments of moderate temperature and pressure. In addition the
prototype must be removed and cleaned after about 8 hours of operation.
The rubber pulleys should probably be changed after 40 hours of operation.
These characteristics of the gauge indicate that the present device will
probably only be useful in laboratory or pilot plant research applications.
If possible, mechanical movement in a device for measuring dust cake
thickness should be avoided. For this reason, investigation of a dual-beam
device is anticipated based on responses to different beta-particle
energies — a device which does not require a rotating shield. A detector
capable of beta-particle spectroscopy will be required as well as either
single- or multi-channel analyzer pulse processing capability. Such a device
should be capable of operation in much higher temperatures and pressures
without frequent maintenance.
ENDNOTES
1. Stephan, D. G., P. T. Bohnslav, R. A. Herrick, G. W. Walsh, and A. H.
Rose, Jr. A New Technique for Fabric Filter Evaluation. Amer. Ind.
Hyg. Assoc. Journal. 19:276-284, August 1958.
2. Mazur, M., J. Mejer, and S. Waratowicz, Radioisotopic Measurement of the
Surface Mass Density of a Dust Layer Deposited on Bag Filter Cloth,
translated from Ochrona Powierza. 10(5): 130-134, 1976.
3. Ellenbecker, Michael J. and David Leith. Dust Deposit Profiles in a
High Velocity Pulse-Jet Fabric Filter. Journal of the Air Pollution
Control Association. 29(12): 1236-1241, December 1979.
4. Gardner, R. P. and R. L. Ely, Jr. Radioisotope Measurement Methods in
Engineering. New York, The Reinhold Publishing Corporation, 1967.
5. Gardner, R. P. and K. Verghese. Analysis and Design of Radioisotope
Gauges. To be published by Plenum Publishing Company.
6. Knoll, G. F. Radiation Detection and Measurement. New York, John Wiley
and Sons, 1979.
7. Hovis, L. S., J. H. Abbott, and R. P. Donovan. Electrically Charged Fly
Ash Experiments in a Laboratory Shaker Baghouse. These proceedings.
139
-------
DIAGNOSING FILTER FABRIC CAPABILITIES WITH
LIGHT SCATTERING AND NUCLEI DETECTING INSTRUMENTATION
Richard Dennis
David V. Bubenick
GCA/Technology Division
213 Burlington Road
Bedford, MA 01730
Louis S. Hovis
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
ABSTRACT
Fabric penetration measurements based upon long term averaging periods,
~hours, do not reveal the probable range of outlet concentrations and size
properties, nor short term effects that play important roles in the selection
of fabric and collector design and operating parameters. Bench scale and
pilot tests at ambient temperatures using a single particle, light scattering
device and a condensation nuclei counter to augment mass and size
determinations by filter and cascade impactor measurements, respectively,
permitted short term resolution, ~seconds, of changes in effluent
properties. Data are presented relating particulate emissions to variations
in inlet concentration, fabric structure, degree of cleaning, and filtration
velocity. The effect of fabric defects and mechanical disturbances of the
dust cake are traced over typical filtration cycles for fly ash aerosols with
woven glass, cotton, and Dacron fabrics.
BACKGROUND
Concurrent mass sampling of inlet and outlet gas streams provides good
estimates of average fabric filter efficiency for the precise aerosol and
operating conditions prevailing during the test. Unfortunately, EPA Method 5
sampling may require several hours to collect weighable amounts of effluent
dust. Additionally, such tests do not delineate transient, ~minutes, changes
in effluent properties related to variations in fabric cleaning action, the
state of the fabric before and after cleaning, the fabric dust loading,
filtration velocity, and certain unique relationships between the type of dust
being filtered and the properties of the filter fabric.
Although field applications of fast response sensing devices such as
whole-cloud or single-particle light scattering detectors may signal rapid
changes in effluent properties (particularly those caused by accidents such as
bag ruptures), they seldom provide quantitative data because of gas
stratification and mixing problems.-'- Insofar as upstream sampling is
concerned, current applications of single particle detectors present serious
140
-------
measurement difficulties, not only because of the sample representativeness
factor, but also because of inability to satisfy the extreme dilution
requirement (3 to 4 orders of magnitude), sampling line losses, and potential
instrument malfunction due to temperature, pressure, corrosion, and
condensation difficulties.1
On the other hand, the same instrumentation can provide invaluable
support in appraising filter fabric capabilities under controlled laboratory
or pilot plant conditions.2.3 The only constraints to such testing are that
(a) the concentrations of the filter effluents fall within the normal working
range (no dilution required) of the sensing device, (b) the aerosol properties
at the sampling point be representative of average conditions, and (c) the
particle losses in the sampling lines be minimal.
Extensive use of a light-scattering, single-particle counter* and a
condensation nuclei counter"*" in prior filtration studies demonstrated a
diagnostic potential far exceeding that anticipated for these highly sensitive
instruments.^3 ^he fact that such measurements could be translated to both
mass and particle number concentrations provided heretofore unavailable
information on the rapid changes taking place in filter effluent properties
over a broad range of simulated field operating conditions at bench and pilot
(single bag) scales. This paper describes experimental systems, calibration
procedures, and specific test results that demonstrate how well the
particulate removal characteristics for various fabrics can be related to
fabric structure, fabric state (new, used, partially cleaned), and basic
filtration parameters such as filtration velocity, inlet concentration, and
dust properties.
EXPERIMENTAL PROCEDURES
Fabric Test Assembly and Particle Sampling Instrumentation
Fabric test panels, 15 cm x 23 cm, were mounted without backing support
in the filter assembly shown in Figure I.3 The system geometry simulated
insofar as practical the typical approach velocity conditions encountered in
the field. Fly ash aerosols, generated by an NBS"^" dust feeder,3 were
injected into the aerosol loop from which the desired quantity was extracted
for filtration at air-to-cloth ratios varying from 0.3 to 2 m/min. The inlet
dust concentrations, ~2 to 8 g/m3, were determined by all-glass filter or
cascade impactor sampling immediately before the fabric panel as well as by
weighing the dust accumulation on the fabric panel. Outlet dust
concentrations were sampled as far downstream as possible to attain optimum
mixing of the effluent. Total gas volumes handled by the system, which ranged
from roughly 0.01 to 0.07 m3/min, were filtered through all-glass (Method 5)
*Model 40-1 Dust Counter manufactured by Bausch & Lomb (B&L), Rochester, New
York.
+Model Rich 100 Condensation Nuclei Counter (CNC) manufactured by Environment
One Corporation, Schenectady, New York.
"^National Bureau of Standards (MBS).
141
-------
filter media except for that fraction required for CNC and B&L analyses.
Limited fly ash filtration tests with single, woven glass bags (Menardi), 3 m
x 16 cm, were also performed to confirm the test panel results.-*
The manufacturers and code designations for the commonly used fabrics
discussed in this paper are: Menardi Southern, Woven Glass 601T (Tuflex);
W.W. Criswell, Woven Glass No. 640048; Albany International, Cotton Sateen No.
960 (Napped); and Dacron, No. 865B (Crowfoot).
Comments on Instrumental Methods
Sampling the filter effluent with all-glass (Method 5) filters provides
an accurate measure of average exit concentrations, but fails to reveal short
term variations regardless of the cause. Use of an in-stack type Andersen
impactor upstream furnished the usual degree of accuracy attributable to this
device under similar field loading conditions. Use of a short, 12 cm
extension, probe did not appear to affect impactor collection.
Although the accuracy of the B&L single-particle light-scattering counter
depends strongly on its application, this instrument can provide time
resolutions down to 0.1 minute insofar as reflecting changes in number
concentration for particle diameters in the 0.3 to 5 ym range. Prior GCA
studies have indicated that mass concentrations derived from B&L data are
usually lower than those determined by parallel gravimetric sampling whenever
a significant fraction of the particles exceeds 5 ym diameter.*• The latter
problem is believed to result mainly from particle settling losses in the
sampling lines and not to defects in the electro-optical system. The CNC is
claimed to detect number (nuclei) concentrations in the 0.0025 to 0.5 ym
diameter range. Although it may not provide an absolute nuclei count, the
fact that the CNC can, with calibration, provide reasonable estimates of mass
concentration, makes it invaluable for tracking rapid changes in effluent
properties. Similarly, the good linear correlation indicated between CNC and
B&L measurements, Figure 2, also suggests that generally parallel
concentration versus fabric loading (or time) curves should be expected in the
micrometer or lower size range.
Special Calibration Procedures
CNC Versus B&L Measurements with Atmospheric Dust—
Background nuclei concentrations in the laboratory area varied from 3 x
to 5 x IQlO n/m^ for the mass concentration range of 25 to 40 yg/m
upon concurrent B&L measurements and an assumed average particle density of 2
g/cnH. For those measurements wherein the mass, but not the particle size
spectrum, underwent changes, a linear correlation was shown between the nuclei
concentrations and those derived from B&L measurements, Figure 2. Thus, rapid
changes in ambient mass concentrations could be estimated from CNC counts
provided that the particle size distribution underwent no significant change.
Since long term (24 hour) high-volume filter sampling in the building area
indicated ambient mass concentrations in the 20 to 100 yg/m-* range, it was
concluded that insofar as atmospheric dust was concerned, B&L and filter
measurements could be equated when greater than 98 to 99 percent of the
aerosolized particles were less than 5 ym diameter. One gram of ambient dust
in the testing area was equivalent to a nuclei concentration of 10^5 n/m^
142
-------
CNC Counts Versus Mass Concentrations Determined by Gravimetric Procedures—
,The test aerosol used in this study consisted of a resuspended coal fly
ash with an aerodynamic mass median diameter (aMMD) of 5.8 ym and a geometric
standard deviation (ag) of 2.5. Concurrent effluent sampling with all-glass
filters and a nuclei counter (CNC) also showed a strong linear correlation
between CNC counts and mass concentration for a coal fly ash aerosol, Figure
2. Note, however, that the condensation nuclei in the coal fly ash represent
a much smaller fraction of the total particulate material than that observed
for atmospheric dust; viz., 5 x 1011 versus 1.45 x 1015 n/g dust,
respectively. Nuclei and mass measurements were compared on the basis of the
average nuclei concentration, the latter obtained by integrating over the time
interval required to collect the mass sample. Both bench scale and single bag
measurements also demonstrate a linear relationship over the outlet
concentration range, 5 x 10~3 g/m3 through 10"1 g/m3. The calibration
curve for fly ash, Figure 2, was extended beyond the actual data points to
outlet loadings exceeding 1 g/m3 for those special situations where no
significant changes in size distribution were noted between up- and downstream
fly ash aerosols.3
Concurrent Filter Drag Measurements
Although the analytical potential of the CNC and B&L devices is the main
theme of this paper, it is emphasized that the application of these
instruments, in conjunction with parallel filter drag determinations for the
dust-fabric combinations of interest, enhances the CNC and B&L diagnostic
capabilities. Several characteristic forms for filter drag versus fabric
loading curves based upon filtration at constant inlet dust concentration and
face velocity are presented in Figure 3. Sketches showing the appearance of
axial yarns (horizontal) and fill yarns (seen on end) demonstrate that fabric
surface depressions are greatest when no free fibers are available to
partially fill the pore regions (Sketch 4) and smallest when a combination of
heavy napping and all-staple yarns provides a relatively even surface (Sketch
2). An initial concave-down form reflects the early pore filling phase during
which the pressure loss increase per unit mass of dust deposit commences at
high levels followed by a gradual reduction to a nearly constant level once
the depressions are filled (Curve 1, Figure 3). The linearity will persist
for some time provided that the surface dust layer itself undergoes no
compression as the fabric loading increases. The readily discernible free or
loose fiber content of napped sateen weave cotton creates a partially
penetrable, spring-like layer that gradually compresses as the surface dust
layer accumulates. Such behavior is therefore reflected by a concave-upward
form for the drag curves due to the progressively higher resistance to gas
flow presented by the underlying fabric (Curves 2 and 3, Figure 3). When
continuous, multifilament yarns alone constitute the fabric weave (Sketch 4,
Figure 3), the initial pore filling or concave-down phase appears to continue
over the complete fabric loading process. However, subsequent decreases in
slope, once the linear level shown in Sketch 1 has been reached, are due to a
progressive collapse of the early pore bridging under the influence of
increasing aerodynamic drag. Consequently, filter penetration approaches the
100 percent level as more and more aerosol "detours" through the pores.
143
-------
DIAGNOSTIC APPLICATIONS, FACTORS AFFECTING FLY ASH COLLECTION
New (Unused) Glass - Fabric Loading and Mechanical Perturbation Effects
A good example of the CNC and B&L diagnostic potential is given by the
concentration measurements shown in Figures 4 and 5. Mass concentrations are
seen to decrease rapidly from an initial value of ~0.1 g/nH to 0.0013 g/m-5
as fabric loadings increase from zero to approximately 120 g/m , Figure 4.
Subsequent interruptions of filtration, because of removal of the filter
sandwich for weighing, caused transient increases in emissions that became
progressively smaller until no discernible effect was noted at the end of the
D filtration interval. These perturbations arose from the checking and
puncturing of the essentially homogeneous dust surface due to filter
handling; Over filtration intervals B through E, the dust cake revealed a
capability to mend itself such that no permanent damage resulted. The time
required to develop the dust deposits for inlet concentrations of 1 to 3
g/m^ at a filtration velocity of 0.61 m/min varied from about 2.5 hours for
the A interval to 5.5 hours for the E interval.
Similar concentration spikes (and related drag perturbations) should also
be expected during the operation of large, multi-compartment baghouses due to
shocks and vibrations caused by rapid damper closings, fan imbalance,
mechanical shaking of adjacent off-line compartments, or the emergency
hammering of a plugged hopper to restore dust removal. The intensity of the
disturbance, the mass of the baghouse (light- or heavy-weight construction),
and presumably the unique adhesion properties for the specific dust-fabric
combination will determine the impact of such mechanical effects on dust
penetration. In the event of minor disruptions, dust emissions will probably
return to the "unperturbed" levels, and no net changes will be indicated in
the form and slope of the drag curves. Conversely, if a heavily-laden fabric
is severely shocked, its dust retention capacity may be seriously impaired.
Figure 4 indicates that the final perturbation, after the fabric loading had
increased to about 950 g/m^ (Region F), caused emissions to rise to levels
characterizing the clean fabric. More importantly, the fabric dust layer no
longer was able to repair itself as indicated for Regions A through D. Actual
effluent concentration fell slightly to a nearly constant level approximately
60 times greater than that observed during the <950 g/m^ loading range.
Although failure to mend itself may have been partially attributable to the
intensity of the mechanical shock, the behavior indicated on Figure 4 also
signifies that dust cake defects are more amenable to repair at low to
moderate surface dust loadings where proportionately lower gas velocities
occur in the pinhole zones. Once the gas velocity through a pore becomes
large enough to prevent any further pore bridging, emission levels tend to
remain constant for a fixed inlet dust concentration. Upon cleaning, however,
former defects will be eradicated in most cases by the freshly deposited dust
layer.
Analyses of concurrent B&L and CNC measurements over the E filtration
interval, Figure 5, show that all B&L size categories over the range >0.3 ym
to >2.0 pm generated parallel curves whose paths were closely traced by the
144
-------
concurrent nuclei measurements. These findings appear to demonstrate that,at
the initiation of the E filtration phase, a maximum in pore closure or
bridging had been achieved for the prevailing inlet concentration and velocity
conditions. The parallel up and down paths displayed by the B&L traces
reflect variations in the mass inlet concentration where, for a constant gas
flow volume through the unblocked pores, the emissions are directly
proportional to the inlet concentrations. Because roughly 50 percent of the
fly ash effluent is composed of particles >5 ym diameter, one cannot translate
directly the B&L data to equivalent mass loadings. However, for those sizes
that appear to reach the B&L sensing zone without significant loss of <5 ym
diameter particles, reasonable estimates of mass concentration appear possible
as discussed earlier in this paper.
Multifilament Dacron - Effect of Minimal Free Fibers on Emissions
The results of CNC and B&L measurements on several new and unused Dacron
test panels showed very high fly ash penetrations for both nuclei and B&L
particles over the complete fabric loading range, Figure 6. In each case, the
initial effluent concentrations not only exceeded by an order of magnitude
their glass and cotton fabric counterparts, but also remained nearly constant
over the ~70 minute filtration interval. Although both CNC and B&L counts
were internally consistent, they were considered suspect until the final
gravimetric efficiencies, ~76 to 81 percent as determined by concurrent
sampling with all-glass filters, were analyzed. The unexpectedly poor
performance of these test panels was attributed to the multifilament character
of the weave whose lack of free fibers allowed for comparatively easy pore
penetration (see Sketch 4, Figure 3). The predicted efficiencies for this
fabric, 83 to 85 percent based upon pinhole concentrations ranging from ~5000
to 9000 openings/m^, confirmed the poor collection determined by gravimetric
sampling.
New (Unused) Sateen Weave Cotton - Fabric Loading and Napping Effect
It has been emphasized that particle size distributions determined
immediately before and after the filter face are nearly identical when coal
fly ash is filtered with woven glass fabrics.3»4 This behavior has been
attributed to the fact that there is negligible removal of particles less than
15 to 20 ym diameter from the flue gas fraction that passes through pinhole
openings during the early minutes of filtration. Since 95 to 99 percent of
the total emissions are contributed by the early penetration phase, the size
properties for the effluent dust are closely described by those of the inlet
dust.
The penetration characteristics change appreciably, however, when a more
efficient fabric (such as a napped, sateen weave cotton) is used, Figure 7.
An increased free fiber population within the pore structure and upon the
fabric surface results in far fewer and usually smaller pinhole-type openings
in the filtering surface. At the initiation of filtration, dust deposits
mainly upon the napped surface with minimal penentration in the direction of
the yarn substrate. Thus, as shown previously by Curve 2, Figure 3, the
pore-filling phase characterized by woven glass fabrics is practically
non-existent.
145
-------
Furthermore, the increasing pressure gradient across the accumulating dust
layer and its supporting nap also leads to a gradual compression of the napped
region. The net result is an increase in the rate of change in drag with
fabric loading as shown by the concave-up form of Curve 2, Figure 3.
According to B&L measurements, particle concentrations in the >1 ym size
class underwent four orders of magnitude decay before the fabric loading had
reached 150 g/m2, while similar, but less rapid reductions in the >0.5 ym
particle category were still occurring at the 400 g/m2 load level. In
contrast, however, nuclei concentrations (constituting the particle class
<0.5 ym) and the B&L particles in the >0.3 ym category underwent a general
levelling off at the 200 g/m2 fabric loading level while displaying the
previously noted tendency to parallel each other. It was concluded,
therefore, that fiber proximity in the cotton filter permitted much more
effective pore bridging than that attainable with woven glass media. The fact
that a low-level penetration persisted for the very small, <0.5 ym, particles
suggested either incomplete pore bridging or cake defects in the form of
minute cracks or perforations. Concurrent mass emission rates for the cotton
fabric, however, were still approximately 15 times lower than those observed
for the woven glass media because of the absence of >1.0 ym diameter
particles.3
Filtration of atmospheric dust through a uniformly distributed fly ash
layer (>200 g/m2) would be expected to show no reduction in particle size
because the parent dust contains few particles, usually <1 to 2 percent,
greater than 1 ym. Figure 7 suggests that most particles in the <1.0 ym
category should penetrate any residual pores with negligible particle
capture. This penetration concept also applies to fly ash filtration with
glass fabrics where, except for the fact that the pores and particles are
larger, the penetration through the pores for the fly ash particles is nearly
100 percent.
Well-Used Glass - Partial Cleaning, Fabric Loading, and Velocity Effects
Figure 8 shows that particulate emissions are strongly dependent upon the
uniformity of the fabric dust layer. Effluent concentrations decreased
rapidly, Test ,71,3 wnen fly ash was filtered with a completely cleaned
Menardi test panel excised from a well-used bag. At all times during this
test, up to a fabric loading of 530 g/m2, the dust was uniformly distributed
on the fabric surface. After cleaning, however, wherein the dust
characteristically separated at the interface region between the surface layer
and the fabric,^ approximately 35 percent of the fabric surface was reduced
to its formerly completely cleaned state. At this point, the only dust
remaining in the cleaned region was that irreversibly retained within the
fiber interstices. When filtration was resumed, the emissions rose to a level
about 200 times greater than that previously noted for a uniformly distributed
layer of 350 g/m2 because a disproportionate fraction of the gas flow was
diverted through the less efficient but much lower resistance filtration
path. The higher gas velocities prevailing at the initiation of Test 72^
explained the excessive penetration while the preferential deposition of fly
ash on the cleaned zone accounted for the gradual improvement in
146
-------
performance. A tabulation of equilibrium emission levels associated with
selected filtration velocities (based on CNC tests with uniformly loaded
filters, Figure 8) suggested that a gradual velocity reduction over the
cleaned region should lead to decreased penetration. However, since
filtration is always initiated at high velocities on a cleaned surface, it
also appeared that the early pore bridging process might not be as effective.
The final concentration excursions noted for Test 72 were attributed to two
factors: possible variations in inlet concentration and/or an enlargement in
pore size caused by some unidentified mechanical disturbance.
CONCLUSIONS
Experimental data presented in this paper demonstrate the capabilities of
nuclei counters and single-particle light-scattering counting and sizing
instruments when used under pilot testing conditions where problems pertaining
to high temperatures, corrosion, and pressure variations can be minimized.
Procedures described in this report enable the researcher and filter
manufacturer to examine fabric potential on the basis of small panel tests or
single bag pilot tests.
The capability to identify unexpected performance characteristics and to
determine, in conjunction with drag measurements and visual (microscope)
observations of the fabric, how fabric design or selection may be improved,
can eliminate unnecessary trial and error procedures in the field.
Considerable data evolving from test procedures discussed in this paper were
used to develop the predictive model for fly ash filtration described in the
filtration literature.-^ 4-
ACKNOWLEDGMENTS
This project has been funded at least in part with Federal funds from the
U.S. Environmental Protection Agency under Contract No. 68-02-1438, Task No.
5. This paper does not necessarily reflect the views or policies of the U.S.
Environmental Protection Agency, nor does mention of trade names, commercial
products, or organizations imply endorsement by the U.S. Government.
REFERENCES
1. Dennis, R., and D. V. Bubenick. Apparent Fractional Efficiencies of
Available Control Techniques. In: APCA Spring Specialty Conference
Proceedings, The Technical Basis for a Size Specific Particulate
Standard, Parts I and II. pp. 129-155. March and April 1980.
2. Dennis, R., and J. E. Wilder. Fabric Filter Cleaning Studies.
GCA/Technology Division. EPA-650/2-75-009 (NTIS PB 240372).
January 1975.
3. Dennis, R., et al. Filtration Model for Coal Fly Ash with Glass
Fabrics. GCA/Technology Division. EPA-600/7-77-084 (NTIS PB 276489).
August 1977.
147
-------
l
8 E
0 10'
ATMOSPHERIC OUST
Ig/m5 • 1.45 X I0"n/m3
FLY ASH
lo/ms -5 XIOMn/ms
SYMBOL COAL FLY ASH
A OUTLET CONC.-8ENCH TESTS
X OUTLET CONC-PILOT TESTS
INLET CONC - ESTIMATED
ATMOSPHERIC DUST
LABORATORY, AIR
i I i
10-4 10-* IO-» IO-' IO°
FLY ASH MASS CONCENTRATION, g/m'
ATMOSPHERIC DUST MASS CONCENTRATION,pj/m* X10"
Figure 1. Exploded view of filter
test assembly.
Figure 2. Calibration curve - Aver-
age nuclei and related
mass concentrations for
GCA fly ash and atmospheric
dust.
0.2 0,4 0.6
FABRIC DUST LOADING, g/m3 xlO'3
WEIGHT*
EFFICIENCY
99 26
99.ae
99.52
99 66
99.63
95 67
400 800
FABRIC LOADING (WI. g
Figure 3. Characteristic drag-
loading curves, dust
filtration with common
fabric weave and yarn
combinations.
Figure 4.
148
Effluent nuclei concentration
versus fabric loading with
coal fly ash filtration at
0.61 m/min on unused Menardi
glass fabric. Test 65.3
-------
J
I02
_
5
2
?
X '°'
£
S
: 2
; 10°'
5
- z
10"
5
2
ID'2
COAL FLV ASH B B L
aWM0.58«m SIZE
CT, • 2 42 M"
• >0 3
A >0 5
11 > 1 0
O >2 0
-
.
• • • *
1
1 A
- A AA A *? » ~
A , K ^> A »
A * „
*
.
-
O
O
° G
O
o
I02
«
o -
x
"E 2
10'
5
-
fO°
s
2
!
'
-V93 V93 V93-
95 X9« °9' O93
X95 "94 x° X"
-
g« ^94
-S^94
COAL FLY ASH size
0 MMO • 5 8^m
CTg • 2 «2 ^m
X B8L > 1 0
0 B8L >2.0
~~ V CNC 00025-0 5 ~~
-
NOTE TEST NUMBERS 93, 94,95 I REFERENCE 3)
500 600 700
FABRIC LOADING (W). g/m2
Figure 5.
Effluent particle concen- Figure 6.
tration versus fabric load-
ing with coal fly ash fil-
tration at 0.61 m/min on
new Menardi glass fabric.
Test 65.3
100 200
FABRIC LOADING [Wt, q/m2
Effluent concentrations ver-
sus fabric loading for coal
fly ash filtration with
Dacron (crowfoot weave) test
panels, face velocity =
0.61 m/min, inlet loading =
-7.5 g/m3.
- ,
'°
«
OUTLET
CONC
9/m X I03
5
55
230
360
FACE
VELOCITY A
m/min i
0.61
' 23 i
1 98 i
2.67 i
L
-
4
21 I
ZOO 400 SOO
FABRIC LOADING (W) g/m
200 400
FABRIC LOADING (W),g/n
~600
Figure 7.
Effluent concentrations
versus fabric loading for
coal fly ash filtration
with napped cotton sateen
test panel, face velocity
=0.61 m/min, inlet
loading = 7.1 g/m3.
Figure 8. Effluent concentration ver-
sus fabric loading for
Menardi fabric and coal fly
ash with uniform and non-
uniform surface dust
loading.3
149
-------
ACID UEWPOINT CORROSION IN PARTICULATE CONTROL EQUIPMENT
By: I.E. Mappes and R.D. Terns, Ph.D*
PEDCo Environmental, Inc.
505 S. Duke St., Suite 503
Durham, North Carolina 27701
K.E. Foster
Technical Support Branch
Division of Stationary Source Enforcement (MD-7)
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
ABSTRACT
A large portion of the corrosion problems observed in particulate con-
trol equipment results from flue gas temperatures falling below the sulfuric
acid dewpoint. A study for the U.S. Environmental Protection Agency (EPA)
included an investigation of 11 fabric filters and 1 electrostatic precipita-
tor (ESP) that had been affected by acid dewpoint corrosion. The study also
included comments of design engineers at seven particulate control equipment
manufacturers on the subject of acid dewpoint corrosion. Results of this
study indicate that acid dewpoint corrosion is a serious problem that can
destroy expensive particulate control equipment after only a few years of
service or reduce particulate control efficiency and equipment availability.
This paper discusses how many corrosion problems can be prevented or amelior-
ated by the use of thermal insulation, hopper heaters, sealing of entry
hatches, control of process gas temperatures, and preheating during process
startup.
INTRODUCTION
In 1978 the President's Council on Environmental Quality conducted a
survey to identify the causes of excess emissions from controlled stationary
sources that had initially been in compliance with emissions regulations.
Inspection of 20 such sources showed that corrosion-related malfunctions in
the control equipment were the primary cause of excess emissions at three
sources and contributing causes at six others(l). Subsequent source inspec-
tions conducted for the Division of Stationary Source Enforcement (DSSE)
confirmed that acid dewpoint corrosion is a common cause of malfunction in
particulate control equipment(2). These inspections also revealed that many
sources do not devote sufficient attention to acid dewpoint corrosion control
when they select particulate control systems. In many cases proper corrosion
control measures have been determined by trial and error, at great expense.
Based on these results, DSSE saw a need to assist particulate control
equipment users and State Enforcement Agency personnel in coping with corro-
sion problems. They commissioned a study to characterize the effects that
'"'Presently with Mobil Research and Development Corp. , Field Services Labora-
tory. P. 0. Box 900. Dallas. Tex?s 75221
150
-------
corrosion can have on the performance and operating life of particulate con-
trol equipment and to assemble guidelines to help reduce the frequency and
severity of excess emissions incidents due to corrosion. The study included
four phases—a review of the technical literature, the identification of
emissions sources where control equipment corrosion was a problem, the in-
spection of a sampling of 18 of the identified sources, and a series of
interviews with several participate control equipment manufacturers. The IS
sources inspected included 11 fabric filters and 1 electrostatic precipitator
(ESP) that had been affected by acid dewpoint corrosion.
The acid dewpoint generally refers to the temperature at which sulfur
trioxide vapor (S03) will combine with water vapor to form sulfuric acid
(H2S04). S03 is found most frequently in flue gases produced by processes
using sulfur-bearing fuels such as coal, coke, and oil. S03 can also be
formed in primary metals-refining processes (e.g., copper smelting) because
the ores typically contain large quantities of sulfur.
Combustion of sulfur-bearing fuel results first in the formation of sul-
fur dioxide (S02), some of this which is further oxidized to SO3 The amount
of S02 converted to S03 depends on many variables, including gas temperatures
in the combustion zone, the configuration of the combustion chamber, and the
availability of oxygen in the flame(3).
Sulfur trioxide and water have a tremendous affinity for each other;
when temperatures are lowered to the dewpoint the two combine rapidly to form
sulfuric acid molecules. The sulfuric acid molecules, in turn, have a high
affinity for water; as they condense they draw additional water molecules
from the gas stream forming a concentrated acid solution. Therefore, when
flue gases containing relatively small concentrations of S03 reach the dew-
point temperature, droplets of concentrated sulfuric acid can condense on the
cooler surfaces. For example, an 82.5 percent sulfuric acid solution (by
weight) will condense at 148°C from flue gases containing as little as
40 ppm (by volume) S03 and 10 percent (by volume) water vapor(4).
Verhoff and Banchero developed an empirical relationship from which the
dewpoint temperature can be calculated when the percentages of water vapor
and sulfur trioxide are known(5).
DP 1.7842+0.0269logPw n-0.10291ogP_n +0.03291ogPu
li^U Q -H._v vju,.
where
T p = dewpoint temperature in degrees kelvin
P = vapor pressure of water in atmospheres
2
P = vapor pressure of sulfur trioxide in atmospheres.
This equation has agreed with most experimental results to within about 7 de-
grees kelvin. In cases where measurement or estimation of S03 and water
vapor concentrations is not practical, gross estimates of the dewpoint can be
151
-------
based on sulfur content of the fuels(6). In either case it is advisable to
add a factor of safety to the estimated dewpoint temperature to allow for
variations in flue gas composition and to compensate for nonuniform flue gas;
temperature throughout the system.
FACTORS THAT CONTRIBUTE TO ACID DEWPOINT CORROSION
The following factors influence the tendency for acid dewpoint corrosion
to occur: typ.es of raw materials and fuels used in a process, temperature
and moisture content of the flue gases, process operating cycles, and opera-
tion and maintenance practices.
Raw Materials and Fuels
Process raw materials are a primary source of sulfur and water in par-
ticulate-laden gases entering control devices. Ores, aggregates, recycled
scrap metals, slags, and other raw materials may contain a variety of unwant-
ed corrosive materials in addition to their primary economically valued
components. For example, iron and copper ores contain high percentages of
sulfur, some of which can be oxidized to S03 during refining processes.
Aggregates used in the production of asphalt concrete typically contribute
significant amounts of water vapor to the flue gases exiting the rotary
dryer. Feedstocks for secondary lead blast furnaces consist primarily of
lead-acid batteries, which contain sulfuric acid. Metallurgical slags, which
are sometimes used as raw materials for rock wool insulation cupolas, can
contain large quantities of sulfur that can oxidize to S03. Fossil fuels.
especially coal, metallurgical coke, and residual fuel oils, also contain
significant quantities of sulfur that: can oxidize to S03 during combustion.
Flue Gas Temperatures and Moisture Content
High temperatures can damage fabric filter bags, protective coatings and
linings, and fiberglass-reinforced plastic components. Gas temperatures can
be reduced by dilution air dampers, quench chambers, or heat exchangers.
Reducing gas temperatures, however, can set the stage for acid condensation.
Therefore, when hot flue gases are to be cooled, it is important to determine
their moisture and S03 contents so that the sulfuric acid dewpoint can be
estimated. An inventory should be taken of all possible sources of flue gas
water vapor including the water vapor added with dilution air or evaporative
cooling and that which forms as a byproduct of fossil fuel combustion. If
analysis indicates that flue gas temperatures will be close to the acid
dewpoint, it is advisable to reduce the amounts of water vapor and/or S03 in
the flue gases or to select an alternative particulate control strategy.
Process Operating Cycles
The operating cycles of a controlled process can sometimes influence the
corrosion activity in a particulate control device. Operating cycles can af-
fect dewpoint corrosion because of the close relationship between process;
variations and variations in flue gas temperatures. In general, the more
frequently a hot process starts up and shuts down or varies between full
production and partial production, the sere frequently the flue gases in the
control device will pass through the ocid dewpoint.
152
-------
Several operating cycles are possible in industrial processes, and each
produces different flue gas temperature cycles. At one extreme is the char-
acteristic cycle of a cast iron cupola, which typically operates for melts
lasting less than 24 hours between startup and shutdown. While a cupola melt
is in progress, there are frequent charges of fuel, scrap iron, and fluxing;
periodic blasts of combustion air through the tuyeres; and frequent iron and
slag taps. Each of these operations affects the temperature of the cupola
flue gases. Wide fluctuations in cupola flue gas temperatures can be temp-
ered by gas cooling devices such as dilution air dampers, quench chambers,
and heat exchangers. It is often difficult, however, to simultaneously
protect particulate control equipment from high-temperature stresses and acid
dewpoint corrosion. At most fabric filter installations, high temperature
protection has priority over dewpoint control because a single high-
temperature excursion can do immediate damage, whereas dewpoint corrosion is
a cumulative problem.
The other extreme in operating cycles is represented by the base-loaded
utility boiler, which runs for months at a time near the maximum firing rate.
Industrial boilers fall somewhere between the two extremes in operation.
These usually operate for periods lasting 5 to 6 days, and their firing rates
vary slowly during this period.
The significance of the operating cycle of a process must be considered
when developing a corrosion control strategy for particulate control equip-
ment. Variability in flue gas temperatures and chemical constituents may be
a determining factor in the selection of a cooling device or a particulate
control device.
Operation and Maintenance Practices
Operation and maintenance practices often affect corrosion activity in
fabric filters and ESP's. Some process parameters affecting dewpoint corro-
sion can be controlled or partially controlled by plant operators. Mainte-
nance practices such as the frequency of equipment inspections and the speed
of repair of deteriorating parts are also important.
In attempting to control the sulfuric acid dewpoint, it is often advis-
able to preheat a cold fabric filter or ESP before introducing moist, sulfur-
bearing flue g.ises(7). For processes having wide fluctuations in tempera-
tures, it is advisable to vary the amount of dilution air or quench spray to
match the changing temperatures, rather than presetting the amount of cooling
for the highest expected temperature. Automatic cooling controls are prefer-
able to manual controls. In processes such as cement kilns or hot-mix
asphalt dryers, operators must avoid the temptation to lower process exhaust
temperatures to save fuel without considering the effects the lower tempera-
tures may have on dewpoint corrosion.
Changes in raw materials sometimes alter corrosion activity in a partic-
ulate control device. For example, a change in fuel from distillate oil to
residual oil can increase the sulfur levels in the gases. A change in the
source of coal can also change the amounts of sulfur introduced into the
153
-------
system. Finally, the moisture content of raw materials and fuels that come
from open storage can vary as rainfall varies.
Maintenance practices that can affect corrosion activity in particulate
control equipment include the frequency of inspection of the various compo-
nents and the speed at which malfunctioning components are repaired. Items
that should be inspected frequently include temperature control instrumenta-
tion, hatches, hoppers, fans and fan vibration sleeves, insulation, fabric
filter enclosures, and ESP enclosures.
CONTROLLING ACID DEWPOINT CORROSION
Most of the manufacturers interviewed during this study agreed that acid
dewpoint corrosion can be adequately controlled in nearly all fabric filter
and ESP applications. The principal means available for this control are
reduction of S03 and moisture levels in the flue gases, control of exhaust
gas temperatures, use of proper thermal insulation, and prevention of cool
air inleakage into the control device. In some cases the use of protective
coatings and special alloys can be helpful; however, the cost of the latter
is usually prohibitive for the major structural components of fabric filters
and ESP's.
Acid dewpoint corrosion is most likely to occur in locations where flue
gas temperatures are the coolest or where steel surfaces are the coolest.
Temperatures in dust hoppers are often cooler because gas detention is longer
than at other locations (which allows more time for cooling) and because the
hoppers have a large surface-to-volume ratio (which increases the rate of
radiant heat loss to ambient air). Thus, hoppers without properly designed
heaters or insulation are frequently affected by acid dewpoint corrosion.
Outer walls (especially corners) of a filter or an ESP are often cooler and
therefore subject to acid dewpoint corrosion (Figure 1). Entry hatches and
hatch frames are also more likely to corrode because of cool air inleakage
and because the hatches are often not as well insulated as the rest of the
structure (Figure 3).
In the 1>. fabric filters studied; dust hoppers, sidewalls, corners,
entry hatches, filter compartment isolation dampers, and "top end" components
such as bag support hardware and bag cleaning mechanisms were affected by
acid dewpoint corrosion. Five fabric filter manufacturers reported that dust
hoppers, top-end components, and bag hardware are the components most fre-
quently affected.
ESP enclosures are subject to cooling effects similar to those in fabric
filter enclosures; therefore, many corrosion problems reported in fabric
filters also occur in ESP's. As in fabric filters, acid dewpoint corrosion
often attacks dust hoppers, corners of the enclosures, access hatches, hatch
frames, and the undersides of roof plates. It can also occur in penthouses
that contain support insulators. At installations where flue gas tempera-
tures are very low, acid dewpoint corrosion can attack structures within the
ESP, such as discharge plates and discharge wires.
154
-------
Reduction of Sulfur Trioxide and Moisture Levels
If there is no moisture inleakage into a control device, the levels of
S03 and moisture in the flue gases depend on the levels of sulfur and water
introduced with the process raw materials, fuels, combustion air, and gas
cooling. The best ways to keep the sulfuric acid dewpoint temperature as low
as possible is to select fuels and raw materials with as low a sulfur content
as possible and to select dry-type flue gas cooling systems whenever possi-
ble.
It should be noted, however, that the use of low-sulfur fuels is no
panacea for acid dewpoint corrosion. Low-sulfur coals and oils generate
sufficient sulfur oxides when burned to cause acid dewpoint corrosion; the
use of these fuels merely lowers the dewpoint temperature, which makes acid
condensation easier to prevent. Economics usually governs the ultimate deci-
sion on the type of fuel to be used. Any cost analysis of available fuel op-
tions should consider the possible ccsts of corrosion damage to the particu-
late control equipment and lost production following corrosion failures as
well as the relative costs of the fuels.
Control of Flue Gas Temperatures
Flue gas temperatures are a function of the process characteristics and
of the flue gas cooling systems. Processes such as cast iron production are
not amenable to flue gas temperature control at the process. Sources such as
these require the use of flue gas cooling devices. Primitive systems such as
a manually operated dilution air damper are usually not sufficient to insure
that flue gases remain above the dewpoint. Control devices with temperature
sensors that automatically control dilution air are preferable. Evaporative
cooling is not recommended for cast iron cupolas using fabric filters, but
recuperative heat exchangers have b ^n used successfully.
In processes where operat .g conditions are relatively stable (e.g.,
cement kilns) it is sometimes possible to control exhaust temperatures by
adjusting the firing rates and the amounts of excess combustion air. In such
processes, it is helpful to have temperature sensors in the control device
that can automatically control the process firing rates or sound alarms in
the plant's control room whenever temperatures in the control device fall be-
low a safe level. The increase in fuel consumption required to maintain high
kiln exhaust temperatures must be weighed against the savings attributable to
reduced corrosion in the particulate control system.
Another important means of controlling acid dewpoint corrosion is to
preheat the control device prior to the introduction of corrosive flue gas-
es(7). For example, a rotary dryer can be preheated by firing the burner at
a reduced rate for a limited time before introducing the wet aggregate. A
cupola can be preheated by firing natural gas afterburners for a limited time
before lighting the sulfur-bearing coke charge. The advantage of preheating
is that the hot corrosive flue gases will not contact cold steel surfaces in
the control device during startup as would occur without preheating. As in
the case of other flue gas temperature control measures, the advantages of
preheating must be weighed against the additional fuel expenses.
155
-------
Thermal Insulation
Thermal insulation is another means of maintaining flue gas temperatures
above the acid dewpoint. The ultimate purpose of insulation is to reduce the
amount of flue gas cooling that takes place between the time the gases enter
the ducting to the control device and the time they exit the stack. At some
installations insulation alone will s;olve acid dewpoint corrosion problems;
in others it is only a supplementary measure-
Insulation will be most effective if a few simple guidelines are follow-
ed(8). Any insulation must be thick enough to impede sufficiently the flow
of heat through the external surfaces of the control device. The insulation
also must cover all exposed surfaces including the dust hoppers, access
doors, and the ductwork. It must be installed in such a manner that it will
remain intact over the life of the equipment. Most types of insulation
should have a protective sheath (usually sheet metal) on the outside to repel
rainwater and to prevent mechanical damage. An insulation blanket must be
sealed at the top and bottom to prevent a "chimney effect" circulation of air
underneath the blanket. Windbreaks can also be helpful in insulating a
fabric filter or an ESP. An example of a properly insulated ESP is shown in
Figure 2.
Among the common errors made in insulating control equipment are the
failure to insulate all exposed ductwork (including reverse air ductwork in
fabric filters), the failure to insulate hatches, and the failure to attach
the insulation securely. Another common error (Figure 4) is to allow struc-
tural members to protrude through the blanket of insulation; these members
radiate heat away from interior surfaces of the control device and foster
condensation at these locations.
Air Inleakage
Unexpected air inleakage can spoil a well-designed acid dewpoint control
program in a fabric filter or an ESP. Inleakage not only allows moisture
from the ambient air to enter the control system, but also causes severe
localized cooling at the point of entry. Both phenomena can contribute to
acid dewpoint corrosion. Inleakage is a self-feeding process—the metal
adjacent to the: initial penetration tends to corrode at an accelerated rate,
which promotes enlargement of the penetration. Inleakage is common in poorly
fitting hatches, hatches with worn seal gaskets, rotary air locks with worn
seals, cracked vibration sleeves, and incomplete welds. It can also occur in
ductwork or at the process. Inleakage should be repaired promptly to avoid
further damage.
Special Alloys and Protective Coatings
The relatively large size of most fabric filters and ESP's makes it un-
economical to use higher priced stainless steels or nickel alloys as the
principal materials of construction. Most structural components of these
devices are constructed of carbon steel or, in some cases, Corten or cement
block. Special alloys are used, however, for some of the smaller components
within these devices, such as filter bag hardware and ESP discharge wires
156
-------
Nearly all exterior surfaces of fabric filter and ESP enclosures receive
a protective coating; however, the interior surfaces of these devices are not
always coated. One fabric filter manufacturer reported that only 10 percent.
of its filters receive internal coatings. The reason that interior coatings
are not always used is that few seem to be very effective on hot sources.
Despite claims of some coatings manufacturers, long-term performance of most
available coatings systems is not good in hot corrosive flue gases. All
control device manufacturers interviewed reported that most coatings fail
when exposed for prolonged periods to acid-bearing flue gases with tempera-
tures above 300 to 400°F. In cases where a coating is appropriate, it is
important that the coating be applied in the proper manner to a properly
prepared surface. In most cases this requires that the coating be applied in
the shop rather than in the field.
ENMOTES
1. Booz, Allen, and Hamilton, Inc. Final Evaluation of North Carolina's
Program to Regulate Air Pollution From Stationary Sources. Prepared
under President's Council on Environmental Quality, Contract No.
EQ8AC015, Bethesda, Maryland, July 1, 1979-
2. PEDCo Environmental, Inc. Unpublished data obtained during SIP audit
inspections under contract to the U.S. EPA, Division of Stationary
Source Enforcement, Contract No. 68-01-4147, Tasks 110, 131, and 137,
1979-80.
3. Balasic, P. J. Electrostatic Precipitator Corrosion Problems on Recov-
ery Boiler Applications. Paper No. 185, presented at the National
Association of Corrosion Engineers, Corrosion/79, Atlanta, Georgia,
March 12-16, 1979-
4. Pierce, R. P. Estimating Acid Dewpoints in Stack Gases. Chemical Engi-
neering 84(4): 125-28, 1977.
5. Verhoff, F. H., and Banchero, J. T. Chemical Engineering Progress, Vol.
70, p. 71, 1974.
6. Katz, J. The Art of Electrostatic Precipitation, Chapter 7. Precipita-
tor Technology, Inc., Pittsburgh, Pennsylvania, 1979.
7. Beggs, T. W. , and U. M. Patankar. Accelerated Baghouse Corrosion in a
Waste Oil Burning Asphalt Concrete Plant. Presented at the 72nd annual
meeting of the Air Pollution Control Association, Cincinnati, Ohio, June
24-29, 1979.
8. Landrum, R. J. Equipment. Chemical Engineering 77(22): 75-82, October
12, 1970.
157
-------
Figure 1. Closeup of a cupola fabric filter cross member showing
that dewpoint corrosion is more severe along surfaces nearest the
cooler outer wall (the outer wall is constructed
of corrugated transite).
Figure 2. Electrostatic precipitator with a complete blanket of
insulation covering the chambers, hoppers and ductwork.
158
-------
Figure 3. Corrosion of fabric filter
access door and door frame. Opening
permits inleakage of cool ambient air,
accelerating condensation and further
corrosion.
Figure 4. Insulated fabric filter
with structural steel protruding
through the insulation blanket. The
exposed steel radiates heat away
from the filter, promoting cold spots
along the interior of the filter.
-------
SECOND GENERATION OF EMISSIONS CONTROL SYSTEM FOR COKE OVENS
By
James D. Patton
Consultant on Coke Side Shed Systems
MikroPul Corporation
Summit, New Jersey U.S.A.
ABSTRACT
The body of information in this paper is directed to coke oven operators
and their management, to the Environmental Control and Protection Agencies,
and labor organizations interested in further protection of their members.
There are a number of areas of concern in pollution control for coke
ovens. We are directing our efforts to properly control all the emissions on
the coke side of the ovens.
In the past four to five years the interest in Coke Side Sheds has been
set aside in favor of a number of concepts relating to "land base systems",
traveling hooded hot cars, etc. Each of these systems looked favorable, but
some have not proven out for a variety of reasons, others have worked well.
Some were not as dependable as expected and for others the maintenance costs
were high. Lately, many of these companies are looking for a system that
meets the required standards, is more dependable and requires a minimum of
maintenance.
The author of this paper is co-inventor of the "Roe-Patton" Coke Side
Shed, U.S. Patent 3,844,901, offered through Great Lakes Carbon Corporation
and promoted by the MikroPul Corporation. This has been the most successful
shed system to date.
The author is also the inventor of the "Patton" Emission Control System,
U.S. Patent 3,972,782, acquired by MikroPul as the latest in shed technology.
Its unique design overcomes the two major objections to the original G.L.C.
shed: fallout of the large particulate during the push and fume escaping the
hood during periods of adverse wind conditions. With the improved air flow,
it proposes to be the cleanest and most efficient "Coke Side Shed" to date.
MikroPul's newest adaptation to the "Patton" system has incorporated the
baghouse filtering system within the coke side shed structure, requiring no
ductwork, reduced energy requirements, and many other cost reducing factors.
The author believes this to be a major breakthrough in new technology and
is worthy of consideration where adaptable.
160
-------
INTRODUCTION
A coke side shed is a special designed structure that is erected on the
coke delivery side of a battery of coke ovens. Its purpose is to capture all
the emissions developed on that side of the ovens. There can be emissions on
the coke side from the following sources:
1. Removal of the coke oven doors.
2. The actual pushing of the incandescent coke from the oven slot
through the coke guide and falling into a waiting quench car.
3. At times there is slight spillage on the bench area that is hand-
shoveled back into the open oven causing emissions.
4. In some cases there could be emissions at the door seals after
replacing the oven doors or around the door frames if the ovens are
older and in need of repairs.
In fact, we believe the coke side shed control system is the only type
system that can completely control this many emission points with one approach.
In order to talk about the second generation of coke side sheds, I must mention
the basic principals of the original shed system. Being the co-inventor of
this system, I would like to compare this original system with my later
developments.
Since MikroPul Corporation has been marketing this first shed system
through Great Lakes Carbon Corporation, it shall be referred to as the MikroPul-
G.L.C. shed system. Great Lakes Carbon owns the patent and MikroPul has the
exclusive sales rights.
For illustrative purposes, we are using the following symbols to aid in
the following slides.
Figure 1
(A) Thermal wall panel.
(B) Thermal heat deflector.
(C) Throat or structure orifice.
(D) Capture and expansion zone.
(E) Evacuation duct.
(F) Filter or cleaning device.
(G) Fallout capture area.
(H) End walls.
161
-------
A THERMAL WALL PANEL
B THERMAL HEAT DEFLECTOR
C THROAT OR STRUCTURE ORIFICE
D CAPTURE AND EXPANSION ZONE
E EVACUATION DUCT
F FILTER OR CLEANING DEVICE
G FALLOUT CAPTURE AREA
H END WALLS
I OVEN BATTERY
J COKE GUIDE
K QUENCH CAR
L LIGHTING INSIDE OF THE SHED
M MODULAR DOME ROOF OR
CONVENTIONAL ROOF
N BENCH AREA
_. . 162
Fig. 1
-------
M
CONVENTIONAL ROOF
304 STAINLESS
STEEL
F DUCT TO REMOTE
BAGHOUSE
MIKROPUL-6.L.C. COKE SIDE SHED
163
-------
(I) Oven battery.
(J) Coke guide.
(K) Quench car.
(L) Lighting inside of the shed.
(M) Modular dome roof or conventional roof.
(N) Bench area.
Figure 2
This is a cross section of the MikroPul-G.L.C. shed system. This is the
system that has operated successfully at four major steel plants since 1975
and is presently being installed on ten additional coke ovens.
(I) is the area where the pulverized coal is heated to 1800°P to 2200°F
for a period of 16 to 24 hours, depending upon the type of coke produced. (J)
is the coke guide used to direct the incandescent coke across the bench area
(N) and into the waiting quench car (K). When this movement occurs, there is a
great thermal mass released. This thermal discharge is confined by the thermal
wall panel (A) which confines the mass and causes it to be driven upwards
toward the throat or orifice (C) . Any heat release coming from the face of the
ovens is directed up to the heat deflector (B) which also directs the thermal
mass into the throat (C). This area (C) is critically sized to cause the gas
velocity to be increased and directed up and over back toward the oven area.
The flow is directed downward and back up in a swirling action. By a constant
controlled air flow through the orifice (C), the mass is contained in area (D)
known as the capture and expansion zone. As the thermal mass is driven up into
this zone, it expands the length of the structure. During this time the
smaller airborne particulate is being evacuated through the evacuation duct
(E) to an offsite filter or cleaning device.
At the same time all this movement is taking place, the natural forces of
gravity are taking place. As the contaminated air is cooled, the heavier
particles fall out. In this design, MikroPul-G.L.C., there is only one area
that actually captures the fallout, that is area (G) on top of the heat shield.
This is good, as it does not fall right back on the people working on the bench
(N). But this only collects about 1/3 of the fallout generated. The balance
falls back through the throat (C) and onto the inclined siding and down the
thermal wall panel (A). By this fallout getting back into the working area and
the normal flow of incoming air going up toward the orifice, it causes a slight
disturbance. This one factor has caused some controversy. Being that this
fallout coming back down is of such size that it's too large to be inhaled by
the workers, but becomes a slight annoyance when it gets into their collars,
etc. In spite of this condition, the men who have worked under a shed prefer it
over no shed at all. It is cooler in the hot weather and they are protected
during the inclement weather. The ends of the structure are extended down as
close to the coke guide (J) and quench car (K) and engine as possible. This
helps keep the emissions confined so they can be directed upwards to the evacu-
164
-------
M MODULAR DOME ROOF
304 STAINLESS STEEL
F DUCT TO
REMOTE
BAGHOUSE-
MIKROPUL-PATTON COKE SIDE SHED
165
-------
ation duct for proper processing. On an extremely windy day there can be some
end blow out.
One area of attention on the MikroPul-G.L.C. shed system is the fallout
that is collected up on top of the thermal heat deflector (B). The accumu-
lation of material in this area must be cleaned out on the average of every
three to six months. This can be done manually or by vacuuming. Aside from
this, the maintenance on this system is almost nil, mainly because it has no
moving parts and the outside siding is of 304 stainless.
The earlier coke side shed systems used MikroPul "Elektrofil" wet elec-
trostatic precipitators as the air cleaning device. In terms of pollution
control, this was a relatively new application with no prior history from which
to draw. The selection of the wet electrostatic precipitator was made prima-
rily due to its ability to tolerate the hydrocarbons in the gas stream and
collect the particulate at acceptable efficiencies. While in every case these
systems performed successfully, the main objections to the use of the WEP were
the initial high cost and the cost of maintenance due to corrosion, plus an
expensive water treatment plant.
An extensive testing program was conducted with pilot Mikro-Pulsaire bag-
houses on existing coke side shed. The results showed that the use of a dry
additive pre-coat, such as limestone, to act as an absorber, would prevent the
bags from blinding. While the principle of dry additive injection has been
well known for many years, its application has been limited primarily to low-
ratio type baghouses. In recent years MikroPul has been successful in applying
it to huge installations in the aluminum industry. The Mikro-Pulsaire reverse
jet type baghouse is now used exclusively on our coke side shed systems and
is widely specified in the competitive land-based system concept.
Figure 3
This is the cross section of the "second generation" of coke side shed
emissions control system. This is to be known as MikroPul-Patton shed design.
MikroPul has acquired complete patent rights of this concept.
This concept is very similar to the original but offers some very impor-
tant improvements that have corrected the problems that exist in original shed
concept. The distance (x) is closer to the oven face (I) permits the flow of
air to pass closer to the oven doors. This gives improved ventilation in the
bench area (N) where the men work. The thermal wall panel (A) is in the same
position to the quench car (K) as in the original, but as the thermal mass is
forced upward, it turns into the throat or orifice area (C) and the thermal
heat deflector (B) comes together closer to the oven. The throat or orifice
(C) is narrow causing the velocity of the thermal mass to increase. It comes
in contact with the modular dome roof, which due to its configuration, quickly
directs the air flow mass over into a true capture and expansion zone (D) .
Since the distance across the dome area (D) is twice the orifice (C) distance,
this slows the air flow down by half and the air mass expanding and traveling
the length of the expansion zone (D) is greatly slowed down. Cooling is taking
place during this period and the natural forces (gravity) are causing the
heavier particles to fall out of the air stream. Different from the original
166
-------
M MODULAR DOME ROOF
304 STAINLESS STEEL
MIKROPUL- RATION COKE SIDE SHED
167
-------
concept, here the completely captured volume is directed downward and the
smaller particles are sucked into the evacuation duct (E) which is connected to
an offsite baghouse cleaning device.
As another improvement, the complete volume taken in the structure along
with the emissions has to pass through zone (D) to the cleaning system. Once
the thermal mass is beyond the peak of the thermal wall panel (A) the particu-
late laden air flow is in complete entrapment and under positive control. The
fallout captured in this system falls down into a hopper for controlled removal
by mechanical or vacuum conveyance.
By comparison with the original concept, it is easy to see the more posi-
tive control we have in this second generation concept. The dome (M) is of
modular design and type 304 stainless. The structure has basically no weight
on the oven buckstays. Doors and buckstays can be serviced with proper equip-
ment. The ends (H) are very similar to the original shed concept but the end
blow through is under better control as the mass of emissions are completely
locked in area (D) and not affected by windy days.
The MikroPul-Patton shed design offers a good positive lighting system
(L) with proper means to service it. This system is adaptable to most oven
installations and offers a more desirable working area for the workers on the
bench (N). The maintenance on this system is almost nil and the reliability
approaching 100% for continuous operations, tops all known systems.
Figure 4
This is a cross section of the latest in technology, the MikroPul-Patton
coke side shed with a series of modular baghouses incorporated within the
structure. As you will notice the air flow is the same as shown in Figure 3,
only the capture and expansion zone (D) comes down lower and connects with the
evacuation opening (E). This is connected through a controlled opening and
enters into the bottom of the modular baghouses (F) . This opening (E) is where
the precoat system is attached to precoat the filter bags. The air volume
travels up through the bags and up through a chamber before going through the
exhaust fans to the atmosphere. These modular baghouses are constructed so any
one can be serviced without a noticeable difference in performance. Say the
structure had 20 bays and 20 baghouses. If one unit had to be shut down for
service, it would increase the evacuation time by approximately six to eight
seconds. This system is the same as Figure 3 where the complete air volume
drawn into the structure has to pass through the filtering devices before
exhausted to the atmosphere. At this point we meet all existing requirements
for allowable emissions.
This concept needs a slight more width than the Figure 3 concept, but the
savings on real estate required and long ductwork to an offsite cleaning device
is considerable. The power demand is almost one-half of the other two systems.
This integral filtering concept is adaptable to many installations, but not
all, as if there is too much in-plant interference it is not practical. For
any new installation, this system should be given much consideration in the
early planning stages.
168
-------
METHOD OF CONSTRUCTION TO AID IN MINIMUM
INTERFERENCE WITH NORMAL
OPERATIONS
\
TO ATMOSPHERE
MODULAR DOMES
ASSEMBLES OFFSITE
MIKROPUL-PATTON COKE SIDE SHED
169
-------
Figure 5
This is a cross section showing how the MikroPul-Patton shed concept is
erected in a simple fashion so as to interfere with normal operations as little
as possible.
As you see in this figure, the basic structural is erected with thermal
wall panel (A) and thermal heat deflector (B) leaving opening (C) open to the
atmosphere during construction. The hoppers as part of (D) and (E) and the end
walls (H) are installed. Then the filtering devices (F), baghouses of modular
design, are fit in place in each bay of the structure. All of this is to be
completed and started up before the final capping of the dome. The dome is
type 304 stainless and is pre-assembled offsite in approximately 30 foot sec-
tions. These sections, being very light but extremely strong, are hoisted into
place. When the last section is fitted in place, the shed concept is totally
operational.
CONCLUSION
The coke side shed emission control concept is only one of a number of
concepts in this area of control. MikroPul, being the forerunner in the shed
concepts, has a number of successful operating installations and many new
orders are in engineering and fabrication stages. The latest association with
the patented "Patton" shed concept provides us with new and improved techno-
logy, enabling us to offer a coke side shed for almost any existing or new coke
oven facility. This is a part of MikroPul's continuing effort to reduce costs
and offer reliable emission control systems to the coking industry. We refer
to these latest developments as "The Second Generation of Coke Side Sheds".
170
-------
EFFECTS OF FLYASH SIZE DISTRIBUTION
ON THE PERFORMANCE OF A FIBERGLASS FILTER
By
William F. Frazier
Wayne T. Davis
The Department of Civil Engineering
The University of Tennessee
Knoxville, Tennessee 37916
ABSTRACT
This study was undertaken to quantify the effects of flyash size char-
acteristics on the pressure drop and efficiency of a conditioned fiberglass
fabric filter typical of the fabrics being utilized in fabric filter parti-
culate collectors on coal-fired boilers. These effects were determined by
conducting a series of filtration tests in which six flyash dusts, each with
a different size distribution, were used to determine the performance. The
range of mass mean diameters was 3.5-13 urn.
Graphs are included showing the experimentally measured relationships
between particle penetration, residual pressure drop and coefficient of re-
sistivity for the six flyash size distributions on the conditioned fiberglass
fabrics.
INTRODUCTION
The fabric filter is one type of high efficiency dust collector whose
performance, as measured by its pressure drop and dust penetration is affec-
ted by the size characteristics of the dust being collected. Collection of
smaller size flyash particles creates the potential for a higher fabric pres-
sure drop due to the formation of a dust cake of more finely divided parti-
cles as well as decreased collection efficiency (or increased penetration)
of the particles. This study was undertaken to quantify the effects of fly-
ash size characteristics on the pressure drop and efficiency of a condition-
ed fiberglass fabric filter typical of the fabrics being utilized in fabric
filter particulate collectors on coal-fired boilers.
Experimental Facility
To accomplish the above objective, a pilot plant fabric filter collector
containing a single 30'- x 11^" diameter filter bag was used. The system
(shown in Figure 1) provided for operation at a constant temperature, humid-
171
-------
ity, and filtration velocity. During the filtration cycle, the test filter
compartment was in a closed loop with an environmental control room which was
utilized to maintain system temperature and humidity. Compartment flow was
maintained by a damper which was controlled by a Dwyer Photohelic on/off
controller. The controller was activated by the pressure differential across
an orifice plate located in the ductwork on the downstream side of the filter
bag.
At the initiation of the cleaning cycle the test filter compartment was
placed in a closed loop with the reverse-air conditioning compartment. This
latter compartment supplied a source of cleaned air to reverse-air clean the
test fabric. The reverse-air conditioning compartment was required to pre-
vent carry-over of dust from the test filter compartment into the control
room or laboratory during the reverse-air cleaning cycle. The reverse-air
flow was determined by the pressure differential across an orifice plate
calibrated with a standard type pitot tube and inclined water manometer, and
was controlled by a manually adjusted damper.
During each filtration cycle, flyash test dust was injected from a
Vibrascrew dust feeder (located in the environmental control room) into the
duct leading to the filter compartment. The flyash utilized in this study
had a mass median diameter of 13.0-14.0 ym and was obtained from the hoppers
of an electrostatic precipitator installed at a nearby pulverized coal-fired
boiler (Kingston Steam Plant, Tennessee Valley Authority). In order to pro-
duce six different size distributions, it was necessary to pass the particle-
laden air leaving the environmental room through various configurations of
precleaners. By controlling the inlet loading with the Vibrascrew feeder and
the particle size distribution with the precleaners, it was possible to pro-
duce six size distributions of flyash as shown in Table 1. Also included in
_the table is the equivalent surface mean diameter, d , for each dust. The
d „ was employed in this study since it has been found to be related to the
resistance through packed beds of granular material (1).
The same style of fabric material was used throughout the test program.
The fabric was a 14.5 ounce/square yard fiberglass woven fabric with a 3 x 1
twill and an air permeability of 30-45 CFM/ft at 0.5" H-O. A new filter
bag was installed in the filter test compartment prior to testing each dif-
ferent flyash size distribution to eliminate any conditioning effects caused
by interaction of the different dusts.
The two performance parameters which were used as direct measures of the
performance were (1) the pressure drop across the fabric and (2) the particle
penetration, Pn, defined as
Pn = 1 - (Efficiency/100%)
The fabric penetration was monitored by extracting (1) an isokinetic
sample through a filter followed by gravimetric determination of the weight
gain on the filter, (2) and an isokinetic sample into a Climet Model 208/210
optical particle counter (OPC) and multichannel analyzer (MCA) (0.3-10.0 ym
range). Measurements using the gravimetric technique were conducted to de-
termine the overall average efficiency of the filter for an entire filtration
172
-------
TABLE 1. PARTICLE SIZE DISTRIBUTION OF FLYASH
Test
1
2
3
4
5
6
Mass Median Diameter
13
9
7
7
4
3
.0
.5
.1
.1
.0
.5
ym
Vim
Vim
Vim
Vim
Vim
Geometric Standard Deviation
2
2
2
2
1
1
.2
.1
.1
.0
.8
.9
9
7
5
5
3
2
d32
.5 ym
.2 ym
.8 ym
.6 ym
.4 ym
.8 ym
cycle. The optical particle counter was used to obtain the temporal resolu-
tion needed to measure changes in the filter penetration during each filtra-
tion cycle (updated every 24 seconds).
The operating conditions for all of the tests were as follows: a fil-
tration velocity, V, of 0.64 m/min (2.1 feet/minute) ; reverse-air cleaning
ratio, R/A, of 0.49 m/min (1.6 feet/minute); a temperature, of 38 C ± 5 C; a
relative humidity, of 40% ± 5%; and a mass loading, C. of 6.9 grams/m3 (nom-
inal) . 1
Analysis of Data
The performance data were analyzed in this study in a manner similar to
the EPA Filtration Model for Coal Flyash with Glass Fabrics (1). Normalized
performance curves were constructed from the data obtained after each filter
had reached a conditioned state by transforming the fabric pressure loss
data to normalized filter drag, AP/V, and time to areal cake density, C_.Vt.
Figure 2 is a schematic of a drag performance curve. The slope of the
linear portion of each performance curve, referred to as the specific resis-
tance coefficient, K2, was determined for each test. The intercept of the
linear portion of the curve with the ordinate was defined as the effective
residual drag, S , for the conditioned fabric. The residual fabric drag,
S , was the valul of initial drag after the fabric had been cycled through
sufficient filtration and cleaning cycles to reach a conditioned state. The
initial slope of the fabric drag performance curve, defined as the condi-
tioned residual specific resistance coefficient (K ), was also determined
for each test. The equation describing the filterrdrag performance curve
was as follows:
S + K W + (S - S ) [ 1 - exp (
R Z E K
-W(Kr-K2)
) 1 (eq. 2)
173
-------
where W = C.Vt and represented the dust cake added to the fabric during each
filtration cycle.
Fabric penetration, Pn, versus time, t, was also recorded for each
fabric after it had reached a conditioned state. Figure 2 also shows a sche-
matic of a fabric penetration performance curve. The relationship between
penetration and areal cake density, C.Vt, was described by the following
equation:
Pn = Pn + (Pn - Pn ) exp (-aW) (eq. 3)
sos
where Pn = penetration at time, t
Pn = steady state penetration
Pns = initial penetration
W °= actual fabric loading above residual (C.Vt)
a = concentration decay function
In this study, the initial penetration, Pn , was measured at one minute
into the filtration cycle. The steady-state penetration, Pns, was measured
near the end of the filtration cycle. The decay constant "a" was determined
for each penetration performance curve.
Results
The above equations 2 and 3 described fabric drag and penetration using
the performance parameters Sg, SR, K.2, Kr, Piio, Pns and a. In this study,
the average value of each performance parameter was determined for each of
the six size distributions that were tested. The results of the filtration
tests have been summarized graphically to illustrate the basic relationships
between the surface mean diameter, d^2) a^d each of the above performance
parameters.
The effects of d32 on the effective residual drag Sg, the residual drag
SR, and the residual resistance coefficient, Kr, are shown in Figure 3. A
least squares fit was determined for the data obtained on each performance
parameter. The relationships illustrated in Figure 3 showed that the param-
eters SR, SE, and Kr were relatively independent of the surface mean diam-
eter, d32, for distributions for which 332, was greater than 6 um. For
values of d32 less than 6 um, however, large changes in the fabric drag par-
ameters were observed. For example, the residual fabric drag, SR, doubled
when d32 was reduced from 4 to 3 um.
Figure 4 illustrates the relationship between K2 and d~32- A least
square fit was performed on this data to illustrate the relationship between
lnK2 and Ind32. The following equation was the result of the least square
fit:
K2 = 10.94 d32 -1.28
where the units on d32 and K2 were micrometers and newton-minutes/gram-meter ,
respectively. This relationship is in qualitative agreement with other re-
174
-------
searchers who have found K£ to be proportional to 332 * (1,2,3).
Figure 5 illustrates the relationship between the parameters PnQ and
Png and the surface mean diameter d32; the values of PnQ and Png were found
to increase as d-^2 was decreased. These data are in agreement with other
research efforts which have shown fabric penetration to increase as the
particle diameter was decreased (1,4,5). It is significant to note that the
reproducibility of the penetration measurements was found to be approxi-
mately one half an order of magnitude due to real variations in the penetra-
tion between successive cycles. The decay constant, a, presented in Equa-
tion 2 was not found to be significantly dependent on A-^2- The mean value
of a for the six different size flyashes tested was 0.04 m3/gram. Best fit
equations were calculated for the relationships between each of the fabric
performance parameters and the surface mean diameter, 532 (See Table 2.)
Figures 6 and 7 illustrate the effect of d^2 on the performance curves
of filter drag and penetration, respectively for the data in this study.
The solid lines superimposed on the data represent the curves obtained by
applying the equations summarized in Table 2.
Discussion
The data obtained in this study were obtained on a specific style of
fabric using six different flyash distributions. The effects of particle
size on fabric performance may be different for other dust/fabric combina-
tions. However, the effects of particle size on fabric drag and penetra-
tion for the fabric and dust used in this study demonstrate the need to
quantitatively evaluate the effect of particle size on fabric performance
for various dust/fabric combinations.
The results of this study indicate that the best fabric performance was
obtained for the flyashes with ^32 greater than 6 vim. Caution should be
used when designing fabric filter systems for flyashes with d32 less than
6 pro, however. A comparison of the flyash with a d32 of 2.8 vim to that of a
flyash with a ^32 of 5.8 urn showed that the penetration increased by approx-
imately two and one half orders of magnitude and the pressure drop increased
by a factor of five.
The relationships developed in this study were developed at an air-to-
cloth ratio of 2.1 CFM/ft2 (0.64 m/min.) and reverse-air cleaning ratio of
1.6 CFM/ft2 (0.48 m/min.). It has been shown previously that both the air-
to-cloth ratio and reverse-air ratio also affect fabric performance for a
single dust (1,3), thus the relationships developed in this study are only
applicable at the operating conditions used during the testing. Additional
research is needed to investigate the combined effects of particle size,
air-to-cloth ratio and reverse-air ratio on fabric filter performance.
Conclusions
This research study has shown the effect of the flyash particle size
distribution on the basic performance of a conditioned fiberglass filter.
175
-------
Although the data are for only one specific fabric, and a specific set of
operating conditions, the fabric is typical of those being applied to full
scale fabric filter collectors on coal-fired boilers in the United States.
The following conclusions were reached in the study:
1 A decrease in the median diameter of the particle size distribution
(as measured either by the mass median diameter or the surface mean
diameter) resulted in significant increases in both the residual
and effective fabric drag, and the residual specific resistance
coefficients of the filter I or MMD < 8 m (d32 1 6 urn) . A change in
MMD from 7.7 ym to 3.5 ym resulted in a 5-fold increase in the
values of these parameters. For MMD 18 ym, the values were found
to be relatively constant indicating only a minor dependence on the
particle size distribution.
2. The specific resistance coefficient, K2 was found to be propor-
tional to d32 to the negative power of 1.28. This is in support
of previous research where K2 was found to be approximately pro-
portional to d32 1(2).
3. A decrease in the median diameter of the particle size distribu-
tion resulted in significant increases in the flyash penetration
for all particle size distributions studied (MMD = 3.5-13.0 urn).
A change in MMD from 13.0 ym to 3.5 ym resulted in greater than a
three order of magnitude increase in flyash penetration. (The
efficiency of the fabric decreased from 99.99+% to less than 95%.)
4. Within the limits of the study, the relationships developed can be
used to predict changes in fabric filter performance resulting
from different inlet size distributions.
Acknowled g ement
The research conducted in this study was funded under a contract with
the Tennessee Valley Authority. The contents do not necessarily reflect
the views of TVA, nor does the mention of trade names or commercial products
constitute endorsement or recommendation for use.
ENDNOTES
1. Dennis, Richard, R.W. Cass, D.W. Cooper, R.R. Hall, Vladimir Hampl,
H.A. Klemm, J.E. Langley, and R.W. Stern, Filtration Model for Coal
Flyash with Glass Fabrics, EPA-600/7-77-084, August 1977.
2. Davis, W.T. and R.F. Kurzynske, "The Effect of Cyclonic Precleaners on
the Pressure Drop of Fabric Filters," Filtration and Separation, Janu-
ary 20, 1979.
3. Strangert, S. and S. Flaktfabriken, "Predicting Performance of Bag
Filters," Filtration and Separation, January/February 1978.
176
-------
4. McKenna, J.D. and R. Weisburg, "A Pilot Scale Investigation of Fabric
Filtration as Applied to Coal-Fired Industrial Boilers," Enviro-Systems
and Research, Inc., P.O. Box 658, Roanoke, Virginia, 24004, USA.
5. Dennis, Richard and John Wilder, Fabric Filter Cleaning Studies, Natio-
nal Technical Information Service //EPA-650-2-75-009, Springfield,
Virginia, 22161, USA.
TABLE 2. RELATIONSHIP BETWEEN FABRIC PERFORMANCE PARAMETERS
s =
K2
Kr
SE
SR
= Sp + K,W + (ST,
IS. £. JL
+10.9 d32 ~1'
=3.6+ 645 exp
= 269 + 3.95 x
= 216 + 2.09 x
Fabric Drag
— W \K- — "-r\/
S\ r i / ^- ^ >
R; L i exp ^ ,
K SE - SR
28 N-min
g-m
f 1 !•> d > N-min
^ J..XZ Q32/)
g-m
, Ou / i on T \ N-min
10H exp (-1.23 d32) ma
4 , 7 N-min
10 exp (-1.11 dQO) — — 3r~
-i/ m
.) («,.
(eq.
(eq.
(eq.
(eq.
2)
4)
5)
6)
7)
where d~« = surface mean diameter (micrometers)
Fabric Penetration
Pn = Pns + (Pn0 - Png) exp (-aW) (eq. 3)
Pn0 = 0.42 exp (-0.81 d~32) (eq. 8)
Pns =0.77 exp (-1.09 d32) (eq. 9)
a = 0.04 mVgram
177
-------
PULSE-JET COLLECTOR
(R/A CLEANING FILTER)
TEMPERATURE
CONTROLLER
LREVERSE AIR
TEST FILTER
COMPARTMENT
FIGURE 1. SCHEMATIC DIAGRAM OF THE FABRIC FILTER TEST FACILITY.
CWt
MAX.
Figure 2. Graphs of Filter Drag and Fabric Penetration
versus Areal Cake Density Added During the
Filtration Cycle for a Conditioned Filter.
178
-------
4000
3000
2000
'"js
C
I 1000
1 800
ac
j? 60°
"? 500
? 400
d
S 300
-| 1 i 1 1 , 1 , , ! r
0 f
J-
s
3 H
10 11 12
10.0
8.0
u 1.0
t 0.8
8
I 0.4
S
"C 0.2
I I I I I I
8 10 20
d^j (Microns)
40 60 80 100
Figure 3. Kr, SE, and SR as a Function of Surface Mean Diameter
Figure 4, Specific Resistance Coefficient (K-)
as a Function of Surface Mean
Diameter (d32)
d32 (MB)
Figure 5. Pn and Pn as Functions of Surface
Mean Diameter (3j2).
179
-------
Figure 6.
Fabric Drag as a Function of Areal Cake Density for Four
Bags Conditioned with Different Sizes of Flyash.
40 60 80
CiVt (grams/m^)
120
Figure 7. Mass Penetration as a Function of Areal
Cake Density for Different Sizes of Flyash.
180
-------
FUNDAMENTAL STUDY OF A FABRIC FILTER WITH A CORONA PRECHARGER
By: K. linoya and Y. Mori
Department of Chemical Engineering
Kyoto University, Kyoto 606, JAPAN
ABSTRACT
The effects of a corona precharger on the performance of a fabric filter
have been studied experimentally in air with controlled humidity. The test
dust, that is fly ash, is precharged and then ducted to a test fabric.
For a small bench scale apparatus, charged dust reduces the pressure
loss across the filter at lower filtration velocity and lower relative
humidity. The collection efficiency of a fabric filter can usually be
improved by using a corona precharger.
Using a larger bench scale apparatus, by which continuous filtering and
cleaning operations can be performed, the electrostatic effects on the
pressure loss across a filter and on the collection efficiency are the same
as for that of a small bench scale. The dust dislodged from a fabric filter
with a corona precharger in a cleaning operation is larger than that that
dislodged without a precharger. However, the residual dust fraction for
charged dust is the same as or Larger than that without a corona precharger.
INTRODUCTION
Dust collectors with the highest efficiency, such as an electrostatic
precipitator or a fabric filter, must be used to meet the requirements of the
particulate emission regulation. However, every collector has both
advantages and disadvantages. Therefore, efforts to develop a combination
type having the advantages of both collectors are necessary. One such
example is the Apitron collector developed in the United States. The
Apitron incorporates a pulse jet fabric filter, preceded by a wire-in-
cylinder type of electrostatic precipitator. However, fundamental studies
have been little concerning with its performance and economy [1-6].
This report, first, describes the preliminary and experimental results
of the filtration performance of the hybrid filter obtained by the use of a
small bench scale test facility, which is called the "small device". This
paper also presents the results of the dust filtration and dislodging
performances using another type of continuously operated hybrid filter called
the "large device".
EXPERIMENTAL APPARATUS AND PROCEDURE
Figure 1 shows a flow diagram of the small experimental device for the
investigation of the effects of a corona precharger on filtration perfor-
mances at various relative humidities and filtration velocities. Test dust
is uniformly fed into a mixer-type disperser by a Micro Feeder (Sankyo Dengyo
Co.), and is ducted to the test filter through both a positive corona
181
-------
precharger and a settling (or mixing) chamber. The dust collection area of a
test fabric is about 50 cm2, and the fabric is polyester (Tetron) felt, the
surface of which is treated in order to dislodge the collected dust easily.
Dust, which penetrates through the test fabric, is completely collected by
a high efficiency glass fiber filter (AP20, Millipore Co.). The electro-
static charge of dispersed test dust is also measured just in front of the
test fabric by use of a Faraday cage with the high efficiency glass fiber
filter.
Fly ash (JIS No.5) is used as the test dust for the small device. The
size distributions of charged dusts, which are measured by a cascade
impactor (226 type, Sierra Co.) just in front of the test fabric, are almost
the same as those of uncharged dusts, and are expressed approximately as
log-normal distributions with a mass median diameter of 5 pm and a geometric
standard deviation of 2.1.
Figure 2 shows a schematic diagram of the large experimental device,
which can be operated continuously. A test filter consists of one bag
which is about 12 cm diameter and 20 cm long. The collection area of a test
bag is about 754 cm2, and the fabric is wool felt. The dust laden gas flows
from the outside to the inside of the test bag, and the filtration velocity
through the test fabric is automatically controlled at 3.0m/min. In order
to sample the inlet dust, a probe is located between the corona precharger
and the test fabric. The concentration and the electrostatic charge of the
inlet dust are measured by the Faraday cage and nearly isokinetic sampling.
The size distribution of the inlet dust is also measured by a cascade
impactor (226 type, Sierra Co.) using the same sampling probe. The dust
concentration of the outlet dust is measured continuously by use of a
aerosol mass monitor (model 51-1111, Piezo balance, Kanomax, Osaka) and a
high efficiency glass fiber filter (GB 100R, Toyo Roshi Co., Tokyo). The
mean concentration of outlet dusts measured by the mass monitor agrees well
with that by a glass fiber filter. The cleaning operation is activated
whenever the pressure loss across the test fabric reaches the predetermined
value of 1.96 kPa. Therefore, the filtration period is variable. The dust
collected on the test fabric is dislodged by using pulse jets. Cleaning jet
pressure is about 390 kPa, and the electromagnetic valve for cleaning is
opened for about 60~ms.
The test dust for the large device is also fly ash collected by an
electrostatic precipitator at a coal-fired power station in Japan. It has
a mass median diameter of 3.6 ym and a geometric standard deviation of 2.7,
measured by a cascade impactor just in front of the test fabric.
CHARACTERISTICS OF A CORONA PRECHARGER
The corona precharger for the "small device" is a kind of a short
electrostatic precipitator, which is constructed with 8 discharge electrodes
of tungsten wire (0.2 mm x 100 mm) and 9 plate electrodes (11 mm x 100 mm).
The distance between the above two electrodes is 10 mm. The velocity of the
air loaded with test dusts is about 17 cm/s in the precharger, and the
resident time is about 0.06 seconds. The field strength is about 4 kV/cm
182
-------
and the corona current per unit length of the wire electrode is 0.35 to 0.75
yA/cm. Test dusts are charged by positive corona. The precharger for the
"large device" is composed of a discharge electrode of tungsten wire of 0.2mm
diameter and a cylindrical electrode of 2.8 cm inner diameter. The length of
the corona precharger is 11 cm. Flyash dust is charged by the negative
corona in the "large device". The field strength is about 5 kV/cm, and the
corona current per unit length of the wire electrode is 5.5 to 9.1 uA/cm.
EXPERIMENTAL RESULTS OF THE "SMALL DEVICE"
Surface Characteristics of the Dust Layer Collected on the Test Fabric
Figure 3 is a photograph of the rugged surface of the charged dust layer
collected on the test fabric. Figure 4 is the smooth surface of the uncharged
dust on the same fabric. The pressure loss of the loaded test fabric in the
case of Figure 3 is lower than that in the case shown in Figure 4.
Pressure Loss
Figures 5 to 8 show the comparisons of the drags of charged and uncharged
flyash for various relative humidity ranges. Except at a high relative
humidity of 85 to 95 %, the charged dust reduces the pressure loss.
Figures 6 and 8 also show the effects of filtration velocities on the
drags. For a relative humidity of 45 to 55 %, the drags without the corona
precharger at higher filtration velocity are lower than those at lower
filtration velocity, as shown in Figure 6, because of pinhole formations in
the dust layer. The rugged surface of the charged dust layer collected on a
test fabric becomes smoother at higher filtration velocity than at lower
filtration velocity. Figure 8 shows a relationship between the drags and dust
loads at a relative humidity of 85 to 95 %. The drags at a filtration
velocity of 6.4 m/tnin are almost the same as those at 2.55 m/min. This means
that there is no pinhole formation at high relative humidity. Furthermore,
the corona precharger does not reduce the pressure loss, regardless of the
filtration velocity.
Collection Efficiency
Figure 9 shows the cumulative penetrations for the same conditions as
those in the pressure loss experiments. The penetration seems to decrease
with lower relative humidity for uncharged dust, but shows no definite
change for charged dusts.
Figure 10 shows the effect of the filtration velocity on the efficiencies.
Unlike the results of the pressure loss, the corona precharger improves
cumulative collection efficiencies even at higher filtration velocity. The
cumulative penetration is higher at a filtration velocity of 6.4 m/min than at
2.55 m/min, because of pinhole formation in the dust layer.
The gradients of the straight lines in a log-log chart of cumulative
penetration vs. dust load are mostly minus unity, which means 100 % collection
efficiency. Therefore, in Figures 9 and 10, the collection efficiencies of
183
-------
higher dust load collected at lower filtration velocity are 100 %, and are
independent of the electrostatic charges of the test dusts. The effect of
electrostatic charge of test dust on the cumulative penetrations becomes
conspicuous at lower dust loads.
Figure 11 shows a relationship between the cumulative penetrations at
100g/m2 dust load and the electrostatic charges per unit dust weight. The
collection efficiency becomes higher as the electrostatic charge per unit
dust weight increases.
Electrostatic Charge of Test Dust
Figure 12 shows the electrostatic charge of flyash dust caused by the
corona precharger. The charge is not always consistent, though the applied
voltage is constant. One of the reasons for this phenomena may be the
change of field strength which is reduced by the amount of dust deposited on
the electrodes. The relationship between the charges and the relative
humidity is shown as a solid line obtained by the least squares method. The
charges decrease as the relative humidity increases. Therefore, this result
indicates that the formation of the rugged surface depends on not only the
electrostatic charge but also on other factors.
Figure 13 shows the electrostatic charge of flyash dust without the
corona precharger, i.e. a natural charge. The charge depends on the relative
humidity. The reason for this may be the leakage of charges.
EXPERIMENTAL RESULTS OF THE "LARGE DEVICE"
Filtration Performance
Figure 14 shows the comparison of pressure losses across a fabric filter
with and without a corona precharger after a start-up. The filtration period
with a corona precharger is about 1.5 times longer than that without a
precharger, at the same operating condition.
Figure 15 shows the cumulative penetrations for every filtration period.
They decrease at the beginning and then remain nearly constant. The
penetrations with a corona precharger are almost one fifth of that without a
precharger, during the first several filtration cycles of a new fabric.
However, the ratio .decreases about 2.5 times after 10 filtration cycles.
Figure 16 shows the cumulative dust emission from the test fabric for
several filtration periods. The solid lines indicate the dust weight
penetrated with the precharger. The broken lines are also the dust weight
without the precharger. The dust emission at the first filtration cycle is
larger than that at other filtration cycles in both operating conditions.
Most of the dust emission takes place during the first few minutes in both
operating conditions. The dust emissions with the corona precharger are
lower and decrease faster than those without the precharger.
Cleaning Performance
Figure 17 shows the comparison of dust loads dislodged from a fabric
184
-------
filter with and without the precharger. The dislodged dust loads decrease
at the beginning and remain constant after about ten cycles. The dust load
dislodged from the fabric with the precharger is larger than that without the
precharger. However, it can't be concluded that the residual dust fraction
for the former case is less than that of the latter case, because the dust
load of each filtration period for the former is larger than that for the
latter.
Figure 18 shows the relationship between the residual dust fraction and
the filtration cycle. At the beginning, the residual fraction of the new
fabric with the precharger is slightly larger than that without the
precharger. However, both residual dust fractions become equal and remain
constant over ten filtration cycles after the start-up.
CONCLUSION
The effects of a corona precharger on the performance of a fabric filter
have been studied experimentally, and the following conclusions are obtained:
1) When the charged fly ash dust is collected on a test fabric at below 75 %
humidity, the rugged surface of the charged dust layer takes place, and
reduces the pressure loss across the loaded test fabric.
2) At the filtration velocity of 6.4 m/min, which is about twice as high as
usual, the rugged surface of the charged dust layer is smoother than that at
the usual velocity, and the pressure loss is not reduced. Furthermore,
pinholes emerge in the dust layer at lower relative humidity.
3) The collection efficiency of a fabric filter is improved by the corona
precharger for wide ranges of relative humidity and filtration velocity. The
collection efficiency gets higher as the electrostatic charge per unit dust
weight increases. It is improved especially at lower dust load by using a
corona precharger.
4) The electrostatic charges of dusts decrease as the relative humidity
increases, because of leakage caused by moisture in the air. The reduced
pressure loss by use of the corona precharger depends not only on the amount
of electrostatic charge of the test dust, but also on other factors.
Furthermore, A bench scale experiments of continuous filtration and
cleaning cycles have been conducted. The following conclusions are obtained:
5) The filtration period of the fabric with the precharger is longer than
that without the precharger at the constant maximum pressure loss across the
fabric.
6) The cumulative penetration of every filtration period with the precharger
is lower than that without the precharger. And the penetrations decrease at
the beginning and remain constant after a few filtration cycles for both
cases.
7) The dust weight dislodged from the fabric with the precharger is larger
than that without the precharger in every cleaning operation. The residual
dust fractions in both cases are nearly the same sand remain constant after
a few filtration cycles.
NOMENCLATURE
c± = inlet dust concentration [g/m3]
185
-------
E
m
md
nf
p
Ap
Ap.
u
w0
A
*
max
= cumulative collection efficiency
= dust load collected on a fabric
= dust load dislodged from a fabric
= number of the filtration cycle
= pulse pressure for cleaning in gauge
= pressure loss across a fabric filter
= pressure loss just before cleaning operation
= filtration velocity
= dust weight penetrated through a fabric
= residual dust fraction on a fabric by weight
= relative humidity
ENDNOTES
[g/m2]
[g/m2]
[ - J
[kPa]
[kPa]
[kPaJ
[m/min]
[ g 1
[ - ]
[ % 1
Helfritch, D.J. Performance of an Electrostatically Aided Fabric Filter.
Chem. Eng. Prog. 73: 54-57, August 1977.
Lamb, G.E.R. and P.A. Costanza. Improving Performance of Fabric Filters.
Chem. Eng. Prog. 73: 51-53, January 1977.
Lamb, G.E.R. and P.A. Costanza. Electrical Stimulation of Fabric
Filtration. Textile Res. J. 47: 372-380, May 1977.
Penney, G.W. Using Electrostatic Forces to Reduce Pressure Drop in Fabric
Filters. Powder Tech. 18: 111-116, 1977.
Penney, G.W. Electrostatic Effects in Fabric Filtration: I. Fields,
Fabrics, and Particles. EPA600/7-78-142a, September 1978
Costanza, P.A., G.E.R. Lamb, and J.B. Dunbar. Electrical Stimulation of
Fabric Filtration (ESFF) for Cotton Dust Control. J. Eng. Ind. Trans.
ASME. 101:65-68, February 1979.
paper filter test fabric
blower
high voltage
power supply
Figure 1.
Experimental apparatus for electrostatic effects in
fabric filtration ("small device") .
186
-------
electromagnetic
valve
air
Figure 2. Pulse jet type bench scale apparatus for
electrostatic effects in fabric filtration
("large device").
Figure 3. Photograph of rugged
surface of deposited dust
with corona precharger,
m=304 g/m2,R.H.=50%.
Figure A. Photograph of smooth
surface of deposited dust
without corona precharger,
m=329 g/m2, R.H.=52%.
187
-------
500i
E 400
c
E
S.
o.
<
200
01
0
1 1 1
felt fabric (B-9650S)
fly ash (JIS No. 5)
u = 2-55m/min
«P= 25-35 */.
without .
corona prechager^x
with corona prechager
_L
_L
100
dust load
200
m (g/m2 )
300
Figure 5. Relationship between
drags and dust loads
on a filter,
R.H.=25-35%.
600
E
"c
£400
.200 -
o»
o
u (m/min)
with corona precharger
without corona precharger
2-55
a
•
640
V
V
fly ash (JIS No. 5)
felt fabric (B-9650S)
«P= 45-55%
Figure 6.
100 200 300 350
dust load , m (9/m2)
Relationship
between drags and
dust loads on a
filter,
R.H.=45-55%.
500
400
a
CL
a.
<
o
k.
•o
200
felt fabric (B-9650S)
fly ash (JIS No.5)
u = 2-55m/min
«P= 65-75%
with corona precharger
without corona precharger
100 200
dust load . m (g/m2)
300
Figure 7. Relationship between
drags and dust loads
on a filter,
R.H.=65-75%.
188
-------
600i
c
E
o
a.
400h
,200
o>
a
u (m/min)
with corona precharger
without corona precharger
255
A
A
6-40
V
V
fly ash (JIS No.5)
felt fabric (B-9650S)
a>
J
>.
J3
* 0-3
0)
Q.
a
3
£
D
U
003
«P(«/.)
25-35
45-55
^5^75
[85-9.5
A
O
a
0
A
with
coro
pre
•
•
•
_A
without
na
charger
O
felt fabric
(B-9650S)
• A
--255
Na5)
V
1 1 1
40 100
300
1000
dust load , m (g/m2)
Figure 9. Relationship
between cumulative
penetrations and dust
loads on a filter
o>
,- 0-5
-Q -s
o^
•° -02
2-
aJ|
£ o> 0-1
ao
a, •-
~ "005
3
1 °
3 0-02
. ! ,
-
^~
.^^
A-s. D
*O
— i —
u = 2-55
m/min
u=6-40
m/min
*P=25-35°/.
45-55
65-75
85-95
>P=45-55°/.
o :
a -
<> .
A .
V.
^x_ 0 v
A 0V\ V 0
: fly ash (JIS No.5) v"" \V ~-
• felt fabric (B-9650S) a "VA '
.with corona y
precharger
1 ' ' ' ' n
01 23456
xlO"
charge per unit dust weight, (C/g)
Figure 11. Effect of electrostatic
charges on cumulative
penetrations .
189
-------
x10"8
-18
^ 7
o
2
O"> 5
5
3*
•O
r
!'
o
1 1 1 1 1 1 III
with corona precharger
fly ash (JIS No.5)
O
T °
J
Cj w
T ^
T O
-
u (m/min) 2-55 6-40
nodular deposit o — •
smooth deposit • T *
"0 20 40 60 80 100
relative humidity, )
Figure 12. Effect of relative
humidity on
electrostatic charge by
corona precharger.
o>
f
."t^
* -6
-i—i—r—i—r
fly ash
(JIS No.5)
T
without
corona precharger
20 40 60 80 100
relative humidity,
-------
0-02
0-01
0-005
o> «—
c -~-
a, - 0-002
0-001
0-0005
wool felt fabric , fly ash
r 9 u=3m/min , APmax=1.96kPa -
\ p = 392kPa , M> =45-55 °/o
without precharger
iV
precharger
15 10 15 20
filtration cycle , nf (-)
Figure 15. Comparison of cumulative
penetrations through test fabric
with and without corona precharger
for every filtration period.
~3
o>
f*
o
o
*— o
rti Jr
C
* -H
£ 1
g>
1
•o
0
x
/ w
i fl
1 U:
C|
9
A
! p
i
i
}k'~~~ -
) 10
1st filtration cycle
ool felt fabric
y ash
:3-Om/min
= 3-3g/m3 . .-.jit, nrorhnrnpr
=45-55 %> : without precharger
=392kPa
1st filtration cycle
10th
20 30 40 50
filtration time , t (min)
Figure 16. Cumulative dust emissions from test fabric
for several filtration periods.
191
-------
~.»U
E
~300
E
T3
O
"200
"O
Ol
TJ
o
(A
XJ
100
wool felt fabric , fly ash
AV u = 3m/min , APmax
A^ " p = 392kPa
A^A.A A A
A
°"b-o^ ° A
without precharger O
i i i
1 5 10
filtration cycle , r
i i
=1-96kPa
. ^=45-55%
A
T~ *^-+±-. r
A A
with precharger
0 0 °
0---°-- --
0 00'
i i
15 20
i, (-)
Figure 17. Comparison of dust load dislodged from test
fabric with and without corona precharger.
70.7
0-6
»•§
4= o
o>
I
0-2
wool felt fabric ,
u=3-0m/min
APmax =1 -96 kPa
p=392kPa
' C"""°" 0
i i
1 5
filtration
1 1 1 —
fly ash
A 0 A °
a,8'67a"°"°2"oaA'.
~~A— : precharger
--/•>--• without
' precharger
10 15 20
cycle , nf ( — )
Figure 18. Comparison of residual dust fraction of test
fabric with and without corona precharger.
192
-------
ECONOMIC EVAULATION FACTORS IN BID EVALUATIONS
A SENSITIVITY ANALYSIS
By: John G. Musgrove, J. E. Shellabarger
Bechtel Power Corporation
P. 0. Box 2166
Houston, TX 77001
ABSTRACT
The trend to evaluate major utility equipment bids on life cycle costs
rather than just on first capital costs has increased due to a greater
awareness of the costs of equipment maintenance and of borrowing money.
The rapidly changing economic climate and changes in interest rates impact
the establishment of a utility's economic evaluation factors.
An analysis has been conducted to determine the extent to which changes
in the basic economic factors (interest rates, demand charge, fuel charge,
escalation rates and plant life) alter the economic evaulation factors used
in competitive bid evaluations. This analysis was made to determine to what
extent these changes might alter the outcome of equipment selection for
major air pollution control systems. In particular, recent budgetary
studies comparing baghouses and cold-side electrostatic precipitators have
been analyzed to determine the sensitivity of the equipment selection to
the economic factors employed.
INTRODUCTION
The trend to evaluate major utility equipment bids by considering life
cycle costs rather than just first capital costs has increased. This trend
is due to a greater awareness of the costs of equipment operation and main-
tenance. In order to better understand the consequences of this trend on
equipment evaluations a parametric analysis was conducted to determine the
sensitivity of the total evaluated cost of air pollution control equipment
to several economic factors provided by utility clients.
FACTORS OF ANALYSIS
Recently equipment evaluation studies have been conducted comparing
the use of fabric filter baghouses and cold-side electrostatic precipitators
for particulate collection on coal-fired utility boilers (1)(2). The equip-
ment capital costs and operating parameters were obtained from budgetary
proposals of equipment suppliers. Engineering estimates were made for
Purchaser-supplied related equipment and services.
193
-------
Equipment costs used in the analysis included:
o Collector capital costs, fully erected.
o Other equipment capital costs.
o Annual maintenance costs.
o Surface area requiring insulation.
o Average electrical load.
o Maximum electrical load.
o Flue gas draft loss.
The economic factors considered in the analysis included:
o Incremental cost of induced $60,000/in. w.g.
draft fan
o Fan power cost 300 KW/in. w.g.
2
o Heat conductivity 31 Btu/hr/ft
o Plant life 20 to 40 years
o Net station heat rate 9000 to 13000 Btu/KWhr
o Average capacity factor 60 to 80 percent
o Fuel charge rate $1.80 to $2.60/10 Btu
o Demand charge $800 to $1200/KW
o Fixed charge 8 to 20 percent
o Discount rate 7 to 15 percent
o Escalation rate, goods and 6 to 14 percent
services
o Escalation rate, fuel 6 to 14 percent
Each of the nine variables was analyzed independently with all other
variables held constant at the midpoint of their ranges. For the variable
being analyzed, five values equally spaced within its range were selected.
Thus, a total of 45 cases were computed. Calculations for each case were
made of the total evaluated present worth of each supplier's offering (3)(4)
Study one contained capital and operating cost data from eight suppliers
each of baghouses and precipitators. Study two contained data from ten
194
-------
suppliers each of baghouses and precipitators. The resulting comparisons
were based on the average values of computed present worth for the two types
of equipment in each study.
RESULTS
Figures 1 through 3 present the results of the calculations of average
total present worth of baghouses and average total present worth of precip-
itators for each study. The calculations were repeated for each variable
at each of the five points within its range. The costs are presented as
differential present worth for rising values of the economic factor analyzed.
The following conclusions from these analyses appear reasonable:
Factors Favoring a Baghouse
Figure 1 presents the results of the sensitivity analysis for the fac-
tors of demand charge ($/KW), discount rate (percent) and fixed charge rate
(percent). Increasing values of discount rate clearly favor selection of a
baghouse, at the rate of $200,000 to $2,000,000 per percent. Increasing
values of demand charge and fixed charge only slightly favor a baghouse, at
the rate of $100 to $1450 per kilowatt and $41,000 to $77,000 per percent,
respectively.
That these factors tend to favor selection of a baghouse is reasonable
when one considers that for collectors of comparable first cost, the bag-
house is more labor intensive, particularly in operation (material and
labor for bag changes) whereas the precipitator is more capital intensive.
It is then logical that increasing values of discount rate (cost of capital)
and fixed charge (on capital) would more heavily penalize the precipitator
and therefore favor the selection of a baghouse. Similarly, since demand
charge is applied against maximum possible electrical consumption and the
precipitator is the higher consumer of electrical power it is reasonable
that increasing values of demand charge would favor a baghouse.
Factors Favoring a Precipitator
Figure 2 presents the results of the sensitivity analysis for the fac-
tors of plant life (years) and for escalation rate for goods and services
(percent). Increasing values of the assumed escalation rate of goods and
services and of the plant life clearly favor selection of a precipitator,
at the rate of $54,000 to $400,000 per year and $54,000 to $2,500,000 per
percent, respectively.
That these factors tend to favor selection of a precipitator is also
reasonable. As previously discussed, a precipitator has a lower operating
expense than does a labor intensive baghouse. Since increasing values of
the escalation rate for goods and services impact maintenance costs for
future years, and since increasing values for plant life extend the time
for which there must be maintenance expenditures, the baghouse is impacted
more highly than is the precipitator.
195
-------
Factors Reinforcing Design Selection
Figure 3 presents the results of the sensitivity analysis for the fac-
tors of fuel charge ($/10 Btu), net station heat rate (1000 Btu/KWhr), plant
capacity factor (percent) and fuel escalation rate (percent). Fuel escala-
tion rate has the greatest impact on total evaluated present worth, $100,000
to $300,000 per percent. The factor of next greatest impact is net station
heat rate, $100 to $125 per Btu/KWhr. Of lesser impgct are fuel charge and
plant capacity factor, $500,000 to $1,500,000 per 10 Btu and $16,500 to
$19,500 per percent.
As can be seen from Figure 3, rising values of these four factors have
opposite impacts for the two studies considered. The explanation for the
phenomenon that increasing values for the four economic factors tends to
reinforce the existing economic selection lies in the portions of the total
present worth cost that are influenced by these four factors. Inspection
of the method of evaluation (4) reveals that all four factors influence
only the present worth penalties for thermal losses, average power consump-
tion and draft loss. Capital costs and annual maintenance costs are not in-
fluenced by these factors. Close inspection of the data from suppliers in-
dicates that only surface area, average electric load and draft loss are
influenced by these variable factors.
Additional analysis, not presented in Figure 3, further substantiates
that increasing values for these four economic factors tends to reinforce
an existing economic selection. If the supplier-provided data for surface
area, draft loss and electrical load in both studies are doubled, the dif-
ferential total present worth based on rising values of the four economic
factors favors the existing design, baghouse or precipitator, by greater
amounts (graphs move outward on Figure 3). Similarly, if the supplier-
provided data are halved, the differential total present worth based on
rising values of the four economic factors continues to favor the existing
design selection but by smaller amounts (graphs move inward on Figure 3).
CONCLUSIONS
o Comparatively high values of demand charge, discount rate and fixed
charge rate cause present worth economic evaluations to favor selec-
tion of equipment having relatively low capital cost and relatively
high operating and annual maintenance costs.
o Comparatively high values of plant life and escalation rate for goods
and services cause present worth economic evaluations to favor selec-
tion of equipment having relatively high capital cost and relatively
low operating and annual maintenance costs.
o Comparatively high values of fuel charge, net station heat rate, plant
capacity factor and fuel escalation rate do not seem to bias present
worth economic evaluations toward equipment of low or high capital
196
-------
cost. Instead, they seem to exaggerate differences in evaluated cost
that exist for comparatively low values of the economic factors.
Use of the above conclusions may reduce the need for budgeting equip-
ment comparisons because inspection of the economic factors to be used
will permit forecasting any biases that will be inherent in such a com-
parison.
ENDNOTES
1. Musgrove, J. G. A Survey of Particulate Collection Devices for Coal-
Fired Boilers. In: Proceedings of the International Coal Utilization
Convention, Houston, Texas, October 17-19, 1978.
2. Particulate Removal Equipment Selection Study for Gulf States Utilities
Company, Roy S. Nelson Power Station - Unit 5, Westlake, Louisiana.
Bechtel Power Corporation, Houston, Texas, November, 1980.
3. Hickman, R. S. Methods and Terms Used in Engineering and Economic
Studies. Interoffice Memorandum, Bechtel Power Corporation, Houston,
Texas, February 3, 1976.
4. Musgrove, J. G. Present Worth Calculation Program. Bechtel Power
Corporation, Houston, Texas, February, 1981.
197
-------
10
00
DISCOUNT
DEMAND CHARGE FIXED CHARGE
-1200 21
7 I
DEMAND CHA
-------
PLANT LIFE ESCALATION
PLANT LIFE, YEARS
ESCALATION OF GOODS & SERVICES. %
O STUUY 1
D STUDY 2
$2,000,000
PRESENT WORTH COST DIFFERENTIAL WITH RISING ECONOMIC FACTORS
FIGURE 2 - FACTORS WHOSE RISING VALUES FAVOR A PRECIPITATOR
-------
14. o ^ tM
rsa
o
o
O STUDY I —• — — H.S.H.R , NWO ITU/lCW HR ._....._ FUf-l CMAROS.J/lt'lTU
DSTUOVZ —»»— FUEL ESCALATION. S _ - CAPACITY FACTOR. >
"X
vA
1
^ f V
. i\ *>
V
\\
\\
to
\^
\
r1
t 7
<(
>
20
<;
1
k 1
>
I
"
^^
, *^^
;
/,
k'
J )
'7/-
T
— JTO ~
5" — -H
n
f
< j
/
p --
_-••* * v "^
i_
n
^ —
5500,000
PRESENT WORTH COST DIFFERENTIAL WITH RISING ECONOMIC FACTORS
FIGURE 3 - FACTORS WHOSE RISING VALUES REINFORCE THE DESIGN SELECTION
-------
FLY ASH RE-ENTRAINMENT IN A BAGHOUSE
WHAT DOES IT COST?
By: J. G. Musgrove
Bechtel Power Corporation
P. 0. Box 2166
Houston, TX 77001
ABSTRACT
The re-entrainment of fly ash from the hoppers of utility baghouses is
acknowledged by baghouse suppliers but is considered by them to be inconse-
quential. A continuing concern about the extent and cost of re-entrainment
in a baghouse designed for a 550 MW utility generating station led to an
analysis of its costs.
The analysis hypothesizes the impact on baghouse operation that would be
caused by re-entrainment and estimates its present worth cost as a function
of increased draft loss across the baghouse compartments. An estimate is
also made for the cost of conducting a scale model test program to determine
the actual degree of re-entrainment to establish a break-even point for cost
of re-entrainment versus costs of conducting the model tests. The analysis
then suggests a method of determining the component costs of implementing
corrections to reduce re-entrainment. From the costs of correcting re-
entrainment, curves of equal cost are presented to suggest when re-entrain-
ment should be determined and corrected and when it should be ignored.
INTRODUCTION
A presentation of baghouse air flow model testing at the 4th Inter-
national Fabric Alternatives Forum (1) demonstrated the clear potential for
chaotic gas flow in the hoppers of bottom inlet baghouses. This chaotic flow
could easily lead to re-entrainment into the flue gas of ash stored in the
hopper and its subsequent redeposition onto the bag.
Bechtel had selected a bottom inlet baghouse for particulate collection
on a large utility boiler and was concerned about the potential for re-
entrainment in its design. Subsequent testing of the gas flow model for the
project indicated similar chaotic gas flow in the hoppers and therefore a
potential for significant re-entrainment problems. The baghouse supplier
repeatedly stated that re-entrainment would not be a problem but failed
either to provide convincing evidence or to perform, at no cost to our client,
additional testing to obtain convincing evidence.
Bechtel, therefore, proceeded to analyze the factors in re-entrainment
and to estimate the cost of re-entrainment to determine if client participa-
tion in further model testing was warranted.
201
-------
RE-ENTRAINMENT - DEFINITION, IMPACT AND DERIVATION OF COST
Bechtel's analysis of the cost of re-entrainment began with defining re-
entrainment as that portion of particulate material stored in the hopper at
the start of a filtering cycle which again enters the flue gas stream and is
recollected by being deposited on the bag surface. Therefore, 10% re-entrain-
ment means that 10% of the material stored in a hopper at the start of the
filtering cycle will be deposited on the bag and add to the filter cake. It
was also defined that a 10% increase in material on the bag produced a 10%
greater pressure drop across the bag surface. These definitions permitted the
analysis to calculate the cost of re-entrainment as a function of draft loss,
a characteristic for which a dollar value had already been obtained (2).
Table I, Parameters of Baghouse Re-entrainment for R=10%, presents the
values for selected baghouse parameters (ash in the hopper, ash on the bag,
compartment AP) assuming the recommended maximum AP prior to bag cleaning is
5 in. w.g. and that the amount of re-entrainment is 10%. Table II, Percent
Draft Loss from Re-entrainment, R=10%, presents, for 10% re-entrainment, the
percentage of compartment draft loss during a cleaning cycle that is due to
re-entrained ash from the hopper.
In all cases, it is assumed one unit of ash enters the compartment during
each cycle, and the Cycle Number is the number of reverse air bag cleaning
cycles since the latest hopper evacuation cycle.
From Table II it can be extrapolated that for any cycle, n, of bag
filtering and cleaning and for an assumed value of re-entrainment, R
(%, expressed decimally), the draft loss in that cycle, W, due to the re-
entrained material may be expressed as:
W = (n-l)R/4 (1)
Similarly, for a design with maximum recommended compartment draft loss,
P (in. w.g.), and a penalty for draft loss, V ($/in. w.g.), the cost of re-
entrainment, C ($), may be expressed as:
CR = PMV(n-l)R/4 (2)
Figure 1, Cost of Re-entrainment, C , presents a family of curves for C
based on the number of bag cleaning cycles between hopper evacuation cycles,
the assumed percent re-entrainment and the product of the draft loss penalty
and design maximum compartment draft loss. From this Figure any customer
which has a draft penalty value, V, and a proposed baghouse design AP, P ,
can determine the cost of re-entrainment for various numbers of bag cleaning
cycles between hopper evacuation cycles.
202
-------
FACTORS IN THE COST OF CORRECTING RE-ENTRAINMENT
Once the cost of uncorrected re-entrainment has been established it
follows that one should determine the cost of correcting the re-entrainment to
determine if correction is economically justified. The costs of correcting
re-entrainment are divided into three categories:
o The cost of a scale model test to determine the extent of re-
entrainment and the selection of a design modification that reduces
re-entrainment.
o The cost of implementation of that modification in full scale.
o The additional draft loss such a modification may itself create.
The cost of building and testing a scale model, C ($), has a fixed cost
based on building the model and a variable cost based on the number of test
configurations examined. For a quarter scale model the fixed cost would be
$50,000 and the variable cost would be $10,000 per configuration (3).
Figure 2, Cost of Model Testing, presents the cost of model testing versus the
number of test configurations examined. The superposition of percent re-
entrainment on number of test configurations follows from the hypothesis that
if one test indicated less than 5% re-entrainment, no further tests would be
conducted, if 10% were indicated a second configuration would be examined to
attempt to reduce re-entrainment to less than 5%, and so on. Therefore, model
test costs as a function of re-entrainment, R, are:
CM = $50,000 + $200,OOOR (3)
The additional costs of correcting re-entrainment, the implementation,
installation and draft loss of a correction have been estimated as follows:
The cost of fabricating and installing a design modification, C ($),
was the product of the cost to modify each hopper, C ($), and the number of
hoppers to be modified, H. C was presumed to range from $1,000 per percent
re-entrainment to $16,000 per percent re-entrainment. These costs were sub-
sequently substantiated by a supplier (4) as reasonable for the design modifi-
cations considered.
CH x H = CIM = $100,OOOR to $1,600,OOOR (4)
The cost of draft loss associated with a design modification, C ($), was
assumed to be 0.5 in. w.g. to correct 25% re-entrainment.
C = $2VR (5)
203
-------
This, too, was confirmed as reasonable by one supplier (4). In order to
assure that all costs are in constant dollars, the capital costs, CM and CIM,
must be multiplied by a present worth factor for capital, here assumed to
be 1.6, to obtain all dollars evaluated in present worth.
CM = $80,000 + $320,OOOR (6)
CIM = $160,OOOR to $2,560,OOOR (7)
Therefore, the total cost of determining and implementing a design modi-
fication to correct re-entrainment, CCR ($), is the sum of CM, CIM, and C^:
C^ = $80,000 + $320,OOOR + $160,OOOR + $2VR (8)
CR
to
C.,-. = $80,000 + $320,OOOR + $2,560,OOOR + $2VR (9)
CR
DECISION CURVES FOR CORRECTING RE-ENTRAINMENT
While it is beneficial to know the cost of re-entrainment for a given
baghouse design and set of economic evaluation factors and the cost of
correcting that re-entrainment, it would be most beneficial to develop curves
of equal cost, C = C , such that one could readily determine if testing and
correction were warranted.
CR = CCR (10)
P V(n-l)R/4 = 80,000 + 2VR + 320,OOOR + C , ,
R = 320,000
((n-l)PM-8)V - 1,280,000 -
(12)
or
n = 320,000 + R(V(8+PM) + 1,280,000 + CTM) ,. ,.
_ M. _ IM (13)
Figure 3, Re-entrainment Decision Curve, C =C H=$160,OOOR, through
Figure 8, Re-entrainment Decision Curve, C =C HS$2,560,OOOR, present the
curves of equalization for varying values or PM, for values of V of
$250,000/in., $500,000/in. , $750,000/in. , and ?l,000,000/in. , respectively,
for values of C of $160,OOOR to $2,560,OOOR.
204
-------
CONCLUSION
From the family of curves, presented in Figures 3 through 8, a decision
regarding whether or not to pursue model testing and corrective modifications
to decrease re-entrainment can readily be made. This decision will be based
on accurate estimates of the cost of re-entrainment and reasonable estimates
of the costs of determining and implementing corrective modifications.
Determination of the optimum duration of time between hopper evacuation cycles
may be made by calculating the time per cycle, t (min.), to reach P ,
according to the equation (5).
PM = k C. t v1<8 ,, .,
M i (14)
7000
or
t = PM 7000 (15)
C. kv1'8
i
ENDNOTES
1. Bowen, P. The Modeling of A Baghouse. In: Proceedings of the 4th
International Fabric Alternatives Forum, Phoenix, Arizona, December 5-6,
1979.
2. Musgrove, J. G. A Survey of Particulate Collection Divices for Coal-
Fired Boilers. In: Proceedings of the International Coal Utilization
Convention, Houston, Texas, October 17-19, 1978.
3. Musgrove, J. G. In: Conference Notes Number 13051-133, Bechtel Power
Corporation, April 23, 1980.
4. Zierak, S. Modified Hopper for Re-entrainment Study. In: Letter to
John Musgrove, September 4, 1980.
5. Jensen, R. M. Baghouse Bid Evaluation. In: Proceedings of the 2nd
International Fabric Alternatives Forum, Denver, Colorado, July 27-28,
1977.
205
-------
TABLE I - PARAMETERS OF BAGHOUSE RE-ENTRAINMENT FOR R = 10%
Cycle
Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Ash in
Hopper at
Start of
Cycle
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
11.0
12.0
13.0
14.0
Ash in
Hopper at
End of
Cycle
0.0
0.9
1.8
2.7
3.6
4.5
5.4
6.3
7.2
8.1
9.0
9.9
10.8
11.7
12.6
Ash in
Hopper
After Bag
Cleaning
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
11.0
12.0
13.0
14.0
15.0
Ash on
Bag from
In Flow
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
Ash on
Bag from
Hopper
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
Total
Ash
on Bag
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2.0
2.1
2.2
2.3
2.4
Maximum
AP This
Cycle
5.0
5.5
6.0
6.5
7.0
7.5
8.0
8.5
9.0
9.5
10.0
10.5
11.0
11.5
12.0
Max. AP
Due to Ash
from Hopper
This Cycle
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
5.5
6.0
6.5
7.0
Avg . AP
Due to Ash
from Hopper,
This Cycle
0.0
0.25
0.5
0.75
1.0
1.25
1.5
1.75
2.0
2.25
2.5
2.75
3.0
3.25
3.5
Avg. AP
Due to Ash
from Hopper,
All Cycles
0.0
0.125
0.25
0.375
0.5
0.625
0.75
0.875
1.0
1.125
1.25
1.375
1.5
1.625
1.75
K)
O
TABLE II - PERCENT DRAFT LOSS FROM RE-ENTRAINMENT, R = 10%
Cycle Number
AP due to
Re-entrainment
AP/5.0 in.w.g.
Avg. AP for
All Cycles
Avg. APD/5.0 iwg
K.
Percent AP
due to R
1
0
0
0
0
0
2
.5
0.1
.125
.025
2.5
3
1.0
0.2
.25
.05
5.0
4
1.5
0.3
.375
.075
7.5
5
2.0
0.4
.5
.10
10.0
6
2.5
0.5
.625
.125
12.5
7
3.0
0.6
.75
.15
15.0
8
3.5
0.7
.875
.175
17.5
9
4.0
0.8
1.0
.20
20.0
10
4.5
0.9
1.125
.225
22.5
11
5.0
1.0
1.25
.250
25.0
12
5.5
1.1
1.375
.275
27.5
13
6.0
1.2
1.5
.30
30.0
14
6.5
1.3
1.625
.325
32.5
15
7.0
1.4
1.75
.35
35.0
-------
FIGURE 1
COST OF RE-ENTRAINMENT, CR
a
I
oc
*' / /
''?<*'
>'/'?
'//
/ s *
'' /
.' /
X
X xxx
^
>''
^
..--•""'^
^^ «*•
/ / t
/ ' /
' f /
/ /
*' / ,1
y /
^ >X /
/ y
/ /
//
s
/ ,X
/
/ ,
/
.,/
**'
.,-•-
1 / '
1 / 1
1 '
' !\
/ 1
/ 1
/ /
•' /
v' /
' *>/
/
s
-vx
.'
^^^"^
1 1
1 1 1
1 1 1
1 1
1 1
1 1
1 1
1 1
1 1
f 1
/
1
/
/
/
/
4
/
/
^
5.0 4.0 3.0 2.0
MAXIMUM VALUE Of PMV - DOLLARS x ID6
1.0
10 15
RE-ENTRAINMENT, R - PERCENT
20
FIGURE 2
COST OF MODEL TESTING
140
120 •
100
80
60
^ 40
UI
a
o
RE-ENTRAINMENT
2 3 4
NUMBER OF TEST CONFIGURATIONS
207
-------
FIGURE 3
RE-ENTRAINMENT DECISION CURVE
C|M = CHH - S160.000R
7.0 60 50 4.0
MAXIMUM RECOMMENDED COMPARTMENT AP. P
tO 15 20
RE-ENTRAINMENT. R -PERCENT
FIGURE 4
RE-ENTRAINMENT DECISION CURVE
9 10 II 12 13
1.0 6.0 5.0 4.0
MAXIMUM RECOMMENDED COMPARTMENT AP. PM - in.*
10 15 20 25
RE-ENTRAINMENT. R -PERCENT
208
-------
FIGURES
RE ENTRAPMENT DECISION CURVE
C,M = CHH = $640,OOOR
7.0 6.0 5.0 4.0 3.0
MAXIMUM RECOMMENDED COMPARTMENT AP, PM - in.n.g.
10 IS 20
RE-ENTRAIHMENT, R -PERCENT
FIGURES
RE-ENTRAINMENT DECISION CURVE
C,M = CHH=$1,280,OOOR
10 11 12 13 14
7.0 6.0 5.0 4.0 30
MAXIMUM RECOMMENDED COMPARTMENT AP,PM - in.n.g.
10 15 20
RE-ENTRAINMENT. R - PERCENT
209
-------
FIGURE?
RE ENTRAPMENT DECISION CURVE
C|M = CHH=$1.920,OOOR
10 II U 13 14
6.0 5.0 1.0 3.0
lAXIMUM RECOMMENDED COMPARTMENT AP, PM - '"•">•«•
10 15 20
RE-ENTRAINMENT, R - PERCENT
FIGURES
RE-ENTRAINMENT DECISION CURVE
C,M = CHH - $2,560,OOOR
10 11 12 13 14
7.0 6.0 S.O 4.0 3.0
MAXIMUM RECOMMENDED COMPARTMENT AP. PM - in.».|
10 IS 20
RE ENTRAPMENT. ft - PERCENT
210
-------
WHY PERFORM MODEL STUDY OF FABRIC FILTER COLLECTOR?
By: William T. Langan
Buell Emission Control Division
Envirotech Corporation
Lebanon, Pennsylvania 17042
Dr. Norman Z. Shilling
Buell Emission Control Division
Envirotech Corporation
Lebanon, Pennsylvania 17042
William A. Van Kleunen
Buell Emission Control Division
Envirotech Corporation
Lebanon, Pennsylvania 17042
Owen F. Fortune
Buell Emission Control Division
Envirotech Corporation
Lebanon, Pennsylvania 17042
ABSTRACT
The industry historically has performed model studies of electrostatic
precipitators to achieve good gas distribution in the field unit. Standards
for electrostatic precipitators are well developed, and field experience has
proven such modeling is important to achieve high collection efficiency for
these particulate collecting equipment. Differences in model study evalua-
tion of fabric filters and precipitators are contrasted in light of the funda-
mental differences between the two types of particulate control devices.
The performance criteria required for satisfactory fabric filter opera-
tion are presented. These criteria are translated into goals to be met
through a model study effort.
Specific results from model studies of several major utility facilities
are reviewed. The impact of the model study results upon the design features
of fabric filter collectors are described. Field data verification of good
particle size distribution within compartments of operational units are
presented to demonstrate the results of proper design for gas flow/particulate
distribution.
INTRODUCTION
Precipitator model studies are performed to obtain the high quality gas
flow distribution required to achieve low outlet emissions. The model
construction is typically one-sixteenth scale of the field unit to ensure
turbulent flow. The ductwork design and gas flow devices are selected to
achieve field gas flow distribution of I.G.C.I, quality. These requirements
are: (1) no velocity measurement is to deviate more than +40% from the aver-
age velocity, and (2) 85% of the velocity measurements are to be within +25%
211
-------
of the average velocity. The predicted influence of gas flow distribution
upon precipitator collection efficiency is presented in Figure 1. As this
figure illustrates, significant performance improvements can be achieved by
reducing high RMS gas flow distributions to I.G.C.I, gas distribution quality.
The cost effectiveness of attempting to obtain even lower RMS gas flow distri-
bution values must be seriously questioned through the evaluation of the
amount of steel for the required gas flow devices, the scale-up correlation
between the field and the model, and the estimated performance enhancement.
Gas flow distribution must be uniform and gas treatment velocity must be low
(typically less than 4.5 ft/sec., depending upon the flyash conductivity.) to
minimize re-entrainment of captured flyash.
The precipitator model study is typically conducted for a steady state
condition. The only transient conditions investigated in a precipitator model
study are dust desposition and sweepage studies. This is fundamentally
different than the fabric filter collector which normally operates in
transient conditions due to the cleaning modes.
To establish criteria that should be utilized for a fabric filter model
study, an overview of potential field problems and their causes are presented
in Table 1. To ensure successful field operation of the fabric filter, the
model study effort should evaluate pressure losses in the ductwork, distribu-
tion of gas flow and/or particulate to baghouse compartments, distribution of
gas flow and/or particulate within baghouse compartments, and distribution of
reverse air within baghouse compartments.
MODEL STUDY APPROACH
The fabric filter collector should be designed for good flow distribution
to ensure high reliability and performance. The design guidelines for good
flow distribution are:
• Relatively constant transport velocity (3000 to 4500 feet per
minute).
• Flow impedance at compartment inlet, to compensate for various
operating modes.
• Low hopper entrance velocity.
• Reverse air distribution over the entire compartment.
These fabric filter designs are based upon extensive fabric filter field
experience and several fabric filter model studies, including: Tennessee
Valley Authority, Shawnee Steam Plant, Units No. 1 through 10; Public Service
of Colorado, Cherokee Units No. 2 and 3; Baltimore Gas & Electric Company,
Crane Units No. 1 and 2; Arizona Public Service Company, Four Corners Units
No. 4 and 5; Plains Electric Generating and Transmission Coop, Inc., Esca-
lante Generating Station, Unit 1, and Board of Light and Power, Marquette,
Michigan, Shiras Unit No. 3.
The model study investigates the transient conditions associated with
212
-------
the fabric filter operation. Typically, the following four conditions are
investigated:
A. "Pressure Limit" or "Time" mode cleaning cycle, at a point
in time when all compartments have just been cleaned. At
this time, all compartments will evidence essentially iden-
tical Pressure drop characteristics corresponding to a 2-3
in. W.G. filter media/residual filter cake pressure drop
characteristic.
B. "Pressure Limit" or "Time" mode cleaning cycle, at a point
in time when all compartments are about to be cleaned. At
this time, all compartments will evidence an essentially
identical filter media/filter cake pressure drop character-
istic of 3-5 in. W.G.
C. Also in "Pressure Limit" or "Time" mode cleaning cycle, but
at a point in time when the first compartment has just been
cleaned. In this short duration situation, one compartment
will demonstrate a minimum pressure drop characteristic,
while all other operating compartments demonstrate the maxi-
mum pressure drop characteristic. This one compartment will
then experience a maximum gas flow rate excursion.
D. In "Integrated" mode of cleaning cycle. At each repetitive
situation when a compartment is being cleaned, there will be
one compartment (the next compartment to be cleaned) which
will evidence a maximum pressure drop characteristic, while
one compartment (the most recent compartment cleaned) will
evidence a minimum pressure drop characteristic.
Model test operating Mode "C" represents an instantaneous situation,
before any equalization of filter cake deposition has occurred. This is
also an exaggerated situation because it includes the maximum amount of
filter cake deposition in all other compartments. Although this situation is
of very short duration during baghouse operation, it does demonstrate excur-
sions of gas flow rate which are approached. The magnitude of the gas flow
rate excursions illustrate why attention must be given to this parameter of
baghouse operation.
Model test operating Modes "A" and "B" represent operating situations
which reflect the natural occurring equalization of filter cake deposition.
Model study test experience shows that there is minimal difference of the
division of gas flow to operating baghouse compartments for either just
cleaned filter bags, or bags with maximum filter cake deposition. The model
program investigates identical simulated pressure drop in all baghouse
compartments so that the effect of the baghouse design configuration may be
investigated, apart from the effect of the filter bag/filter cake pressure
drop.
The "Integrated" mode bag cleaning cycle, as illustrated by model test
operating Mode "D", illustrates reduced gas flow rate excursion in the just
213
-------
cleaned compartment. .However, this is again a short duration situation,
before any natural equalization of filter cake deposition.
The operation of the 1:16 scale model at scaled flow rates of ambient
temperature air allows the model to be operated in fully-developed turbulent
flow regime (as defined by a Reynolds number >4000), at all pertinent model
locations, and at all load gas flow conditions. This is a necessary require-
ment to insure dynamic similitude between the model and prototype installation
operating conditions.
The division of volumetric flow rate between baghouse compartments is
determined by measuring the pressure drop created by perforated plate, which
simulates the filter media/filter cake pressure drop, and for which the
pressure drop vs. volumetric flow rate characteristic is known.
Investigations of gas flow distribution within baghouse compartments are
performed utilizing flow visualization techniques. This procedure permits
optimization of such parameters as: Baghouse compartment isolation valve
size and location, internal baffles and gas flow directing vanes within the
baghouse compartment hoppers, and the configuration of baghouse inlet and
outlet gas flow manifolds. A superior procedure for performing flow visuali-
zation investigations is the Neutrally-Buoyant Helium Bubble technique.
Gas flow distribution tests are performed in the high-velocity transport
ductwork systems. Typical locations in the model for these investigations
are: Air heater outlet ductwork, air heater outlet crossover duct, entrance
to baghouse, discharge from baghouse, baghouse bypass ductwork system, and
inlets to I.D. fan.
Incremental pressure drop surveys are performed in the model for pertin-
ent model configurations of simulated filter media/filter cake pressure drop,
number and arrangement of operating baghouse compartments and operating load
flue gas flow rates. Model data consists of static pressure measurements and
the contributing effect of the computed velocity pressure at the model test
volumetric flow rate. Model test results are projected to the full-scale
installation operating conditions. Typical locations for incremental model
measurements are: Discharge from air heaters, baghouse system inlet perform-
ance test port location or baghouse inlet, within baghouse compartment
hoppers (entrance to baghouse compartments), discharge from baghouse compart-
ments (downstream of simulated filter media/filter cake pressure drop),
baghouse outlet or baghouse system outlet test port location, and inlets to
I.D. fan.
These measurements are performed at both normal baghouse operation and
baghouse bypass system operation.
Particulate deposition and distribution tests are performed in the model
for pertinent baghouse configurations and expected flue gas flow rates. In
assessing these parameters, the testing procedure employs a test particulate
of appropriate density and particle size distribution characteristics,
to simulate the dynamics of particulate at prototype operating conditions.
Model tests include observations of dust deposition within the transport
duct and within the baghouse
214
-------
for the full range of expected flue gas flow rates. Model Investigations are
also performed to survey particulate deposition within the baghouse inlet
manifold and in the operating baghouse compartment hoppers.
RESULTS
The design areas modified, based upon model study results, include:
hopper baffling, outlet poppet valves, inlet poppet valves, bypass poppet
valves, and bypass ductwork.
The impact of proper hopper baffling is illustrated by a comparison of
Figures 2 and 3. Clearly, proper hopper baffling, as shown by Figure 3,
minimizes the probability of detrimental dust re-entrainment.
The impact of proper poppet sizing is illustrated by a comparison of the
compartment gas flow distribution shown by Tables 2 and 3. By modification
of the outlet poppet dimensions, the gas flow distribution RMS was improved
significantly.
Particulate particle size distribution measurements have been performed
for filter cake obtained from an operating baghouse. Filter cake samples
were extracted from various operating baghouse compartments, as well as from
various locations within an operating compartment. The purpose of these
determinations was to verify that the distribution of particulate is uniform
in a properly designed baghouse. Table 4 illustrates particulate particle
size distribution for various compartments. Table 5 illustrates the particle
size distribution of particulate within a compartment. These data verify that
excellent particulate distribution was achieved in the operating baghouse.
SUMMARY
Early model studies have been important in impacting the fabric filter
design to ensure high performance levels. The need to perform model studies
for future plants decreases since the results of past model studies and.field
experience are incorporated in Engineering design standards. The focus of
future fabric filter model studies will be more cost-effective designs, as
well as even lower maintenance designs.
ILLUSTRATIONS
Figure 1 - Precipitator efficiency vs. % RMS
Figure 2 - Hopper without baffles
Figure 3 - Hopper with baffles
Table 1 - Potential fabric filter field problems and associated causes
Table 2 - Division of gas flow between compartments - not optimized
Table 3 - Division of gas flow between compartments - optimized
Table 4 - Particle size distribution - between operating baghouse compartments
Table 5 - Particle size distribution - within operating baghouse compartment.
215
-------
99.99 - -
80
Predicti9n Of Effect Of Flow
Uniformity On Precipitator Efficiency
Figure 1
216
-------
Figures
Flow Distribution Within
Baghouse Hopper Illustra-
ted By Flow Visualization
217
-------
Problem
Excessive
Pressure Drop
oo
Short Bag Life
Excessive
Emissions
Fabric Filter
Potential Cause
• Pressure Losses In Ductwork
• Maldistribution Of Gas Flow And/Or
Particulate To Baghouse Compartment
• Maldistribution Of Gas Flow And/Or
Particulate Within Baghouse Compartment
• Maldistribution Of Reverse Air Within
Baghouse Compartment
• Excessive Sub-Dewpoint Operation Of Unit
(Normal Operation/Reverse Air Cleaning)
• Abrasion Due To Improper Inlet Conditions
• Excessive Sub-Dewpoint Operation Of Unit
(Normal Operation/Reverse Air Cleaning)
• Maldistribution Of Gas Flow And/Or
Particulate To/Within Baghouse Compartment
• Ineffective Utilization Of Available Cloth
Area Due To Flow/Particulate Maldistribu-
tion To/Within Baghouse Compartment
• Damaged Bags Resulting From Mechanical Or
Operational Problem
Table 1
-------
All Outlet Poppets At 5'-6" Diameter
Simulated Bag ATP = 2.0IN.W.G.
8 Of 10 Compartments In Operation With Clean Bags
0>
D>
2
5
<
E
2
UL
c
Q
^
05
>
o>
Q
Compartment
1
2
3
4
5
6
7
8
9
10
Test
No. 1
Off
Off
+ 8.80
+ 7.84
+ 3.68
+ 2.64
-4.00
-5.12
-7.52
-6.32
Test
No. 2
Off
+ 19.04
Off
+ 9.60
-0.56
-1.76
-1.76
-5.04
-9.76
-9.76
Test
No. 3
+ 8.48
+ 11.36
+ 3.44
+ 4.48
Off
Off
-5.20
-6.32
-8.72
-7.52
Test
No. 4
+ 5.76
+ 7.68
-0.32
+ 1.76
-0.32
-2.24
Off
-5.60
Off
-6.72
Test
No. 5
+ 6.48
+ 10.40
+ 1.44
+ 1.44
+ 1.44
+ 0.40
-4.96
-16.64
Off
Off
RMS 6.10% 9.42% 7.35% 4.70% 7.51%
Table 2
Outlet Poppets In Compartment 1 & 2 - 4'-0"
Outlet Poppets In Compartment 3 & 4 - 4'-6"
Outlet Poppets In Compartments 5-10 - Unchanged 5'-6"
Simulated Bag ATP= 2.0IN.W.G.
8 Of 10 Compartments In Operation With Clean Bags
CD
O)
CO
CD
<
E
o
£
c
o
CO
CD
O
Compartment
1
2
3
4
5
6
7
8
9
10
Test
No. 6
Off
Off
+ 4.96
+ 4.96
+ 4.96
+ 2.80
-3.84
-3.84
-6.16
-3.84
Test
No. 7
Off
+ 7.28
Off
+ 5.28
+ 3.12
+ 3.12
-1.28
-5.84
-7.04
-4.64
Test
No. 8
-1.52
+ 2.88
+ 2.88
+ 4.00
Off
Off
-0.32
-1.52
-3.76
-2.64
Test
No. 9
-2.48
-0.40
+ 0.72
+ 0.72
+ 3.76
+ 2.72
Off
-1.44
Off
-3.60
Test
No. 10
-3.20
-1.04
+ 1.12
-1.04
+ 5.20
+ 3.20
-1.04
-3.20
Off
Off
RMS 4.52% 5.09% 2.70% 2.34% 2.79%
Tables
219
-------
Fly Ash Samples From Filter Bags
In Various Operating Baghouse Compartments
Baghouse Mean Particle
Compartment/ Diameter, D,
Sample I. D. Microns
1/1 5.7
1/2 II
1/3 H
2/1 6.0
2/2 5.8
2/3 5.9
3/1 6.8
3/2 6.1
3/3 5.8
4/1 5.9
4/2 5.9
4/3 5.7
5/1 6.2
5/2 6.2
5/3 5.5
6/1 5.8
6/2 6.2
6/3 5.7
7/1 6.2
7/2 6.6
7/3 6.4
8/1 6.1
8/2 5.9
8/3 6.1
9/1 6.1
9/2 6.0
9/3 6.6
10/1 6.5
^3 Table 4 l;i
Fly Ash Samples From
Filter Bags In Operating
Baghouse Compartment
Baghouse Mean Particle
Compartment/ Diameter, D,
Sample I. D. Microns
7/1 6.4
7/2 6.7
7/3 6.9
7/4 6.6
7/5 6.6
7/6 6.4
7/7 6.6
7/8 6.3
7/9 6.5
7/10 8.0
7/11 7.0
7/12 6.3
7/13 6.3
7/14 7.0
7/15 Table 5 6.1
220
-------
EXPERIENCES OF A SMALL INSULATION MANUFACTURER IN MAINTAINING
COMPLIANCE WITH AIR POLLUTION CONTROL REGULATIONS
By: R. L. Hawks
PEDCo Environmental, Inc.
505 South Duke Street, Suite 503
Durham, North Carolina 27701
ABSTRACT
The implementation of stringent regulations has created a hardship on
small businesses who have little technical expertise in the area of environ-
mental control. This paper presents the experiences of a small blown-insula-
tion producer that has been in operation in Southeast United States since
1976. Numerous difficulties with air pollution control equipment have
hindered this plant from achieving full production capacity and have led to
a tiring series of sparring exercises with local regulatory officials. Many
of the problems could have been avoided had the equipment been designed
better and had the supplier given better instructions to company management
responsible for its operation. The author contends that the equipment sup-
pliers have a special obligation to educate small companies regarding impor-
tant design factors and operation and maintenance procedures. The author
further believes that regulatory agencies should provide a thorough permit
review of smaller pollution sources and assist them in developing technical
expertise in air pollution control.
INTRODUCTION
The enactment of the Clean Air Act in 1970 has resulted in major eco-
nomic and technical impacts on American business. The setting and implemen-
tation of emission limits has created a particular hardship for small
business.
Most small businesses do not initially have the technical ability to
design, operate, or maintain air pollution control equipment. Nor do most
of these small sources have technical staffs versed in the disciplines of
industrial ventilation or control equipment design. Therefore, they must
rely on local regulatory agencies, control equipment manufacturers, and
vendors to provide them with knowledge regarding air pollution control
requirements.
Historically, the regulatory agencies have not provided technical
assistance in design, operation, or maintenance, either because of legal
restraints or their own limited manpower and experience. Thus, the design
of abatement equipment at small sources has been based on limited experi-
ence recommendations of vendors, or in some cases, guesswork. Unfortu-
nately the equipment installed in this manner has been susceptible to
operation and maintenance problems, and most importantly, the small firms
have frequently been subjected to compliance penalties due to excess emis-
sions.
221
-------
This paper describes the experiences of one small business in its
attempt to maintain compliance with emission limits with improperly designed
air pollution control equipment and process variables that were not fully
appreciated.
PLANT DESCRIPTION
The plant, a conventional mineral wool manufacturer, uses a cupola and
blow chamber to produce blowing wool for insulation. There are two
32-in.-diameter water-cooled cupolas. These cupolas are charged with coke,
trap rock, copper and steel slags, and limestone. The flue gases from the
cupolas are diluted with ambient air and exhausted by an induced-draft fan
through a reverse-air/shaker-type fabric filter.
This privately owned plant employs approximately 100 persons and oper-
ates on a 24-h schedule for periods up to 60 days without cupola outages for
major maintenance. Staffing consists of production personnel, maintenance
crews, and management. The plant does not have sufficient operating funds
to employ a staff engineer.
CUPOLA PARTICIPATE CONTROL EXPERIENCES
Construction of the plant was completed in August 1976, and the firm
filed a construction permit application with the local air pollution agency
to operate two cupolas and two blow chambers. Because the uncontrolled
particulate emissions from these sources would not comply with emission
limitations set by the agency, the plant specified low-energy scrubbers to
control blow chamber emissions and fabric filter systems to control cupola
emissions.
The fabric filter specified for each cupola was a reverse-air/shaker
design with a total cloth area of 16,000 ft2 in four isolatable compart-
ments. The filter was designed to operate at an air volume of 20,000 acfm
and at an air-to-cloth ratio of 1.25 acfm/ft (gross). The system was equip-
ped with dilution air cooling to reduce gas temperatures to below 500°F and
with silicon/graphite-finished fiberglass bags. The plant installed a
limestone precoat system to increase collection efficiency and remove sulfur
dioxide (S02) emissions from the gas stream, which later proved to be an
unnecessary and costly investment.
The local regulatory agency reviewed the filter design and issued an
operating permit. The agency's review was superficial and considered only
the air-to-cloth ratio and the compatibility of the fabric with the expected
gas temperature range. Also, it was assumed that emissions would be less
than 1.0 Ib/h, based on an expected collection efficiency of 99.5 percent.
The agency did not evaluate the variability of the cupola operation, composi-
tion of charge materials, or potential for filter failure due to process
influences. The regulatory agency included a stack test stipulation in the
permit, which required an EPA Method 5 test 30 days after plant startup.
A local contractor installed the filter under the direction of the
filter manufacturer's field representative. The erection period was
222
-------
lengthy and plagued by bad weather and inexperienced installation person-
nel .
After a shakedown period the plant conducted the required particulate
stack test. Emissions were determined to be 9.85 Ib/h, which complied with
the regulatory limit of 13.0 Ib/h. However, the efficiency of the fabric
filter on line 1 was only 91 percent, much less than the expected 99.5 per-
cent.
Shortly after the initial particulate stack test, the filter static
pressure drop began to increase sharply. At the direction of the manu-
facturer, the plant increased the intensity and frequency of cleaning. This
vigorous cleaning soon caused failure of the fiberglass bags and increased
filter maintenance, which resulted in periods of excess emissions. The
reason for the abnormally high static pressure drop was not determined, only
the symptom was treated.
A review of the particulate stack test showed that the average exit gas
temperature was lower than expected (136°F) and below the gas stream mois-
ture dewpoint of 160°F. A check of the inlet gas stream temperature con-
firmed the low temperature and suggested that this was a typical operating
pattern for the cupola. The cupola exhaust is diluted with ambient air and
quenched from approximately 1200°F. The amount of dilution air was not
measured, nor was any provision made to regulate the amount. This unregu-
lated dilution allowed the system to operate far below the design tempera-
ture of 450°F. It is believed that the fabric filter manufacturer was not
aware of this poor temperature control system during the design phase.
Further measurement indicated that the cupola gas stream had a predict-
able temperature profile that resulted in the flue gases passing through the
dewpoint each 20 min (Figure 1). The temperature increases as the cupola
charge is consumed and then decreases sharply when a batch charge is added.
The baghouse inlet temperature ranged between 130° and 230°F over this
period. Each time the system dropped below the dewpoint, the limestone
precoat on the bags would adsorb moisture; when the gas temperature in-
creased, this precoat hardened to a cake that could not be removed easily
without damaging the fabric (Figure 2). Due to the resulting high static
pressure drop, cupola production capacity was limited.
Once the reason for the high pressure drop was identified, the plant
reduced the use of limestone and limited the dilution air so as to maintain
the temperature above 250°F. These changes reduced the static pressure drop
of the system. Cleaning intensity also was reduced; however, the rate of
bag failures continued to increase. Based on the assumption that the fail-
ures resulted from previous damage, a complete bag change was made, but the
new bags continued to fail in unexpected numbers.
It must be noted that plant personnel were inexperienced in the care
and installation of fiberglass bags. As a result, the bags were seriously
abused during installation. Some were tied in knots to allow access to
defective bags, some were dragged through the access doors, and others were
walked on (Figure 3). Discussion with plant personnel indicated that
223
-------
neither the filter manufacturer nor the bag supplier had provided instruc-
tions concerning the delicate handling required for fiberglass bags.
Mistreatment of the bags did not totally explain the high failure rate.
The bags deteriorated in the lower 25 percent of their length and lost
tensile strength. The failures did not appear to be associated with high
temperature damage, but was more representative of chemical damage.
The slags contain trace amounts of fluorides, which are emitted as
hydrogen fluoride under reducing conditions. These fluorides were eventu-
ally determined to be partially responsible for the failure of the fiber-
glass bags. The plant purchases slags from several sources, but does not
have the option to limit purchases based on fluoride content. The obvious
solution was to use a bag fabric resistant to fluorides. Because the
fabrics available for high-temperature fluoride exposure were either expen-
sive or subject to attack by the sulfur oxides in the gas stream, the plant
eventually was forced to limit gas temperatures to a maximum of 350°F so
that polyester bags could be used. These bags have provided good life under
these conditions.
CUPOLA HYDROGEN SULFIDE EXPERIENCES
The maintenance of the fabric filter has been a burdensome and hazard-
ous experience for the company because of the high levels of hydrogen sul-
fide present in the gas stream. This had not been anticipated by plant
personnel. The system was equipped with isolatable compartments to allow
on-line maintenance. Unfortunately, the butterfly dampers for compartment
isolation did not seal and no compartment air-purge system was provided. At-
tempts to conduct on-line maintenance have on two occassions resulted in
hospitalization of maintenance personnel.
Neighborhood complaints concerning odors forced the plant to determine
the level of its hydrogen sulfide (H2S) and S02 emissions. Six stack test
runs using EPA Methods 11 and 6 indicated maximum H2S emissions of 46.7 Ib/h
(500 ppm) and S02 emissions of 1.0 Ib/h. The H2S levels were highly varia-
ble and related to slag sulfur content. Individual test emission rates
varied between 16.2 and 66.05 Ib/h, and slag sulfur content varied between
1.84 and 3.02 percent. The State agency subsequently issued a notice of
violation based on a boundry line H2S concentration limit. Using worst case
meteorological conditions, the agency calculated that the ambient concentra-
tion would be exceeded slightly at the maximum tested H2S emission rate. The
company was successful in overturning this citation since the regulation was
vague regarding required emission limits.
The plant is now considering three options: 1) installation of posi-
tive sealing isolation dampers, 2) installation of a compartment air-purge
system, and 3) use of air line respirators during internal maintenance.
SUMMARY
In retrospect, the unfortunate experiences described could have been
avoided. Many of the operating problems were the direct result of design
224
-------
deficiencies in the fabric filter itself or in its application to the
process. The gas-temperature control system supplied by the plant did not
provide adequate control of dilution air. The fabric filter manufacturer's
isolation dampers operated poorly, and the filter medium was specified
without regard for flue gas composition. Also, the manufacturer did not
provide training or instruction in the installation or maintenance of the
filter bags. Many of the bag failures were traced to improper installation.
The regulatory agency's permit review staff chose not to make a
rigorous evaluation of the cupola fabric filter system which may have identi-
fied some of the problems. Furthermore, the agency's unbending enforcement
posture distracted plant management personnel at the very times where
efforts were being made to expeditiously resolve the technical problems.
225
-------
N3
N3
600
500
400
o
UJ
300
200
100
10
i CHARGE i
r~ CYCLE ~n
DESIGN TEMPERATURE
WITH GLASS BAGS
UPPER LIMIT FOR
POLYESTER BAGS
LOWER LIMIT FOR
POLYESTER BAGS
MOISTURE DEWPOINT
20 30 40 50
TIME MINUTES
60
70
80
90
Figure 1. Temperature profile of cupola emissions entering fabric filter showing influence
of charge cycle.
-------
Figure 2. Photograph showing limestone precoat cake that could not be
removed from the interior of fiberglass bags.
Figure 3. Photograph of typical treatment of fiberglass bags
during installation.
227
-------
ADVANCED FABRIC FILTER TECHNOLOGY
FOR DIFFICULT PARTICULATE EMISSIONS
By: Heinz P. Beutner, Ph.D., President
Interel Corporation
7108 South Alton Way, Bldg.M
Englewood, Colorado 80112
ABSTRACT
The paper discusses fabric filter technology developed by H. Luhr of
West Germany specifically for control of difficult submicron emissions, e.g.,
from metallurgical furnaces, electric arc furnaces, and glass and brick
kilns.
The baghouse uses a dust conditioning drum, located in the hopper in
place of a screw conveyor, to pretreat the gas entering the baghouse with
dust returned from the bags. This approach increases the average particle
size of the dust and reduces the required frequency of bag cleaning. The
drum also can serve to hold additives in the baghouse for removal of gaseous
components, such as HC1, HF, and S03. The baghouse utilizes horizontally
mounted, flat Twin-Bags and either low pressure reverse air or pulse jet
cleaning.
Another element of the Luhr technology is an air-to-gas heat exchanger,
which is designed to cool high temperature exhaust gases to the temperature
acceptable in the baghouse. The gas cooler, which is equipped with an auto-
matic cleaning mechanism, can be designed as an integral part of the baghouse
and dust conditioning system.
INTRODUCTION
Luhr Staubtechnik of Germany has achieved international recognition in
the pollution control field for the technology it pioneered and perfected in
the filtration of submicron dust emissions. Many references on a variety of
difficult dust applications testify to the success of this technology.
Interel Corporation is the U.S. licensee of Luhr.
Submicron dust emissions, such as metallurgical fumes and emissions
from glass furnaces, arc furnaces, incinerators, other high-temperature
sources, have long been recognized as difficult control problems. Typical
practice for fabric filtration of submicron fumes has been the application
of a shaker-type baghouse with very low air-to-cloth ratios, such as 1:1
or 1.5:1. In many cases, precoating of the fabric after each cleaning was
found to be necessary to prevent fabric blinding and to hold outlet dust
loading within acceptable limits.
The Luhr technology utilizes several specialized approaches to the
control of furnace emissions, including:
228
-------
+ a fabric filter system with horizontally mounted flat bags
(Twin-Bags) and either reverse air or pulse jet cleaning
+ dust conditioning and supplementary dust addition in a
conditioning drum
+ gas cooling in a gas-to-air heat exchanger with heat recovery
As a result of application of the principle of dust conditioning, the
Luhr baghouse can filter submicron furnace emissions at low pressure drop and
air-to-cloth ratios ranging from 3:1 to 6:1, depending on the specific
application. The high air-to-cloth ratio reduces the fabric area and size of
the baghouse required for a given application; the lower fabric area reduces
bag replacement cost. Combined with credits for heat recovery, substantial
operating cost savings are often feasible.
Baghouse Design
The Luhr baghouse is based on dust collection on the outside surface of
the bag, which is held by a support cage. The baghouse utilizes flat "Twin-
Bags" up to 6.5 ft. long with fabric areas to 8.3 ft2 each.
A unique aspect of the Luhr baghouse design is the horizontal arrange-
ment of the Twin-Bags as shown conceptually in Fig. 1. This design allows
for easy bag replacement from the front walkway through full-sized access
doors to the clean air space. The design also permits easy visual detection
of leaking bags. Also, it is easy to temporarily block the clean gas exit of
a leaking bag,
The Twin-Bags are sealed to a perforated wall by means of a spring-loaded
snap mechanism. The bag attachment is free of stress from the weight of the
bag, bag cage, and dust compared with typical hanging systems. This design
results in improved baglife. At the rear wall, the bag cage is held by a
simple support.
The Twin-Bags are cleaned either by reverse air flow or by pulse jet.
Reverse air is supplied by a traveling plenum, which is positioned in front
of one vertical row of bags at a time while the clean air outlets of the two
adjacent rows of bags are blocked off to prevent redeposition of the dust
removed from the center row (type MWF filter). The flow of reverse air is
950 cfm at 20" WG for each 116 ft2 of fabric area, equal to an air-to-cloth
ratio of 8.2:1.
The pulse-jet design uses stationary air pipes with compressed air
supplied to one or several vertical rows of bags at a time (type DF filter).
The reverse air and the pulse-jet systems are also available with off-line
cleaning.
Dust Conditioning
The dust conditioning drum, which has been developed by Luhr over a
period of years, is an effective means of dust preconditioning. Installed
229
-------
Dirty Gas
Dust Discharge
FIGURE 1
LUHR TWIN-BAG FABRIC FILTER — GAS FLOW ENTERING FROM ABOVE AND BACK
230
-------
generally in the gas inlet plenum, the drum is made of perforated metal and
partially filled with ceramic balls(Fig. 2 6- 4). In operation, the drum
holds agglomerated dust returned from the bags while the entering gas is
forced partially through and over this dust as in a fluidized bed. Fume
particles in the gas stream adhere to larger particles in the bed, and this
agglomerated dust is carried by the gas flow to the filter and returned to
the bed upon cleaning of the bags. Excess dust flows in the slowly rotating
drum toward the discharge valve. The pressure drop added by the drum is
from 0.4 to 2 inches, depending on application.
In addition to promoting dust agglomeration, the drum serves as a
buffer to prevent temporary condensates and other gas components from affect-
ing the filter cloth. Such components as oils, acids, or water can be
absorbed in the dust of the conditioning drum. Intermittent rotation of the
drum breaks up any cake formation. The drum can be seen as a pretreatment
of the gas by dust discharged from the filter.
The net effect of the use of a dust conditioning drum in a filter
system is a larger particle size dust on the filter bags and longer intervals
between required cleanings. The lower frequency of cleaning can substantially
improve baglife. As an alternate, we can apply higher air-to-cloth ratios in
many applications compared with operations without the dust conditioning
drum.
The concept is extended in the use of the conditioning drum for holding
additives in the gas stream, such as for removal of acid mist or reaction or
absorption of HC1, HF, or S02 (dry scrubbing). The coarse additive material
is continually milled in the drum and exposing new surface areas to the gas.
This process improves the reactivity of the additive in addition to sub-
stantially increasing the residence time of the additive in the gas. The
average residence time of an additive increases from 20 minutes to several
hours with the use of the conditioning drum.
Another important benefit is the even flow distribution along the length
of the filter (or cooler, see below) as a result of the pressure drop caused
by the conditioning drum. The motion of the drum prevents localized deposits
or caking and the constant pressure in the inlet plenum produces an even
flow over the length of the drum into the lower pressure filter housing,
Gas Cooling
Submicron dust emissions are generally generated in high-temperature
processes, such as metallurgical furnaces, glass furnaces, kilns, and combus-
tion processes. An integral problem of any emission control process for
these types of applications is the cooling of the exhaust gases to the tem-
perature required for fabric filtration.
The design of the Luhr Flat-Tube air-to-process gas heat exchanger is
based on cooling air passing through the inside of horizontally mounted flat
tubes (Fig. 3). The process gas flows outside and between the tubes in the
up or down direction, i.e., perpendicular to the cooling air flow. Swinging
chains between the flat tubes prevent the buildup of deposits on the tube
231
-------
Dust Conditioning Drum
Dirty Gas
Dust Discharge
T
FIGURE 2
LUHR TWIN-BAG FABRIC FILTER
GAS FLOW ENTERING FROM BELOW THROUGH DUST CONDITIONING DRUM
232
-------
High Temperature Gas Inlet
f
(V I
^r \
\
Cooling
Air
/
IV I
— "N J
-1/7
\
1
V
i 1
\
)
\ .
}
}
/
X* >>
— ^
\
1
1 \
\ T T
i
i
1 i i
I 1 '
i > »
M, S
k. r-l 1_
1
:
1
=
_
=
i
r=>
h>
r^
l—Js,
Heated Air
r=t> Exit
r>
r=>
f=t>
Low Temperature
Gas Exit
Dust Discharge
FIGURE 3
LUHR FLAT-TUBE GAS COOLER
233
-------
Bog house
<^
Clean Gas
T T Tf
Cooler
Heated Air
Exit
Dirty Gas
Dust Conditioning Druqi
Dust Discharge
FIGURE 4
LUHR COOLER/BAGHOUSE COMBINATION
DIRTY GAS ENTERING FROM BELOW THROUGH DUST CONDITIONING DRUM AND COOLER
234
-------
walls. The chains are attached to a shaft that is moved by a gear motor on
the outside of the cooler.
Generally, the gas flow through the cooler unit is designed to maintain
a low velocity of under 2400 ft/min. This approach holds the pressure drop
through the unit in the range of 1-3 inches and causes larger dust particles
to settle out. The cooler, therefore, performs the functions of a presepara-
tor. In addition, the cooler assists agglomeration of submicron particulate
matter. The low velocity also prevents abrasive wear on the cooling tubes.
A large excess of cooling air is normally supplied by axial fans, which
keep the heat exchanger tubes at a relatively low temperature. The normal
air flow is about 10 cfm/ft^ heat exchange area.
The cooler is sized using heat exchange coefficients, which take into
account the equilibrium coating of dust experienced in each type of applica-
tion. Typical heat exchange values are between 2 and 4 Btu per hr. per ft2
area and °F temperature differential.
The primary purpose of the Luhr cooler design is to cool hot process gas
to the maximum safe operating temperature of the fabric selected for the
baghouse. The cooler allows fine control over the gas exit temperature
through on/off control of several groups of fans. The large mass of the
cooler prevents sudden temperature changes in the exit gas.
The temperature control and the prevention of deposit buildup in the
cooler are major advantages compared with simple U-tube duct cooling.
Compared with water spray cooling, the indirect heat exchanger does not lower
the dew point through addition of water vapor to the gas and, of course,
eliminates the problems of spray nozzle maintenance, water droplet carryover
to the baghouse, and consumption of cooling water.
Substantial space and cost savings are possible with a design that
locates the filter housing on top of the cooler housing. The dust removed
from the filter bags falls through the cooler into the common hopper. The
system has one gas inlet plenum into the hopper and one gas outlet plenum
from the filter clean air space (Fig. 4). This approach generally is feasible
when the number of cooler and filter modules can be made equal.
Heat Recovery
In many typical applications of the cooler, the cooling gas flow is
about 3 to 5 times the process gas flow. This means that the average tempera-
ture increase in the cooling air is about 1/3 to 1/5 of the temperature loss
of the process gas. For a process temperature decrease from 100°F to 250°F,
the average cooling gas temperature increase would be in the range of 150°F
to 300°F.
The cooler can be modified to allow recovery of heated air in the 500°F
to 600°F temperature range by multiple passes of the cooler air (Fig. 5).
Because of the lower heat transfer efficiency, the cooler generally must be
sized larger to allow the use of multiple pass/counter-current air flow.
235
-------
Cross-Over Duct
Cooling Air Exit
Process Gas
900° F
1
1
i
1
t
I
[
i
i
i
t
I
L
!
J
r
r
H
i
i
ft
U
ft
U
ft
u
Process Gas
250° F
16000 SCFM Air 500-600°
Process Gas Inlet
Air Row
Process Gas Outlet
•Heated Air Outlet
Ambient Air Inlet
FIGURE 5
HIGH TEMPERATURE HEAT RECOVERY THROUGH
MULTIPLE PASSES OF COOLING AIR THROUGH COOLING TUBES
236
-------
The air in the 500°F to 600°F range can be re-used as preheated combus-
tion air in the burners of the furnace for a 10% to 12% fuel savings. Other
uses of the heated air include drying of raw material and space heating
during winter months. Many furnace applications draw substantial air from
the plant area. In this case, the entire cooling air from the cooler can be
used as heated make-up air for the plant during the winter months.
Cost Projections
The reverse air or the pulse-jet baghouse with dust conditioning drum is
generally in the price range of $12 to $15 per square foot of fabric. Based
on an air-to-cloth ratio of 5:1, the price is equal to $2.40 to $3.00 per
cfm of gas processed (1980 prices; equipment only).
For many metallurgical or other types of submicron dust applications,
the high-ratio baghouse competes in initial cost with the shaker-type bag-
house applied at much lower air-to-cloth ratio. The high-ratio baghouse
offers the advantage of substantially lower operating costs as a result of
lower fabric area. Typically, the number of bags required and the cost of
bag replacement is 1/2 to 1/3 that of a shaker baghouse.
In most furnace applications, the high-ratio Luhr baghouse can be com-
bined with the gas-to-air heat exchanger. In this case, operating cost
savings, compared with systems using air dilution or water spray cooling,
can be substantial even before credits for heat recovery for low-temperature
air. Any incremental investment required to achieve recovery of high-
temperature air has a payout of under one year.
Installation Experience
Numerous installations of the Luhr baghouse with horizontal bags and
dust conditioning drum are operating throughout Europe in the secondary
metal industry on brass, copper, lead, aluminum furnaces, as well as on
furnace emissions in the ferrous foundry industry and on glass furnaces.
A number of systems are operating now on brick kilns for the purpose of
hydrogen fluoride absorption by dry scrubbing with lime.
In North America, we presently have in operation a system on a lead
rotary smelting furnace at Canada Metal Company in Calgary, Alberta,at an
air-to-cloth ratio of 3.0:1; and a smaller baghouse unit on an aluminum
induction furnace at H. 0. Forgy & Sons in Jackson, TN, at an air-to-cloth
ratio of 3.7:1. Both installations very successfully demonstrate the
effectiveness of dust conditioning.
Gas cooler installations in the U.S. number some two dozen in a large
variety of applications.
237
-------
DEVELOPMENT OF GUIDELINES FOR OPTIMUM BAGHOUSE
FLUID DYNAMIC SYSTEM DESIGN
By: David Eskinazi
Gerald B. Gilbert
Dynatech R/D Company
99 Erie Street
Cambridge, Massachusetts
Robert C. Carr
Electric Power Research Institute
Palo Alto, California
ABSTRACT
In recent years, the utility industry has turned to fabric filters as
an alternative technology to electrostatic precipitators for particulate
emission control from pulverized coal-fired utility boilers. One aspect of
baghouse technology which appears to be of major importance in minimizing the
size, cost, and operating pressure drop is the development of ductwork and
compartment designs which achieve uniform gas and dust flow distribution. A
recently com-pleted experimental modeling program focused on developing design
guidelines for optimizing the fluid mechanic performance of baghouses. Tasks
included evaluation of the current technology of baghouse systems, formulation
of the appropriate modeling techniques for analysis of the flow of dust-laden
gas through the collector system, and extensive experimental analysis of
fabric filter duct system design. A matrix of geometric configurations and
operating conditions was experimentally investigated to establish the charac-
teristics of an optimum system, to identify the fluid mechanic integrity of
current designs, and to validate the development of new ideas and designs.
Experimental results indicate that the design of the inlet and outlet
manifolds, hopper entrance, hopper region below the tubesheet, and the com-
partment outlet have not been given sufficient attention. Unsteady flow
patterns, poor velocity profiles, recirculation zones, and excessive pressure
losses may be associated with these regions. It is evident from the results
presented here that the fluid mechanic design of fabric filter systems can be
improved significantly.
INTRODUCTION
Increased emphasis by regulatory agencies on the application of high
efficiency particulate control devices to coal-fired boilers has increased the
popularity of fabric filter systems in the electric utility industry. In June
1980, 37 units, representing 3500 MW, were in operation. This increasing
popularity and the tendency towards larger units have been the primary cata-
lysts for the initiation of this program.
The focus of this program was to develop design guidelines and pre-
sent technical information which will aid in the critical assessment of mani-
fold and hopper designs in fabric filter systems. The necessity for this type
of program manifests itself in the relatively severe penalty for high pressure
losses in terms of the power requirements and boiler derating. Therefore,
pressure loss reductions were strongly considered in this investigation, as
were dust and flow balance among compartments and bags within a compartment.
238
-------
This program has made substantial progress in developing guidelines
and assessing the fluid dynamic performance of utility fabric filter systems.
During the program, a state-of-the-art review of literature was completed,
eight installations were visited, a system model and a single compartment
model typical of utility fabric filter systems were constructed for
experimentation, and the data collected from several configurations were
analyzed. Emphasis has been directed toward providing information which can
be used in the preparation of engineering specifications for baghouses. Of
primary importance in this program was an increased knowledge of modeling
field installations in a laboratory environment. Through design,
construction, instrumentation, and experimentation, a procedure for modeling
baghouse systems was developed and refined. In addition, the scaling laws for
flyash trajectories were developed, and an appropriate model dust was
selected.
DEVELOPMENT OF THE MODELING CRITERIA
Selection of Models
The experimental work conducted in this investigation was completed
using models which are representative of the variety of configurations,
geometries, and operating conditions currently employed in the field. Two
separate models were used in the investigation of the fluid mechanic perfor-
mance of fabric filter systems: 1) a system model, and 2) a single compart-
ment model. Each model was selected and designed to be representative of all
fabric filter systems rather than of a particular unit, although some inferen-
ces can be drawn regarding individual sites.
The system model (approximately 1/16 scale) was designed to investi-
gate the performance of the inlet and outlet manifold systems. A plane of
symmetry is assumed at the centerline of the baghouse along the manifold
system such that a ten-compartment baghouse is simulated with a five-
compartment model and a half-width manifold system (see Figure 1). The model
has been designed with sufficient flexibility to allow most inlet and outlet
manifold systems to be investigated.
The details of the performance criteria in an individual compartment
are examined in the single compartment model (approximately 1/4 scale), which
possesses the same flexibility as the system model. This apparatus is geome-
trically scaled from representative dimensions and included a hopper inlet,
hopper, tubesheet, compartment region with 35 fiberglass bags, and a compart-
ment outlet. Together these two models provide the opportunity to investigate
a maximum number of configurations at a minimum cost. In this paper, only
system model results are discussed.
Measurement and Analysis
During the experimental portion of the program a variety of measure-
ments and analysis are required to document the performance of each configur-
ation. These include flow visualization, velocity profiles at particular
locations in the models, flow and dust balance among compartments, and pres-
sure losses through the models. Whenever possible this data is normalized by
average values. An analysis of pressure loss is one of the best assessments
of the fluid mechanic performance of these particulate control devices. It is
customary to report total pressure losses as non-dimensional loss coefficients.
These loss coefficients are defined as the total pressure loss between two
239
-------
locations divided by a representative velocity head in that region, usually
the velocity head at the upstream location.
Four loss coefficients are discussed in this report. Losses
described by these coefficients occur (1) from the inlet manifold just
upstream of any take-off to just under the simulated tubesheet in that par-
ticular compartment; this coefficient is referred to as the hooper inlet loss
coefficient (based on the average velocity head at the hopper inlet), (2) from
the baghouse inlet flange to a station just upstream of the final take-off
before the downstream end of the inlet manifold — the inlet manifold loss
coefficient (based on the average velocity head at the baghouse inlet flange),
(3) from above the bag region in any one compartment to the outlet manifold
just downstream of the corresponding outlet manifold entrance for that com-
partment — the compartment outlet loss coefficient (based on the average
velocity head through the com-partment outlet), and (4) from a point just
downstream of the first (most upstream) outlet manifold entrance to the
baghouse outlet flange — the outlet manifold loss coefficient (based on the
average velocity head at the baghouse outlet flange). The loss coefficients
associated with typical hopper inlets and compartment outlets can be improved
substantially by the addition of flow distribution devices.
RESULTS
Significant improvement can be made in the fluid mechanic performance
of fabric filter systems. The objective of this type of analysis is to mini-
mize duct pressure losses, improve dust and flow balance, and insure reaso-
nable flow patterns throughout the system to avoid operating problems and
improve flyash removal efficiency.
Optimizing the performance of baghouse systems includes evaluating
the following performance criteria:
9 the uniformity of ash loading and velocity profiles through the
system,
• minimum manifold, control valve, and system pressure losses,
Q even flow and dust balance among compartments and bags within a
compartment,
Q minimum dust accumulation in the inlet manifold,
® minimum turbulence levels in the hopper, and
® optimum transition designs for the geometry from 1) the inlet manifold
to the hopper, and 2) the upper compartment to the outlet manifold.
Each of these objectives was examined, although the extent to which equipment
modifications can be implemented depends on an economic analysis of suggested
changes.
Description of Configurations
Two configurations have been examined using the system model (see
Figure 2). System model experiments were useful in understanding the behavior
of (1) the inlet and outlet manifolds in terms of general flow patterns and
pressure losses, (2) inlet manifold-hopper inlet transitions and compartment
outlet-outlet manifold transitions in terms of general flow patterns, hopper
inlet velocity profiles and pressure losses, and (3) dust and flow balance
among compartments. Each configuration was tested twice: once as base con-
figuration without modifications and once with modifications. These modifi-
240
-------
cations included either flow improvement devices, valving systems, or a
combination of both. Flow improvement devices are uncommon in existing field
manifold systems. The design velocity levels in the model were obtained from
average field values at five critical locations in the fabric filter system.
These levels are nominally (a) 60 ft/sec at the baghouse inlet and outlet
flanges, (b) 5 ft/sec in the region just upstream of the simulated tubesheet,
and (c) 40 ft/sec at the hopper inlet and compartment outlet.
Inlet Manifolds
The general flow patterns in the inlet manifold are uniform and
steady. Figure 3 is a sketch of the flow patterns in the inlet manifolds of
Configurations 1 and 2. A recirculation zone is observed at the end of the
inlet manifold. This type of flow pattern is expected in designs which include
an extension region beyond the last manifold take-off. Flyash accumulations
are common in this region due to lower velocity levels and the horizontal
deposition surface. As long as the structural integrity of this region to
support flyash accumulations is considered in the design, no maintenance
problems would be expected. Also, it is evident from the sketches and a fun-
damental knowledge of fluid mechanics that a separation region will be
generated at the sharp edges as the flow passes into individual take-offs.
A more detailed comparison of the flow patterns in each configuration
reveals that the character of the flow differs in the two geometries. In
Configuration 1 the flow to each compartment is taken off in fairly well
defined layers; that is, it is relatively easy to trace smoke introduced at a
particular location at the model entrance to the compartment(s) it reaches.
In contrast, due to increased mixing in the Configuration 2 inlet manifold,
this same behavior is not observed.
The results of a pressure loss analysis indicate that the losses
along the inlet manifold are minimal. The inlet manifold loss coefficient for
each configuration is approximately 0.25.
Modifications to improve flow patterns or to examine the effects of
valving systems were made for each configuration. These modifications were
included to examine their effect on other regions in the fabric filter system;
no effect was expected regarding the inlet manifold.
Inlet Manifold - Hopper Transitions
The design of the inlet manifold-hopper transition is important to
the performance of fabric filter systems since it affects flow patterns in the
hopper and the trajectories of flyash particles entering the bag region. Also,
this region can be associated with abrupt changes in geometry and hence signi-
ficant flow separations. These flow separations are strongly related to
pressure losses. Three basic inlet manifold-hopper transitions are commonly
used, although the classification system is dependent on the detail
considered. They are:
• a 90° elbow from the bottom of the inlet manifold to the side of
the hopper,
• an inclined or horizontal transition from the side of the inlet
manifold to the side of the hopper, and
» a transition from the bottom of the inlet manifold directly to the
upper corner of the hopper
241
-------
Virtually all hopper inlet systems are some permutation of these classes. They
can be circular or rectangular in cross-section, or offset on the side wall of
the hopper. Four valving systems are commonly used: 1) poppet valves, 2) but-
terfly valves, 3) louver dampers, and 4) a more sophisticated butterfly-vane
combination. The inlet manifold-hopper transition in Configurations 1 and 2
were selected to investigate experimentally the behavior of two these general
classes.
Flow separation plays a significant role in the fluid mechanic beha-
vior of this region. A penalty in the form of a pressure loss is incurred
when the flow over a sharp corner separates. The separation zone is carried
downstream, which reduces the effective cross-sectional area of the duct and
increases local velocity levels above design values. These higher velocities
(velocity heads) at the hopper inlet give rise to additional losses as the
dust-laden gas enters the hopper. Due to the similarity of flow patterns in
this region with the flow in an idealized sudden expansion, one inlet velocity
head will be lost as the gas enters the hopper. Configuration 2 is a good
example of this behavior. In Configuration 1 a similar event occurs; however,
the two successive 90° bends in different planes lead to the generation of a
vortex in the elbow. In addition to creating unfavorable flow patterns, the
vortex is associated with additional pressure losses. This vortex, or
swirling flow, rotates the separation region and skews the velocity vector at
the hopper entrance. A velocity traverse at the hopper inlet quantifies the
character of the flow in this region (see Figure 4). The separation zone is
seen as a low (or zero) velocity region in the traverse. Also, velocity
levels in non-separated regions are more than double their design values.
Each configuration was modified either to reduce the magnitude of
flow separation, lessen the swirling, or examine the effect of a control valve
on the flow patterns in this region (refer to Figure 3). In Configuration 1,
the modifications to Compartments 1-4 (Compartment 1 is nearest the baghouse
inlet flange) included the addition of a poppet valve, a poppet valve with a
square grid just below the floor of the inlet manifold, a butterfly valve
aligned parallel with the mainstream flow in the inlet manifold and skewed
slightly towards the hopper, and a butterfly valve aligned normal to the
mainstream flow in the inlet manifold with and without a vane at the upstream
edge of the manifold take-off, respectively. The geometry of the hopper
inlet region of Compartment 5 remained unchanged. The poppet valves in the
first two compartments were located at a distance greater than l.OD from the
floor of the inlet manifold, (where D is the diameter of the poppet valve).
Preliminary experiments indicate that as long as the poppet valve is at a
distance greater than l.OD, the flow in the manifold or the take-off is
unaffected. This was observed in the modified configuration. The flow pat-
terns in this type of inlet manifold-hopper transition are highly unsteady due
to the flow separation and are associated with significant swirling due to
the two successive 90° bends as described previously. The square grid in the
second take-off reduced some of the swirling; however, the unsteadiness
remained unchanged. The butterfly valve aligned with the mainstream flow again
reduced the swirling by directing the flow through the take-off. The but-
terfly valve aligned normal to mainstream flow in the inlet manifold was most
successful in reducing the swirling* The addition of a vane efficiently
reduced the separated flow region, almost eliminating the unsteadiness. The
hopper inlet velocity traverses for the compartments with (1) a poppet valve
and a grid and (2) a butterfly valve aligned normal to the mainstream flow in
the inlet manifold with a vane are shown in Figure 5. A comparison with the
unmodified arrangement (Figure 4) indicates the improvement.
242
-------
In Configuration 2 the modifications were intended to reduce the
separation caused by the sharp corner (refer to Figure 3). These modifica-
tions for Compartments 1-5 included a single vane at the upstream sharp
corner, a rounded corner with no vanes, a rounded corner with one vane, and a
single vane with the sharp corner, respectively. Each modification reduced
the flow separation by varying degrees. Clearly, an assessment of the best
modification would require an associated economic analysis. However, it
appears that a single vane or a rounded corner gives significant improvement
for a minimum of modifications. The best hopper inlet velocity profiles are
observed with two vanes or a rounded corner with a single vane. Two hopper
inlet velocity profiles are included to quantify the improvement * These pro-
files are for a single vane with and without a rounded corner (see Figure 6).
Again, a comparison can be made with the unmodified arrangement in Figure 4.
The analysis is not complete without quantitative pressure loss
results since they can result in significant costs over the life of a power
plant. A comparison of the hopper inlet loss coefficients for the two con-
figurations with and without modifications is presented in Figure 7. The
loss coefficients through similar inlet manifold-hopper transitions (basic
configuration) decrease as the progression is made downstream along the inlet
manifold. This decrease can be explained by (1) the component of the velocity
vector directed into the take-off increases as the flow proceeds downstream
along the inlet manifold; therefore, the severity of the separation region at
the sharp corner is reduced, and (2) the velocity level further downstream
along the inlet manifold decreases which also reduces the separation.
Therefore, unmodified and modified transitions must be compared compartment by
compartment. The losses in the basic arrangement for Configuration 2 are
higher than these for Configuration 1 because of the severity of the
separations. However, the losses in Configuration 2 with modifications are
lower than those in Configuration 1 because a single 90° bend, which is rec-
tangular in cross-section, is more easily adapted to vaning; Clearly, the
pressure losses of these transitions can be improved significantly.
Compartment Outlet - Outlet Manifold Transitions
The compartment outlet—outlet manifold transition is another region
in which pressure losses can be reduced through more careful designs. A
reasonably uniform velocity profile is observed in the upper bag region due to
the substantial pressure losses across the fabric. Two basic compartment
outlet transitions are currently employed:
» a 90° elbow from the top of the compartment to the side of the
outlet manifold,
» a side compartment outlet with a prechamber before the entrance
to the outlet manifold (usually from the top).
The cross-section of outlet transitions may be rectangular or circular and in
most cases are offset with respect to the centerline of the compartment.
Three basic valving systems are used: 1) poppets, 2) louver dampers, and 3)
flapper dampers. Configuration 2 is not considered typical of current field
installations, but was selected because of its simplicity. Both of these con-
figurations have side compartment outlets.
Again, the primary pressure losses are caused by turning the flow
across sharp edges. The compartment outlet loss coefficients for Configur-
ations 1 and 2 are 2.6 and 1.5 respectively, for the basic arrangements.
243
-------
Non-uniform velocity profiles along the outlet manifold make accurate measure
ments of wall static pressure difficult to obtain. The compartment outlet
loss coefficients for Configuration 1 are larger since this outlet geometry
includes three consecutive 90° bends as compared with two bends for
Configuration 2. Similar improvement can be made with the addition of vanes
and/or rounded corners. In the current program, no attempt has been made to
improve the compartment outlet loss coefficients.
Outlet Manifolds
The general flow patterns in the outlet manifolds of unmodified
Configurations 1 and 2 are slightly unsteady and non-uniform. Again, a recir-
culation zone exists at the beginning of the outlet manifold. The charac-
teristic of the flow in Configuration 1 is such that sequential jets layer the
flow along the outlet manifold, but stronger flow is observed on the floor of
the manifold. Configuration 2 flow patterns are characteristic of stronger
mixing along the manifold. A convenient method of summarizing the outlet
manifold flow patterns is to compare the fraction of points in a traverse at
the baghouse outlet flange which fall between + 10% of the average velocity.
In Configuration 1, 24% of the points in the baghouse outlet flange traverse
fall between +10% of the average velocity. In Configuration 2, 64% of the
points meet this criterion.
To gain a better understanding of the pressure losses along the
outlet manifold, a simple theoretical model based on the momentum equation was
applied to the outlet manifold as a control volume. The results of this ana-
lysis indicate that significant losses along the outlet manifold are due to
the effort required (loss in kinetic energy) for the mainstream flow, upstream
of any incoming jet, to turn that jet coincident with the outlet manifold
mainstream velocity vector. In both configurations each compartment produces
a jet normal to the mainstream flow in the outlet manifold. This loss can be
reduced by the addition of solid surfaces (vanes) to redirect the flow towards
the baghouse outlet flange.
Modifications were made along the outlet manifolds of each of these
configurations to examine this effect and improve the velocity profile at the
baghouse outlet flange. A different set of vanes was designed and installed
in each configuration. The vane geometry and a sketch of the general flow
patterns are shown in Figure 8. A contoured vane is included at the beginning
of the outlet manifold in each case. In Configuration 1 a different two vane
set has been installed in Compartments 1-3. Each set directs the outlet flow
to a corresponding portion of the baghouse outlet flange cross-section. No
vanes were installed in the outlet manifold near the Compartment 4 exit. In
Configuration 2 an identical two vane set was installed in Compartments 1-4.
These vanes extend across the height of the outlet manifold entrance from each
compartment but not the full height of the outlet manifold.
The results of experiments conducted with Configuration 1 showed a
dramatic improvement. The outlet manifold loss coefficient for Configuration
1 without modifications is 1.6. This was reduced to 0.6 by the addition of
the vanes described. Also, the uniformity of the velocity traverse at the
baghouse outlet flange was greatly improved (all the points in the traverse
fall between +10% of the average velocity). The outlet manifold loss coef-
ficient for Configuration 2 without modification is 1.1. With the addition of
vanes this value increased to 1.7. The uniformity of the baghouse outlet
flange velocity traverse did not improve. At present, this is not fully
244
-------
understood, and further experiments are needed. It is difficult to understand
why the addition of vanes would increase the losses. The flow into the mani-
fold in Configurations 1 and 2 is from the top and side of the outlet
manifold, respectively. The nearer half-width manifold wall (imposed by the
plane of symmetry assumption) may offer some turning effect for the normal jet
without vanes. This plane of symmetry assumption may "over idealize" the
impinging jet behavior with an outlet manifold side entrance geometry. Also,
the character of the unvaned flow in Configuration 2 is to produce some
swirling in the outlet manifold. This may limit the accuracy of the wall sta-
tic pressure measurement by underestimating the outlet manifold loss
coefficient. In summary, it appears as though the character of the flow has
been changed by the addition of vanes. These results indicate that while flow
improvement devices remain an effective means of reducing the pressure losses
through improved flow patterns, a complete understanding of the fluid mechanic
behavior of these systems in necessary in selecting appropriate designs.
Related Analysis
An understanding of the flow and dust balance among compartments is
necessary to characterize fully the performance of these devices.
Intuitively, a reasonable balance among compartments is required if the total
fabric surface area is to be used effectively. Further work will be needed to
quantify the limits of a reasonable balance.
The flow balance among compartments was measured for each unmodified
configuration at a moderate fabric pressure drop simulation (2.7 inches of
water, field). In addition, Configuration 1 was selected to document the
effect of increasing fabric pressure drop (equal among all compartments) from
1.4 to 4.5 inches of water (field). Figure 9 summarizes the flow balance
results for increasing fabric pressure drop simulation in Configuration 1.
For a moderate fabric pressure loss (2.7 inches of water), the flow balance
among compartments in Configuration 1 is +_ 7.5% of the mean flow per compart-
ment. At the same fabric pressure loss the flow balance among compartments in
Configuration 2 is + 3.5% of the mean flow per compartment (see Figure 10).
It is evident from the graph in Figure 9 that as the fabric pressure loss per
compartment (equal in all) increases, the flow balance improves. As we have
seen from an analysis of loss coefficients along the inlet and outlet
manifolds, the outlet manifold losses are significantly greater than the inlet
manifold losses. Therefore, the flow balance is distributed according to the
shortest path along the outlet manifold. That is, the flow balance adjusts
itself such that the highest flow is observed in the compartment associated
with the shortest outlet run. This is evident in Figure 10. This effect has
been con-firmed by flipping the inlet manifold to simulate a baghouse with
inlet and outlet flanges on opposite sides of the collector. This analysis is
limited since in practice the fabric pressure loss does not increase uniformly
across all compartments simultaneously. Additional work will be required with
non-uniform fabric pressure losses to understand flow balance fully.
The dust balance among compartments is dependent, to some extent, on
the flow balance. Figure 10 indicates that the dust balance among compart-
ments is + 11% and + 24% of the mean dust loading per compartment for
Configurations 1 and 2, respectively. This difference is not fully
understood; however, the more uniform layering in the inlet manifold of
Configuration 1 is undoubtedly part of the reason.
245
-------
A critical evaluation of any flyash simulation experiments is manda
tory, and caution must be applied in interpreting the results. Each set of
results is valid for the particular flyash simulated through the scaling laws
for flyash trajectory modeling. The variety of flyash characteristics makes it
difficult to generalize these results. Clearly, particle size distributions
which differ greatly from the simulated distribution will yield different results.
CONCLUSIONS
Experiments conducted in this program indicate that significant
improvement in the fluid mechanic performance of fabric filter systems is
possible. The primary objective of this effort is to upgrade the performance
of baghouses through experimental modeling to improve existing designs and to
begin to develop future designs.
It has been demonstrated that experimental modeling of these par-
ticulate control devices is a reasonable approach to optimizing the perfor-
mance of fabric filter systems. Experimental modeling is particularly
attractive since it offers flexibility, ease of geometrical modifications, and
a partially controlled environment not available in field installations. The
reduced size and minimal cost of experimental models offer more efficient
troubleshooting and optimization of designs.
Although a complete set of design guidelines requires additional
work, many significant results have been obtained furthering the development
of improved designs. These improvements will primarily be reflected in a
reduction in system pressure loss and perhaps fewer maintenance requirements,
which translate into a reduction in operating costs.
The inlet and outlet manifold systems and compartment transitions are
regions in which the greatest improvements can be made. Abrupt changes in
geometry in the absence of flow improvement devices are common. The loss
coefficients for typical inlet manifold-hopper inlet transitions are typically
high. It has been demonstrated that these can be reduced with the addition of
flow improvement devices and more careful design of each component in the
system. Similar improvements can be made regarding the compartment outlet-
outlet manifold transition loss coefficients. Reduction in pressure loss
along the inlet manifold will be minimal since reasonably low values currently
exist. However, improvements along the outlet manifold maybe possible by
directing compartment outlet jets entering the outlet manifold towards the
baghouse outlet flange.
Hopper inlet designs can be greatly improved by emphasizing lower
hopper inlet velocities since the kinetic energy associated with the incoming
flow to the hopper is essentially lost due to the sudden expansion. The addi-
tion of flow distribution devices will aid in controlling flow separations and
will therefore reduce pressure losses. Careful monitoring of ash levels in
the hopper can minimize the reentrainment of previously removed flyash.
Additional work is necessary to understand the reentrainment process if the
ash is allowed to accumulate in the hoppers.
NOTE: Many details have been omitted from the text due to
space limitations. A more detailed version of this
paper is available from the authors upon request.
246
-------
Figure 1
SYSTEM MODEL GENERAL ARRANGEMENT
(ALSO CONFIGURATION 1)
Figure 2
CONFIGURATION DESCRIPTION,
SYSTEM MODEL
PERFORATED
PLATES
SEPARATED FLOW
INLET MANIFOLD FLOW PATTERNS RECIRCULATION ZONE
(Side View)
INTENDED AS A SKETCH ONLY
RECIRCULATION ZONE
SEPARATED FLOW
INLET MANIFOLD FLOW PATTERNS
(Top View)
247
Figure 3
INLET MANIFOLD FLOW PATTERNS
(NO MODIFICATIONS)
(CONFIGURATIONS 1 & 2)
-------
Fi gur e 4
HOPPER INLET VELOCITY PROFILES
(NO MODIFICATIONS)
(VELOCITY/AVERAGE VELOCITY) -
AVERAGE DESIGN VELOCITY: 40 ft/sec
Figure 5
HOPPER INLET VELOCITY PROFILES
(WITH MODIFICATION, CONFIGURATION 1)
(VELOCITY/AVERAGE VELOCITY) -
AVERAGE DESIGN VELOCITY: 40 ft/sec
1.1 0
1.8 0 very unsteady
0
0
2.0 1.1 ° „
2.1 1.3
2. 1 1.6
2.0 1.7
POPPET VALVE
WITH A GRID
1.3
1.2 1
1.0 °
S.5 "•' °
'•' 1.0 0.3
1 .
Kfi 1.4 1.2 0,8
1.6 0 9
1.6 1-6 I-* 09
1.7 1.6
1.6
1.5
1.7
1.5
Configuration 1
0.2
0.2
0.3
0. 3
0.3
0.3
very unsteady
2.5
2.0
Configuration 2
(mean flow into page)
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
Figure 6
HOPPER INLET VELOCITY PROFILES
WITH MODIFICATIONS, CONFIGURATION 2
VANE AND
BUTTERFLY VALVE
0.9
0.6
0.2
0.3
O.S
0.7
'•" 09 0.8 0.5
1.1 i., i.« , Oo.8°-6
1.0 . '-°
1.3 '**
1.6
1.6
',.6 -.7 l.«
1.5 1.6
1.4 1.4
(mean ffow into paper)
Figure 7
HOPPER INLET LOSS COEFFICIENTS
CONFIGURATIONS 1 & 2 WITH AND WITHOUT
MODIFICATIONS
ONE VANE
Fraction o
ONE VANE
AND ROUNDED
CORNER
Fraction
0.2 1.7 0.6 1.3 0.7 1.4 . 3 .2
0.2 1.5 1.1 1.6 1-5 1.4 .3 .2
0.5 1.4 1.0 0.5 0.5 1.0 .2 . T
f points wNch are + 10% of the average: 0. 15
0 0
1.4 1.3 1.1 1.3 1.2 1.2 1.1 1.0
1-3 1.3 0.4 0.7 1.1 1.1 1.1 1.0
of points which are + 10% of the average: 0. 35
(mean flow into oaoe)
COMPARTMENT NUMBER
Configuration 1 :
Basic Arrangement
Modified Arrangement
Modification
Configuration 2:
Basic Arrangement
Modified Arrangement
Modification
1
4.2
5.2*
poppet
valve
5.2
3.1
one
vane
2
3.6
4.7*
poppet
valve
with a
grid
4.6
2.1
two
vanes
3
3.3
4.3
butterfly
valve
aligned
parallel
with the
inlet flow
5.1
2.4
rounded
corner
4 5
3.0 2.9
3.6/2.4 2.8
butterfly none
valve aligned
perpendicular
to the inlet
flow /with vane
3.8 2.9
1-5 1.4
rounded one v
corner S
one vane
(The first compartment is the one nearest the baghouse inlet flange.)
* questionable data point
248
-------
Figure 8
OUTLET MANIFOLD FLOW PATTERNS
(WITH MODIFICATIONS)
(CONFIGURATIONS 1 & 2)
OUTLET MANIFOLD
(Side VIev.)
SEPARATED FLOW REGION
OUTLET MANIFOLD
(Top View)
INTENDED AS A SKETCH ONLY
Figure 9
FLOW BALANCE AMONG COMPARTMENTS
(CONFIGURATION 1)
Figure 10
DUST BALANCE AMONG COMPARTMENTS
(CONFIGURATIONS 1 & 2)
1.0 2.0 3.0 4.0 5.0
Fabric Pressure Drop Simulation
(inches of waCet, field)
COMPARTMENT NUMBER
Configuration 1
Percent of Dust in Each Compartment 21.1 21.7 20.8 19-6 16.8
(by mass)
Percent of Flow in Each Compartment 22.2 21.1 20.0 18.9 17.9
Configuration 2
Percent of Dust in Each Compartment 24.9 21.7 19.5 18.5 15.4
(by mass)
Percent of Flow in Each Compartment 20.8 20.7 19.2 20.0 19.4
4. 5 inches of water
2.7 inches of water
1.4 inches of water
Percent of Flow in Each Compartment
COMPARTMENT NUMBER 1
Fabric Pressure Drop Simulation (field)
21.26
21.50
22.06
20.32
20.51
21.12
19.78 19.V9 19.15
20.05 19.43 18.51
19.96 18.92 17.94
(The first compartment is the one nearest the baghouse inlet flange.)
249
-------
THEORETICAL ASPECTS OF PRESSURE DROP REDUCTION
IN A FABRIC FILTER WITH CHARGED PARTICLES
By: Ta-Kuan Chiang, Eric A. Samuel, Kris E. Wolpert
Buell Emission Control Division
Envirotech Corporation
Lebanon, PA 17042
ABSTRACT
Various mechanisms have been presented in the past to explain the reduced
pressure drop observed in a conventional fabric filter augmented with parti-
cle precharging. Such mechanisms involve dendrite formations, polarizations,
and electric contact potentials. In this paper, yet another mechanism due to
nonuniform dust profile is presented. Artificial step and bell-shape profiles
are modeled to show the reduction of pressure drop without the assumption of a
more porous cake. Laboratory experiments with metallic screens were used to
verify the proposed mechanism.
INTRODUCTION
New concepts for fine particle control (1) were generated from time to
time in the past. For fabric filtration, the concept was to enhance the
fabric filtration with either charged particles (2), or noncharged particles
in an electrostatic field (3). Reportedly, with electrostatic augmentation,
the fabric filtration efficiency was increased considerably; and yet coinci-
dent with this improved filtration efficiency was an observed reduction of
pressure drop. Improved filtration efficiency was easily understood due to
an additional collection force introduced by the charged or noncharged parti-
cles in a nonuniform electrostatic field; whereas the experimentally observed
reduction of pressure drop became a puzzle and a center of dispute. Various
mechanisms were offered in the past to explain the reduced pressure drop.
Notably, such mechanisms under investigations included dendrite formations
(4), polarizations, and electric contact potentials (5,6); all of them lead
to a more porous cake formation. In this paper, without the assumption of a
more porous cake, another mechanism of nonuniform dust deposition caused by
the fundamental behavior of charged particles in a combined flow field and
electrostatic field is introduced.
MECHANISM OF REDUCED PRESSURE DROP
It is known that fundamentally the entrained particles in a flow system
follow the air flow lines attaining same velocities as the air flow. Mean-
while, due to particle inertia, small particles follow the flow lines more
readily than the large particles whenever there is a change of flow velocity.
With this fundamental behavior of the particles, the nonuniform velocity dis-
tribution in a conventional bag filter, maximum at the entrance of the bag and
zero at the other end, results in a particle size stratification. Since the
resistance coefficient is sensitive to the porosity of dust deposits and the
volume-to-surface ratio of the particles, size stratification will contribute
250
-------
to changes in the resistance of the dust cake which in turn, due to the nature
of air seeking the least resistance, redistributes the air velocity and parti-
cle size again. As this interaction between the flow and the dust goes on
during the filtration, different filter velocities, amount of dust collected,
and deposit structure occur at various locations on the filter surface. Since
areas of low resistance will handle higher than average air volumes, and vice
versa, consequently more dust is deposited on the low resistance areas, less
on the areas of high resistance, and a self-balancing system exists. The
result is a cake with a nonuniform porosity but a uniform areal density with
actual filtration velocities much higher than the design velocity, and conse-
quently a higher pressure drop.
When the particles are charged, an additional force of electrical origin
is acting at the particles throughout the space via an externally applied
electric field or a self-space-charge field of the charged particles. This
additional force results in another particle drift velocity following the
electric field lines in addition to the aforementioned velocity following the
flow lines. When the charged drift velocity is much greater than the air
velocity, for instance highly charged particles in a strong electric field,
the particle trajectory is then completely governed by the electric field
lines resulting in a field-controlled size stratification and deposition rate
instead of a flow-influenced size stratification and a self-balancing deposi-
tion rate.
From the knowledge of electrostatic precipitation, the charged drift
velocity for particles greater than 0.2 ym in diameter in the field charging
regime is directly proportional to the particle size and approximately to the
field square. Thus, the field-controlled size stratification in a conven-
tional bag configuration will be more or less orderly along the axial direc-
tion starting with the largest particles at the entrance of the bag and end-
ing with the finest particles at the other end. The field-controlled areal
density will, however, be nonuniform. Unlike the self-balancing effect in a
flow influenced system, this field-controlled size stratification and non-
uniform areal density maintains the initial distribution of cake resistance
invariant throughout the entire filtration cycle leading to an overall
observed pressure drop reduction. Mathematical proofs using artificial depo-
sition profiles appended with this paper further support this postulation.
The cause of reduced pressure drop in a charged filtration system is thus
believed, primarily because the charged particles follow preferentially the
field lines. There is no need to assume the formation of a more porous cake
caused by other phenomena.
Quantitatively, from the experience of electrostatic precipitation, the
mean drift velocity experienced in an electrostatic precipitator is commonly
quoted as 5 cm/s or approximately 10 ft/min transverses to a gas velocity of
the order of 1.5 m/s (5 ft/s); and an appreciable precipitation is generally
observed at the first meter of the first field. Relative to a conventional
bag designed at a nominal air-to-cloth ratio of lxlO~2 m/s (2 ft/min), and
an aspect ratio of 30, the maximum gas velocity at the entrance of the bag
is 6x10"-'- m/s (4 ft/s) ; filtration velocity at the bag surface is only 1x10"^
m/s (2 ft/min). It is seen that the radial drift velocity relative to the
maximum entrance velocity is in line with the electrostatic precipitator
251
-------
practice; and it is also approximately a factor of 5 larger than the mean
nominal filtration velocity. It appears reasonable to assume that the parti-
cles if adequately charged would certainly follow the field lines rather than
the flow lines, resulting in an electric field-controlled nonuniform areal
density together with an electric field-controlled particle size stratifica-
tion, both along the bag axis with the heaviest deposition and the largest
particles at the entrance and decreasingly lighter and finer toward the other
end.
LABORATORY EXPERIMENTS
Laboratory experiments were designed to verify the postulated mechanism.
To eliminate possible effects of dendrite formation, polarization and contact
potential, two metallic screens were used to simulate the fabric filtration.
The first screen was a coarse mesh; the second screen was a fine mesh. Both
screens were electrically grounded. D.C. corona discharge was used for par-
ticle charging. Two configurations, a patch and a cylindrical, were tested.
Test conditions were all at high dust loadings of the order of 2xlO~2 Kg/m3
(10 grains/ft3), and air-to-cloth ratios of 2.54xlO~2 to 3.66xlO~2 m/s (5 to
7 ft/min). Instantaneous pressure drop was continuously monitored and
recorded.
Typical results obtained at room temperature are illustrated in Figures
1 and 2. Pressure drop reduction was observed for all tests and for all con-
figurations. Exceptions were noted, however, when tests were conducted on
high resistivity dusts. Due to a back corona problem, high resistivity
particles were inadequately charged; consequently, the particles were control-
led predominantly by the air flow lines instead of the electric field lines.
On the other hand, when an appreciable reduction of pressure drop was observed
for the patch configuration, a nonuniform deposition profile on the first
screen as shown in Figure 3 was also visually observed. Dust deposition was
concentrated opposite the corona wire and tapered off qualitatively according
to the electric field distribution. Failure to show appreciable pressure drop
reduction for high resistivity dusts, and a nonuniform deposition profile
accompanied by an appreciable reduction of pressure drop were experimental
evidences that were needed to support the postulated mechanism of pressure
drop reduction.
DISCUSSIONS
Study of the fundamental interaction of charged particles in a combined
flow field and electrostatic field indicates that qualitatively there exists
a strong possibility of nonuniform dust deposition in an electrostatic aug-
mented fabric filtration. Particle deposition, unlike a conventional fabric
filtration, is predominantly controlled by the electric field lines instead
of the flow lines. Consequently, fewer particles are brought to the least
resistance areas where larger than average air volumes would pass. Thus, the
least resistance areas would remain almost unchanged during the filtration,
resulting in an overall reduction of pressure drop as reported. With this
fundamental reasoning, it is thus concluded that the first order mechanism of
pressure drop reduction in an electrostatic augmented fabric filtration is
due to the electric field controlled dust deposition — a nonuniform dust
252
-------
profile; other mechanisms leading to a more porous cake formation are not
needed. Due to the nature that the charged particle drift velocity is size
dependent, there also exists an orderly stratified particle size distribution
constituting the nonuniform dust deposition. It is further believed that
this electric field-controlled particle size distribution enhances the effect
of reduced pressure drop by maintaining an orderly distributed cake porosity
throughout the filtration cycle.
To further confirm the postulated mechanism, laboratory experiments with
metallic screens, thereby eliminating the possible phenomena of dendrite for-
mation, polarizations and electric contact potentials, verified the existence
of nonuniform dust profile coincident with the reduction of pressure drop.
Also verified was an unobservable reduction of pressure drop when high resis-
tivity dust particles, hence back corona and inadequately charged particles,
were used as test particles.
Finally, the postulated mechanism is theoretically verified by modeling
an artificial step and a bell-shape dust profile mathematically based on the
fundamental laws of conservation of mass, conservation of volumetric flow
rate, and equal pressure drop of parallel flow. Mathematical modeling indi-
cate that there always exists pressure drop reduction relative to a uniform
dust distribution; the relative magnitude of pressure drop reduction depends
on the degree of nonuniformity.
SUMMARY
The observed pressure drop reduction in a conventional fabric filter
augmented with particle precharging can be explained completely by a non-
uniform dust profile without introducing the possible existence of a more
porous cake. Fundamental reasoning of the behavior of charged particles in
a combined flow and electric field, experience of electrostatic precipita-
tion, laboratory experiments with metallic screens, and mathematical modeling
of nonuniform dust profile all consistently support the nonuniform profile
mechanism.
REFERENCES
1. Ariman, T. (Compiler). Proceedings - Symposium on New Concepts for Fine
Particle Control. EPA-600/7-78-170 (NTIS PB-292 095), U.S. Environmental
Protection Agency, Research Triangle Park, NC, August 1978.
2. Helfritch, D.J., and T. Ariman. Electrostatic Filtration and the Apitron
- Design and Field Performance. In: Proceedings - Symposium on New Con-
cepts for Fine Particle Control, Ariman, T. (Compiler). EPA-600/7-78-170
(NTIS PB-292 095), U.S. Environmental Protection Agency, Research Tri-
angle Park, NC, August 1978. p. 286-304.
3. Lamb, G.E.R., P.A. Costanza, and D.J. O'Meara. Electrical Stimulation
of Fabric Filtration, Part II: Mechanism of Particle Capture and Trials
with a Laboratory Baghouse. Textile Research Journal, Textile Research
253
-------
Institute, October 1978. p. 566-573.
4. Oak, M.J., and D.A. Saville. Dendrite Structures in Electrically Stimu-
lated Filtration. (Presented at the Joint Symposium on. Fibers, Electro-
statics and Filtration, Princeton, NJ, November 14-15, 1979). p. 2.
5. Penney, G.W. Collection of Electrically Charged Particles in Filters
J.APCA. 26(1):58, January 1976.
6. Penney, G.W. Electrostatic Effects in Fabric Filtration: Volume 1.
Fields, Fabrics, and Particles (Annotated Data). EPA-600/7-78-142a
(NTIS PB-288 576), U.S. Environmental Protection Agency, Research
Triangle Park, NC, September 1978.
8 3-
H
Patch Configuration
%" Mesh + 400 Mesh
Both Electrically Grounded
•35 KV On Wires
200 250
Time (Minutes)
Figure 1. Laboratory Test Results -
Patch Configuration
254
-------
100 Mesri! Outer Screen
Both Electrically Grounded
Air-To-Cloth Ratio 345X10'inWs
|6.8tt/min)
Inlet Particle Loading, 2.08 X 10'' kg/m'
(9.1 Grains/fl')
Figure 2. Laboratory Test Results -
Cylindrical Configuration
Figure 3. Nonuniform Dust Profile,
Laboratory Test -
Patch Configuration
255
-------
APPENDIX
Mathematical Modeling of Nonuniform
Dust Profile and Pressure Drop
Nomenclature
C, dust loading
U, mean filtration velocity, or air-to-cloth ratio
^t, filtration time
A, fabric area
Ap, pressure drop
Apjj, pressure drop of a uniform deposition profile
K2, specific resistance coefficient
h, dust thickness
W, areal density
&, linear dimension of filtration area
P, particle density
£, cake porosity
Subscripts a and b denote parameters in deposition regions a and b
respectively.
256
-------
Step Profile
Let the dust profile be a discontinuous step as shown in Diagram A,
I U,C
„ . ' Ub'Cb
T"
fa I7777T77777l\\\\\\\\l ^b DIAGRAM A
•a
1.
From conservation of mass ,
(C U At) A +(C. U At) A, = (CUAt)A (1)
a a a b b a
From conservation of volumetric flow rate,
U A -HI. A, = UA (2)
a a bo
From parallel-flow consideration,
AP = APV (3)
a b
From Carman-Kozeny equation, assuming the particle properties are the
same in the two regions, i.e., K2 = K2, Equation (3) leads us,
C U 2 = CO. 2 (4)
a a b b
Assuming the charged drift velocity results in different deposition
rates, namely,
(5)
where n > 1 according to Diagram A.
Solving Equations (4) and (5) , we have,
U = nU, (A /A, ) (6)
a b a b
Solving Equations (I) and (5), we have,
CUA (7)
Let A = A, = A/2, and from Equations (2) and (6) ,
a "b 257
-------
Ub = 2U/(n+l)
From Equation (7),
C,U, = 2nCU/(n+l)
b b
Substituting Equations (8) and (9) into Equation (3), we have,
2 2
AP, = 4nIC CU At/(n+l)
b 2
(8)
(9)
(10)
Let APrj =
Equation (1) shows,
be the pressure drop for a uniform dust profile,
AP, = 4nAP/(n+l)'
b U
(11)
From Equations (11) and (5), for a uniform profile, n = 1, and thus,
AP]-, = APy. For n > 1, Table 1 illustrates that the pressure drop ratio,
, is always less than one.
TABLE 1. PRESSURE DROP REDUCTION DUE
TO A STEP DUST PROFILE
n
2
3
4
5
6
AVApu
0.889
0.750
0.640
0.556
0.490
Bell-Shape Profile
Let the dust height be hQ at center, and 6 be the weighted distribution.
The bell-shape profile as illustrated in Diagram B can be presented by the
expression,
Ua(x)
DIAGRAM B
258
-------
r „ ?-i-1
h(x) = h (l+(Bx/£) )
o
Assuming the distribution is uniform in the Z-direction, from conserva-
tion of mass, we have,
2/ w(x)dx = 2(£/B)/\ p(l-e){l+(Bx/£)2r1d(Bx/£)
w O O
= CuAt(2£)
This normalization provides,
h = BCuAt{p(l-e)tan~1B}~1
Therefore, the profile can now be represented by,
h(x) = BcuAt{p(l-e) (l+(Bx/£)2)tan~1B}~1 (12)
and the pressure drop Apb(x) is,
Ap, (x) = K.W(x)U. (x)
D 2 b
= K2p(l-e)h(x)Ub(x)
Substitution of Equation (12), we have,
AP, (x) = BK-CUAt'U. (x){(l+(Bx/A)2)tan~1B}~1
b 2 b
or,
U, (x) = uB~1{l+(3x/£)2>(tan~1B) (AP,(x)/AP)
L> 13 U
From conservation of volumetric flow rate, and using the fact of
parallel flow that Apj-^xJ's are equal for all x's, we have,
£
2U£ = 2f U, (x)dx
o b
= / UB"1(tan~1B)(AP /AP ) {l+(Bx/£)2}dx
o b U
Carrying out the above integration, and rearranging the terms, we have,
Ap /AP = SBtan^B/O+B2) (13)
b U
259
-------
Table 2 illustrates how the pressure drop ratio varies with
TABLE 2. PRESSURE DROP REDUCTION DUE TO
A BELL-SHAPE
DUST PROFILE
6 APb/APu
1
2
3
4
5
6
Larger than 6
0.4
0.3
Smaller than 0.3
0.955
0.774
0.600
0.476
0.390
0.198
Approaches 0.190
0.998
0.999
Approaches 1.000,
a uniform profile
260
-------
EXPERIMENTAL CORRELATION OF DUST CAKE POROSITY,
AIR-TO-CLOTH RATIO AND PARTICLE-SIZE DISTRIBUTIONS
By: Ta-Kuan Chiang
Buell Emission Control Division
Envirotech Corporation
Lebanon, PA 17042
Ronald L. Ostop
Department of Public Utilities
City of Colorado Springs
Colorado Springs, CO 80903
ABSTRACT
Experimental values of the mean specific resistance coefficients obtained
from a full-scale utility baghouse and a slipstream Research and Development
pilot baghouse were used to obtain the cake porosities using Carman-Kozeny
equation. Log-log plot of the calculated mean porosity vs. the particle Rey-
nolds number using volume-to-surface mean diameter indicated a functional
dependence of minus two-thirds power to exist.
INTRODUCTION
Since the completion of the main baghouse and the "slipstream" pilot
Research and Development (R&D) baghouse for Martin Drake Unit No. 6, the City
of Colorado Springs, Colorado, an intensive R&D product optimization program
was performed on site to evaluate and to advance the "State-of-the-Art" of
fabric filtration for particulate and gaseous pollution control. This paper
presents experimental correlation of dust cake porosity, air-to-cloth ratio
and particle-size distribution found under the particulate control program.
Martin Drake Unit No. 6 is an 85-megawatt pulverized-coal-fired (PC)
utility boiler with a flue gas volume of 189 actual cubic meters per second
(400,000 ACFM) at full load. The main baghouse provided for the stack emis-
sion control is a conservative design using reverse air cleaning. Bag size
is nominally 0.305 meters (12 inches) in diameter by 9.44 meters (30 feet)
long with an effective cloth area of 8.454 square meters (91 square feet) per
bag. The design air-to-cloth ratio is 9.398xlO~3 meter per second (1.85 feet
per minute) with all compartments in operation, and 1.128xlO~2 meters per sec-
ond (2.22 feet per minute) with two compartments out of service for mainte-
nance or cleaning. The pilot R&D baghouse, provided for testing at higher
air-to-cloth ratios without compromising the ability of the main baghouse to
continue its basic stack emission control function, is a single compartment
with a total of 16 full-size bags identical to the bags used in the main bag-
house. A schematic flow diagram of the test setup is illustrated in Figure 1.
EXPERIMENTAL APPROACH
The test program was designed to determine the particle loading and par-
ticle size distribution simultaneously at the main baghouse and the slipstream
261
-------
pilot R&D baghouse and to correlate their effect to the baghouse performances.
After the establishment of the correlation of particle loading tests under
identical operating conditions, extended range of tests, such as higher air-
to-cloth ratio, would be performed at the pilot baghouse only. Analysis of
the extended series of tests together with the simultaneous tests would then
provide us data to develop a baghouse model that would be useful and practical
in predicting filter operation throughout potential operation regimes, and to
recommend the optimum operation regime for the main baghouse.
Since the important parameters which characterize a filter performance
are the flow pattern, the dust particle properties, and the interactions be-
tween the flow pattern and the particle dynamics, test data obtained simul-
taneously at the main baghouse and the pilot R&D baghouse would be meaningful
only if the condition of dynamical similarity between them was satisfied and
maintained. In order to establish this dynamical similarity, the pilot R&D
baghouse was thus purposely designed to accommodate full-size bags and bag
spacings identical to the main baghouse. With this design, unlike laboratory
bench scale or other small scale pilot tests, the flow Reynolds number in the
pilot R&D baghouse and the main baghouse at the same air-to-cloth ratio would
then be identical at the entrance of the bag, along the bag axis and at the
bag surface leading to similar interactions between the flow pattern and the
particle dynamics. Test data obtained from the pilot R&D baghouse of this
program would thus be more representative, and can be scaled directly or
generalized for a typical full-scale utility baghouse operation.
EXPERIMENTAL RESULTS
A total of 28 tests were conducted under this program. Among them, 13
tests were conducted simultaneously on the main baghouse and the slipstream
pilot R&D baghouse. Air-to-cloth ratio covered a wide range from 7.620xlO~3
to 2.235xlO~2 meters per second (1.5 to 4.4 feet per minute).
As field testing proceeded and field data generated, it was found that
the particle size distribution for the first eight tests was not log-normal
and deviated from a normal utility PC boiler operation. Examination of cas-
cade impactor results indicated that all the inlet samplers were flooded with
more than 10 mg in the first few stages, and yet all the outlet samples were
much less than 10 mg for reliable weighing. Inlet samples from absolute fil-
ter in accordance with EPA Method 5 for the remaining tests were then ana-
lyzed by laboratory sedimentation analyzers to establish the inlet particle
size information. Further screening of test data based on the acceptable
criteria of isokinetic sampling rate resulted in 10 valid tests as summarized
in Table 1. The accustomed practical units were adopted to formulate Table
1. Conversion of practical units to MKS units are,
u, air-to cloth ratio (A/C) 1 ft/min = 5.080xlO~3 m/s
c, dust concentration 1 grain/actual ft3 (ACP)
= 2.290xlO-3 Kg/actual m3 (ACM)
AP, pressure drop 1 in. H2o = 249 N/m2
262
-------
K2, specific resistance 1 in.-H20-min-ft/lb
coefficient = Ixlo4 N-S/Kg-m
From Table 1, although variations of particle loading and particle size
distribution were found between the main baghouse and the pilot R&D baghouse,
the ultimate results derived from this series of tests indeed characteristi-
cally demonstrated normal baghouse operations; namely, collection efficiencies
of baghouses were both at 99.8 to 99.9 percent, nearly a constant for A/C of
1.6 to 2.4 for utility PC boilers without precollectors.
EXPERIMENTAL CORRELATION
By validating all the test results to within ilO percent isokinetic samp-
ling and by adapting laboratory particle size analyses, confidence was thus
established for the key measurements which were needed for experimental corre-
lations. These key measurements were:
• The air-to-cloth ratio,
• The inlet and outlet particle loading, and
* The inlet particle size distribution.
Using these key measurements together with bag pressure drop, AP, and
time, t, strip chart recordings, the universally accepted Carman-Kozeny equa-
tion was then used to calculate the mean cake porosity. To avoid needless
arguments on the validity of Carman-Kozeny equation, this calculated porosity
was defined as £2, corresponding to K2, which might or might not represent
the real cake porosity if actual measurement was carried out. Table 2 listed
the key parameters and their corresponding mean cake porosities, £2-
Following Carman-Kozeny equation, particle volume-to-surface mean dia-
meter, dvs/ was used. Plotting of Jin £2 vs- ^n ^vs as illustrated in Figure
2, apparently failed to show consistent functional relationship between £2
and dyg. However, inspection of Table 2 indicated that the product of the
air-to-cloth ratio, u, and dvs might be a better choice. To generate a dimen-
sionless plot, the product of udvs was normalized with gas density, pg, and
viscosity, U, representing the particle Reynolds number Re2« To generalize
the plot representing baghouses at various aspect ratios (bag length/bag dia-
meter) , the air-to-cloth ratio, u, was further referred back to the air
velocity, Ug, at the bag entrance. The final plot of £n£2 vs. in Re2 as
shown in Figure 3 resulted in a straight line with a slope of minus two-thirds
covering the ranges of experimental variation. Analytically, the experimental
correlation found on the average was
£2 - 0.1945 (Re2)"2/3 (1)
Substituting Equation (1) into Carman-Kozeny equation, the mean specific
resistance coefficient, K2, was found,
K2 - 2.448xl04(Pg/pp)(Pgug2/y)(1-0.1945(Re2)~2/3> (2)
263
-------
where pp was the particle density and Kozeny constant of five was incorporated
into the constant. Use of consistent units in Equation (2) provided us a tool
to estimate K2 either in practical engineering units or MKS units.
DISCUSSIONS
The physical significance of the particle Reynolds number can be inter-
preted the same way as the flow Reynolds number encountered in the fluid
mechanics. In particular, the particle Reynolds number is a measure of the
relative magnitude of the inertial forces of the gas flow and the particulate
flow. The higher the particle Reynolds number indicates that particles would
tend to follow the flow relatively readily. On the other hand, the opposite
is true at lower particle Reynolds number. Consequently, at higher particle
Reynolds number, particles follow readily with the flow seeking the least
resistance resulting in a uniformly higher specific resistance coefficient;
whereas at lower particle Reynolds number, particles are reluctant to follow
the flow seeking the least resistance, leaving most particle collections to
remain at the high resistance areas with least air velocity, and least par-
ticle collections to remain at the least resistance areas with the highest
air velocity. Contrary to the flow and particle interaction at high particle
Reynolds number, flow and particle interaction at low particle Reynolds num-
ber leads to, and more or less maintains, a nonuniform cake resulting in an
overall lower mean specific resistance coefficient.
Understanding the physical significance of the particle Reynolds number,
we have thus provided a sound physical foundation for the experimental corre-
lations of Equations (1) and (2) . Inspection of Equation (2) indicates that
the upper limit of K2 at high particle Reynolds number would be simply asso-
ciated with the fluid acceleration. This is seen to be in agreement with the
principle of fluid mechanics when the dust cake solidity, (1-E2) / approaches
the limit of unity; it should be noted that, according to Carman-Kozeny equa-
tion, the limiting K2 approaches infinity when the cake solidity approaches
unity. On the other hand, the lower limit of K2 approaches zero at low par-
ticle Reynolds numbers when the dust cake solidity approaches zero. Physi-
cally, zero dust cake solidity means strictly that there would not be any
collection of particles on the fabric filter, and of course, the experimental
correlations and the Carman-Kozeny equation would consequently break down at
this extremely low particle Reynolds number. Therefore, the applicable cri-
teria of Equations (1) and (2) for particle Reynolds number less than the
experimental particle Reynolds number of 0.25 is dependent on the collection
efficiency of fabric filtration. Furthermore, for particle Reynolds number
less than 0.1, experimental correlations found here would be certainly invalid
for the reason that a negative cake solidity, or a negative mean specific
resistance coefficient, simply does not have any physical meaning. Apparently,
for particle Reynolds number less than 0.1, or for fine particulate control,
a different approach would certainly be needed.
SUMMARY
With the condition of dynamical similarity satisfied at a "slipstream"
pilot R&D baghouse, and a utility baghouse, good correlation of baghouse per-
formance was found between the two baghouses. After the verification of
264
-------
baghouse performance between the two, testing proceeded at higher air-to-
cloth ratios on the pilot R&D baghouse only. A total of 28 tests covering
A/C of 1.5 to 4.4 were performed. Among them, 10 tests covering A/C of 1.6
to 2.4 were validated for experimental correlation of cake porosity, air-to-
cloth ratio and particle size distribution. Using Carman-Kozeny equation and
experimental K2, mean cake porosities, E2, at various A/C and particle size
distribution were then calculated. Log-log plot of the mean cake porosity
and particle Reynolds number based on volume-to-surface mean diameter indi-
cated that there existed functional dependence of minus two-thirds power.
Physical significance of the particle Reynolds number was fully discussed to
extend the applicable criteria of the finding outside its experimental range.
Limiting value of K2, according to the established experimental correlation,
was found to be finite and proportional to the fluid kinetic energy, when the
cake solidity, l-£2, approached the limit of one at high particle Reynolds
number. Contrary to the Carman-Kozeny equation, the limiting K2 approaches
infinity when the cake solidity approaches one. With a finite limiting K2,
proportional to the fluid kinetic energy, it appears that the experimental
correlation presented in this paper provides a better prediction of K2 than
the Carman-Kozeny equation.
At low particle Reynolds number, the experimental correlation indicated
that there exists a critical particle size, or a critical air velocity, at
the mean cake porosity of one, or K2 of zero. To stay away from this physi-
cally meaningless regime from the viewpoint of cake filtration, the lower
limit of particle Reynolds number was set at 0.1 with a caution that the fil-
tration collection efficiency has to be high enough such that the inlet parti-
cle size distribution is a representative measure for the particles constitu-
ting the dust cake. It is further suggested that other approaches should be
taken to extend the current finding for particle Reynolds number below 0.1.
By referring the particle Reynolds number to the bag entrance, we have
generalized the experimental correlation to any aspect ratio. Thus, knowing
particle size distribution, the experimental correlation presented here pro-
vides the confidence to predict baghouse performance for utility applications
in particular, and other applications in general, provided that the particle
shape is not much different from that of the fly ash.
265
-------
Figure 1.
Test Setup
50 100
Figure 2. Mean Cake Porosity (e2) vs.
Particle Volume-to-
Surface Mean Diameter
266
-------
Figure 3. Mean Cake Porosity (£2)
vs. Particle Reynolds
Number (Re2)
267
-------
TABLE 1. VALIDATED TEST RESULTS
Test
No.*
IP
2M
3P
4M
5M
6M
7P
8M
9P
10P
1
1
1
1
1
1
1
1
1
2
u
.613
.691
.554
.616
.683
.637
.731
.651
.801
.437
1
2
1
1
1
1
1
1
1
1
c
.8086
.1145
.9999
.8021
.5305
.7596
.5184
.9914
.4104
.4566
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
n
998
999
999
999
999
999
999
998
998
999
d 4
m
19.
19.
25.
14.
14.
21.
31.
21.
16.
15.
0
3
4
0
0
0
0
5
5
0
a
3.33
2.67
2.42
2.71
3.03
2.33
1.84
2.52
2.97
3.19
d
vs
9-2
11.9
17.2
8.5
7.6
14.7
25.8
14.0
9.1
7.6
K/
12.5
15.1
15.1
18.1
19.1
20.7
21.7
23.1
25.7
36.2
* Not in chronological order; P for pilot baghouse, M for main
baghouse
+ dm, mass median diameter, in ym; a, standard deviation
# K2, in.-H^O-min-ft/lb, derived from AP-t strip chart recordings
and c, u in Table 1
268
-------
TABLE 2. CALCULATED MEAN CAKE POROSITIES (£2)
Test
No.
IP
2M
3P
4M
5M
6M
7P
8M
9P
10P
1
1
1
1
1
1
1
1
1
2
u
.613
.691
.554
.616
.683
.637
.731
.651
.801
.437
d
vs
9.
11.
17.
8.
7.
14.
25.
14.
9.
7.
2
9
2
5
6
7
8
0
1
6
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
£2
479
392
321
458
468
318
228
316
397
391
K2
12.
15.
15.
18.
19.
20.
21.
23.
25,
36.
5
1
1
1
1
7
7
1
7
2
2
2
2
1
2
2
2
2
2
2
*
PP
.043
.157
.038
.958
.167
.064
.038
.170
.017
.137
* Particle density, in grams/cm^
269
-------
MODEL FOR DUST PENETRATION THROUGH A PULSE-JET FABRIC FILTER
By: D. Leith, M.J. Ellenbecker
Harvard School of Public Health
Department of Environmental Health Sciences
Boston, Massachusetts 02115
ABSTRACT
Presently, dust penetration through a pulse-jet filter cannot be inter-
preted or predicted well. This paper presents a model which considers pene-
tration straight through the filter and penetration by seepage. Previous
studies have considered penetration by the straight through process; however,
a comparison of data from the literature with results from the present model
shows that seepage and not straight through penetration accounts for virtual-
ly all penetrating dust. Although insufficient information is presently
available to use the model to predict penetration, the model does show
trends that should occur with changes in filter operating variables such as
filtration velocity and pulse pressure, and suggests areas in which further
research is necessary.
INTRODUCTION
Pulse-jet filters have captured a substantial portion of the fabric fil-
ter market(1) and the efficiency with which they operate is of strong inter-
est to regulatory officials, to industrial users who must meet emission
regulations, and to equipment manufacturers who supply these filters with a
performance guarantee. In spite of many attempts to model penetration
through fabric filters,(2-5) there is presently no satisfactory way to pre-
dict dust penetration through a pulse-jet cleaned fabric filter. This is not
because of insufficient interest in the problem.
Performance characteristics which must be considered when modeling pene-
tration include: (1) particle collection by clean fibers in a new fabric,
(2) particle collection by the dust deposit accumulated on and in these fab-
rics, and (3) retention of this dust so it does not seep through the fabric
during the rather violent cleaning cycle.
Penetration Straight Through a Clean Fabric
Particle collection by isolated, clean cylinders or fibers has been well
studied, and adaptation of single fiber collection models into models for
particle penetration through an array of clean fibers are well known (&)
Fraser and FoleyW have made use of this work to describe penetration
through a clean woven fabric and Hampl and Rimberg(2) have adapted clean
fiber bed theory to a study of particle penetration through clean felt of the
kind used in pulse-jet cleaned filters. The latter found that penetration of
0.35 to 1.1 ym particles through clean, new industrial felts ranged from 20
to TO/, at typical pulse-jet filtration velocities, penetrations that are
much higher than the U or less generally found for intermittently
270
-------
cleaned(7-9) industrial filters using fabrics well conditioned with dust.
Although agreement is excellent between penetration theory and data for clean,
new felt fabrics(2), the performance of new felt in the laboratory is clearly
different from that of a well conditioned felt when used in an industrial
pulse-jet filter.
Penetration Straight Through a Dust Deposit
The accumulation of dust in and on a fabric can affect dust collection
in several ways. At low dust accumulation, before a continuous dust deposit
forms on the fabric surface, dust particles collect on previously collected
particles adhering to the fibers, as well as on the fibers themselves. Be-
cause of the greater collecting efficiency of the deposited dust particles
for collecting additional particles, the apparent efficiency of individual
fibers increases as dust collects upon them so that fabrics conditioned with
dust have lower penetration than beds(6) or fabrics(3,4) composed of new,
clean fibers. Fraser and Foley(4) modeled the effect of dust accumulating on
woven fabrics by assuming the effect equivalent to a reduction of fiber
diameter.
At high dust holdings, when the dust deposit is continuous and uniform
on the fabric surface, the deposit acts as a filter cake and efficiency im-
proves further. Cooper and Hampl(3) studied dust penetration straight
through a well-developed dust deposit using an entirely theoretical model
under the assumptions that impaction, interception(sieving), and diffusion
are the important particle collection mechanisms. They determined that es-
sentially no particles larger than 1 ym or so should pass through the dust
deposit unless pinholes are present in it. Although models for penetration
straight through a dust deposit are not as well developed as those for pene-
tration straight through a clean fiber bed or a clean fabric, theory suggests
that a continuous, unflawed deposit should collect virtually all particles
from the aerosol passing through it. Theory predicts that penetration
straight through a deposit should decrease with increasing particle diameter
and increasing filtration velocity(3) for particle diameters larger than a
few tenths micrometer.
Although penetration data must be carefully examined with reference to
the conditions under which the tests were made, tests on new, clean felts
by Hampl and Rimberg(2) show the decrease in penetration with increasing
particle diameter and filtration velocity predicted by straight-through
theory,and bench-scale experiments by Afify and Mohamed(lO) and by Hall(ll)
show that penetration through an uncleaned felt becomes very low when a sub-
stantial dust deposit builds up on it. These conclusions generally confirm
the Fraser-Foley(4) and Cooper-Hampl(3) theories that virtually no dust pene-
trates an established dust deposit. However, experiments on uncleaned fab-
rics are not relevant to the effects of seepage caused by filter cleaning.
Penetration by Seepage
Seepage penetration is defined as a process whereby collected particles
later pass through the filter, i.e., seepage is a failure to retain collected
271
-------
dust rather than a failure of collection. In pulse-jet cleaned filters, most
seepage is thought to occur when the conditioned fabric strikes its support-
ing cage at the end of a cleaning pulse(12). Rapid deceleration of the
residual dust deposit causes some particles or agglomerates to separate from
the support (fabric fibers or other dust particles). As normal filtration
flow resumes, some of these particles or agglomerates are swept through the
fabric and into the cleaned gas stream.
Seepage penetration is likely to increase whenever the separation force
between the collected dust and its substrate increases. For example, large
particles or agglomerates with considerable inertia are likely to break loose
more easily than small particles when the fabric strikes its supporting cage
smartly at the end of a cleaning pulse. It is reasoned that high filtration
velocity, which drives the fabric back on to its cage at a faster rate,
causes it to hit with greater impact, and thereby aggravates seepage.
In contrast, straight-through penetration theory predicts penetration should
decrease with increasing particle size and with increasing filtration velo-
city (2-4) for particles larger than a few tenths micrometer in size. This
is opposite to the trends expected from seepage.
PENETRATION MODEL
A reasonable model to describe dust penetration through a pulse-jet-
cleaned filter should consider both straight-through penetration and seepage
penetration. The rate at which dust passes through the filter can be ex-
pressed in terms of outlet mass flux, the mass of dust that penetrates per
unit fabric area per unit time:
N = N + N (1)
st se v '
in which N represents the total outlet mass flux, Nst is the outlet flux due
to straight-through penetration, and Nge is the flux due to seepage.
The total straight-through flux, for particles of all sizes, will depend
on penetration for particles of each size in the inlet aerosol:
fl
N=vc.PdG (?)
st i J st ^'
in which v is superficial filtration velocity, c± is the overall mass concen-
tration for particles approaching the filter, Pgt is the fraction of particles
of a particular size that pass straight through the filter, and DG is the
mass fraction of all incoming particles which are of that size.
If the relationship between particle diameter and straight-through flux
is needed, then relationships among particle diameter, straight-through pene-
tration, and inlet dust size distribution must be known. However if an
overall value for straight-through flux is sufficient, the overall mass frac-
tion of the inlet dust that passes straight through the filter F~ can be
used in place of the integral in Equation 2. ' st'
272
-------
As discussed above, dust that seeps through a pulse- jet-cleaned fabric
may be comprised essentially of dust driven through the fabric when the fab-
ric strikes its supporting cage as a cleaning pulse ends and filtration re-
sumes. Deceleration of the dust held by the fabric as it strikes the cage
may cause a separation force sufficient to free some dust which normal fil-
tration air can then sweep through into the cleaned gas stream. If so, then
seepage flux will be:
dH
(3)
in which w is the mass areal density of the total dust deposit on the fabric
as it strikes its cage, t is the time interval between cleaning pulses, e is
the fraction of particles or agglomerates of a particular size in the dust
deposit that are separated from the dust deposit as the fabric strikes the
cage, and dH is the mass fraction of all particles or agglomerates in the
dust deposit that are of that particular size.
If the relationship between particle diameter and outlet flux is re-
quired, the relationships among e, dH, and particle or agglomerate size must
be known. If an overall mass outlet flux is sufficient, an overall mass-
weighted average for e can be used, ~e~, in place of the integral in Equation
(3) Insufficient data are presently available to allow evaluation of the
integral in Equation (3) and in the following discussion ~e will be used.
The overall mass fraction of dust separated from a conditioned fabric,
e, can be related(14) to the applied area-specific force causing the separa-
tion, FQ/A. The exact relationship between fraction of dust separated and
o
separation force applied is not known, and is presently under investigation
at this laboratory . Pending the outcome of this work, it will be assumed
that these variables are proportional :
k*F
Impulse-momentum considerations say that the area-specific separation force
applied, F /A, times the time over which this force acts, t , will equal the
momentum of the area-specific dust deposit as it begins deceleration upon
striking the cage:
F t
= wv
(5)
Substitution of Equations (4) and (5) into Equation (3) yields:
k*w2v
kw v
se t* t t
in which the unknown proportionality constant, k*, and the time necessary to
273
-------
stop the fabric as it strikes its cage, t*, are combined in a new constant, k,
which should depend on the dust separation and retention characteristics of
the fabric and fabric elasticity.
Equations (2) and (6) can be substituted into Equation (1) to give
an expression for overall outlet flux through the pulse- jet cleaned filter:
_
N = c.v P + k *—*- (7)
i st t
EXPECTED TRENDS IN OUTLET FLUX
Equation (7) predicts the effects of changes in filter operation on
overall outlet flux, N. First, consider the case in which straight-through
penetration is much more important than seepage; that is, the first term on
the right side of Equation -i-7) is considerably larger than the second:
N = c. vP (8)
i st
In this instance, Equation (8) shows that decreases in inlet dust concentra-
tion, c-^, should cause a decrease in outlet flux. However, Dennis(9) has de-
termined from field and laboratory studies(7,15) that the mass emission rate
or flux from a fabric filter is essentially constant despite changes in in-
let concentration.
Straight-through theory predicts that factors which decrease overall
penetration, Pg|-, such as increasing diameter for particles larger than a few
tenths micrometer, or increasing dust deposit areal density should also de-
crease outlet flux. However, Leith and First(8) determined that large par-
ticles penetrate a pulse-jet filter as readily as small particles, at vari-
ance with this prediction. Data(16,17) discussed later in this section of
the report indicate that more dust penetrates if the dust deposit areal den-
sity increases, also at variance with this prediction.
The effect of increased filtration velocity on straight-through outlet
flux is unclear. Equation (8) suggests that outlet flux should increase
with velocity, v, if all else remains constant; however, increased velocity
should also decrease penetration, Pgt, for particles collected by impaction
so that the net effect of velocity changes on outlet flux is uncertain. In
summary, straight-through penetration theory predicts trends generally at
variance with those determined experimentally for changes in pulse-jet filter
operating conditions such as inlet dust concentration, particle diameter,
and areal density of the dust deposit.
Next, consider the case in which seepage is much more important than
straight-through penetration, that is, the second term on the right side of
Equation (7) " is considerably larger than the first:
2
M • I W V
N = k — (9)
274
-------
Equation (9) predicts that outlet flux should decrease with decreases in
the areal density of the dust deposit, w, in marked contrast to straight-
through theory, but in agreement with experimental data reported by Ellen-
becker(16) and by Leith et al.(17). Equation (9) also indicates that de-
creases in filtration velocity should decrease outlet mass flux, in agreement
with several experimental studies(8,18-20) using pilot scale and full scale
pulse-jet filters. Dennis(9), Hall(ll), and Loffler(13) have shown that out-
let dust concentration from a pulse-jet cleaned filter increases substan-
tially but momentarily, immediately after cleaning. Modifications to the
cleaning pulse which ease the cleaned bag back to its supporting cage gently
so that it hits with less impact have been shown to decrease penetration com-
pared to that found with normal pulse-jet cleaning(21). These results are
expected if seepage associated with pulse-jet cleaning causes penetration.
Leith and First(10) measured the relative amounts of straight-through
flux and seepage flux for fly ash penetrating a pulse-jet filter using un-
treated but conditioned polyester bags at superficial filtration velocities
from 50 to 150 mm/s. Outlet flux was very low for the straight-through
mechanism under all conditions, but the seepage mechanism increased with in-
creasing filtration velocity. In general, the conclusion was reached that
seepage accounted for essentially all the dust that passed through the pulse-
jet filter and was particularly important at high filtration velocities.
Although Equation (9) is presently inadequate to describe the depend-
ence of outlet flux on particle diameter, this relationship can be incor-
porated into the equation when information adequate to allow integration of
Equation (3) becomes available. It should be clear that penetration by
particle size will be greatly influenced by filtration velocity, dust loading,
pulse pressure, etc. In summary, penetration trends predicted by seepage
theory seem generally to agree with trends observed in experiments conducted
over a range of pulse-jet filter operating conditions. These trends require
precise quantification to make possible a fully comprehensive mathematical
model of a pulse-jet cleaned filter. These tests are presently under way at
this laboratory.
COMPARISON OF MODEL WITH DATA
If seepage alone accounts for nearly all the dust penetration from a
pulse-jet cleaned fabric filter, Equation (9) predicts that a log-log plot
of outlet dust flux versus the group w^v/t should yield a straight line with
a slope of unity and air intercept at log(k). Data(16,17) adequate to check
this prediction are summarized elsewhere(16,17). Bags made from
untreated polyester felt and Gore-tex felt were tested in a three-bag pulse-
jet filter at superficial filtration velocities from 50 to 150 mm/s. Outlet
dust concentrations were measured and total outlet flux calculated. The
areal density of the dust on the bags was measured by weighing the bags and
subtracting the weight of the clean, new fabric. The interval between clean-
ing pulses was 60 s per bag in each test.
2
Figure (1) is a log-log plot of outlet dust flux, N, against w v/t.
The data for both bag types scatter about lines with a unit slope as Equa-
275
-------
tion (9) indicates will occur when outlet flux caused by straight-through
penetration is negligible and seepage is caused by deceleration of the bag as
it hits the supporting cage. A least squares fit for Equation (9) to the
data shown in Figure (1) gives values of k = 0.0077 m-s/kg for untreated
polyester fabric with coefficient of determination, rz =0.92, and k =
0.034 m-s/kg for the Gore-tex fabric with r2 = 0.96.
The higher value for constant k found for Gore-tex fabric should not
necessarily be interpreted to mean that penetration in an operating filter
will be higher using Gore-tex fabric, as this fabric retains less dust. Al-
though Equation (9) shows that outlet flux increases with increasing k, it
decreases with the square of areal dust density, w. Inter-fabric comparisons
require knowledge of both factors.
DISCUSSION
To predict outlet mass flux using Equation (9) , values are needed for
filtration velocity, v, time between cleaning pulses, t, constant, k, and
areal density of the dust deposit, w. Although velocity and time between
cleaning pulses are ordinarily known, the other two parameters might not be
known. At present the only way to determine constant k is to evaluate data
as done in Figure (1) . Our data to the present indicate that constant k
depends on fabric type, although it is likely to depend as well on the char-
acteristics of the dust collected, pul'se pressure, filtration velocity, and
other parameters.
The areal density of the dust deposit on the fabric is also likely to be
unknown. The magnitude and distribution of the dust deposit on an operating
nonwoven bag can be measured with difficulty(22). A procedure for calculating
the areal. density of the dust deposit on a pulse-jet cleaned filter has re-
cently been developed as part of a theoretical model for pressure drop(14).
This procedure, however, requires other information concerning the resistance
of the fabric and the specific resistance of the dust deposit which may not
be available. Regrettably, it is not presently possible to use Equation
(9) to predict quantitatively the outlet flux from a pulse-jet-cleaned fab-
ric filter; further work is necessary.
However, the model is useful to predict the direction of outlet flux to
changes in pulse-jet filter operating variables. Equation (9) shows that
outlet flux increases with increasing filtration velocity. Increased pulse
pressure should decrease the areal dust density, w; if so, the outlet flux
should also decrease although sufficient dust must remain on the filter to
assure good collection of incoming dust. Equation (9) suggests that in-
creasing t, the time between pulses to each bag, should decrease emissions,
but this might also increase the areal density, w, of the dust deposit on the
276
-------
bag. An increase in areal density should increase emissions so that the net
effect of pulse interval on outlet flux is unclear. Changes to the -size dis-
tribution of the dust, to the relative humidity or temperature of the gas
stream, or to the electrostatic properties of the dust or fabric may all af-
fect constant k; the relationships between these variables and this constant
are presently unknown.
1
VI
f
E
2
z"
X
-^
u.
(0
<
H
1-
0
^
^
»-
IO'5
8
6
4
•?
c.
.g
'°8
4
2
,0-7
I 1 III 1 1 1 i i
n
/'_
_
-
-/ u
O J3 xO
/ y/^
/o /b
o/ / n
~ ^X °u ° -
h/ :
^o o QORE-TEX, k=0.034
D UNTREATED POLYESTER,
k= 0.0077
1 i i i i 1 i till
IO"5 2 4 6 8 IO"4 2 468 ID*3
v»2v kg2
FIGURE 1. TOTAL OUTLET MASS FLUX, N, VERSUS w v/t, FOR BAGS
MADE FROM GORE-TEX AND UNTREATED POLYESTER FELT
ci
dG
dH
k
k*
N
N
N
st
jse
•st
NOMENCLATURE
O
dust inlet concentration, kg/m
mass fraction of particles of a particular size in the inlet dust, di-
mensionless
mass fraction of particles or agglomerates of a particular size in the
dust deposit, dimensionless
area-specific force acting to separate the dust deposit from its sub-
strate, N/m^
constant, see Equation (6) m-s/kg
constant, see Equation (4) m^/N
total outlet mass flux, kg/m^-s
outlet mass flux due to straight-through penetration, kg/m^-s
outlet mass flux due to seepage penetration, kg/nr-s
mass fraction of a particular size that passes straight through the fil-
ter, dimensionless
277
-------
t time between cleaning pulses, s
t* time during which dust deposit decelerates, s
v superficial filtration velocity, m/s „
w areal density of dust deposit on and in the fabric, kg/m
e fraction of particles or agglomerates of a particular size in the dust
deposit that are separated from the dust deposit, dimensionless
"e mass fraction of the dust deposit that is separated from the dust de-
posit, dimensionless
REFERENCES
1. Frey, R.E. Types of Fabric Installations. J. Air Poll. Control Assoc.
24:1148, 1974.
2. Hampl, V. and D. Rimberg. Aerosol Penetration of Felt Filter Media.
(Presented at Annual Conference of Gesellschaft fur Aerosol-Forschung E.V.
(Association for Aerosol Research), Bad Soden, Germany, Oct. 26, 1974).
3. Cooper, D.W. and V. Hampl. Fabric Filter Performance Model. In: Con-
ference on Particulate Collection Problems in Converting to Low Sulfur
Coals. Report EPA-600/7-76-016. NTIS, Springfield, Virginia, 1976.
p. 149-185.
4. Fraser, M.D. and G.J. Foley. A Predictive Performance Model for Fabric
Filter Systems. 1. Intermittently Cleaned Single Compartment Systems.
(Paper 74-99 presented at 67th Annual Meeting of Air Poll. Control Assoc.
Denver, Colorado, 1974).
5. Dennis, R. , R.W. Cass, D.W. Cooper, R.K. Hall, V. Hampl, H.A. Klemm, J.E.
Langley and R.W. Stern. Filtration Model for Coal Fly Ash with Glass Fab-
rics. Report EPA-600/7-77-084. NTIS, Springfield, Virginia, 1977.
6. Davies, C.N. Air Filtration. New York, Academic Press, 1973.
7. Dennis, R. and J. Wilder. Fabric Filter Cleaning Studies. Report EPA-
650/2-75-009. NTIS, Springfield, Virginia, 1975.
8. Leith, D. and M.W. First. Performance of a Pulse-Jet Filter at High Fil-
tration Velocity. 1. Particle Collection. J. Air Poll. Control Assoc.
27:534, 1977.
9. Dennis, R. Collection Efficiency as a Function of Particle Size, Shape,
and Density: Theory and Experience. J. Air Poll. Control Assoc 24-
1156, 1974.
10. Afify, E.M. and M.H. Mohamed. Collection of Efficiency and Pressure Drop
of Needle Punched Filters. J. of Engineering for Industry. 98:675, 1976.
11. Hall, R.F. Getting the Best Results from Needlefelts in Dust Filtration.
Filtr. and Sep. 14:350, 1977.
278
-------
12. Leith, D. and M.W. First. Performance of a Pulse-Jet Filter at High Fil-
tration Velocity III. Penetration by Fault Processes. J. Air Poll. Con-
trol Assoc. 27:754, 1977.
13. Lb'ffler, F. Separation Efficiency and Pressure Loss of Filter Materials
of Different Structure, at Differing Conditions. Staub, Reinhalt. Luft.
30(12):27, 1970.
14. Leith, D. and M.J. Ellenbecker. Theory for Pressure Drop in a Pulse-Jet
Fabric Filter. Atmospheric Environment. 14:845, 1980.
15. Dennis, R., G.A. Johnson, M.W. First, and L. Silverman. How Dust Col-
lectors Perform. Chem. Eng. 59:196, 1952.
16. Ellenbecker, M.J. Pressure Drop in a Pulse-Jet Fabric Filter. Sc.D.
Thesis. Harvard School of Public Health, Boston, Massachusetts, 1979.
17. Leith, D., M.J. Ellenbecker, M.W. First, J.M. Price, A. Martin and D.W.
Gibson. Performance of a High Velocity Pulse-Jet Filter II. EPA Publi-
cation 600/7-80-042. Research Triangle Park, North Carolina, 1980.
18. Billings, C.E., M.W. First, R. Dennis, and L. Silverman. Laboratory
Performance of Fabric Dust and Fume Collectors. AEC Report No. NYO-1590
(revised), 1961.
19. McKenna, J.D., J-C. Mycock, and W.O. Lipscomb. Performance and Cost
Comparisons Between Fabric Filters and Alternate Particulate Control
Techniques. J. Air Poll. Control Assoc. 24:1144, 1974.
20. Mohamed, M.H., E.M. Afify and J.W. Vogler. Needle Punched Fabrics in
Filtration. (Paper presented at Technical Symposium - Nonwoven Product
Technology, International Nonwovens and Disposables Association,
Washington, D.C., 1974).
21. Leith, D., M.W. First, and D.D. Gibson. Effect of Modified Cleaning
Pulses on Pulse-Jet Filter Performance. Filtr. and Sep. 15:393, 1978.
22. Ellenbecker, M.J. and D. Leith. Dust Deposit Profiles in a High Velocity
Pulse-Jet Fabric Filter. J. Air Poll. Control Assoc. 29:1236, 1979.
279
-------
PERFORMANCES OF DUST LOADED AIR FILTERS
By: C. Kanaoka, H. Emi, M. Ohta
Department of Chemical Engineering
Kanazawa University
Kanazawa 920, Japan
ABSTRACT
It has been realized that dust load in an air filter enhances particle
collection and pressure drop across the filter. However, it was almost imposi-
ble to predict its effects on both properties properly because of complexity
in the structure of deposited particles.
In this study, growing process of particles on a fiber was simulated and
it was utilized to the calculation of a collection efficiency of a dust loaded
fiber. Experimentally particle agglomerates on a fiber, and collection effi-
ciency and pressure drop of a filter were measured for various filtration con-
ditions and times. Effects of loaded mass and filtration conditions on a sin-
gle fiber collection efficiency and pressure drop were discussed through the
comparison of those results. Finally estimation method for performances of
dust loaded air filter has been proposed.
INTRODUCTION
Researches on the collection mechanisms inside a clean filter where no
particles were captured on the filter, have been studied from various view
points. These advances, however, have not yet reached the stages whereby the
design and opertation of a fibrous air filter can be made on a rational basis,
since mass of dust in a filter changes the performances of filters.
There are very few studies focused on these phenomena. Experimentally,
two different kinds of approaches have been attempted, e.g., microscopic and
macroscopic approaches. The former is the method to observe growing processes
of dendrites on afiber(l,2,3) and is suitable to the analysis of the growing
process and the effect of collection mechanisms precisely.
While the latter is more practical method in obtaining the relations be-
tween collection efficiency and corresponding pressure drop and dust load in
a filter as well as filtration time(4,5). Yoshioka et al. have obtained exper-
imental correlation between collection efficiency of dust loaded fibern and
accumulated mass of particles per unit filter volume m as, am
namAloKT 1 + Xm (1)
where, X is the collection efficiency raising factor and they have obtained 5
m3/kg as the value of \ regardless of experimental conditions.
Theoretical approach is more difficult to describe the process because of
the lack of knowledges on the phenomena. Recently, Payatakes and Tien(6) and
Payatakes(7,8,9,10) have derives deterministic expressions for growing process
of particle dendrites for several collection mechanisms using a set of succes-
Sly?9d:ir?^n!;ial e^a*1on^ On the other hand, Wang et al.(ll) and Kanaoka et
al(l
-------
tion technique to estimate the process has been proposed. Then single fiber
collection efficiency was evaluated using simulation results. Further it was
utilized to predict performance of a fibrous filter. Obtained results were
compared with experimental single fiber collection efficiency. Estimation
method for pressure drop across a filter based on experimental observation was
also proposed.
SIMULATION OF PARTICLE DENDRITES ON A CYLINDRICAL FIBER
The simulation has been performed in Kuwabara's cell shown in figure 1 by
assuming that his equation in equation (2) holds even at dust loaded condition
and dendrites have no effects on upcoming particles.
2K
ln(x
where K is the hydrodynamic factor and is defined as
1, . a2 3
K - - 2lna + a - 4 - ^
2)]
(2)
1 5
Thus, particle motion can be expressed by,
stf= u -
V
(3)
(4)
where U and V are dimensionless velocity vectors of fluid and particle,
respectively.
In actual filtration, spatial distribution of aerosol particles is random
at a microscopic level, even when time average concentration is kept constant.
In other words, each particle passes through a plane, which is normal to mean
flow direction, at an arbitrary time and position (y,z) of the plane as shown
in figure 1.
Calculations were carried out by following procedures;
1) Construct three dimensional mesh network, dimensions of which were deter-
mined equal to particle diameter to be able to store one particle in it as
shown in figure 2.
Kuwabora
cell
Figure 1 Schematic diagram of particle motion
in Kuwabara's cell
281
-------
4) C^inuTstepltt^one of dendrites reaches to the outmost layer of the
5) ReJJStksteps 3 and 4 several hundred times to obtain stochastically suffi-
C1SiLlations°werencarried out for various filtration conditions but packing
density of the filter a and particle density Pp were fixed at a=0.03 and Pp=
lg/cm3, respectively.
Simulation Results
Figure 2 is a typical simulation result. The numbers in each particle
show the order of deposited particles. As seen from the figure, most of parti-
cles are captured on deposited particles rather than on the fiber surface and
further their order is random. The shape of the particle dendnts simulated
is similar to the observations shown in figure 3. Since growing process is
random, number of stored particles in each mesh were counted and averaged and
were plotted against circumferential angles. Calculated number distributions
for R=0.1 were shown in figure 4- It is remarkable that maximum deposition
appears at some certain angles and it increases as Stokes number decreases.
Same tendency was observed in experiments as shown in figure 5.
Single fiber collection efficiency with M particles on a fiber of unit
length nTTM can be evaluated by equation (5) knowing the average imcormng par-
ticles during the collections of M-th and (M+l)st particles NM, since mass of
particles in unit volume of a filter is expressed by equation (6).
nTIM - VNM (5)
m = MppR2 (6)
Figure 6 shows the increases in calculated efficiency with deposited mass.
As seen from the figure, it increases almost linearly with m which is the same
as previously reported tendency(5) but value of slopes X varies with filtra-
tion conditions. Table 1 shows the calculated values of A. A shows the highest
at St=0 and lowest at St=°° for any interception parameter R. This means that
collection efficency increases rapidly with dust load m in small St region.
TABLE 1. COLLECTION EFFICIENCY RAISING FACTOR A
R
0.
0.
0.
St
05
1
2
0
56
14
3
.0
.1
.8
.09
0.4 0.6 1.
45.1 16.
12.2 - 2.
1.82 0.
0 1.4
4 0.873
89
783
2.0
0
0
0
.238
.253
.144
oo
0.
0.
0.
0769
0696
0482
282
-------
Figure 2 Cross sectional view of mesh network
and a typical simulation result
,**.
*7.00x10 |
-------
2.0
^-, 1.5
1.0
05
Figure 4
Figure 5
Figure 6
st=o.o
10
in
« 6
t_
O
Q.
St'l.O
R = O.I
i.o
Trad]
1.0
6 [rod]
iy2
Change in average number of deposited particles
on a fiber with time
i.o
0.8
0.6'.
•§• '0.4
0,2
0.0
S-r -0.85
t=20hr
—, w-0.028
1 .U
0.8
•3 0.6
0}
O
1 0.4
z
0.2
O.O
Sf = 1.70
t =2 hr
i/ = 0.031
-
Tlv
90
0 30 60 90 0 30 60
0 Cdeg] & Cdeg]
Experimental distribution of collected particles
on a fiber
0.5 |.o
m Ckg/m3]
Collection efficiency of a single fiber
with dust load
284
-------
MEASUREMENT OF COLLECTION EFFICIENCY OF A DUST LOADED FILTER
Figure 7 is a diagram of the experimental apparatus used. Sodium chloride
and methylene blue particles were used as test aerosols. They were generated
by an ultrasonic nebulizer and a vibration orifice monodisperse aerosol gener-
ator(TSI Inc., Model 3050). Glass fiber mat, which is composed of three thin
differential filters and fiber diameter of which is llym, was used as a test
filter. After aerosol was deionized with radioactive isotope 21flAm, it was
introduced to the test filter. Escaped aerosol particles from it were recovered
with backup absolute filter. Pressure drop and aerosol concentration were mon-
itored during the experiment. For given set of filtration conditions, experi-
ments were carried out repeatedly but with different exposure times. Accumu-
lated mass in each differential filter and backup filter were determined by
colorimetric and titration methods.
Single Fiber Collection Efficiency
Figure 8 shows typical relations of m and CVt. Although values of m scat-
ter in large CVt, m increases with CVt and filtration velocity V as a whole.
Since collection efficiency of a differential filter is expressed by equation
(7), Em was obtained through graphical differentiation. Further single fiber
collection efficiency was calculated from equation (8) assuming that it holds
for a thin differential filter.
(7)
dm
am
cm " d(CVtrA
IT (1 - a) Df , ,, p v /o\
~ ' 4a AX ln(1 ' V (8'
The calculated values of ^jlm ?rom the above procedure was normalized by
the experimental initial efficiency nfl and it was plotted against m in figure
9. In the figure, ^jlm/^TI increases almost linearly with m and slopes of
these lines A decreases with increasing filtration velocity V. These trends
coincide with the simulation qualitatively but not quantitatively. This is
probably because reentrainment of particles is not taken into account in the
simulation.
1. Aerosol generator
2. Plenum chamber
3. 21(1Am
4. Test filter
5. Absolute filter
6. Orifice and manometer
7. Valve
8. Vacuum pump
9. Light scattering photometer
10. Pressure transducer
Figure 7 Experimental apparatus H- Recorder
285
-------
ot=0.03 jx-O.5cm
Sodium chloride
Figure 8
0 1.0 2.0 3.0x10"
CVt [kg/rrfl
Experimental relation of dust load and
total inlet mass of particles
0
1.0
Figure 9
0.5
m Ckg/m33
Experimental collection efficiency of a fiber
0
Figure 10
23456
cvtxio2 tkg/nf]
Comparison of experimental collection efficiency
of dust loaded filter with predicted ones
286
-------
Since it has become clear that single fiber collection efficiency is approxi-
mated with a linear function of m, time dependency of collection performance
of dust loaded filter was estimated by following procedure.
When riam is a linear function of m, filter efficiency Em is expressed
analytically as exp(-XACVt)
L ~ '
exp(-AACVt) + exp(AX) -
The estimated filter efficiency using the equation was compared with the
experimental results with stearic acid particles in figure 10. Value of 6m3/kg
determined from experimental correlation, was used as X in the calculation.
As seen from the figure, predicted curves agree well with the experimental
result at arbitrary total imcoming particles and filter depth.
ESTIMATION OF PRESSURE DROP OF A DUST LOADED FILTER
Drag theory, which is said suitable to the analysis of pressure of high
porosity bed like a fibrous filter, was used in this study.
Based on the idea of the drag theory, pressure drop across a filter can
be expressed by following equation.
(10)
here, Fm denotes drag force of fiber of unit length and it is defined by,
j_ O'f V / 1 T \
It is evident from the above equations that CQ™ and Dfm are the only terms
that change with deposition of particles on a fiber. Consequently, they were
investigated experimentally using model filters(Df=30ym, 2h=190ym) and methy-
lene blue particles(Dp=lym). Figure 11 and 12 show experimental correlations
of them with m. They increase smoothly with m and approximated by following
equations.
Dfm = Df(l + 0.28m1'28) (12)
Com = CD(1 + 0.064m) (13)
here, Cn is the drag coefficient of a clean fiber and following experimentally
obtained correlation was used.
CD = 24.8/Re (14)
Substituting equations (11)^(14) to equation (10), pressure drop across
a filter comprised of 44 model filter layers was calculated and was compared
with experimental result in Figure 13. Predicted pressure drop agrees very
well with the experiment. Although proposed method has been tested for the
pressure drop of a model filter, this method could be applied to the estima-
tion of real filter at arbitrary condition knowing changes in Dfm and Com as a
function of dust load m.
CONCLUSION
Collection process of particles and pressure drop of a dust loaded filter
has been studied both theoretically and experimentally.
Growing process of particle dendrites on a fiber was simulated by Monte
Carlo method. It was related to single fiber collection efficiency. As a result,
it was expressed with a linear function of dust load and that of a clean fiber.
Plausibility of the simulation was verified experimentally.
Further, estimation method based on drag theory has been proposed and
287
-------
1.5
o
f
E
—• i o
o!
df « 30pm , df * I.Opm
H =240pm, 2h>190pm
v » 0.5 m/sec
ot = 0.0156
Figure 11
0 0.5 1.0 1.5
m tkg/m'3
Experimental relationship between Dfm/Df and m
1.3
"1.2
CJ
1.0
df - 30 pm, dp- 1.0 pm
H • 240pm, 2h- 190pm
o(* 0.0156
v »• 0.5 m/s«c
Com/Ceo- 1+ 0.064m
3.0
Figuer 12
0 1.0 2.0
m [kg/m3]
Experimental relationship between
CDrn/Cn and m
05
key
0
df -
H =
L •
v =
Oi
No
'
Experimental data
Calculated
30pm ,dp • 1.0 pm
240pm, 2h « 190pm
0.24cm
0.5 m/sec
0.0156
of layers 10
i 1 l
«-°-^
__L J 1
~~l 1 1 ;
/
/ '
y*
i
0
001
" 3.0
key_
Experimetol data
Calculated
df " 30 jjm,dp ' I.O jjm
H •24Ojim,2h-l9Ojim
L • .06 cm
v • O.5 m/sec
« ' 0.0156
No of ayers 44
0.02 0.03 0.04 0.05 0.06 0.07
Cvt [kg/m2D
25l
001
005
Figure 13
OOZ 003 OO4
Cvt Ckg/m23
Comparison of experimental pressure drop of dust loaded
filter with predicted ones
288
-------
tested with experimental observations with a model filter. Evolution of filter
efficiency and pressure were estimated using proposed methods.
NOMENCLATURE
A constant used in equation (9)(=4n a/firD-pO-a)}) [-]
C particle concentration a [kg/m3]
CQ drag coefficient of a fiber [dyne/m]
D diameter [ym]
F drag force of a fiber of unit length [dyne]
h half distance of the center of two fibers in a model filter [ym]
M,m number and mass of collected particles in unit filter volume[l/m3][kg/m3]
NM average generated particles during the collection of M-th and (M+l)st
particles [-]
AP pressure drop [dyne/cm2]
R interception parameter(=Dp/Df) [-]
Re Reynolds number(=Dfupf/y) [-]
St Stokes number(=ppDpU/9yDf) [-]
u,U,V average and x, y directional fluid velocity [m/sec]
t,x,y,z time and Cartesian coordiantes [sec][ym]
a packing density of a filter [-]
n single fiber collection efficiency [-]
K hydrodynamic factor [-]
X collection efficiency raising factor [m3/kg]
p density [kg/m ]
^ stream function [-]
Subscripts
f fiber
I interception
M,m M particles and dust loaded
p particle
T inertia
REFERENCES
1) Leers,R.: Staub, 50, 402(1957)
2) Billings, C.E.: Ph.D. dissertation, Calif. Inst. Techno!. (1966)
3) Bhutra, S. and Payatakes, A.C.: J. Aerosol Sci., 1_0, 445(1979)
4) Kimura, N.and linoya, K.: Chem. Eng. Japan, 28, 729(1964)
5) Yoshioka, Y., Emi, H. and Sone, H.: Chem. Eng. Japan, 33, 1013(1969)
6) Payatakes, A.C. and Tien, C.: J. Aerosol Sci., 7_, 85(1976)
7) Payatakes, A.C.: Filtration and Separation, J3_, 602(1976)
8) idem: Powder Technol., 14, 267(1976)
9) idem: AIChE J., 23_, 192~fT977)
10) idem and Gradon, L.: Chem, Eng. Sci., 35,, 1083(1980)
11) Wang. C.S., Beizaie, M. and Tien, C.: AIChE J. 23_, 879(1977)
12) Kanaoka, C.,Emi, H. and Myojo, T.: Chem. Eng. Ronbunshu, -5_, 535(1978)
13) idems: Chem. E. Symposium Series No. 59, 3:4/1(1980)
14) idems: J. Aerosol Sci., Jl_, 377(1980)
15) Kuwabara, S.: J. Phy. Soc..Japan, Ij4, 527(1959)
289
-------
ELECTROSTATICALLY ENHANCED FABRIC
FILTRATION OF PARTICULATES
by
1
Teoman Ariman
and
Stuart T. McComas
INTRODUCTION
Two main concerns are of importance in reducing the dust level in the gas
leaving any process; the efficiency of collection and the energy cost in
attaining a prescribed level of cleanliness in the exhaust gases. Both of
these factors depend on the filtration system employed. The energy consump-
tion of a fabric filter air pollution control system is a function of the
pressure drop across the fabric-dust cake combination and the work required
to remove collected dust from the fabric. Pulse jet systems, for example,
consume about an equal amount of energy in cleaning as in over-coming pressure
drop. The total cost may be on the order of 30c/cfm-year.
Fabric filtration systems have been employed in industry for over a
century with relatively few technological modifications [1,2]. Indeed, there
have been in recent years only a few advances of note: development of pneu-
matic cleaning techniques, introduction of fabrics capable of withstanding
higher temperatures, and the addition of microporous membrane on the collec-
tion side of a felt or fabric backing. For the most part, there was in the
past little incentive for improvement—the collection of the systems was
satisfactory and the units were inexpensive to purchase. However, with the
recent substantial increase in energy costs attention should now be given to
operating expenses (e.g., energy consumption) and also the possible recircu-
lation of air for further energy conservation. As a result, the filtration
systems of yesteryear may not be the best approach for future applications.
In examining the utility of employing electrostatics to fabric filtration
systems, a literature review was conducted. The results of this review, which
are presented below can be summarized as follows:
Associate Dean and Director of Petroleum and Energy Research Institute, The
University of Tulsa, Tulsa, Oklahoma 74104
2
Professor of Mechanical Engineering Department of Aerospace and Mechanical
Engineering, University of Notre Dame, Notre Dame, Indiana 46556
290
-------
From both the theoretical and experimental standpoints, the application
of electrostatics to fabric filtration systems tends to augment the already
comparatively high value of collection efficiency (with the obvious exception
of a case in which particles and fabric-dust cake are unipolarly charged).
However, work is needed to advance the state-of-the-art in formulating and
predicting pressure losses across the fabric-dust cake system for specific
application. [2-8]
In general, a fabric filter is made up of a woven or felted textile in
the shape of a cylindrical bag or flat supported envelope. The textile
material is contained in a metal housing having inlet and outlet gas connec-
tions, a dust storage hopper, and means for cleaning the fabric periodically.
Woven fabrics consist of parallel rows of yarns in a square array. Open
spaces between adjacent yarns are occupied by projecting fibers. Felted
fabrics consist of close, randomly intertwined fibers compacted to provide
fabric strength. In operation, dusty gas passes through the filter normal to
the fabric and dust particles, and, at the start of filtration deposits on
individual fibers and yarn surfaces. Particles are captured by such mechanisms
as 1) direct interception, 2) inertial impaction, 3) diffusion, 4) electro-
static attraction, 5) gravitational settling, and 6) sieving.
Additional particles then deposit and accumulate on already deposited
particles forming filamantous aggregate structures which project into the
gas stream. As deposition and accumulation continue, openings between yarns
and individual fibers become occupied by aggregates and reduced in size. If
this process continues for a sufficient time, a more or less continuous dust
cake forms. Particle collection then occurs by mechanisms normally associated
with porous granular media. Thus the fabric filtration process may be con-
sidered to consist of at least three distinct phases: (a) initial or early
deposition, when depositing particles land on individual fibers, fibrils or
filaments of the yarns; (b) intermediate deposition when particles accumulate
on previously depositied particles, (c) the continued deposition of particles
on a matrix similar to a granular layer, leading to the formation of a filter
cake.
The membrane covered fabric, trade name Gore-Tex, is purported to behave
in a manner different than described above for plain fabric bags. The manu-
facturer states that this bag resists dust cake buildup because the filtra-
tion occurs on the surface of the membrane and that the release characteris-
tics are superior. The result, as they state, is extended operation at a
near clean bag condition.
When dealing analytically with the collection of dust from an air stream
past a cylindrical fiber, inertia, interception, sedimentation and diffusion
of particles are generally taken into consideration. Electrostatic effects
are mostly neglected although they often occur and may play an important role.
[8] However, for essentially isolated spheres and cylinders comparisons of
the collection efficiency for electrical forces with that due to other mech-
anisms show that for low velocities, the electrical collection may be clearly
dominant. Experimental confirmation of this prediction has been obtained [5].
Past results have shown filter efficiency to increase inversely with velocity
and directly with fiber charge. An extensive review of the subject of the
electrical behavior of aerosols is given by Whitby and Liu [9].
The present study was initiated to evaluate the performance of an electro-
statically augmented bag filter system using dust samples from the foundries
of the General Motors Corporation. A comparison of charged and uncharged
291
-------
cases were made using a different air flow rate for each sample using dacron
fabric filter bags. The performance of a Gore-Tex bag was also investigated
with and without electrostatically charging the particles.
TEST APPARATUS AND PROCEDURES
All tests to evaluate the effect of electrostatic charge on bag filter
operation and effectiveness were carried out using an electrostatically aug-
mented single bag test unit supplied by the Apitron Division of American
Precision Industries, Inc. This unit basically is a combination of a single
unit cylindrical electrostatic precipitator and a single fabric filter bag
in series. Several modifications to the apparatus were made by Notre Dame
personnel including a continuous dust feeder system and sampling tubes and
absolute filters located at the inlet and exit of the unit. Figure 1 is a
schematic drawing of the apparatus as modified.
Inlet air to the test apparatus was infused with dust particles by means
of a dust feeder shown in the lower left hand corner of the figure. Air at
a regulated pressure of 12 to 18 psig was applied to a distribution tube
located inside the chamber from which small jets of air impinged on the sur-
face of the dust pile causing a dust storm to form. The number and diameter
of the jets along with the supply air pressured governed the rate of dust
delivery to the air in the upper portion of the dust feeder (storming chamber).
In order to retain a relatively unpacked surface when dealing with sticky dust,
a mechanical stirring device was added to this chamber. The pressure in the
chamber was maintained at or near atmospheric by means of a cloth covered
vent located on the top of the chamber.
The mixture of air and dust was drawn into the ionizing wire tube by
means of an air jet vacuum pump and then through the one square foot area of
fabric filter mounted on a metal cage. A calibrated orifice mounted just
prior to the jet pump was used to determine the volumetric flow rate with an
inclined water manometer being used to measure the pressure drop across this
flow measuring device. The pressure drop across the fabric bag filter was
measured by means of pressure taps located just prior to the lower metal
cylinder and in the plexiglas outer shell. A U-tube water manometer was used
to measure this pressure difference.
An aspirated sampler tube was placed in the inlet dirty air line. The
air and dust sample passed through an absolute filter in order to measure the
rate of dust flow. The volumetric air flow rate through the sampler and
filter was monitored. A sampling tube and an absolute filter were placed in
the exhaust line prior to the flow measurement orifice and jet vacuum pump in
order to determine the amount of dust remaining in the air after filtration.
A small diameter tungsten wire located in the lower cylinder was used to
create an electric corona and thus to charge the incoming dust particles.
This wire was supported at the top of this cylinder by the teflon blow pipe
thus insulating that end of the wire from the metal cylinder. The radial
distance between the ionizing wire and the metal cylinder is approximately
2.8 centimeters. The lower end of the wire was connected to an electric
terminal which passed through a ceramic insulator. This terminal was con-
nected to an adjustable high voltage supply with the metal cylinder being
grounded to that unit.
A high pressure air jet pulse was used to clean dust off the lower col-
lector tube. Air was delivered to the blow pipe located at the top of this
tube by means of a three way solenoid valve. When cleaning was in progress,
292
-------
OUTLET
TEST PORT
AP TAP
ORIFICE
APTAP
PLEXIGLAS OUTER
SHELL (REMOVABLE)-*-
BAG 8 CAGE
MAGNEHELIC
PRESSURE
GAGE
BLOW PIPE
NEEDLE
VALVE
3 WAY
SOLENOID
VALVE
SEALED WATER
JACKET
COLLECTOR TUBE
IONIZING WIRE
INLET TEST PORT
FLUE GAS
FROM AFBC
COLLECTOR BIN
WITH ACCESS
DOOR a DRAIN
PLANT
SUPPLY
110 V
PLANT
SUPPLY
110 V
COMPRESSED
AIR
REGULATOR
SHUT-OFF VALVE'
PRESSURE TANK
Figure 1. Test Apparatus - Modified Apitron Single Bag Test Unit
293
-------
the air supply to the jet pump was interrupted and thus the through flow of
dirty air to be filtered. At the same time, the solenoid valve located in
the 6 psi pressure air line connected to the plexiglas outer shell was opened
to ensure a reversal in pressure difference across the bag thus snapping the
bag in order to remove collected dust from the bag. The cleaning time was
preset by the timer at 0.18 seconds. The frequency of cleaning is controlled
by varying the setting of the timer which actuated the two solenoid valves.
The dust cleaned from the filter bag and lower cylinder were collected in a
bin located at the bottom of the apparatus.
The plexiglas outer shell attaches to the lower portion by means of a
bolted flange. In order-to replace the fabric filter bag, this outer shell
was removed and the old bag was detached from the metal frame which held it
in place. A new bag was slid down over the frame and hose clamps were used
to seal it to the frame at top and bottom.
At the start of each test run, the dust feeding chamber loaded with the
appropriate dust was weighed. After one or two hours had ellapsed, the run
was interrupted in order to reweigh the chamber to determine the amount of
dust which has been delivered to the test unit. This procedure was repeated
throughout the running of a test. A test was started by opening the valve
which allowed pressurized air to be delivered to the jet vacuum pump and the
air flow rate was set. The opening of the main compressed air line valve also
applied pressure regulated air to the storming chamber and the tube connected
to the top of the outer shell. The timer was actuated at the same time and
if the particles were to be charged, the high voltage source was adjusted to
the appropriate level.
During the test run the pressure drop across the bag filter was recorded
every one-half hour on the average. In certain cases where this pressure drop
was changing relatively rapidly, recordings were made more frequently. When
it was determined that the bag fabric had been sufficiently seasoned, namely,
that no appreciable increase in pressure drop took place thereafter, the inlet
and outlet samplers were actuated and recordings of flow through each were
recorded. For each of these two,samplers, the absolute filter was weighed
prior to its insertion into the holder. Upon completion of the sampling
period the absolute filters were again weighted. Because of the relatively
high dust loading, the inlet sampler was operated for about 10-15 minutes.
The exit sampler was operated for one and one-half hours on the average due
to the small rate of dust accumulation on that absolute filter. The amount
of dust removed from the dust feeder chamber was also determined during the
sampling period.
RESULTS
Tests were run with air containing no dust in order to determine the
clean bag pressure drop of the dacron bag and the Gore-Tex bag. The results
are indicated in Table 1. The flow rate area through each bag is one square
foot.
As indicated from this data, the clean bag pressure drop for Dacron
material is negligible for the flow rates used in the present investigation.
On the contrary, the results for the Gore-Tex or microporous membrane coated
bag are very significant.
294
-------
Air Flow Rate
(CFM)
5
6
7
8
8.5
Table 1. Clean bag AP
Dacron
0.04
0.04
0.05
0.06
AP (inches water)
Gore-Tex
3.1
4.0
5.0
6.0
6.5
A summary of all the test runs are given in Table 2. Many preliminary
runs had been made in order to establish the operating characteristics of the
unit using a silica dust. An earlier attempt was made to connect the inlet
pipe to the collector bin and place the air jets inside that bin to create a
supply of dust laden air. This attempt was made in order to try to have a
continuous supply of dirty air without the necessity of reloading a supply
chamber. Because of the difficulties caused by the air pulse cleaning this
method had to be abandoned and the storming chamber as previously described
was fabricated.
Test runs number one through five were made with course dust supplied by
GM. The source of this material is unknown. The characteristics are that
there is some fine dust in this sample but much of it is very large particles.
During test run one it was found that the pressure drop across the bag kept
on increasing up to values on the order of 18 inches of water. Visual obser-
vation indicated insufficient snapping of the bag and thus a continuous build-
up of a dust cake occurred. The frequency of bag snapping and the pressure
of air delivered at the top of the unit during cleaning were varied in order
to establish limits on these parameters in order to obtain proper cleaning of
the bag. By this method, periods between cleanings of 30 seconds was estab-
lished as the time to be used for all tests. Air at a pressure of 6 psig was
established as being sufficient to properly snap the filter bag.
Test number two with course dust, an air flow rate of 4 CFM and no charge
showed a very small pressure drop across the bag. For this reason no charge
in case was run at this flow rate. At a flow rate of 8 CFM, test number three,
and no charge the pressure drop was 3.1 inches of water and a collection effi-
ciency of 99.22%. This test was run with the same bag as used at 4 CFM thus
some seasoning of the bag had occurred. Test four repeated the conditions but
with a new bag. The inlet dust loading was very much lower in this case and
so was the pressure drop. The test was again repeated but with charging at
12 KV. The pressure drop was only one-tenth the value for the no charging
test number 3 and the collection efficiency increased even though the inlet
loading was somewhat higher. Because of the large size of some of the course
particles, much of this dust may never have reached the bag. Instead, gravity
effects may have prevented the large particles from being carried along with
the air. This could contribute to the relatively low pressure drops measured
using this particular dust.
Test runs numbered six through 11 were made using dust from the foundry
cleaning room. This material contained quite a bit of metal particles and
295
-------
TA.BLE *
SUMMARY OF TEST DATA
MODIFIED APITRON SINGLE DAG UNIT
NJ
vO
Tent
No.
1
2
3
4
5
6
7
«
9
10
11
12
1]
14
15
16
17
IS
19
20
21
22
Material
C
C
C
C
C
CRD
CRD
CRI>
CRD
CRD
CRD
AMI)
AMD
A»OI
AMD
AMD
AMD
AMD
AND
AMD
AMD
AMU
Air flow Rate
(CFM)
8
4
8
8
8
5
8
5
8
8
•
5
10
5
8
5
8
8
8
8
7.8
8
Charge
(Kllovolta)
0
0
0
0
12
0
0
12
12
0
12
0
0
0
0
12
12
12
8.3
0
0
11
AP Acroaa Bng
(Indira Hnter)
1.8
0.1
3.1
0.5
0.3
5.0
10.4
1.7
2.3
10. 5
2.8
2.4
4.4
5.0
6.5
0.2
2.5
4.6
6.8
10.8
9.8
8.6
Inlet Dtint Loud
(Rm/cn.Ct.)
0.23
0.11
0.27
0.0')
0.37
0.68
0.21
0.82
0.57
0.18
0.12
0.02
0.02
0.22
0.10
0.17
0.08
0.25
0.18
0.22
0.25
0.19
collection err
m
_
99.40
99.72
99.22
99.96
99.95
99.98
99.99
99.99
-
99.93
-
-
99.96
99. 9J
99.99
99.99
99.54
99.75
99.33
99.48
99.54
Tent Determined Required Snipping Time and Prensure-
Ncv D Ban
Sane Bng na 2
Ncv D Dag
New D lt.ig
New D Bag
Same Bng as 6
Hew D Bag
Snme Bag aa B
Snme Bag aa 8 6 9
Same Bag aa 8 -*• 10
New D Bag
Snme Bng na 13
New D Bag
Snme Bng aa 15
New D Bag
Snme Bag as 17
New D Bag
Same Rng aa 24
Same Bag aa 24 4 25
New C Bag
New G Bag
Diint Miter In I
C - Coiirae
CRII - Clcmilng Room
AMI! - Arc Melt Kiirnnce
Bag Material
D - Hncron
G - Gore-Tcx
-------
seemed to readily accept an electric charge.
5 CFM without charge and the test completed.
to 8 CFM using the same bag and a second test
for the charge case at 5 and 8 CFM. At 5 CFM
The flow rate was first set at
Then the flow rate was increased
run. This procedure was repeated
the charge reduced the pressure
drop by a factor of 3 and at 8 CFM this value was reduced by a factor of over
4. The reduction in each of these cases occurred even though the inlet dust
loading was larger for the charge case. The variation in pressure drop with
time for these four tests are shown in Figure 2.
After running at 5 and 8 CFM
with charge, the charge was removed
and another test run. The pressure
drop value increased and leveled
off after 12 hours to a value near
that of the previous no charge case
with just a slightly smaller inlet
loading. Reapplying the 12 KV
charge at 8 CFM flow rate for the
same bag reduced the pressure drop
back down to 2.8 inches of water
after 10 more hours of running.
This is not quite as low as the
previous charged case even though
the inlet loading was much lower.
Figure 3 indicates the pressure drop
variation with time for these two
tests. A difficulty arose in all
charge runs with cleaning room dust
in that dust would buildup on the
lower ceramic insulator causing
an electric short from the wire
to the outer case. A deflector
had to be placed above the insu-
0000
LO-O-DJ
|-a-cran-
TJMT (HRS)
FXGUSI 2. DACROK BAG CLEARING ROOK DUST
lest 6
lest 7
lest e
lest 9
5 CFM
8 CTH
5 CFM
8 CTH
No Charge
No Charge
12KV Charge
12m1 Charge
Inlet Load
Inlet Load
Inlet Load
Iclet Load
0.21 gu'ft,
0.62 gm.'it,
0.57 gn.'ff5
later to reduce this buildup making
it possible to run the previous cases. The collection efficiency for all test
runs using this dust was extremely high and was better or the same when the
particles were charged except when the charge was reapplied in test number 11.
Test runs using arc melt furnace dust are number 12 through 20 using a
dacron filter bag. This dust did not seem to accept a charge as readily as
the cleaning room dust and it was extremely fine and stickly. A stirring
mechanism was required in the dust feeder when this dust was used in order to
keep the dust surface loose. Otherwise the air jets would dig a hole and com-
pact the surface of the dust resulting in very little being entrained in the
chamber air leading to extremely low inlet dust loading. In general, this
dust tended to stick to everything. It would also buildup on the ledges of
the frame holding the fabric filter bag and would not release when the bag
was snapped. Any bag system using a frame requires a design which will prevent
this from occurring otherwise excessive pressure drop will result. Tests 12
and 13 were run prior to addition of the mechanical stirrer. The low inlet
loading thus obtained resulted in lower pressure drop then for the higher
loadings of runs 15 and 16 also with no charge which were made after the addi-
tion of the mechanical stirrer. The addition of a 12 KV charge reduced the
pressure drop very significantly for 5 CFM and by about 2 1/2 times at 8 CFM.
297
-------
The collection efficiency was
extremely high for all four runs
and no effect of charge was noted.
The pressure drop versus time for
these four tests, numbers 14 to 17,
are shown in Figure 4.
The strength of the electric
field was varied in order to deter-
mine the resultant effect. A
voltage of 12 KV was first applied,
then 8.5 KV and next no charge.
These are tests 18 through 20. The
pressure drop did increase as volt-
age was decreased. The highest effi
ciency occurring at 8.5 KV is
unexplained.
The last set of tests, number
21 and 22, were run using Gore-Tex
bags made from material supplied by
W.L. Gore and Associates, Inc.
Without charge, the flow rate was
first set at 8 CFM but had to be
reduced to 7.8 CFM due to the in-
IIMI (HRS)
Figure 3. DACROK BAG - CLEAKTSG ROOM DUST
BAG SEASOK IX TEST 8(9
Test 10
Test 11
6 CFK
8 CFM
Ho Charge
12m' Charge
Inlet Load
Inlet Loac
0.1B gm/ftr
0.12 fB/ft
ability of maintaining the higher flow rate was maintained at 8 CFM. The inlet
load was about the same as used to test the dacron bags. A slight reduction in
pressure drop was obtained for the charged case as compared to the uncharged
case but the inlet loading was also slightly lower. The Gore-Tex bag did reach
a constant pressure drop situation in a shorter time than did the dacron bag
but at values near or above those for dacron bags. The pressure drop versus
time for the Gore-Tex bag tests are
shown in Figure 5. It should be
noted that the clean bag pressure
drop for the Gore-Tex bag was higher
than for the seasoned dacron bag.
The collection efficiency was approx-
imately the same as those for dacron |
bags. *
The effect of charge seems to f
be very much smaller for the Gore-Tex I
membrane surface bag as compared to I
the dacron bag. This is probably |
due to the dust being filtered at the |
membrane surface where only small
pores exist. Under this surface con-
dition the effect of charge cannot
change the dust cake buildup charac-
teristics as is the situation with a
fabric filter without this special
coating.
•
•
% O 0 O
1
•
•
I
•
>
. o o-o-o-l
j
X
— ** —
• v
i;
D
D
r°
X
* X
X
ho DircH
3
TIHE (HRS)
FIGURE 4. DACROK BAG - ARC KELT ITOSACE DUST
• Test 14 5 CFK
X Test 15 8 CFM
0 Test 16 5 CFK
D lest 17 8 CIH
Ho Charge Inlat Load 0.22 pi ft,
Ho Charge Inlet Load 0.10 gi/ft"
12m1 Charge Inlet Load 0.17 gB.'ft*
12FV Charge Inlet Load 0.08 gs/ff"
298
-------
x x
TIME (HRS)
FIGURI 5. GORE-IB! BAG - ARC MELT FDIWtCE BEST-
X Tut 21 « CTH Ho Cbirge lal.t Lo«d
D Tut 22 8 cm 12 KV Ch*r(« Inlet Laid
0.25 |
0.19 I
CONCLUSIONS
A test program carried out on
the Apitron single bag test unit
using three different dust samples
supplied from General Motors Cor-
poration foundries indicate that
electrostatically charging the |
particles decreases the pressure *
drop across the fabric filter bag f
as compared to the uncharged case. =
Tests run to study the effect of I
variation of voltage applied to 1
the ionizing wire indicate that I
pressure drop across a dacron
fabric filter bag decreases with
increased voltage. The collection
efficiency was high for the fabrics
used and for this test unit and the
charging of particles, in almost
all cases, caused an increase in
collection efficiency.
The pressure drop decrease
when particles were charged prior
to filtration through a dacron bag was significant. However, only a slight
effect was found for the arc melt furnace dust when using the Gore-Tex
material. Also, the clean bag pressure drop for the Gore-Tex bag was quite
high when compared to that for dacron bags. There did not seem to be any
advantage, and possibly a disadvantage, in using the Gore-Tex material.
A buildup of dust material was found to occur on the insulators holding -
the ionizing wire. If the dust particles are highly electrical conducting
this caused an electrical short across the insulator and a subsequent over-
load on the high voltage supply. A final design for operational size bag
house filter incorporating electrostatic charging of the particles must take
into account this problem. Dust deposits on the insulator from the inflowing
stream of air and dust and when the air pulse cleaning blows dust down past
these insulators. Locating the lower insulator below the incoming air stream
would help. Also, both insulators should be protected by means of some type
of deflector.
REFERENCES
1. Billings, C.E. and Wilder, J., Handbook of Fabric Filter Technology, Volumes
I-IV, GCA-TR-70-17-G, prepared for NAPCA (EPA) (1970.
2. Butterworth, E., Manufacturing Chemist, 65, Feb. (1964).
3. Helfritch, D.J. and Ariman, T., "Electrostatic Filtration and the Apitron
Design and Field Performance," Novel Concepts, Methods and Advanced Technology
in Particulate/Gas Separation, Teoman Ariman, Editor, University of Notre Dame,
1979.
4. Ariman, T., Rao, K.S., Yang, K.T. and Hosbein, R.L., "Collection of Dust by
Fabric Filtration in an Electrostatic Field," Proceedings of the Second Annual
Environmental Engineering and Science Conference, p. 555 (1973).
299
-------
5. Ariman, T. and Helfritch, D.J., "Pressure Drop in Electrostatic Fabric
Filtration," Proceedings of the Second Symposium on the Transfer and Utiliza-
tion of Particulate Control Technology, volume III, Particulate Control
Devices EPA-600/9-80-039c, (1980).
6. Ariman, T., "Electrostatic Fabric Filtration in Industrial Dust Control -
A Review," UND-DOE-EFF, TR. No. 1, April (1979).
7. McComas, S.T. and Ariman, T., "Collection Efficiency and Pressure Drop in
Dust Collection by Fabric Filters in an Electrostatic Field," University of
Notre Dame, Final Report for General Motors Corporation, Notre Dame, Indiana,
November (1979).
8. Davies, C.N., Air Filtration, Academic Press, New York (1973).
9. Whitby, K.T. and Liu, B.Y.H., Aerosol Science, Academic Press (1966).
300
-------
A STAGGERED ARRAY MODEL OF A FIBROUS FILTER
WITH ELECTRICAL ENHANCEMENT
By: Frank Henry
Department of Mechanical Engineering
Brunei University
Kingston Lane
Uxbridge, Middlesex UBS 3PH United Kingdom
Teoman Ariman
Associate Dean
College of Engineering and Physical Sciences
The University of Tulsa
Tulsa, OK 74104
ABSTRACT
The staggered array of parallel circular cylinders (fibers) is proposed
as a model for a fibrous filter. The electrical enhancement of the model is
accounted for by the method of images. Stokes1 approximation is utilized in
the analysis of viscous flow around fibers. The predicted pressure drops and
collection efficiencies are compared to experimental data. The model is shown
to over predict the pressure drop but give reasonable predictions of collec-
tion efficiencies in certain cases.
INTRODUCTION
The theory of fibrous filtration is essentially concerned with the pre-
diction of two important parameters. These are: the filter collection effi-
ciency, and the resistance experienced by the gas stream as it passes through
the filter. Theoretically, if the flow field within a fibrous filter could
be mathematically described such parameters could be precisely calculated.
This is assuming that all other transport mechanisms, such as gravity,
inertia, diffusion, and electrical forces, could be described with equal pre-
cision. However, the random distribution of the fibers within a filter makes
it impossible to analytically describe the flow field within a real filter.
Various amplified models of fibrous filters have been proposed. An
account of the history of filter modelling is provided by Davies (1). The
most common model consists of an isolated cylinder in an otherwise undisturbed
flow. Neglecting the influence of the surrounding fibers can only be justi-
fied, however, for especially designed excessively porous filters. An
unsolved theoretical problem with this model is the lack of a closed form sol-
ution to the viscous flow around an isolated cylinder. While the isolated
cylinder model has been quite useful in illustrating the importance of the
various transport mechanisms present in a filter, it has not been possible to
relate the results to real filters without the use of empirical constants.
Kuwabara (2) developed a cell model of Stoke's flow over an array of
parallel circular cylinders. This model has been used by several researchers
(3,4,5) as a fibrous filter model. It has the advantage over the isolated
301
-------
cylinder model of addressing the problem of the influence of the surrounding
fibers. However, the Kuwabara model has been shown to have significant short-
comings (6). Another model of flow over parallel cylinders proposed by
Happel (7) has similar deficiencies. Yeh (5) constructed a numerical model
of flow over a staggered array of cylinders to investigate Kuwabara's model.
He provided results for a particular array configuration and for a packing
density of 0.1. Hence, it appears that the use of a staggered array of par-
allel cylinders as a model for a fibrous filter has not been fully investiga-
ted. This paper attempts to address this problem by solving for the flow over
various configurations of parallel cylinders arrays. The effect of an exter-
nal electrical field is also considered in the analysis. The predicted pres-
sure drop and collection efficiency of the model are compared with experimen-
tal data.
THE NUMERICAL MODEL
The finite difference technique was used to solve Stokes' flow over an
array of parallel cylinders shown in Figure (1). Polar coordinates were
used in the two regions close to the cylinders and Cartesian coordinates were
utilized in the connecting outer region.
The biharmonic Stokes' equation was split into two Poisson equations,
i.e.
V2^ = -oo (1)
and
V2w = 0 (2)
where ij; = stream function, and to = vorticity. The finite difference
analogues of equations (1) and (2) were solved using the Gauss-Seidel itera-
tive technique. For any interior node, (i,j), and Cartesian coordinates,
this can be given as
where the superscript k denotes the iteration step and
Ax.Ax. 1 Ay.Ay. - (*/
1 i—X j 3~1
. = 2/[Ax.L(Ax1 + Ax,.^^)]
302
-------
- 2/[Ayj(Ayj + Ay^)] (7)
and
Ei,j = 2/[Ayj-l(Ayj + Ayj-l)]
Similar expressions can be given for equation (2) and for equations (1) and
(2) in polar coordinates.
The symmetry of the array requires the top and bottom boundaries of the
solution domain to be both streamlines and lines of zero vorticity. The
values of the vorticity and streamfunction on the vertical boundaries were
calculated by employing the periodicity of the solution. The streamfunction
was also known on the cylinders' surface as they are streamlines. However,
on these surfaces the vorticity had to be calculated using an approximate for-
mula , i.e.
u - 2(ip .. - ij; )/Ar2 (9)
s,n vys+l,n rs,n' s v '
where, n is the nodal index in the angular direction and s denotes the cylin-
der surface.
It was found that the numerical solution was insensitive to any reason-
able initial condition. However, the pressure drop, through the vorticity,
was found to be sensitive to the nodal density. A more detailed description
of the numerical model can be found in Henry and Ariman (6) .
ELECTRICAL ENHANCEMENT
The collection efficiency of a fibrous filter can be substantially in-
creased by applying a high voltage (usually of the order of 10 kv/cm) across
the filter. The electric field is usually aligned with the flow direction.
See, for example, the experimental work of Bergman et al (8). Using the
method of images, Henry and Ariman (9) investigated the influence of the
neighboring fibers on the electric field of an enhanced filter. As the
strength of successive images quickly decreases it is only necessary to con-
sider first-order-images.
For the configuration of thirteen cylinders shown in Figure (2) the com-
plex potential, W, can be approximated by,
W(Z) = c(>e -f i^ = - EQZ + arj Eo f ( ^ - r2 \ J^^ (10)
1=1 .1=1 J J
303
-------
where
4> = electric potential
ib = electric streamfunction
e
E = electric field intensity
a = (ef- l)/(ef+ 1)
e = dielectric constant of fiber
Z. = position of the ith cylinder
r = cylinder radius
Zl. = position of first-order image of cylinder '.j' in
cylinder 'i' .
A charged dielectric particle entering an electric field such as that given
by equation (10) will experience two distinct forces. The first, known as
electrophoresis , is due to the interaction between the particle charge and
the electric field. The second, known as dielectrophoresis, is due to the
interaction between the particles' polarizability and the divergence of
the field. Note that electrophoresis will be present even if the dielectric
constant of the fiber is unity i.e., the lines of electric force are parallel,
whereas, in this case, dielectrophoresis will not occur. Also, as dielectro-
phoresis is proportional to the cube of the particle radius it is less impor-
tant for smaller particles. Because of the above and the fact that a stream-
function for the dielectrophoresis force does not exist, it was decided to in-
clude only electrophoresis as a first approximation to an enhanced fibrous
filter.
The streamfunction for the electrophoresis force ip can be shown to be,
lj;c = q Y Im[W] (11)
where q = particle charge, y = mechanical mobility of particle
and Im[W] = imaginary part of equation (10).
Hence, the streamfunction for the trajectory of an inertialess particle ty
neglecting gravity, diffusion, fluid inertia, hydraulic interaction, dielec-
trophoresis and the image effect of the particle charge and the cylinder, can
be given by superposition as,
1>t = ^c + 4» (12)
where ty and ^ are given by equations (1) and (11), respectively.
304
-------
RESULTS
Pressure Drop
For Stokes' flow, the pressure p is given by,
Vp = - y V x to (13)
where y is the fluid viscosity.
The pressure drop over the solution domain, Ap , shown in Figure (1) was
calculated from the predicted vorticity field usingCSimpson's rule (6). The
total pressure drop, Ap , over an array of finite thickness, h, could then be
estimated by
(14)
where, M = number of cells = h/2b cos(J>
Davies (1) defined a dimensionless pressure drop coefficient, f(c), for
fibrous filters which is only a function of the packing density. This can be
written as,
Aptrf
f(c) = TrM (15)
through the use of dimensional analysis and experimental data for pressure
drops over real filters, Davies found f(c) to be given by
f (c) = 16c1'5(l + 56c3) (16)
Curves of f(c) predicted by the numerical model and equation (16) are shown
in Figure (4). Experimental results derived from the works of Yeh (5) and
Whitby (10) are also included. The pressure drop over the array of cylinders
predicted by the numerical model is considerably higher than the experimental
measurements. Davies' equation (16) can be seen to be mid-way between the
experimental results and the numerical prediction. It appears that a change
in the relative orientation of the cylinders, i.e. a change in cf>, has a neg-
ligible effect on the predicted pressure drop.
A possible explanation for the difference between the numerical and
experimental values of pressure drop is that real filters do not have equi-
spaced fibers. The work of Yano and Kieda (11) shows that the drag created
by two parallel cylinders decrease as the distance between the cylinders'
centers decreases. Hence, if the fibers in a real filter are grouped in
bundles the pressure drop can be expected to be lower than if the fibers were
equi-spaced. As the packing density is only a measure of the volume of the
fibers relative to the volume of the filter it unfortunately, does not give
any indication of how the fibers are distributed.
305
-------
Collection of Inertialess Particles
If it can be assumed that the particles follow the streamlines described
by equation (12), the amount of particles that are intercepted by a cylinder
(fiber) can be estimated. For a specific particle size there will be a crit-
ical trajectory or streamline that just grazes the cylinder. Below this line
all particles will deposit on the cylinder, but particles travelling on higher
streamlines will miss the cylinder. This is schematically illustrated in
Figure (5). Assuming the particles are homogeneously distributed within the
array, the amount of particles entering the solution domain or cell shown in
Figure (5) will be proportional to the cell height, 2bsin. The amount of
particles captured by the cylinder will be proportional to the height of the
critical trajectory as it enters the cell, y*. Hence, the cell penetration,
P , can be estimated by,
2bsincf» - y* y*
P = _ £_ = i _ - £ — (17)
c 2bsin4> 2bsin<{> v '
The total filter efficiency, E, can be given as,
E=1-P-1-PM (18)
c
where, P is the total penetration, and M is the number of cells in the filler
thickness, h, i.e.
M = h/2b cos<{> (19)
Equation (18) can be rearranged to give,
h$
E = 1 - e (20)
where, the penetration function, $ is given by
log(Pc)
2b cos*
Figure (6) shows the dependence of $ on the electric-flow velocity ratio, G,
where G = qyE /E . From these curves and equation (20) , an estimate of the
collection efficiency of an enhanced fibrous filter can be made.
Figure (7) shows the influence of the angular orientation on the collec-
tion efficiency of the model. Two experimental points due to Whltby (10) for
a commercially available paper filter [c = .096] were also plotted. Whitby's
filters were not electrically enhanced but it is possible that, due to nat-
ural charges, electrostatic attraction between particles and fibers was pre-
sent. Also, it appears that, in some cases, the particle sizes and veloci-
ties used in his experiments were such that inertial effects were significant
Hence, it would be expected that the model prediction would be lower than
306
-------
those given by Whitby. This would suggest that the 60 degree model over-
predicts the collection efficiency of a real filter. Figure (7) also shows
the substantial increase in collection efficiency that can be expected if a
filter is electrically enhanced.
Figure (8) compares the predicted efficiency using the 45 degree model
and the experimental results of Yeh (5). Yeh's measurements are of particular
interest because the filter used was specially manufactured using dacron
fibers of uniform diameter. However, the fiber distribution was random. The
experimental results show the influence of diffusion, which is not included
in the model. The smaller particles are more subject to random Brownian
motion and can be expected to deviate from the predicted streamline pattern.
Yeh's particles were made electrically neutral and hence, there were no elec-
trical effects present. In contrast the work of Bergman et al (8) shown in
Figure (9) is for naturally charged particles collected by a fibrous filter
polarized with an electric field of 10 kv/cm. In the numerical model the
average particle charge, q, was estimated by employing a formula given by
Nelson et al (12) i.e.
q = 6(d /CQ1/3)1/2 (22)
P s
where, d is the particle diameter and C is the concentration (1%) of the
sodium chloride solution from which the particles were created. It can be
seen that the model underpredicts the efficiency for the larger particles.
This is probably due to the neglect of dielectrophoresis in the model.
CONCLUSIONS
The staggered array model has been shown to overpredict the pressure
drop of real fibrous filters. It is thought that the discrepancy is largely
due to the nonuniform distribution of fibers within a real filter. Particle
deposition on individual fibers has been shown to be a function of the angu-
lar orientation of the surrounding fibers. While the 30 and 45 degree arrays
predict comparable collection efficiencies the 60 degree array gives much
higher values. The experimental data of Whitby (10) appears to suggest that
the 60 degree array model overpredicts the collection efficiency of real
filters.
It must be remembered that the staggered array model is a two-dimension-
al approximation of a complex three-dimensional flow field and that only a
limited number of transport mechanism were considered. Hence, it can not be
expected that the model will give results that agree precisely with experi-
mental measurements on real filters. The model should be viewed as the next
step on from the isolated cylinder models. This model can describe the
dependence of the collection efficiency on fiber orientation which is an
important feature and can not be achieved by the cell models of neither
Kuwabara (2) nor Happel (7).
307
-------
ACKNOWLEDGEMENT S
This paper is a partial outcome of the research program supported by the
National Science Foundation under Grant No. ENG-77-0938. The second author
acknowledges the continuous interest and advice of Dr. M. S. Ojalvo, Director
of the Particulate and Multiphase Processes Program of the NSF.
ENDNOTES
1. Davies, C.N. Air Filtration. Academic Press, 1973.
2. Kuwabara, S. J. Phys. Soc. Japan. 14:527, 1959.
3. Henry, F., and T. Ariman. A Cell Model of Aerosol Filtration by Fibrous
Filters in an Electrostatic Field. Technical Report, UND-AFEF TR No. 3,
University of Notre Dame, 1979.
4. Pich, J. Aerosol Science. Davies, C.N. (ed.). Academic Press, 244,
1966.
5. Yeh, E.G. Fundamental Study of Aerosol Filtration by Fibrous Filters.
Ph.D. Dissertation, University of Minnesota, 1972.
6. Henry, F., and T. Ariman. A Numerical Study of the Kuwabara Model.
Technical Report, UND-AFEF TR No. 5, University of Notre Dame, 1980.
7. Happel, J. Am. Inst. Chem. Eng. J. 5:174-177, 1959-
8. Bergman, W., H. Hebard, R. Taylor, and R. Lum. Electrostatic Filters
Generated by Electric Fields. UCKL-81926, 1979, presented at the Second
World Filtration Congress, London, England, September, 1979.
9- Henry, F., and T. Ariman. The Effect of Neighboring Fibers on the
Electric Field in a Fibrous Filter. Technical Report, UND-AFEF TR No. 4,
University of Notre Dame, 1980.
10. Whitby, K.T. Am. Soc. Heating, Refrig. Air Cond. Engrs. J. Sept. 56-65,
1965.
11. Yano, H., and A. Kieda. J. Fluid Mech. 97:157-179, 1980.
12. Nelson, G.O., W. Bergman, H. Miller, C.P. Richards, and A.H. Bierman.
Enhanced Filtration. UCID-16949-76-3, 1977.
308
-------
Polar region
Solution domain
Rectangular
region
•• x
Figure 1 Cylindrical Fiber Array
and Solution Domain
o
G
O O
O
Im
o
Re
o
o
o o
Figure 2 Thirteen Fibers
Array for Electrical Field
Cylinder 13
First order
imoges
Figure 3 First-Order Images
v- Numerical solution
\ domain
Particle
2.0
+ 45 Array
60° Array
x YEH-Experimental
O WHITBY-Experimental
0 0.02 0.04 0.06 a08 0.10 0.12 0.14 0.16
PACKING DENSITY - c
Figure 4 Dependence of Pressure Coeffi-
cient on Fibrous Filter Packing Density
and Comparison With Related Experimental
Work
Figure 5 Numerical Solution Domain
and Critical Trajectory
309
-------
-1.0
001
ELECTRIC-FLOW VELOCITY RATIO, 6
Figure 6 Dependence of Penetration
Function on Electric-Flow Velocity
Ratio
0 0.05 0.10 0.15 0.20 0.25
NONDIMENSIONAL PARTICLE RADIUS rp/rf
Figure 1 The Effect of Angular Orien-
tation of Fibers on Collection Effi-
ciency
40
30
20
UJ
o
UJ
5 ,0
- YEH- Experimental
U0= 7.523 cm/s
r(
5.65
h = 0.706 cm
c = 0.0493
Numerical
model
45° array
I
0.025 0.050 0-075 0.160 0.125
U0« 66 cm/s
E0= 10 kV/cm
c = 8.13x10
r. . 2.49 jire
50
0-05
0.25
Figure 8 Comparison of the Predicted Filter
Efficiencies with the Experimental Results
of Yeh (5)
310
Figure 9 Comparison of the
Predicted Filter Efficiency
with the Experimental Results
by Bergman et al (8)
-------
AEROSOL FILTRATION BY A COCURRENT
MOVING GRANULAR BED: PENETRATION THEORY
By
Thomas W. Kalinowski and David Leith
Department of Environmental Health Sciences
Harvard School of Public Health
Boston, MA 02115
ABSTRACT
A penetration model for aerosol filtration by a cocurrent moving bed of
granules has been developed. The model incorporates straight-through pene-
tration and reentrainment of previously collected dust due to granule motion,
both mechanisms having been found significant in experiments. The model, an
extension of classical clean granular bed theory, utilizes the familiar con-
cept of a single granular coefficient for collection and proposes a similar
coefficient for reentrainment.
Reentrainment was found in experiments to be a function of particle size
and other factors such as gas velocity, granule velocity and the extent of
intergranular dust deposit. The single granule reentrainment coefficient for
moving granules, nR» was found to depend upon the product of intergranular
dust deposit and the square root of particle diameter (K/d ) for particle
diameters between 0.16 and 5.5 pm. P
INTRODUCTION
Granular bed filters are of interest for collection of particles from
hot gas streams. A recent review of granular bed filtration theory and
experiments is provided by Yung et al. (1). A common feature of practical
granular bed filters is bed motion for granule cleaning and recharge. Most
previous experiments and theory concerned stationary clean beds. These stud-
ies are useful to explain results from clean-bed experiments but have not
been extended to more practical, loaded-bed cases.
Recent theoretical studies to model loaded granular beds consider
enhanced particle collection by previously captured particles as well as
reentrainment of collected particles (2, 3). Hydrodynamic factors affecting
particle collection by already-deposited particles or by dendrites near the
surface of a granule are complex. Factors affecting reentrainment are also
poorly understood. However, reentrainment in moving granular bed filtration
experiments significantly affects overall bed efficiency (4, 5). A penetra-
tion model for a moving granular bed filter must account for reentrainment
related to granular motion.
PENETRATION MODEL
A previous study of Cocurrent Moving Granular Bed Filter (CMGBF) per-
formance showed that some dust penetrates straight through, and some captured
311
-------
dust is reentrained due to granule motion (5). Penetrating dust from these
two sources had different size distributions. This paper considers these two
separate penetration mechanisms and discusses a model that is an extension of
the classical clean granular bed filtration model, to which it reduces when
reentrainment is negligible.
Mass Balance on the Aerosol
In the CMGBF, clean granules and dusty gas are added continuously at the
bed top, then pass downward cocurrently and exit at the bed bottom. A unique
feature of the CMGBF is the controlled formation of an intergranular dust
deposit to enhance filtration of submicron particles. A mass balance on the
aerosol in the differential volume AdZ is:
Rate In = Rate Out + Accumulation
Q C = Q (C + dC) + E - R
(1)
Solution of equation 1 requires expressions for the collection rate, E, and
reentrainment rate, R.
Particle Collection Rate
Consider particle collection analogous to classical clean bed theory
except that the single granule collection coefficient, ru, accounts for the
effect of dust adhering to the granule as well as for the granule itself.
The collection rate, E, can be expressed as the product of the cross-
sectional area swept clean by these collectors and the particle flux through
the differential volume:
TTd 2/4
V
E- [Adz(l -
(2)
The porosity, e, in the flux expression is the effective porosity
accounting for reduced voidage due to the intergranular dust deposit. The
effective porosity can be estimated by
Particle Reentrainment Rate
Consider a particle reentrainment rate, R, based on removal of dust
already collected and adhering to the granule surface. The mass of collected
dust in the differential volume is:
(1 - £Q)A dZ
K
(4)
Next define a gas parcel as the gas volume equivalent to the volume of a
single granule. The number of gas parcels per second passing each granule
is:
312
-------
(5)
A dimensionless single granule reentrainment coefficient, n , will now
be defined as the fraction of all dust adhering to a single granule that is
reentrained from that granule by a single gas parcel passing it. The reen-
trainment rate in the differential volume is the product of equations 4, 5,
and the single granule reentrainment coefficient:
3 (1 " Eo) Vg
R ~ T ^ 1— P/-iK r^A dZ (6)
2 e dG G R
Substituting equations 2 and 6 into equation 1 gives an expression for
the mass balance on the aerosol:
(1 - e ) (ri C - UpPpK)
dC = -| —2- G A U G dZ (7)
Before equation 7 can be integrated across the depth, Z, of the bed, an
expression for K in terms of dust concentration, C, is required.
Equilibrium Mass Balance
The CMGBF is designed for continuous operation, and at equilibrium all
dust which enters the bed must leave in the aerosol stream or on the gran-
ules. Assuming dust entering on the granules is negligible, an equilibrium
mass balance on total dust in the aerosol and on the granules yields:
Qgci = Y + VGK (8)
Solving equation 8 for K produces:
K = (Q [C± - C])/(QGPG) (9)
Substituting equation 9 into equation 7 produces:
3 (1 ~ Eo) \ \
dc = - 1-—^[G(nG + ^ J) - v± ^]dz (10)
G G G
If porosity, e, is independent of bed depth, equation 10 can be inte-
grated readily. This approximation is valid when the equilibrium dust
deposit is established near the bed surface so that it does not vary appre-
ciably through the bed depth. Pressure drop data indicate that a stable
deposit is formed within two or three centimeters of the bed surface (6).
Integration of equation 10 yields:
313
-------
T; exp [" I
p " Xp u f. c. u u i\ y (11)
Pt""" v-T?
This is the theoretical expression for particle collection and reentrainment
in a cocurrent moving granular bed filter. Equation 11 is for steady state
operation with a continuous dust feed at constant concentration.
If no reentrainment occurs (n + 0) equation 11 reduces to the classical
granular bed penetration expression (7):
Pt_ . = exp[- TT ——-i— n,J (12)
11 o Z. £• d (j
R r*
\j
To demonstrate the effect of reentrainment on bed performance, consider
the situation when the bed is very deep, z-*00; in this case:
Pt
(13)
Even if particles are collected efficiently, ri -*1, significant penetration
may occur if reentrainment is important, TL >0.
Discussion of Theory
Although many theories are available to describe clean granule effi-
ciency, no adequate theories are presently available to predict the single
dirty granule efficiency, TI , or to predict the single granule reentrainment
coefficient, r\ . As a first approximation, a lower bound for efficiency of
the conditioned granule, nr» might be calculated from theory for clean,
unconditioned granules. Reentrainment should depend upon the ratio of sepa-
ration forces to adhesion forces for particle-to-granule and particle-to-
particle interactions. Adhesion forces include van der Waals, electrostatic,
and surface tension capillary forces. Separation forces should be related to
air drag and mechanical shear due to granule slippage. A review of the lit-
erature suggests that these forces, for other than the most simple configura-
tions, are poorly understood and cannot be predicted accurately (6).
EXPERIMENTS
Experiments were performed to quantify the coefficients n and n in
equation 11 and to investigate the functional relationship of factors affect-
ing reentrainment. The CMGBF experimental apparatus consisted of a 203-mm
diameter column, granule control mechanism, circumferential gas exit screen
and simultaneous upstream and downstream isokinetic samplers, as described
elsewhere (5). Samples for particle size analysis were collected on 0.2 ym
pore diameter Nuclepore filters. Photomicrographs were obtained by scanning
electron microscope (SEM), and particles were sized and counted with a Zeiss-
MOP image analyzer.
314
-------
High density (PG = 3250 Kg/m3) alumina granules with a diameter of aver-
age mass of 2.1 mm were used for all runs. The test aerosol was electro-
statically precipitated fly ash from a utility boiler fired with pulverized
coal and had a count median diameter,of 0.3 un.
Twenty runs with the CMGBF were conducted at ambient temperature and
pressure in a "central composite design" experiment for three control varia-
bles: V , superficial gas velocity; V , granule velocity; and K, the ratio
of collected dust mass to granule mass within the conditioned bed at steady
state (8). Nominal factor levels are shown in Table 1. All experiments
described here were conducted at constant bed depth and granule size; effects
of these factors were reported elsewhere (5).
TABLE 1. EXPERIMENTAL CONDITIONS
Block I
Run
No.
18
17
16
20
15
19
*
K
1.3
3.7
1.3
3.7
2.5
2.5
V
0.026
0.026
0.049
0.049
0.037
0.037
V §
g
261
139
139
261
200
200
Block II
Run
No.
13
12
9
11
14
10
*
K
1.3
3.7
1.3
3.7
2.5
2.5
V
0.026
0.026
0.049
0.049
0.037
0.037
V
139
261
261
139
200
200
Run
No.
4
6
3
7
2
1
5
8
*
K
0.5
4.5
2.5
2.5
2.5
2.5
2.5
2.5
Block III
V
0.037
0.037
0.019
0.056
0.037
0.037
0.037
0.037
V§
JSL
200
200
200
200
100
300
200
200
K = intergranular dust deposit ratio (% by weight)
„ V = granule velocity (mm/sec)
V = superficial gas velocity(mm/sec)
8
(Bed depth = 130 mm and granule diameter = 2.1 mm for all runs)
Mass penetration and pressure drop data for all runs shown in Table 1
are reported elsewhere (6).
Penetration data by count in particle size categories between 0.08 \m
and 7.3 ym were obtained for the eight runs in Block III of Table 1 and the
third run of Block II. Fractional penetration data were obtained for these
nine runs for the conditions of (1) an initial clean moving bed when dust
315
-------
feed began, (2) a normal cocurrent steady state bed at the end of at least
two bed volume replacements and (3) straight through penetration of the sta-
tionary loaded bed. Particle size distributions were also obtained for the
particles reentrained from the moving bed when dust feed had stopped.
RESULTS
Fractional Penetration
Average fractional penetration data for Runs 1 through 9 are shown in
Figure 1 for the moving and stationary loaded bed cases. Penetration in the
submicron range was approximately the same whether the loaded bed was moving
or not; however, penetration in the larger particle sizes was clearly greater
when the loaded bed was moving. The relatively greater variability in the
penetration data for the normal moving bed is due to the influence of control
variables, other than particle size, not separated in this crude average but
discussed below. Straight through penetration was influenced much less by
these factors as indicated by reduced variability in the lower curve in
Figure. 1.
Particle size data from the photomicrographs clearly showed that the
reentrained particles were larger than the particles penetrating straight
through the stationary loaded bed. Very small particles (approximately less
than 0.3 Vm) were not found on the photomicrographs of reentrained particles,
suggesting that small particles were not reentrained. Although agglomerates
were present in the reentrained particles as observed on SEM photomicro-
graphs, the proportion of agglomerates in any size category was not signifi-
cantly different from the proportion of agglomerates in the same size cate-
gory either upstream of the CMGBF or in the particles penetrating straight
through. The fly ash was well-combusted and consisted of glassy spheres
which normally do not agglomerate appreciably. Thus, it appears that most
particles penetrated or were reentrained without significant additional
agglomeration.
Single Granule Reentrainment Coefficient
Reentrainment due to granular motion is exemplified by the difference
between penetration when the bed is operating normally and straight through
penetration immediately after the bed is stopped. The increased penetration
due to reentrainment in larger particle sizes can be seen in Figure 1.
Equation 11 with n = 0 was used to calculate the single dirty granule
collection coefficient, i) , from the straight through penetration data in
different particle size categories. The empirical collection coefficient,
n , calculated in this fashion may contain the effects of simple air scour-
ing, bounce and reentrainment of particles unrelated to granular motion. On
an overall mass penetration basis, simple air scouring of particles from the
loaded, stationary bed was generally found to be negligible at the deposition
velocity (5).
With values of HG, £ and bed geometry, the single granule reentrainment
coefficient, nR, was obtained from equation 11 using penetration data for the
316
-------
25-
20-
g
c
o
'jo 15-
1
0)
a.
10-
5 -
0-
0
G) Normal Penetration - 2 S.E.
•J" Bed Moving
/
A Straight Through Penetration i 2 S.E. ,
Bed Stopped
<
C
^
7
' ^=-\
, ^
/
f/
A
_L
I I I
1 0.15 0.2
_,
"~^ 7
x
-
f
)
___
r-
"^\^ N
'
>
(
' S.
s,
i . — n —
\,
7 -\^^ T
A
^ ^
\^l
j.
iii iii i i
— —. A
A
±
I I I i i
0.5 1.0 1.5 2.0 5.0 7.
Particle Diameter (pm)
Figure 1. Average Penetration versus Particle Size
for Run Nos. 1 through 9.
-------
bed under normal operating conditions. Calculated values of r\ for the nine
runs with alumina granules are presented in Table 2. A negative value of TL
results when the penetration straight through the stationary loaded bed
exceeds the penetration in the moving bed case and occurs where measurement
error for the two penetration measurements overlap.
TABLE 2. CALCULATED SINGLE GRANULE REENTRAINMENT COEFFICIENT, r\*
K
Particle Diameter, ym
Run No.
1
2
3
4
6
7
9
5
8
Average
0.16
1.31
9.61
14.9
5.61
47.3
49.1
0.0
2.15
22.3
16.9
0.35
-13.4
-7.99
-8.01
2.47
3.97
22.0
-3.42
-6.57
15.5
0.51
0.69
-2.50
-21.3
9.73
2.18
80.9
18.6
9.71
-4.56
8.73
11.3
1.4
3.64
2.63
28.7
-8.46
88.3
82.1
13.6
3.66
40.9
28.3
2.7
1.46
30.7
24.7
18.8
82.9
49.0
33.3
19.9
58.0
35.4
5.5
11.4
(58.3)+
(91.6)+
19.5
37.9
45.8
64.0
52.7
20.7
36.0
Conditions: e0 = 0.4; e from eq. (3); p, = 1.95 g/cm3;p, = 0.90 g/
Z = 130 mm; d = 2.1 mm (alumina) p
„ , „. OVj
3.
* listed values are n_ x 10 .
R
+ Pt straight through =0 .". n estimated and not included in regression.
The anomalously high values of n in the smallest particle size cate-
gory in Table 2 suggest that particles 0.16 Vm in diameter were reentrained,
but this appears to be an artifact of the calculation procedure which assumes
the difference between normal penetration and straight through penetration is
due solely to reentrainment. Photomicrographs of particles reentrained from
moving loaded beds, after dust feed had ceased, show a complete absence of
very small particles approximately less than 0.3 Urn.
The effects on nR of particle size, d (ym); intergranular dust deposit,
K (% by weight); gas velocity, V (mm/sec)f and granule velocity, V (mm/
sec), were investigated by multiple regression. The resulting equation for
prediction of n for moving alumina granules over the range of variables in
Table 1 is:
318
-------
V
nR = (-23.2 + 6.56 K v^T - 1.22 tC^)2 - 4]2 + 849 VQ) x 10 (15)
The multiple correlation coefficient, R2, for equation 15 is 49% and all
coefficients except the intercept are highly significant (P <0.01). Stan-
dardized residuals were approximately normally distributed.
The most significant term in equation 15 with respect to percent of
variance explained is the product K vd~, indicating a positive interaction
between K and J&~. The second term inpequation 15 indicates a parabolic
function for suplrficial gas velocity with reentrainment first increasing
with gas velocity but then decreasing at the highest gas velocity, which may
be explained by greater removal of reentrained particles by subsequent impac-
tion on other collectors. Finally, granule velocity is seen to have a posi-
tive effect on reentrainment in the form of an incremental increase in a, in
the various particle sizes.
As defined here, reentrainment depends on granule motion. For a CMGBF
with circumferential screen gas outlet, factors affecting intergranule shear
at the wall should also strongly affect reentrainment; for example, increases
in bed depth, granular bulk density, column diameter and fluid pressure drop
should all increase reentrainment. However, variations in these factors were
not examined in the present study. A previous study demonstrated increased
reentrainment with increased bed depth (5). Equation 15 considers the effect
of variables perhaps more closely related to resuspension of dust freed from
granules by wall shear, rather than freeing of the dust from the granules in
the first place.
These results show that reentrainment of collected dust strongly affects
penetration in a CMGBF with a circumferential screen gas outlet as tested
here. Alternate gas outlet designs, such as that used by Rudnick and First
(9) or a sheath of clean granules at the gas exit, which minimize intergranu-
lar shear and granule-wall shear at the gas outlet should minimize reentrain-
ment and reduce overall penetration appreciably.
ACKNOWLEDGEMENT
This work was supported by the National Science Foundation, grant
ENG7 7-26975. Alumina granules were supplied by Electrorefractories and
Abrasives Division of Ferro Corp., East Liverpool, Ohio.
NOMENCLATURE
A = cross-sectional bed area perpendicular to gas and granule flow, cm2
C = particle concentration in gas stream, g/cm3
C. = particle concentration in gas stream inlet, g/cm3
Co = particle concentration in gas stream outlet, g/cm3
D = bed diameter, cm
K = mass ratio of dust in bed, g dust/g granules
Pt = (C0/C.)» penetration fraction
Q = gas volumetric flowrate, cm3/s
Q? = granule volumetric flowrate, cm3/s
(j
319
-------
V = superficial gas velocity, cm/s
6
V = granule velocity, cm/s
Z = bed depth, cm
dZ = differential bed depth, cm
dp = granule diameter, cm
d = particle diameter, cm
e = clean bed porosity, fraction
e = loaded bed intergranular porosity, fraction
p = granule density, g/cm3
p = loose bulk density of granules, g/cm3
p, = loose bulk density of dust, g/cm3
TU = single dirty granule collection coefficient, ditnensionless
a, = single dirty granule reentrainment coefficient, dimensionless
ENDNOTES
1. Yung, S. L., R. Patterson, R. Parker and S. Calvert. Evaluation of
Granular Bed Filters for High Temperature High Pressure Particulate
Control (prepared by Air Pollution Technology, Inc., San Diego, Ca.)
U.S. Environmental Protection Agency Report No. 600/7-79-020,
January 1979.
2. Payatakes, A. C. Advances in Dendritic Deposition of Aerosols by
Inertial Impaction and/or Interception. Proceedings of the 2nd
World Filtration Congress, London, September 18-20, 1979.
3. Goren, S. L. Appendix B: A Mathematical Model for the Moving
Granular Bed Filter. In: Granular Bed Filter Development Program,
Theoretical Analysis of Granular Bed Filtration Principles and Per-
formance Prediction (prepared by Combustion Power Corporation, Menlo
Park, Ca.). U.S. Department of Energy, Report No. FE-2579-18,
January 1978.
4. Wade, G., H. Wigton, J. Guillory, G. Goldback and K. Phillips.
Granular Bed Filter Development Program Final Report (prepared by
Combustion Power Corporation, Menlo Park, Ca.). U.S. Department of
Energy Report No. FE-2579-19, April 1978.
5. Kalinowski, T. W. and D. Leith. Aerosol Filtration by a Cocurrent
Moving Granular Bed: Penetration Mechanisms. Proceedings of 73rd
Annual Meeting of the Air Pollution Control Association, Montreal,
June 22-27, 1980.
6. Kalinowski, T. W. Aerosol Filtration by a Cocurrent Moving Granular
Bed. Sc.D. Thesis, Harvard University, 1981.
7. Tardos, G. I., N. Abuaf and C. Gutfinger. Dust Deposition in Granu-
lar Bed Filters: Theories and Experiments. J. Air Poll. Control
Assoc. 28:354-363, April 1978.
8. Cochran, W. G. and G. M. Cox. Experimental Designs. Wiley New
York, 1957.
9. Rudnick, S. and M. W. First. High-Efficiency Filtration in Cocurrent
Moving Granular Bed. Proceedings of the 73rd Annual Meeting of the
Air Pollution Control Association, Montreal, June 22-27, 1980.
320
-------
FUNDAMENTAL EXPERIMENTS ON A GRANULAR BED FILTER
By: K. linoya and Y. Mori
Department of Chemical Engineering
Kyoto University, Kyoto, 606 Japan
ABSTRACT
The filtration performance of granular packed beds has been studied
experimentally. The filter media are silica sand and glass beads of various
sizes, ranging from 1 to 5 mm. The test dust is calcium carbonate, and the
filtration velocity is 20, 30 and 40 cm/s.
The additional pressure loss due to the collected dust load does not
depend on the filtration velocity, because of pin-hole formation at higher
velocities. The collection efficiency is improved at higher velocities and
for heavier dust loads collected in the beds. The dust distributions in the
filter media are also measured by use of a chemical analysis.
INTRODUCTION
A granular bed filter is a particulate control device to remove dusts
from hot gas streams. Most previous studies have been concerned with clean
packed beds [1-4]. The fundamental performance of packed beds must be
studied for dust loaded conditions, while moving granular bed filters of
various configurations are investigated for particle collection.
This paper presents the results of fundamental experiments on a granular
packed bed filter, the granules of which are retained by a stainless steel
screen. It describes the effects of filtration velocity and the diameter of
granules on the filtration performance as well as the dust distributions
collected in the bed.
EXPERIMENTAL APPARATUS
A schematic diagram of the experimental apparatus is shown in Figure 1.
The test dust, calcium carbonate, is uniformly fed into a mixer type dis-
perser by use of a table feeder (Micro Feeder, Sankyo Dengyo Co., Tokyo), and
is suspended in the air flow.
The packed bed consists of a 150 mm square acrylic column, and is
horizontally retained between stainless steel screens having 0.8 mm openings.
The total thickness of the packed bed is 128 mm. The bed is divided into five
sections perpendicular to the gas stream. The length of the divided bed
section just above the inlet screen is 0.3-0.35 em, and the lengths of the
other four sections are about 25 mm, respectively.
The air with suspended dust flows upwards in the packed bed, and the
dust penetrated through the packed bed is collected on a high efficiency
glass fiber filter (GB 100R, Toyo Roshi Co., Tokyo). The gas flow rate is
measured by an orifice meter, and the pressure loss across the packed bed is
321
-------
continuously recorded on a chart.
EXPERIMENTAL PROCEDURES
Experimental studies on the packed bed have been conducted in order to
investigate the effects of filtration velocity and granule size on the
pressure loss across the packed bed and on the collection efficiency. The
properties of the granules used in the test are given in Table 1. The used
granules are washed with water and dried before the next test. The inter-
granular void of clean beds is 0.366 through all test runs. The effect of the
thickness of the packed bed is investigated by measuring the dust weight
collected in each divided bed section after filtration, instead of by
changing of bed thickness.
The particle size distribution of the test dust is measured by a
cascade impactor (226 type, Sierra Co.) just in front of the packed bed. It
is expressed approximately as a log-normal distribution with the mass median
diameter of 3.5 ym and the geometric standard deviation of 3.5.
The dust weight penetrated through the packed bed is determined from the
weight of the dust collected on a high efficiency back-up filter. The
method for measuring the dust weight collected in each sections of the packed
bed is different for glass beads from that for slica sand. When the filter
media are silica sand, after a test run the packed bed is divided into two
sections, one of which is about 0.35 cm thick section with the inlet screen.
The weight of the collected dust is determined by subtracting Wo from Wf ,
where Wf is granules plus dust weight in each section after a run, and Wo is
the granule weight in each section after cleaning.
When the filter media are glass beads, after a test run the packed bed is
divided into five sections. Calcium carbonate collected in each section is
dissolved into hydrochloric acid. The weight of the dissolved dust is
determined from the concentration of the calcium ion in the solution by a
titration method using EDTA ( ethylene diamine tetraacetie acid) .
RESULTS AND DISCUSSION
Pressure Loss
Figure 2 shows the effect of the thickness of the packed bed on pressure
loss through the clean packed bed of glass beads of 1.1 mm diameter.
Pressure loss is proportional to the thickness of the packed bed.
Figures 3 and 4 are the relationships of pressure loss per unit bed
thickness and unit filtration velocity against filtration velocity, when the
bed is clean. The broken lines in Figures 3 and 4 are calculated by use of
the following Ergun equation [5],
Apo 150 2 (1-Eo)2 1.75
2 -
q2 o + . Q o 2
yS2 - + —pS __ U2
0
322
36 e03 6 e3
-------
where the granules are assumed to be spherical, the diameters of which are
uniform and equal to the mean diameter of the granules. The solid lines in
Figures 3 and 4 are calculated by using the equation obtained by Macdonald
et. al. [6], which has the modified experimental constants of 180 instead of
150 as its in the right hand first term of Equation 1, and of 1.8 instead of
1.75 in the second term. Our experimental results appear to agree with the
equation obtained by Macdonald et. al..
The additional pressure loss is defined as the value of subtracting the
pressure loss of the clean bed from the one of the dust loaded bed. Figures 5
and 6 show relationships between the additional pressure loss and the dust
load. The dust load is defined as the dust weight collected in the packed
bed divided by the filter area. The additional pressure loss is proportional
to the 1.7th power of the dust load, and is not affected by the filtration
velocity. The latter phenomena shows that the drag of the dust layer
deposited into the bed decreases as the filtration velocity increases. The
following experiment has been tried in order to confirm the above phenomena.
Figure 7 shows the relationship between the additional pressure loss and the
filtration velocity at a dust load in the packed bed. Point 1 in Figure 7
expresses the additional pressure loss for 0.84 kg/in2 dust load at 20 cm/s
filtration velocity. Then the filtration velocity is increased without dust
feed. Points 2, 3 and 4 in the same figure are the additional pressure
losses at the velocities of 30, 40 and 45cm/s, respectively. When the
velocity decreases from 45 to 20 cm/s, the pressure losses decrease from the
former values at the same velocities. That is, points 5, 6 and 7 are the
pressure losses at 40, 30 and 20 cm/s, respectively. Finally, when the
filtration velocity increases from 20 to 45 cm/s, the pressure losses have
the same values as before from points 7 to 4. Therefore, the constant
pressure losses at points 1 to 4 indicate the formation of pin-holes in the
dust layer.
Collection Efficiency
Figures 8 and 9 show the relationships between the cumulative penetration
through the packed bed and the dust loads. The filter media in Figure 8 are
glass beads, and that in Figure 9 is silica sand. The cumulative collection
efficiency becomes higher as the filtration velocity increases, and the
instantaneous penetration becomes promptly lower as the dust load increases.
The cumulative penetrations for various granule sizes becomes lower with
smaller glanules, and the higher dust load in the bed improves the collection
efficiency remarkably, as shown in Figure 10. Figure 11 shows the relation-
ship between the cumulative penetrations and the granule size at a dust load
of 0.2 kg/m2. The penetration with silica sand seems to be slightly lower
than that with glass beads. This improvement may be due to the irregular
shape of granules.
Figure 12 shows the dust distribution collected in the filter media,
that is the effect of the thickness of the packed bed on the penetration.
Most of dusts are collected within the several layers of granules from the
inlet screen of the packed bed. The penetration decreases almost exponential-
ly after the thin inlet layers as the bed thickness increases. Figure 13
323
-------
shows relationships between the cumulative penetration and the dust load for
various bed thicknesses, which are calculated by use of the dust distributions
collected in the filter media. The penetrations of the 0.3 cm thick bed, i.e.
about one layer of the granules, has a minimum value because of reentrainment.
On the other hand, the collection efficiency becomes higher as the dust load
increases when the bed thickness is more than 2.8 cm.
Figure 14 shows the effect of the granule size on the penetration for
two kinds of the thickness. Reentrainment becomes dominant at higher dust
load for the thin bed of 2.9 or 5.2 mm glass beads. However, penetration
becomes lower even at higher dust load for the 0.3 cm thick bed of 1.1 mm
diameter granules. All penetrations becomes lower at higher dust load for the
5.3 cm thick bed, independent of the granule size.
Figure 15 shows the effect of the filtration velocity on the collection
performance for two thicknesses. The penetrations do not depend on the
filtration velocity for the 0.3 cm thick bed. Considering no effect of the
filtration velocity on the additional pressure loss as mentioned before, it
may be concluded that the pin-hole formation in the dust layer does not
increase the penetration at higher velocity for the thin packed bed. On the
other hand, the filtration velocity has an effect on the penetrations for the
5.3 cm thick bed. Figure 16 shows the relationship between the cumulative
penetration and the dust load in the 0.35 cm thick bed with silica sand. The
penetrations also do not depend on the filtration velocity.
CONCLUSIONS
The filtration performance of granular packed beds has been studied
experimentally, and the following results are obtained:
1) The equation by Macdonald et.al. or the Ergun equation can be used to
predict the pressure loss through the packed beds with clean granules.
2) The additional pressure loss due to the collected dust load does not
depend on the filtration velocity. The reason may be pin-hole formation in
the dust layer deposited in the bed at higher filtration velocities.
3) The collection efficiency is improved at higher filtration velocities,
for heavier dust loads collected in the beds and with smaller size of
granules.
4) When the thickness of the packed bed becomes comparative to the granule
size, the cumulative penetration has the minimum value and does not depend
on the filtration velocity for both filter media.
5) The cumulative penetration decreases almost exponentially as the bed
thickness increases over a few cm. This means that the log-penetration law
is valid.
NOMENCLATURE
Dc = diameter of granule of packed bed [cm]
Dp = diameter of dust particle rcmi
E = collection efficiency F - 1
L = thickness of packed bed rcmi
m = dust load collected in packed bed [g/m2]
Ap = pressure loss through dust loaded packed bed [pa]
324
-------
= pressure loss through clean packed bed [Pa]
Sc = surface area per unit volume of granules, =6/Dc,
if granules are uniform size spheres [cm2/cm3]
u = filtration velocity [cm/s]
e0 = intergranular void of clean bed [ - ]
Vi = coefficient of air viscosity [g/cm-s]
p = air density [g/cm3]
Pp = dust particle density [g/cm3]
ENDNOTES
1. Miyamoto, S. and H.L.Bohn. Filtration of Airborne Particulates by
Gravel Filters. J. Air Poll. Control Assoc. 24:1051-1054, November 1974.
2. Schmidt, E.W., J.A.Gieseke, P.Gelfand, T.W.Lugar, and D.A. Furlong.
Filtration Theory for Granular Beds. J. Air Poll. Control Assoc. 28:
143-146, February 1978.
3. Gutfinger, C. and G.I.Tardos. Theoretical and Experimental Investigation
on Granular Bed Dust Filter. Atmos. Env. 13:853-867, June 1978.
4. Lee, K.W. and J.A. Gieseke. Collection of Aerosol Particles by Packed
Beds. Env. Sci. Tech. 13:466-470, April 1979.
5. Ergun, S. Fluid Flow through Packed Columns. Chem. Eng. Prog. 48:
89-94, February 1952.
6. Macdonald, I.F., M.S.El-Sayed, K.Mow, and F.A.L.Dullien. Flow through
Porous Media - the Ergun Equation Revised, Ind. Eng. Chem. Fundam.
18:199-208, March 1979-
Table 1. GRANULE CHARACTERISTICS
Glass Beads Silica Sand
Mean Diameter
11 2 Q S 2 " • ? 9
D [mm] -
P50L
Size Range
, , 1.0-1.2 2.3-3.4 5.0-5.6 2.0-2.4
D [mm]
Density 2^5 2^9 2^Q 2_52
p [g/cm ]
325
-------
Manometer
Orifice
Blower
Figure 1. Experimental apparatus of packed granular
bed filter.
14
s.
(A 1
O
Oi
glass beads
(Mmm)
"0 5 10 15
thickness of packed bed.
L (cm)
Figure 2. Effect of thickness of
packed bed on pressure
loss through packed bed
of glass beads.
silica sand
Dc = 2 2mm
L=12-8cm
O: experimental data
°'60 20 40 60 70
filtration velocity . u (cm/s)
Figure 3. Relationships between
pressure loss and
filtration velocity
(silica sand).
326
-------
glass beads Macdonald eq.
-Jas
0 20 40 60
filtration velocity , u (cm/s)
Figure 4. Relationships between
pressure loss and
filtration velocity
(glass beads).
s.
0-2
i
a.
ai
0-05
003
filter media: sand
. CaC03
L =12 8 cm
£. = 0366
0-1 02 05 1 3
dust load , m (kg/m2)
Figure 6. Relationship between
additional pressure loss
and dust load (silica
sand).
0
2
"o 0-5
a.
VJ
0-2
o
a.
• 0-1
a
^-OS
0-02
0-01
0.005
» »| ' r—
. glass beads
CaC03
Dc = 2-9 mm
L =12-8 cm
~ e. = 0-366
•
,
- o/c
: AA7"
" n9^
- /<*
O /
f /°
^ 1 1 1 1 , I,
i i i t M| i i /
/
a
t\ n
**•' QJ^ n
-------
LJ
' 20-
o
o
0>
CJ
D
3
u
glass beads (2.9mm)
CaC03
L=12.8cm
e0 =0-366
u =20cm/s
0-05
0-1
0-2
dust load
0-5 1
m (kg/m2)
Figure 8. Relationship between cumulative penetration and
dust load (glass beads).
15
'10
Q. V
> "~
u
f i Uer media : sand (2 2mm)
CaC03
L=12.8cm
= 0 366
0-1 , 0-2 0-5 1 23
dust load , m (kg/m2)
Figure 9. Relationships between cumulative
penetration and dust load (silica sand).
328
-------
30
'20
UJ
o
210
I 7
O
* c
c 5
di
I 3
3
I 2
CoC03
= 20cm/s
L = 12-8cm
e0=0 366
... l
0-1 0-2 0-51 2 <
dust load , m (kg/m2)
Figure 10. Cumulative penetration for packed bed of various
granule sizes.
UJ
i
o
o
-10
o
'•5 7
Is
3
3
I 2
glass beads
CaC03
L=12-8 cm
e0=0-366 /*
silica sand
•100
1
granule size , Dc (cm)
Figure 11. Relationships between
cumulative penetration
and granule size at dust
load of 0.2 kg/m2.
UJ
8 30
o
2 10
a* .
> 3
3
O
1
glass beads (1-1 mm)
CaC03
u = 21cm/s
e. = 0366
O
A
a
0
m(kg/m2)
00463
0-0706
0-113
0-317
0 5 10 13
thickness of packed bed , L (cm)
Figure 12. Effect of thickness of
packed bed on cumulative
penetration.
329
-------
60
glass beads (2-9 mm)
CaC03
u =20cm/s
e.=0-366
0-04
0-2 0-5
dust load . m
Figure 13. Cumulative penetration of packed beds for various
thicknesses.
60
0-03 0-05
0-1 02
dust load
0-5 1
m (kg/m2)
Figure 14. Effects of granule size on cumulative penetration of
packed beds for two thicknesses.
330
-------
glass beads (2 9mm)
CaC03
0-04
0-1 0-2
dust load
0-5 1
m (kg/m2)
Figure 15. Effect of filtration velocity on cumulative
penetration of packed beds for two thicknesses.
. 100
* 6
.> o
r
30
silica sand(2-2mm)
CaC03
0 A
L=0-35cm
e.=0366
u(cm/s)
O 20
A 30
a 40
0-07 0-1
0-2
dust load
0-5
m (kg/m2)
Figure 16. Relationship between cumulative
penetration and dust load for 0.35 cm
thick bed with silica sand.
331
-------
DRY DUST COLLECTION OF BLAST FURNACE EXHAUST GAS
BY MOVING GRANULAR BED FILTER
By : A. Wakabayashi, T. Sugawara, S. Watanabe
Kobe Steel, Ltd.
Kobe, Japan
ABSTRACT
For the purpose of attaining more effective power recovery of blast fur-
nace top gas energy by dry dust collection, a pilot plant of one module of a
plant of commercial size was installed to ensure performance and reliability
in the field operation, where it proved to have sufficient applicability in
various operating conditions.
This report is to succeed the first report of the small scale fundamen-
tal test which was published at the first symposium on the Transfer and
Utilization of Particulate Control Technology held in Denver, July 24 - 28,
1978.
INTRODUCTION
As for the dust collection device for blast furnace top gas, it is es-
sential to ensure stable performance and reliability in long time operation.
Therefore, we installed the pilot plant device having the same size and con-
struction as one module of the full size plant which consists of multiple
modules of cylindrical vessels to treat the whole volume of gas from a blast
furnace and to contain the high pressure gas.
The pilot plant test was performed in two stages in the field of our
KAKOGAWA steel making plant. The first stage test entered into operation in
March, 1979 by inducing the top gas from No.l blast furnace, and continued
until January, 1980. The test was carried out for the filteration velocity
across the inside of the filters of 0.2 to 0.37 m/sec. By reviewing the re-
sult of the first stage test, we concluded that :
1) The filteration velocity can be increased up to 0.4 m/sec or more, to
make an equipment smaller and attain a smaller installation space and
a more reasonable cost in commercial use.
2) Longer operating duration should be preferable to prove the reliability
of the equipment.
Then, the second stage test started in July, 1980 by inducing the top gas
from No.2 blast furnace to attain the above mentioned objects and continued
until January, 1981. The test result showed the acceptable performance for
particle collection efficiency on the increased filteration velocity of 0.4
m/sec and justified the reliability of the device.
Specifications of Pilot Plant
Specifications of the pilot plant are shown in Table 1. in the first
332
-------
stage test, the ranges of the operating factors such as gas velocity across
the fiJ.ter, dust load in the media, and media size were decided according to
the result of the preceding fundamental test, and in the second stage test,
those factors were decided based on the test results in the first stage test.
TABLE 1. SPECIFICATION OF TEST FACILITIES
Gas induced to Test Device
Gas Capacity
Gas Pressure
Gas Temperature
Filteration Velocity
Inside Filter/Outside Filter
Layer Thickness
Inside Filter/Outside Filter
Media Size
Inside Filter/Outside Filter
First Stage Test
No.l Blast Furnace
Top Gas
40000 to 60000 Nm3/h
1.7 kg/cm2g
average 140°C
0.2 ~ 0.37/
0.12 ~ 0.23 m/sec
200/200 mm
1.2 ~ 2.5/
1.2 ~ 2.5 mm
Second Stage Test
No.2 Blast Furnace
Top Gas
40000 to 80000 Nm3/h
2.1 kg/cm^g
average 120°C
0.4 ~ 0.6/
0.3 ~ 0.45 m/sec
200/400 mm
1.2 - 2.5/
1.2 ~ 2.5 mm
Description of Pilot Plant
The view of the pilot plant is shown in Figure 1; and Figures 2 and 3
show the gas flow and the layout of the plant.
Gas Flow
Figure 2 shows the gas flow from the blast furnace to the pilot plant.
The dust-laden top gas from which large particles are collected previously
by the dust catcher is supplied to the pilot plant by-passing the venturi
scrubber. As shown in Figure 3, the gas from the dust catcher first enters
the multi-cyclone in which coarse particles are removed from the gas stream,
and flow into the vessel of the granular bed filter device, and then flow
through the inner and the outer granular bed filters arranged in the cylin-
drical vessel. The particulate in the gas stream is captured in the granular
media, and the cleaned gas exists from the vessel.
Gas flow capacity is controlled by the flow control valve to balance the
flow rate of the main line and the by-pass line.
Shut-down valves are provided at the inlet and the outlet of the device
to shut down the gas stream from the device at any time. In the case of
shut-down, the top gas in the device is replaced with nitrogen gas to avoid
333
-------
water vapor condensation in the gas. The nitrogen gas in the device is re-
placed with air at the time of inspection inside of the device. For the use
of the replacement of the gas in the device, the breather valves (Vi to V7)
and the inlet ports of air and nitrogen gas are provided at the suitable po-
sitions in the device.
Moving of Granular Media
Figure 3 shows the layout of the pilot plant. The two staged moving
granular bed filters are located in equi-central multi-angular double rows,
and the granular media in the filters moves slowly and continuously downward
by gravity, controlled by table feeders. The setting of the table feeders
can be regulated independently to control the speed of the media movement in
the inner and the outer filter. The granular media in the outer filter moves
more slowly than that in the inner filter to capture the finer particles.
The gas seal valve assemblies are provided at the top and the bottom of
the vessel so that the granular media is fed to and discharged from the ves-
sel without gas flowing out from the vessel.
The dust-laden granular media is discharged from the bottom of the gas
seal valve, and conveyed to the vibrating screen by the belt feeder and the
bucket elevator to remove the dust from the media. The cleaned media is
stored in the granule storage tank and conveyed to the top of the vessel by
the bucket elevator and the vibrating feeder, and then recycled back to the
top of the filter through the gas seal valve.
Test Results
In filteration of granular bed filter, filteration velocity and dust
load in the media make the strongest effect on the performance such as col-
lection efficiency and pressure drop across the filter. Therefore, the test
was performed on the various conditions of filteration velocity and moving
speed of the media.
Filteration Performance
Figures 4 and 5 show the effect of the filteration velocity on the dust
concentration at the outlet of the granular bed filter in the first stage
test and the second stage test. In the first stage test, 200 mm - thick fil-
ters were applied for both the inner and the outer filter, and in the second
stage test, 200 mm - thick filter was applied for the inner filter and 400 mm
- thick filter was applied for the outer filter. The data obtained from both
tests demonstrate a considerable decrease in the dust concentration at the
outlet of the granular bed filters in the second stage test. The second
stage test data indicates that the specified dust concentration of 5 mg/Nm3
or less at the outlet of the granular bed filters is attained at the filter-
ation velocity of 0.4 m/sec at the inside of the filters. The range of the
dust concentration at the inlet of the multi-cyclone and of the granular bed
filters were 2 to 5 g/Nm3 and 0.2 to 0.5 g/Nm3 in both tests. The data in
both tests indicate that the dust concentrations at the outlet of the granu-
lar bed filters are not affected by the inlet values.
334
-------
In filteration of granular bed filters, collection efficiency and pres-
sure drop across the filter are affected by dust load in the media, there-
fore, we kept in each test the optimum moving speed of the media, which was
obtained in the beginning.
Figure 6 shows the particle size distribution at a few points in the
system by dust sampling in the second stage test, and Figure 7 shows the
average collection efficiency in the granular bed filters in each particle
size. The data indicates that particles over 10f{ m are almost perfectly
captured and even finer particles than lOu m are collected efficiently by the
two-staged granular bed filters.
In both tests, the relationship between the pressure drop across the
filters and the collection efficiency was not observed.
Gas Temperature
Figure 8 shows the photograph of a record of the gas temperatures. As
shown in the record, the gas temperature at the inlet of the granular bed
filters fluctuates widely according to the operation of the feeding system
at the top of the blast furnace, but the fluctuation of the gas temperature
at the outlet of the granular bed filters is moderated by the heat capacity
of the granular media. This phenomenon benefits the energy recovery system
of the top gas in commercial use in the field, because the temperature fluc-
tuation of the gas into the energy recovery turbine is moderated and the
stable load is obtaineded from the turbine.
Reliability of Device
During the test period of two years,' the device has been confirmed to
endure for long-time operation.
Conclusion
For dry dust collection of blast furnace top gas by a pilot plant, the
granular bed filter proved acceptable applicability in spite of wide fluctu-
ation of gas volume, gas temperature, and dust load in gas stream.
335
-------
Fiqure 1 View of Pilot Plant
Pilot Plant of Granular Bed Filter
/Blast Fur
To Gas Holder
Electrostatic Precipitator
Generator
Figure 2 Gas Flow to Pilot Plant
336
-------
«• Drafting
(to Bag Filter)
Level Switch
Gas Seal Valves
.. Level Switch
Shut-down
Bucket Elevator
Vibrating Sceen
Drafting
Gas Seal Valves
Pi : Pressure Gage
Ti : Thermometer
Gas Seal Di : Dust Concentration
Valves Measuring Place
Vi : Breather Valve
P.M.: Gas Flow Meter
Feeder
Figure 3 Layout of Pilot Plant
337
-------
tfl ^-,
0.2 0.3
Pilteration Velocity
at Inside of filters
0.37
(m/sec)
Figure 4 Effect of Filteration Velocity
on Dust Consentration at Outlet
of Granular Bed Filters
Filteration Velocity
at Inside of filters (m/sec)
0-6
Figure 5 Effect of Filteration Velocity
on Dust Consentration at Outlet
of Granular Bed Filters
338
-------
c
01
u
id
QJ
flj
W
W
(C
s
•H
4J
(0
r-(
3
U
at Outlet of
Granular Bed
Filters
at Inlet of
Granular
Bed Filters
at Inlet of
Multi Cyclone
2 4 6 810 20 40 60 80100150
Particle Size (u m )
Figure 6 Particle Size Distribution
at Each Place of System
0)
4J
Collection Efficiency
in Two Staged Granular Bed
(
IUU
98
96
94
92
90
-
// / / /
'
2 4 6 8 10
Particle Size (n m )
Figure 7 Collection Efficiency in.
Two Stage Granular Bed Filters
in Each Particle Size
339
-------
CO
O DJ *^d
hh rt- H
C
OHO
H 3 ft
JH I—' C
a fD (D
C rf rt
•H H-
DJ 51) O
^ g,3
w o
fD Q) Hi
Qi rt
t O fD
H- C 3
M rt T3
rt H fD
fD fD i-i
M rr D>
cn rf
i-s
fD
-------
IRON AND STEEL AIR POLLUTION CONTROL
USING MAGNETIC SEPARATION
By: D. C. Drehmel
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
C.E. Ball
Research Triangle Institute
Research Triangle Park, NC 27709
C.H. Gooding
Clemson University
Clemson, SC 29631
Introduction
The separation of process streams according to magnetic susceptibility
has been an area of expanding research and development. During the last
decade well over 100 patents, reports, and technical papers have been
published throughout the world. Some of the application areas are
mineral beneficiation (1,2); taconite beneficiation (3); coal de-ashing
and desulfurization (4-10); wastewater treatment (11,12); steel mill
wastewater treatment (13,14); municipal wastewater treatment (15,16);
blood component separation (17); and catalyst recovery (18). While
application of magnetic separation to water pollution problems has
received extensive investigation, application to air pollution problems
is a recent development. Among the major industrial groups with significant
air pollution problems, the iron and steel industry has sources which
are difficult to control by conventional devices and which have particulate
emissions with high specific magnetizations. Specific sources which
would be most favorable for magnetic separation are the basic oxygen
furnace, open hearth furnace, and the scarfing machine. Other possible
sources for application are the blast furnace, electric arc furnace, and
the sinter machine. The results reported in this paper are for a field
test using a side stream of the exhaust from the sinter windbox.
Application of magnetic separation to air pollution control could
take many forms. Possible variations include:
a) Permanent vs electromagnet.
b) Dipole vs quadrapole magnet.
c) Collection by filtration vs diversion to a secondary
collector.
341
-------
d) Gas flow parallel or perpendicular to magnetic field lines.
e) If the collection is by filtration, collection by fibers vs
collection by granules.
f) On-line vs off-line cleaning of the collector.
With so many possible combinations, it will take extensive research
and development to fully assess the applicability of magnetic separation
to air pollution control. The subject of this paper is a stainless-.
steel deep-bed fiber filter enhanced by a magnetic field from a dipole
electromagnet. The filter is cleaned by off-line reverse air flow.
Theoretical considerations and general applicability of this type of
system have been discussed in detail in previous papers (19,20). For
the sinter plant field tests, the magnetic separation equipment was
installed on a 13 m trailer and trucked to the plant. In order to
provide continuous operation during periods of off-line filter cleaning,
this mobile pilot plant actually consisted of two sets of magnetic
separation equipment as described below.
Apparatus
The mobile pilot plant was designed for a nominal flow capacity of
5100 m /hr (3000 CFM) based on the previous experimental work. The
dirty gas enters the trailer via a 0.317 m ID, 316 stainless steel pipe
(12 in., Schedule 5) and passes by test ports through which samples can
be drawn to determine the size distribution and concentration of the
inlet dust. The gas is then directed to one of two functionally identical
magnetic separation devices. One magnet was constructed by Magnetic
Corporation of America (MCA), Waltham, MA, and the second magnet was
originally constructed by Sala Magnetics, Inc., Cambridge, MA, for use
in the laboratory pilot plant. This second magnet was later modified by
the addition of new pole pieces and a stand purchased from MCA. Each of
the magnetic filters consists of an iron-bound solenoid surrounding a
canister that measures 0.432 m ID by 0.305 m long. The canisters are
filled with an appropriate amount of magnetic stainless steel wool as
dictated by the test program. Each of the magnets can be energized to
provide an applied field of up to 0.5 tesla throughout the canister
volume. The magnets are energized by DC modular power supplies manufactured
by Controlled Power Company (Troy, MI) and purchased from MCA.
After passing through the filter, the cleaned gas travels past
another set of test ports and exits the trailer. The exterior pipe size
is reduced to 0.266 m ID (10 in., Schedule 5). The gas passes through
an orifice and an induced draft blower and is then exhausted to the
atmosphere through an 11 m high stack.
The filters are cleaned by backflushing with compressed air provided
by a Worthington Model 7-1/2 EDBR two-stage compressor (Worthington
Compressors, Inc., Holyoke, MA) that is mounted to the underside of the
trailer. The compressed air tank associated with each filter has a
volume of approximately 0.28 m . To pulse the filter, the compressed
air is released through an 8 in. (nominal pipe size) Galigher Delta
valve (Galigher Company, Salt Lake City, UT). The Galigher valves are
342
-------
pneumatically actuated pinch valves with an equivalent throat diameter
of approximately 20 cm and a fully open C rating of 1300. Each valve
consists of a pair of identical elastomeric diaphragms contained within
a cast aluminum, split housing. The diaphragms are closed by introducing
compressed actuator air into the chamber between the housing and the
diaphragms. The valve can then be opened very rapidly by exhausting the
actuator air through two large ports provided in the housing. The
actuator air was controlled to a pressure approximately 140 kPa (20 psi)
higher than the cleaning air and was released through 1-1/2 in. (nominal)
Model 168S poppet valves manufactured by Kay Pneumatics (Commack, NY).
Dynamic flow calculations conducted during the design phase indicated
that the Galigher valves should be able to achieve the required C of
750 within 0.05 seconds of the initial diaphragm separation and tXus
release the cleaning air in a pulse sufficient to clean the filters.
The agglomerated dust that is flushed off the filter is removed
from the cleaning air by a Kirk and Blum Size 6, Type C7 cyclone (Kirk
and Blum Manufacturing Company, Cincinnati, OH). The exhaust from the
top of the cyclone is routed back to the dirty gas stream and into the
operating filter. Dust can be removed from the cyclone during operation
through the double seal formed by two 6 in. (nominal) Norris butterfly
valves (Dover Corporation/Norris Division, Tulsa, OK) that are mounted
in line at the bottom of the cyclone hopper. Pneumatic vibrators are
mounted on the walls of the dust hopper to aid in dust discharge.
Results
The test program was conducted to give results on the effects of
suspected parameters and on reliability during long term testing.
Parameters which were directly varied include filter density and depth,
applied magnetic field strength, and gas velocity. Resultant observed
values include temperature, pressure drop, outlet concentration, total
collection efficiency, and collection efficiency by particle size. By
statistical analysis of the data it was determined that outlet concentration
and collection efficiency were significantly correlated to pressure drop
but not to magnetic field strength or velocity. Magnetic field strengths
tested were 0.25 and 0.50 tesla and velocities ranged from 3.0 to 8.8 m/s.
Failure to obtain a significant correlation does not imply that these
variables do not affect the performance but only that this approach was
not conclusive. The next step was to average data taken with varying
field strength and velocity in order to determine the effect of other
parameters. In particular, one may note the change of average pressure
drop and outlet concentration with changing filtration media. The stainless
steel wool filter bed was either 0.2 or 0.3m long and was packed into the
canister described above with a void fraction of 0.990 or 0.986. For
the shorter length of bed, lowering the void fraction quadrupled the
pressure drop and halved the outlet concentration. Similarly for the
larger void fraction, increasing the bed length more than quadrupled the
pressure drop and more than halved the outlet concentration. It is
possible to further interpret the relationship between outlet concentration
and pressure drop by curve fitting the individual data points. The
343
-------
apparent relationship is:
c = k (p)-°-5 (1)
where: c is the outlet concentration,
k is a cons tant , and
p is the pressure drop.
Although large changes in pressure drop are associated with bed character-
istics, a significant range of pressure drops can be found for a given bed
and the same general relationship between outlet concentration and pressure
drop can be found.
A complete listing of the results of the performance characterization
is shown in Table 1. Two levels each of the parameters filter length,
filter density, and applied field provide the basic eight experimental
conditions which were tested at varying superficial velocities. The
resultant pressure drops ranged from 9 to 114 cm W.C. and the collection
efficiency from 46.6 to 94.5%. As expected the short low-density
filter gave the low pressure drops and low collection efficiencies while
the highest collection efficiency was found with the long high-density
filter. Not all collection efficiencies with this latter filter were
high because the efficiency tended to decrease with increasing gas
velocity. This is the opposite trend seen with the short low-density
filter. Other filter combinations did not show clear trends with velocity.
It is proposed that, as the filter becomes more efficient, the role of
velocity shifts from promoting collection with increased impaction to
degrading collection with increased reentrainment of collected material.
Discussion of Results
In the preceding section it was noted that the results were correlated
with pressure drop. Pressure drop is more usually a dependent variable
and is a function of filter density, F; filter length, L; and superficial
gas velocity, V. In the case of these data it was found that pressure
drop was significantly correlated to the product of F, L, and V . To
further define the relationship between the results and filter properties,
the experimental results were compared to a model developed by Gooding (21) .
The equation for this model is :
E = 1 - exp - Yc P
n2 s (i-F)
where: E is the collection efficiency,
Y is the collision radius,
P is the probability of particle retention, and
S is the radius of the collecting wire.
344
-------
Table 1. RESULTS OF PERFORMANCE CHARACTERIZATION
u>
Filter
Length
(m)
short
(0.2)
long
(0.3)
Filter
Density
(metal fraction)
low
(0.010)
high
(0.014)
low
high
Applied
Field
(tesla)
low
(0.25)
high
(0.50)
low
high
low
high
low
high
Superficial
Velocity
m/s
3.6
4.4
3.8
3.9
4.9
7.9
3.7
5.8
7.4
4.0
5.8
.7.9
3.8
5.9
7.7
3.9
6.0
6.9
3.0
6.4
8.8
6.5
Filter
Pressure Drop
cm W.C.
9
11
9
11
17
32
18
41
66
25
64
76
38
44
79
31
72
114
89
102
109
76
Collection
Efficiency
%
56.6
51.9
48.7
46.6
51.1
66.1
85.6
83.9
86.2
86.7
79.6
74.4
78.2
90.5
91.2
81.5
87.2
87.7
94.5
83.4
75.7
76.1
-------
The collision radius, Y , is a dimensionless value derived from solution
to a trajectory model for magnetic filtration. In this trajectory
model four parameters—A, G, K, and W— describe the behavior of the part-
icle. Parameter A accounts for the relative magnitudes of the magnetic field
induced in the ferromagnetic wire and the applied field. After magnetic
saturation of the steel wool filter, increased applied field would cause
A to decrease from unity. However, for these experiments, the steel
wool was at or below saturation where A may be taken as constant at unity.
Parameter G accounts for the effect of gravity, which is small _^
for these experiments. The calculated value for G was less than 2 x 10
and G was taken as zero.
The parameters W and K account for the effects of magnetic field and
inertia on collection and are defined as:
W - u V* H 2 (3)
P v2
P o
K = 2b2 p V (4)
p o
9 Sri
where: y is the magnetic permeability of a vacuum,
X* is the effective magnetic susceptibility of the particle,
H is the magnetic field,
Si
p is the density of the particle,
V is the fluid velocity upstream of the wire,
b is the particle radius, and
TI is the fluid viscosity.
With G taken as zero and A as unity, the effect of W and K on the collision
radius, Y , is shown in Figure 1 as contours of constant values for Y
for varying W and K. As the magnetic field decreases, W decreases an§
the value of Y is identical to conventional inertial impaction in potential
flow. As the magnetic field increases, the value Y can be significantly
increased by as much as an order of magnitude. The physical significance
of this is that the individual collection fiber may collect particles
that would have bypassed the fiber without the magnetic field. In fact
the fiber may cover a radius as great as many times its physical radius.
Given the collection efficiency of the individual fiber, the model
assumes a randomly packed filter which gives an active projected area of
2/r, times the fiber area given as 2hS Y where h is the wire length and
S is the wire radius. By integrating over the entire length of the
346
-------
filter and using a plug-flow reactor mass balance, Equation (2) is
obtained except for the P. The factor P was developed after analysis of
in-house pilot scale data as shown in Figure 2. The physical significance
of P is that it removes the assumption that all particles which reach
the fiber are permanently collected. The model without P predicts well
for small particle sizes but overpredicts collection efficiency at
larger particle sizes. The loss of efficiency at larger sizes may be
attributed to reentrainment of agglomerates or lack of retention of
impacting large particles. This latter case was analyzed by computing
the energy needed to cause particle bounce. A theoretical critical
impaction angle was fitted to pilot data by adjusting parameters within
certain bounds. The model with bounce correction agreed with some data
as shown in Figure 2 but in general underpredicted efficiency at high gas
velocity or large particle size.
An alternative approach was to assume that reentrainment and bounce
would be some function of W and K. Analysis of the in-house pilot scale
data showed that P was directly related to W and inversely to K. The
field test data were also analyzed by this approach and the best fit was
obtained when P was described as follows:
P - 2 ( 10 W2 + ^ ) (5)
Using Equation (5) for P, the comparison between the model and experimental
data is shown in Figure 3. Agreement is good but some collection efficien-
cies are over or under predicted by as much as 30% (the dashed lines give
boundaries of 30% over and under prediction).
With regard to ultimate application of magnetic filtration, these
tests are not conclusive. Although three tests had efficiencies in
excess of 90%, the lowest pressure drop for these was 44 cm W.C. The
high pressure drop needed for good collection is an ultimate consequence
of the low iron content and poor magnetic properties of the sinter plant
particulate matter. Values of W ranged from 0.0077 to 0.083 with K
values in the range from 0.16 to 0.46. Figure 1 shows that W has no
impact on Y over much of the experimental range. Only when W is large
and K is small is Y significantly different from that predicted for
impaction only. As described in Equation (3), W is related directly to
magnetic susceptibility and magnetic field. Although the magnetic field
could be increased, the cost could outweigh the savings in pressure
drop. However, magnetic susceptibility could be increased 20 to 30 times
by applying magnetic filtration to particulate from an electric arc
furnace or basic oxygen furnace instead of that from a sintering machine.
The corresponding increase in W of 20 to 30 times yields an increase in
Y of approximately 4 to 10 times. Using Equations (1) and (2), the
ultimate result is that the pressure drop could be decreased by a factor
of approximately 20 and still maintain 90% collection. In other words a
less-dense shorter filter could be used to give a collection efficiency
greater than 90% and a pressure drop less than 5 cm W.C. By similar
calculations a 99% collection efficiency should be achievable with a
pressure drop less than 20 cm W.C. This hypothesis will be tested in
347
-------
a pilot demonstration on an electric arc furnace.
The pilot demonstration on an electric arc furnace will take place
at the Georgetown Steel Corporation plant in Georgetown, SC. The
waste gas stream to be treated is collected from three EAFs with direct
shell evacuation and canopy hoods above. A slipstream of gas will be
taken from a duct upstream of the fabric filter serving the EAFs.
The mobile pilot plant is a continuous HGMF system which is designed
for a nominal flow capacity of 3,060 m /hr (1,800 cfm) . The pilot plant
contains a SALA-HGMS Carousel Model 120-05-00 (Sala Magnetics, Inc.,
Cambridge, MA), which is a continuous device that incorporates a magnet
head and a cleaning station mounted 180° apart on a rotating carousel.
The magnet coils are split into a mirrored-saddle configuration to allow
the carousel to be rotated through the magnetized zone by a variable
speed drive. The carousel contains 48 removable cassettes which can be
loaded with filter material to a depth of 0.4,7 m (5.8 jn.). The magnet
head encloses an active face area of 0.085 m (133 in. ) in the direction
of fluid flow. The magnet head is designed to provide an applied field
from 0.0 to 0.5 tesla. The cleaning station is designed to provide a
pulse of air to clean the collected particles from the filter material
once it has left the magnetized zone.
The agglomerated dust cleaned from the filter material with the
cleaning air pulse is sent to a Kirk & Blum size 4, Type C5 cyclone.
(Kirk & Blum Manufacturing Company, Cincinnati, OH). Exhaust from the
top of the cyclone is then recycled into the dirty gas stream.
The test program is designed to test the effects of four parameters
on collection efficiency, and the reliability of the equipment during
long term operation. The four parameters which will be varied are
filter packing density, filter type (steel wool, expanded metal), applied
field, and gas velocity.
After the performance characterization tests are completed, long
term testing will be started. During this test, the pilot plant will be
operated continuously for 500 hours. The pilot plant tests are now
scheduled to run from June to December 1981.
Conclusions
Pilot scale field tests of magnetic filtration were conducted at a
sintering plant. The 5,100 m /hr magnetic filter was examined under
varying conditions of an extensive test matrix and in a long term (450
hour) test. No fundamental difficulties were encountered either during
long term testing or at high gas velocities up to 8.8 m/s. High collection
efficiencies (greater than 90%) were achievable but generally at the expense
of high pressure drop (greater than 44 cm W.C.). It is felt that magnetic
filtration would not be cost effective at high pressure drop even though
the high superficial velocity is attractive for compactness of the
control device. Analysis with a theoretical model predicts that the
348
-------
pressure drop could be reduced by at least an order of magnitude while
maintaining high collection efficiency if the specific magnetization of
the dust were more than an order of magnitude higher. Previous in-house
pilot plant work has shown that both basic oxygen furnace (EOF) and
electric arc furnace (EAF) dust provide the high specific magnetization
indicated as necessary by analysis in this study and both EOF and EAF dust
have been collected with 99% efficiency in in-house work. It is concluded
that magnetic filtration would be cost effective for application to EOF and
EAF control even though it is not cost effective for sinter plants. To
test this conclusion, a pilot demonstration of magnetic filtration on an
EAF is being planned for the near future.
References
1. Kelland, D. R. , "High Gradient Magnetic Separation Applied to Mineral
Beneficiation," IEEE Trans. Magn., Mag-9. 307 (1973).
2. Murray, H. H., "Beneficiation of Selected Industrial Minerals and Coal
by High Intensity Magnetic Separation," IEEE Trans. Magn.,
Mag-12, 498 (1976).
3. Kelland, R. R., and E. Maxwell, "Oxidized Taconite Beneficiation by
Continuous High Gradient Magnetic Separation," IEEE Trans. Magn.,
Mag-11, 1582 (1975).
4. Ergun, S. , and E. H. Bean, "Magnetic Separation of Pyrite from Coals,"
Report of Investigation No. 2718, U. S. Bureau of Mines,
Pittsburgh, PA (1968).
5. Trindale, S. C., and H. H. Kolm, "Magnetic Desulfurization of Coal,"
IEEE Trans. Magn., Mag-9, 310 (1973).
6. Vives, D. L,, L. J. Hirth, and W. H. Summerlin, "Direct Reduction and
Magnetic Beneficiation of Alabama Brown Ore with Lignite,"
IEEE Trans. Magn., Mag-12, 490 (1976).
7. Maxwell, E., I. S. Jacobs, and L. M. Levinson, Magnetic Separation of
Mineral Matter from Coal Liquids, EPRI AF-508, Electric Power
Research Institute, Palo Alto, CA (1977).
8. Maxwell, E., and D. R. Kelland, "High Gradient Magnetic Separation in
Coal Desulfurization," IEEE Trans. Magn., Mag-14, 482 (1978).
9. Maxwell, E., D. R. Kelland, and I. Y. Akoto, "High Gradient Magnetic
Separation of Mineral Particulates from Solvent Refined Coal,"
IEEE Trans. Magn., Mag-12, 507 (1976).
10. Liu, Y. A., G. E. Crow, C. J. Lin, and D. L. Vives, "A Pilot-Scale Study
of High Gradient Magnetic Desulfurization of Solvent Refined Coal
(SRC)," IEEE Trans. Magn., Mag-14, (1978).
349
-------
11. Mitchell, R., G. Bitton, and J. A. Oberteuffer, "High Gradient Magnetic
Filtration of Magnetic and Non-Magnetic Contaminants from Water,"
Separation and Purification Methods, 4^ 267 (1975) .
12. Petrakis, L. , and P. F. Ahner, "High Gradient Magnetic Separations in
Water Effluents," IEEE Trans. Magn., Mag-14, 491 (1978).
13. Oberteuffer, J. A., I. Wechsler, P. G. Marston, and M. J. McNallan,
"High Gradient Magnetic Filtration of Steel Mill Process and
Waste Waters," IEEE Trans. Magn., Mag-11, 1591 (1975).
14. Harland, J. R. , L. Nilsson, and M. Wallin, "Pilot-Scale High Gradient
Magnetic Filtration of Steel Mill Wastewater," IEEE Trans. Magn.,
Mag-12. 904 (1976).
15. DeLatour, C., and H. H. Kolm, "Magnetic Separation in Water Pollution
Control-II," IEEE Trans. Magn., Mag-11, 1570 (1975).
16. Yadidia, R., A. Abeliovich, and G. Belfort, "Algae Removal by High
Gradient Magnetic Filtration," Envr. Sci. and Tech., 11, 913
(1977).
17. Melville, D., F. Paul, and S. Roath, "High Gradient Magnetic Separation
of Red Cells from Whole Blood," IEEE Trans. Magn., Mag-11, 1701
(1975).
18. Whitesides, G. M., C. L. Hill, and J. Brunie, "Magnetic Filtration of
Small Heterogeneous Catalyst Particles. Preparation of Ferri-
magnetic Catalyst Supports," Ind. Eng. Chem. , Process Des. Dev.,
^5, 226 (1976).
19. Drehmel, D. C., and Gooding, G. H., "High Gradient Magnetic Particle
Collection," AIChE Symposium Series 74- (175) (1978).
20. Gooding, G. H., and Drehmel, D. C., "Application of High Gradient
Magnetic Separation to Fine Particle Control," JAPCA, 29
(5) (May 1979).
21. Gooding, C. H., "High Gradient Magnetic Filtration of Fine Particles
from a Gas Stream," Ph.D. Thesis, North Carolina State University,
Raleigh, N. C. (1979).
350
-------
CO
Oi
101
LU
LU
Q.
O
LU
10-
i i i 11
T
G = 0
A = 1
°.°.°.o o
10
r2
10'
10°
101
102
STOKES NUMBER (K)
«^ • XXl»fc-%^ I v>^l w ll_rb_i • fl«f
Figure 1. Contour map of collision radius, Y
* i •t••!
103
-------
99.9
99.8
93.5
90
u
t 70
UJ 60
1"
t
40
ill 30
20
10
5
2
1
EMPIRICAL
CORRECTION
BOUNCE CORRECTION
BOF DUST
Run No. 03021
F = 0.0080
L = 0.15m
V = 8.7 mis
Ba = 0.30 T
2 0.5 1.0 2 6
PARTICLE DIAMETER, ion
Figure 't. Comparison of models for magnetic filtration.
10
100
90
80
§ 70
t-
> 60
UJ
u 50
u.
UJ
s «°
h-
D 30
U
20
10
o
..-Boundaries of 30% ' /
over and under / /
prediction / • y*
/ • y
/ •/ r*
/ ^^**** xx
/ s * • S
/ / /
/ -/• s
/ /* /
/ / /'
// /
- ,y/ oi
- ty' p - 2 do w + »ji ,
fl"
r \ i i i i i i i i i
0 10 20 30 40 50 60 70 80 90 100
EXPERIMENTAL EFFICIENCY. Percent
Figure 3. Comparison of data and model.
352
-------
TECHNICAL AND ECONOMIC EVALUATION OF TWO NOVEL
PARTICULATE CONTROL DEVICES
By: R.R. Boericke, J.T. Kuo, K.R. Murphy
General Electric Company
Energy Systems Programs Department
1 River Road
Schenectady, NY 12345
ABSTRACT
A system-level comparison study is made for two novel particulate re-
moval devices, based on both technical and economic considerations. The two
novel devices are an acoustic agglomerator and an electrocyclone. These de-
vices are considered in combination with conventional cyclones, electrostatic
precipitators, and a baghouse.
Analytical models for the electrocyclone and the acoustic agglomerator
have been used to establish preferred geometries and operating parameters,
and to predict the performance of these devices in two specific applications:
a 663 MWe Pressurized Fluidized Bed (PFBC) power plant and a 475 MWe conven-
tional Pulverized Coal (PC) power plant. It is shown that both the electro-
cyclone and the acoustic agglomerator appear economically attractive for PFB
power plants, but are not competitive with electrostatic precipitators for
conventional pulverized coal burning plants.
INTRODUCTION
Two novel methods for dust removal are the electrostatically enhanced
cyclone (or electrocyclone), and acoustic agglomeration. The electrocyclone
combines electrostatic and inertial forces to achieve improved performance
and larger (more reliable and economical) unit size compared to conventional
cyclones. The acoustic agglomerator uses the oscillating flow of an acoustic
field to cause particle collision and agglomeration. The resulting large
particles are then more readily separable by conventional methods.
The application of these novel devices to pressurized fluidized bed com-
bustion (PFBC) is of special interest. PFBC is an advanced method for direct
combustion of coal in an environmentally acceptable manner and is particular-
ly suited to central station power generation. The major technology issues
in the development of PFB combined cycle power plants are associated with the
combustor/gas turbine interface. The hot combustion gas contains significant
amounts of fine dust composed of coal ash and attrited sorbent bed materials.
This dust, impacting on the turbine blades at high velocity, can rapidly
erode the metal and cause premature failure.
Figure 1 shows the dust load and size distribution leaving the combustor
exit, as well as current estimates of the dust load required to protect the
turbine (1). Also shown on Figure 1 is the maximum dust load of 0.03 lb/10"
Btu allowable under the EPA New Source Performance Standard. Note that if
353
-------
the dust is sufficiently free of coarse particles, the turbine can tolerate
total dust loadings of several hundred parts per million, well in excess of
the 22 ppm required to meet the EPA limit.
The fact that the gas turbine can tolerate dust loads in excess of the
EPA limit suggests that a relatively simple cleanup device (such as a cyclone
or electrocyclone) could be used for turbine protection, together with con-
ventional (low-temperature, low-pressure) cleanup equipment (e.g., baghouse)
downstream of the turbine to remove the fine dust particles. An important
economic and technical issue is whether the electrocyclone can also, with
appropriate design, provide good removal efficiency in the respirable dust
range, and thus eliminate the need for supplementary stack gas cleanup. Sim-
ilarly, the acoustic agglomerator may enable a HTHP cleanup device suitable
for turbine protection to also meet the EPA limit by agglomerating the fine
respirable dust particles into coarser ones which can then be removed up-
stream of the turbine.
EQUIPMENT DESCRIPTION
Electrocyclone Concept
The electrocyclone concept is a synthesis of two significant improvements
to conventional cyclones which offer promise of meeting the cleanup require-
ments in a PFB combined cycle system. These two features, the air shield and
electrostatic augmentation have been combined to yield a new cyclone concept,
the electrocyclone, with significantly enhanced performance.
The General Electric electrocyclone concept is illustrated in Figure 2.
The electrocyclone incorporates a clean air shield adjacent to the exhaust
duct which prevents short circuiting of dust into the outlet. The clean air
is provided by an axial flow separator/electrostatic charger device posi-
tioned in the exhaust duct of an upstream cyclone. This axial flow component
uses a high voltage corona electrode to charge the dust particles and improve
the separation efficiency. The vortex field in the axial flow component is
provided by the swirling motion naturally present in the exhaust of a conven-
tional reverse flow cyclone. Within the electrocyclone the separation of
dust is strongly augmented by another high-voltage field applied with an elec-
trode along the axis of the cyclone.
A principal advantage of the electrocyclone concept is that a large dia-
meter unit can be built to give the same efficiency as a small conventional
cyclone. The large unit is possible because the electrostatic separative
force is independent of the cyclone diameter in contrast to the centrifugal
force which decreases as the cyclone is made larger.
The physics of operation of the electrocyclone based on studies with a
two-dimensional electrovortex apparatus (3), have been incorporated into a
mathematical cyclone model, developed previously by GE (4). This original
three-region cyclone model was modified to accommodate electrostatic effects
and the air-shield feature of the GE electrocyclone concept. This was
354
-------
achieved by adding two new regions, resulting in a five-region model (2, 5,
6). Complete turbulent mixing is assumed within each region, and analytical
expressions for the cyclone collection efficiency are developed from the re-
sulting differential equations. The new model also includes improvements for
molecular slip and corona charging of small (respirable) particles.
Acoustic Agglomeration Concept
Acoustic agglomeration (AA) has been used successfully in the past as an
aerosol preconditioner. Most of the past experiments on AA were summarized
in an excellent work of Mednikov (7). More recently, the subject was re-
viewed by Shaw (8). AA can be applied effectively for the removal of airborne
dust particles when it is used as a preconditioner in series with more con-
ventional devices such as cyclones and filters. The dust laden gas (or aero-
sol) first enters the AA chamber. As it passes through the chamber, the gas
is subjected to the action of the sound waves radiated by the sound genera-
tor, which can be a mechanically vibrating piston, a siren, or a whistle.
In the AA chamber, smaller particles (diameter <10ym) agglomerate into larger
ones which can be effectively removed by conventional cyclones.
A model for predicting the performance of acoustic agglomerators has
been developed independently by Prof. David T. Shaw at SUNY-Buffalo, who
served as consultant to GE in performing the present study. This model in-
corporates Shaw's most recent findings on the dramatic improvement in agglo-
meration at sound levels above 160dB due to the onset of acoustic turbu-
lence, and is documented in Reference 2.
APPLICATION TO POWER GENERATION SYSTEMS
Application to PFB Combustion System
The design of a 663 MWe PFB combined cycle reference plant has been es-
tablished by General Electric in considerable detail (9). The plant consists
of three PFB/gas turbine modules which generate a total of 154 MW and a sin-
gle steam turbine generating 530 MW. The net plant output is 663 MW. Table
I summarizes the gas cleanup requirements. The cleanup equipment must oper-
ate at 1650°F and 10 atmospheres and be capable of handling 155,000 acfm gas
flow in each of the three gas turbine modules. The dust load leaving the
combustor is about 19,000 ppm and consists of virtually all the coal ash and
about a third of the dolomite in the form of fine solids worn from the larger
fluidized bed particles by abrasion. The gas cleanup systems for the PFB
application are designed to meet both the gas turbine protection requirement
and the EPA particulate emission standard of 0.03 Ib/lO^ Btu, as shown pre-
viously in Figure 1.
The electrocyclone design selected for the PFB system was based on para-
metric studies carried out with the five-region performance model. The unit
is 12 feet diameter, and the design details are shown in Figure 3. The size
of the electrocyclone was chosen such that the total flow of a PFB module can
be handled by four electrocyclones. The design flow rate for each electro-
cyclone is 38750 acfm.
355
-------
TABLE I. CLEANUP SYSTEM REQUIREMENTS (PER PFB/GAS TURBINE MODULE)
Gas Flow 1-81 x 106lb/hr
(503 Ib/sec)
1650°F
10 atm
155,000 acfm
Dust Load
Coal Ash (100%) (9.6%) (515,300 lb/hr)/3 16,500 Ib/hr
Dolomite (32.5%) (173,000 lb/hr)/3 18,700 Ib/hr
Total Dust 32,200 Ib/hr
Concentration 19,000 ppm
The individual acoustic agglomerator designed for the PFB system is an
18-inch diamter, 33 feet long cylindrical chamber. Each acoustic agglomer-
ator is designed to handle 4080 acfm of gas flow. Nineteen acoustic agglo-
merators are packed into one unit, and two units are required per PFB/gas
turbine module. Figure 4 shows the configuration of the assembly and the
packing arrangement of the acoustic agglomerators. Gas flows into the acous-
tic agglomerators from the bottom through the flow distribution plate and
exits from the top. The sound generators are air horns installed at the top
of each agglomerator chamber. The sound frequency is 900 Hz and the sound
level is 160dB. The length of the acoustic agglomerator is designed for one
second of residence time. The 900 Hz frequency was chosen to avoid exciting
lower resonant frequencies in the ductwork or structure of the power plant.
These natural frequencies are generally on the order of 300 Hz. The sound
intensity of 160dB (lw/cm2) was chosen such that the particle agglomeration
processes would be dominated by turbulent interactions.
Analysis of the grade efficiency obtainable with an electrocylcone based
on the analytical model studies reveals that it is impractical to design an
electrocyclone system to meet the EPA particulate limit. Instead, a three-
stage configuration is employed to satisfy the gas turbine requirement, to-
gether with a baghouse downstream of the turbine to reduce the particulate
emission to the EPA limit.
A combination of acoustic agglomerators and electrocyclones gives very
high collection efficiency in collecting particles of all sizes, readily
meeting the EPA limit as shown-in Figure 5. In this application, acoustic
agglomerators increase the collection efficiency of the 3-stage electrocyclone
system from 98% to 99.92%. Figure 6 shows a typical layout of the gas clean-
up system including the acoustic agglomerator units and electrocyclones.
356
-------
Application To Pulverized Coal System
The gas cleanup systems for the pulverized coal system are designed to
operate at 310°F and 1 atm. The reference plant is a 475 MW unit patterned
after TVA's Bull Run Plant (10) and has a total flue gas flow of about 1.46 x
10^ acfm. The estimated dust concentration is approximately 2 .6 gr/scf (4700 ppm)
The electrocyclone for the PC system is a 12-foot size design with a nom-
inal gas flow of 9.12 x 10 acfm. Sixteen parallel units in each of three
stages are needed to handle the total gas flow. The internals and geometri-
cal proportionality of the electrocyclone are similar to the one shown in
Figure 3 except that the refractory and stainless steel liners are not re-
quired. Since the operating temperature is relatively low, electrical insu-
lation for the high voltage electrodes will be less of a problem.
The acoustic agglomerator designed for the PC system is a 12-inch dia-
meter tube 58 feet long. The length of the tube was designed for 1 second
residence time. The design acoustic frequency is 860 Hz at a sound level of
160dB. The sound frequency and intensity are chosen for the same reasons as
given for the AA design for the PFB system. Figure 7 shows the acoustic
agglomerator system configuration and the tube arrangement.
Several particulate removal systems including combinations of acoustic
agglomerators, conventional cyclones or electrocyclone and electrostatic pre-
cipitators were considered for the PC systems. Currently, most PC plants are
built with electrostatic precipitators to control the plant particulate
emission. Predictions based on the Deutch equation showed that with acoustic
agglomerators, about 67% less plate area is required to meet the EPA emission
standard. If conventional cyclones are used with acoustic agglomerators, a
system consisting of a large (4 foot) size first stage cyclone followed by
one stage of 10-inch multiclones could meet the EPA limit. The estimated par-
ticulate collection efficiency for this system is 99.45%. With electrocy-
clones alone, performance calculations showed that the number of stages (about
8 in series) required to meet the EPA requirement is impractical. With the
acoustic agglomerators, only 3 stages of 12 foot electrocyclones (per path)
are necessary to meet the EPA standard. The calculated performance for the
three alternate systems (acoustic agglomeration plus electrostatic precipita-
tion, conventional cyclones, or electrocyclone) is shown in Figure 8.
SYSTEM COST ESTIMATES
Component Costing
The cost estimates given here were developed by first estimating the cost
of major components and then combining these elements into a complete system.
Whenever possible, component costs were obtained from established vendors. In
some cases (e.g., acoustic horns, B.C. power supplies) a cost allowance was
added to upgrade an existing commercial unit to the desired specifications.
Where no vendor cost data was available (i.e., electrocyclone, multiclone, and
agglomerators) a design drawing was prepared. From that design drawing the
quantities of various materials were computed. Unit costs for these materials
357
-------
(i.e., carbon steel at $1.00 per Ib), were used to develop the cost for that
entire item. For conventional cleanup equipment, historical cost data pub-
lished by EPA has been employed (11).
System Costing
Accurate estimates of installed cost for air pollution control equipment
require either detailed engineering drawings or an extensive data base. The
advanced designs under consideration in this study afford neither; thus, an
alternate approach was taken. This was to assume that installation costs for
particulate control parallel those for S02 scrubbers. The Tennessee Valley
Authority has done extensive cost estimates for these S02 cleanup systems.
In a recent presentation (12), they showed that the total capital investment
could be ratioed to the equipment costs by a factor of 4.38. Using this
method, costs for the capital investment of the various particulate cleanup
options have been developed and are discussed in the following paragraphs.
Table II summarizes the installed costs for two particulate cleanup
systems. System A comprises three stages of electrocyclones at high pressure
and temperature, coupled with a conventional baghouse located downstream of
the turbine as a final cleanup device. System B considers acoustic agglomer-
ators to shift the particle size distribution so that electrocyclones can
remove sufficient particles to meet the EPA NSPS. This option offers a saving
of $9 Million installed cost, and considerable reduction in space required
since the agglomerator is much smaller than the bag filter system.
TABLE II. 663 MWe PFB PARTICULATE CLEANUP SYSTEM COSTS
CLEANUP SYSTEM
A. Electrocyclone/Baghouse
B. Acoustic Agglomerator/
Electrocyclone
ITEM
Electrocyclones
Baghouse
TOTAL
Agglomerator
Electrocyclones
TOTAL
EQUIPMENT
COST
$12,700,000
4,600,000
12,700,000
INSTALLED
COST
$56,000,000
29,000,000
$85,000,000
$20,000,000
56,000,000
$76,000,000
Table III summarizes system costs for a 475 MWe pulverized coal plant.
System A, the baseline cleanup system, is either an electrostatic precipitator
or a bag filter system. For comparison, System B comprises an acoustic agglo-
merator coupled with a smaller ESP. The costs shown in Table IV show that
the reduction in precipitator size and cost is not sufficient to offset the
capital cost of the acoustic agglomerator. The cost of an electrocyclone
system has not been shown in Table III as it offers no advantages for a pul-
verized coal application relative to a baghouse or precipitator.
358
-------
TABLE III. 475 MWe PULVERIZED COAL PARTICULATE CLEANUP SYSTEM COST
CLEANUP SYSTEM
A. Precipitator
OR
Baghouse
B. Acoustic Agglomerator/
Precipitator
ITEM
ESP
Baghouse
Agglomerate r
ESP
TOTAL
EQUIPMENT
COST
$5,100,000
INSTALLED
COST
$15,700,000
24,700,000
$22,100,000
8,100,000
$30,200,000
SUMMARY AND CONCLUSIONS
The application of an electrostatically agumented cyclone to both a PFB
and a conventional pulverized coal power plant has been investigated. This
investigation revealed that three stages of electrocyclones would reduce the
concentration of particles larger than 5\im to levels sufficient to protect
the gas turbine from erosion in the PFB case. However, the NSPS limit on
particulates could not be met in either the PFB or the PC cases with realis-
tic assumptions regarding size, configuration or number of electrocyclones in
series.
System cost studies show that the electrocyclone continues to offer an
attractive cleanup system for the PFB application even if a baghouse must be
added downstream of the gas turbine in order to meet the NSPS limit. The in-
ability of the electrocyclone, as presently conceived, to effectively remove
particles in the respirable range effectively eliminates it from considera-
tion for pulverized coal power plant flue gas cleanup.
Studies of acoustic agglomeration have shown that at sound levels above
160dB (in the acoustic turbulence regime) a gas residence time of one second
in the agglomerator chamber is sufficient to effectively agglomerate the res-
pirable aerosol particles in either the PFB or PC cases. An agglomerator
chamber coupled to either a conventional cyclone system (roughing cyclone plus
two stages of 10" multiclones) or to 3 stages of 12-foot diameter electro-
cyclones could effectively meet the NSPS particulate limit. Alternatively,
the acoustic agglomerator could reduce the plate area required in an electro-
static precipitator by 67%.
Cost studies of acoustic agglomeration applied to the PFB power plant in-
dicate a potential cleanup system cost saving of 13% through elimination of
the baghouse otherwise required with the electrocyclone base case. The po-
tential cost saving with acoustic agglomeration is contingent on the develop-
ment of a suitable low-cost acoustic driver, as no suitable commercial equip-
ment presently exists.
Cost studies of the PC power plant indicate that acoustic agglomeration
with cyclones, electrocyclones or ESP is non-competitive with a conventional
359
-------
electrostatic precipitator or with a baghouse. This finding is consistent
with the earlier study by Hegarty and Channon (13), despite a reduction in
gas residence time made possible by the higher acoustic intensity.
ACKNOWLEDGEMENT
This work was supported by the Environmental and Safety Engineering
Division of the U.S Department of Energy under Contract DE-AC02-79EV10143,
Dr. Fred E. Witmer was the Project Technical Manager.
REFERENCES
1. Boericke, R.R., Hantman, R., and Kuo, J. "Assessment of Gas Turbine Ero-
sion by PFB Combustion Products," proceedings 6th International Confer-
ence on Fluidized Bed Combustion, Atlanta, GA, April 9-11, 1980.
2. "Novel Methods for Respirable Dust Removal," Final Report, Contract DE-
AC02-79EV10143, General Electric Co., Schenectady, NY, Sept., 1980.
3. Dietz, P.W. "Electrostatically Augmented Collection in Vortical Flows,"
(see this conference proceedings).
4. Dietz, P.W. "Collection Efficiency of Cyclone Separators," Report No.
79CRD244, Dec., 1979, General Electric Co., Schenectady, NY.
5. "Advanced Cleanup Device Performance Design Report (Task 4.3, Volume A -
Cyclone Theory and Data Correlation," Report FE-2357-70, Contract DE-AC21-
76ET10377, Oct., 1980, General Electric Co., Schenectady, NY.
6. Boericke, R.R. "Electrocyclone for High Temperature High Pressure Dust
Removal," paper 81-0393, AIAA 19th Aerosapce Sciences, Jan. 12-15, 1981,
St. Louis, MO.
7. Mednikov, E.P. Acoustic Coagulation and Precipitation of Aerosols, transl.
from Russian by C.V. Larrick, Consultants Bureau, New York, 1965.
8. Shaw, D.T. Recent Development of Aerosol Science, Wiley-Interscienee,
1978* Chap. 13, pp. 279-319.
9. "CFCC Development Program, Commerical Plant Design Definition," General
Electric Co., Energy Systems Programs Department Report FE-2357-28, March,
1978 (Contract EX-76-C-01-2357).
10. "Bull Run Steam Plant," Tennessee Valley Authority No. 38, pp. 27-152
1967.
11. "Particulate and S02 Emission Control Costs for Large Coal Fired Boiler "
EPA-450/2-78-007, Feb., 1978.
12. Stephenson, C.D., and Torstrick, R.L. "The Shawnee Lime-Limestone Com-
puter Program" presented at EPA's 5th Industry Briefing, Dec., 1979.
13. Hegarty, R., and Shannon, L.J. "Evaluation of Sonics for Fine Particu-
late Control," EPA 600/2-76-001, Jan., 1976.
360
-------
id*
1000
z
o
100
oc
ui
a.
V)
10
BED EFFLUX
.03 LB/I06BTU
EPA LIMIT
22PPM
FIRESIDE H
CASCADE TESTS
PROJECTED
TURBINE
TOLERANCE
10 100 1000
PARTICLE SIZE -MICRONS
AXIAL
FLOW
COMPONENT
INLET SCROLL
ELECTROCYCLONE
CONVENTIONAL
CYCLONE
Figure 1. Allowable Dust Load to Prevent
Turbine Erosion and to Meet EPA
New Source Performance Standard
1/2 CARBON \|
STEEL PLATE
5" LIGHTWEIGHT
CASTABLE REFRACTORY
1/2" 310 STAINLESS
STEEL PLATE
Figure 2. Electrocyclone
Configuration
51' 4"
EXPANSION JOINT is" DIA. AGGLOMERATORS
6'-l7|-"DIA GAS FLOW
CHANNEL
3-12 DIA. GAS FLOW
CHANNEL
5 LIGHTWEIGHT CASTABLE
REFRACTORY
i"-3IO STAINLESS STEEL
PLATE
SECTION
Figure
-5'4"O.D.
Figure 3. Electrocyclone Design
for PFB Application 361
4. Acoustic Agglomerator
Configuration for PFB
Application
-------
§100
CURVE A-WITH ACOUSTIC AGGU3MERATOR
CURVE B-WITHOUT ACOUSTIC AGGLOMERATOR
*^
PROJECTED TURBINE
TOLERANCE
ACOUSTIC
AGGLOMERATOR
UNIFLOW CYCLONE
fj^t
f
r: f
WH
6-
p.
..-
X
f
(*>
V
/
•. £
ii
1
A
$
H
I
A
V,'
/•
||, ,1 STACK
j GAS TURBINE GEN.j GAS COOLERS
ELEVATION
I 5 10 100 1000
PARTICLE SIZE - MICRONS
Figure 6. Equipment Arrangement For PFB System With
Acoustic Agglomerator and Electrocyclone
Figure 5.
Three-Stage
Electrocyclone
System
Performance
r®
XE
Mj
i
i
| j
- SOUND GENERATOR
,-AIR INLET
/,-AIR OUTLET
• 5'-
1
5-0"
58-
1^
5'-0"
L_
5'
0"
2l'-0"
0"
— 12" 1.0. AG
y CARBON
0"
\-FLOW CHANNEL
END ELEVATION
224- 12 'I.D. CARBON STEEL
/ AGGLOMERATORS PER
/ MODULE -4 MODULES REO'D.
/ ^224-l2"DIA.GAS FLOW
// CHANNELS PER MOOULE-
// 4 MODULES REO'D.
0 0000000.0. 00. 000
8? 5
8? •#
&8 j?
i| 38
?W!i?888o888o8!l8o88S88S8g8ol
Ifi'-O"
SLOMERATOR ENLARGED
STEEL PLATE SECTION f^\
Figure 7. Acoustic Agglomerator
Configuration for
Pulverized Coal
Application
JLVERIZED COAL COMBUSTION
DUST EFFLUX
CURVE I - ACOUSTIC AGGLOMERATION
PLUS ELECTROSTATIC PREC1PITATO*
CURVE 2 • ACOUSTIC AGGLOMERATION
PLUS Z STAGES Of CONVENTIONAL
CYCLONES
CURVE 3 - ACOUSTIC AGGLOM. PLUS 3
STAGES OF IZ-FOOT ELECTROCYCLONES
EPA LIMIT
• 30 PPM
362
PARTICLE SIZE, dj, (MICRONS)
Figure 8. Particulate Removal
Performance for
PC Plant Using
Acoustic
Agglomeration
with ESP,
Cyclones, or
Electrocvclones
-------
THE ELECTROSCRUBBER* FILTER
APPLICATIONS AND PARTICULATE COLLECTION PERFORMANCE
By: David Parquet
Combustion Power Company, Inc.
A Weyerhaeuser Company
Menlo Park, California
ABSTRACT
The Electroscrubber Filter, used to remove particulate in dry form from
gas streams, is a simple but dramatic improvement to Combustion Power
Company's DRY SCRUBBER. Achieving "the best of both worlds" of the DRY
SCRUBBER granular filter and electrostatic collection, the ELECTROSCRUBBER
Filter utilizes a moving bed of filtering media and a pneumatic media recir-
culation and dust collection system, allowing self cleaning operation on
a continuous basis. The concept development and principle of operation is
discussed along with emission test data from selected applications.
INTRODUCTION
In 1974, Combustion Power Company started up on a small wood fired
boiler its first DRY SCRUBBER. Since that first unit, CPC has installed
collectors on numerous other applications. More significantly, as it
pertains to the ever more stringent particulate emissions codes, a second
generation filter was developed that retained the good points of the DRY
SCRUBBER while much improving particulate collection performance, replacing
the maintenance prone mechanical equipment of the former unit with a
simple air blower, and reducing the cost of the unit as well. This new
filter has only a single moving part and includes an electrostatic grid to
augment the basic collection phenomena of the granular filter. The
ELECTROSCRUBBER Filter can accept higher inlet loadings than the DRY SCRUBBER
and will provide outlet loadings lower than New Source Performance Standard
levels. High collection efficiencies are attainable on fine, submicron
particulate.
ELECTROSCRUBBER FILTER PRINCIPLE OF OPERATION
Conceptually, as shown in Figure 1, the ELECTROSCRUBBER Filter consists
of a cylindrical vessel containing two concentric louvered cylindrical tubes.
The annular space between the tubes is filled with pea-sized gravel media.
The particulate-laden exhaust gas enters the filter through appropriate
breeching and is distributed to the filter face by the plenum section formed
by the inner louver cylinder. Dirty gas is passed through the filter media
at velocities ranging from 100 to 150 feet per minute and the particulate
is removed from the gas stream by impaction with the media. Clean gases
exit to downstream breeching or directly to the atmosphere through a free-
standing exhaust stack that can be added to the top of the ELECTROSCRUBBER
Filter unit.
* Registered U.S. Patent & Trademark Office
363
-------
MEOIA/DUST
OE ENTRAPMENT
CHAMBER
COMPRESSED.
To prevent a filter cake from
forming on the face of the filter,
and the resulting potential plugging
problems and high pressure drop, the
filtering media is continuously, but
slowly (six to ten feet per hour),
moved downward in plug or mass flow.
The resulting churning action across
each louver opening prevents a filter
cake from forming. To provide com-
plete cleaning of the louver face,
the louvers are designed so that some
of the media is pushed through each
louver opening thus preventing any
bridging or buildup of particulate
material.
The particulate-laden media is
continuously removed at the bottom of
the ELECTROSCRUBBER Filter where it is
transported by a pneumatic conveying
system to the media/particulate de-
entrainment section of the system.
The action of the media being trans-
ported vertically in the pneumatic
lift pipe separates the particulate
from the media so that the particu-
late can be pneumatically removed
from the de-entrainment section for
pneumatic transport to a particulate
separation and storage silo. The clean media then drains by gravity from
the de-entrainment section and is returned to the top of the ELECTROSCRUBBER
Filter unit for recycling.
The ELECTROSCRUBBER Filter module is designed to operate full of media
at all times. To accomplish this, the media overflow vessel, as shown in
Figure 1, is used to assure that the freeboard section of the ELECTROSCRUBBER
Filter module remains full and to eliminate any differential media flow down
to the two media return pipes. Excess media is returned from the media level
control section by the media overflow pipe to the media inventory hopper.
Media from this hopper is continuously "trickle" fed back into the media
lift pipe to provide an oversupply of media to the media overflow vessel to
assure that the ELECTROSCRUBBER Filter module is always completely full of
media. Air is piped from the media lift air blower into the lower portion
of the media overflow pipe to provide purge air up this pipe which prevents
any residual dust from re-entering the top of the ELECTROSCRUBBER Filter.
An electrical conductor, configured in the form of a cage, is positioned
within the media bed of the ELECTROSCRUBBER Filter as shown in Figure 2. A
high voltage is applied to this conductor and the electrical field generated
between the conductor and the inlet and outlet louvers enhances the collection
of the particulate in the gas stream, greatly increasing the efficiency of
Figure 1 ELECTROSCRUBBER Filter Flow
Schematic
364
-------
the basic granular filter system. (See
Figure 3). This enhancement phenomenon
relies on the fact that industrial pro-
cesses produce particulate with slight
positive or negative charges. Put
simply, the electrical field greatly
increases the probability that the
particulate will impact with a piece
of media in its long, tortuous path
through the filter. As also seen in
Figure 3, electrostatic augmentation
of particulate collection is especi-
ally pronounced on the fine, submicron
particulate.
The ELECTROSCRUBBER Filter pro-
vides extremely high collection effi-
ciency with very low pressure drop,
collects particulate in a dry form and,
because of the large heat sink provided
by the filtering media, the propagation
of an explosion or fire is eliminated
under all conditions within the
ELECTROSCRUBBER Filter. And, because
it is not required that the charge of
the particulate be changed to allow
collection (as in an electrostatic pre-
cipitator), the ELECTROSCRUBBER Filter
is insensitive to such particulate
characteristics as resistivity. Power
requirements for the electrostatic
grid are typically only 10 to 20 watts
per 1000 acfm.
APPLICATIONS HISTORY
As of March, 1981, DRY SCRUBBERS
and ELECTROSCRUBBER Filters have had
experience on five separate applica-
tions including boilers burning wood,
salt-laden wood, coal, or combination
fuels; lime kilns; coke calciners;
municipal solid waste incinerators; and
steel sinter machines. As shown in
Table 1, a total of 117 separate filters
have been or are being installed at 25
sites for a total of over 8 million
acfm.
ELECTRIC
FIELD
NOZZLE FORMED BY
ROCK CLUSTER
PARTICLE IMPACT
i- '/~ EFFECT OF
.1 •< FIELD ON -t
V A PARTICLES
LAU
'O,
EFFECT OF
FIELD ON-
PARTICLES
ELECTROSTATIC
GRID CHARGED
TO 20,000 VOLTS
NEGATIVE
Mechanical Impaction Mechanism
of Particulate Collection in a
Granular Filter Augmented by
Electrostatic Field
Figure 2 ELECTROSCRUBBER Filter High
Voltage Particulate Collection
Concept
15K VOLTS
10K VOLTS
I I I I
I I I I I I
.3 A .5 .7 1 23
PARTICLE AERODYNAMIC DIAMETER [MICRONS!
Figure 3 Typical Effect of Voltage
on Fractional Efficiency of Par-
ticulate Collection. (Example is
from a Wood Boiler ELECTRO-
SCRUBBER Filter
In the Application Summaries (AS)
that follow, selected test data has
been presented with the intention of
365
-------
TABLE 1. DRY SCRUBBER & ELECTROSCRUBBER FILTER INSTALLATION SUMMARY
Application
Quantities Total Gas Flows (acfm)
Sites/ In In In
Filters Operation Startup Construction
Total
Boilers
Wood 13/48 2,040,000
Wood (salt) 2/6 300,000
Coal 2/4 346,000
Coal/wood 2/28
Lime Kilns 2/3 262,000
443,000
546,000
223,000
960,000
3,029,000
523,000
346,000
960,000
262,000
Coke Calciners
Cooler
Kiln Gas
MSW Incinerator
Steel Sinter Plant
2/2 40,000
1/2
1/2
3/22 781,000
28/117 3,769,000
95,000
220,000
90,000
1,740,000
628,000 3,689,000
135,000
220,000
90,000
2,521,000
8,086,000
presenting the reader with a cross-section of ELECTROSCRUBBER Filter capa-
bilities. Note that when the term "DRY SCRUBBER" is used, it refers to
the filter schematically shown in A.S. No. 1. The granular filter section
is identical to that already described. The DRY SCRUBBER, however, the
predecessor of the ELECTROSCRUBBER Filter, utilized a vibrating screen-type
dust removal system and a bucket elevator to transport the media back to
the top of the module filter area.
A.S. No. 1 is a DRY SCRUBBER which had an electrostatic grid retrofitted
into the media cavity on a wood fired boiler application. A.S. No. 2 depicts
full scale pneumatic ELECTROSCRUBBER Filter performance on the same applica-
tion. A.S. No. 3 and No. 4 show ELECTROSCRUBBER Filter fine particulate
performance as demonstrated with a mobile 500 acfm test unit and with a
stationary 7000 acfm test unit, respectively, on two diversely different
applications. A.S. No. 5 compares the performance of a DRY SCRUBBER with a
grid retrofit operating on a pulverized coal boiler to the performance of the
ELECTROSCRUBBER Filter 7000 acfm test unit. A.S. No. 6 provides a similar
comparison for a lime kiln application. These summaries clearly portrav the
ELECTROSCRUBBER Filter as a versatile, high performing system fully capable
of meeting the difficult requirements of diverse applications.
366
-------
APPLICATION SUMMARY NO. 1
APPLICATION:
MAXIMUM DESIGN GAS FLOW:
DUST COLLECTOR:
WOOD FIRED BOILER
50,000 acfm
1 DS400 DRY SCRUBBER RETROFITTED WITH ELECTROSTATIC
GRID IN SEPTEMBER 1978
The DRY SCRUBBER granular filter
was invented in 1973 and was installed
on this wood boiler application in
1974. At that time, lenient pollu-
tion codes did not require the strin-
gent outlet loadings of today's
requirements. During the last seven
years, however, the ever more diffi-
cult requirements caused CPC to re-
evaluate and improve its technology.
After extensive lab tests the first
improvement, the electrostatic grid,
was developed and retrofitted into
this first DRY SCRUBBER unit in order
to demonstrate its capability on an
operating system. (The second improve-
ment, the pneumatic media recirculation
and ash collection system, was not
retrofitted into this unit). As shown
below, the performance of the grid was
marked: outlet loading reduced by a
factor of five to 0.01 gr/dscf and
outlet opacity reduced by a factor of
eight to clear stack opacity. Grid
power consumption was typically only
400 to 1000 watts.
ELEVATOR
Retrofit of Electrostatic Grid into
DRY SCRUBBER with Mechanical
Media Recirculation of Ash
Collection System
GRID VOLTAGE. K VOLTS
GRID VOLTAGE. K VOLTS
Outlet Loading vs. Grid Voltage and Single Pass Opacity vs. Grid Voltage
Inlet Loading and Inlet Loading
367
-------
APPLICATION:
CAPACITY:
MAXIMUM DESIGN GAS FLOW:
DUST COLLECTOR:
STARTUP:
APPLICATION SUMMARY NO. 2
WOOD FIRED POWER BOILER
550,000 pph @ 30% MOISTURE HOGGED FUEL
420,000 acfm @ 345 F
3 ES250-4 ELECTROSCRUBBER FILTERS
SEPTEMBER 1979
After an extensive economic tradeoff analysis against alternative parti-
culate collection technologies and a test and demonstration program that in-
cluded the grid retrofit program discussed in A.S. No. 1, this ELECTROSCRUBBER
Filter system was designed for and installed on the (then) largest 100% wood
waste fired boiler in the world. These filters included the pneumatic media
recirculation and ash collection system that had been in development in CPC's
test labs since 1974. After startup in 1979, the filter system has NEVER
caused the boiler to go down in over a year and a half of operation. Per-
formance is shown below for good fuel (normal wetness, summer quality) and
for poor fuel (reclaimed, high moisture). The results show trends not
typical of other particulate collectors, that is, outlet loading remains
constant even though steaming rate and inlet loading are increasing. The
emissions from this ELECTROSCRUBBER Filter system have been tested both by
the user and the EPA with similar results.
(gr/DSCF@12%CO2)
0.6
f Design Loading-,
H
Good Fuel
Poor Fuel
400
Steaming Rate 10a Ib/hr
(gr/DSCF
0.05
0.04
0.03
0.02
0.01
0.0
912% CO,)
1 1 1 1 1 1
Q Good Fuel
I 1 Poor Fuel
High Ash & Grit
,A Grid Voltage Off ^^7 in Fuel
1 / of
/
f s y^ ©_
1 ! 1 1 1 1
300 400 500 60
Steaming Rate 103 Ib/hr
1980 Test Data: Inlet Loading vs.
Steaming Rate
1980 Test Data: Outlet Loading vs,
Steaming Rate
368
-------
APPLICATION SUMMARY NO. 3
APPLICATION:
TEST DATE:
SALT LADEN WOOD FIRED BOILER TEST PROGRAM
FINE PARTICULATE COLLECTION (50% <1.6 /O
MARCH 1980
In March 1980, CPC sent its 500
acfm ELECTROSGRUBBER Filter mobile
test unit for demonstration testing
on a wood boiler fired with hogged
bark from logs floated in sea water.
The application is a particularly dif-
ficult one in that the particulate is
primarily salt (60+% salt, 40+% ash)
and has a very high proportion of sub-
micron material. As has been reported
elsewhere, the salt-laden ash was a
troublesome material for the DRY
SCRUBBER. First, simply stated, the
vibrating screen used on those filters
did not thoroughly remove collected
particulate from the media, thus re-
cycling it back into the filter area
and causing buildup. Second, the
large portion of submicron material
was difficult for the filter to cap-
ture, resulting in rather high outlet
opacities. The curves and table
below demonstrate the effectiveness of
the improvements made to the ELECTRO-
SCRUBBER Filter. Since a large portion
of the particulate was submicron, the
effectiveness of the electrostatic
grid was particularly evident.
RUN 7A
RUN 8
RUN 10
0
CO
z
I
UJ
-J
o
10
20 30
GRID VOLTAGE, KV
Outlet Opacity vs. Grid Voltage
SUMMARY OF PARTICULATE EMISSIONS TEST
Run
No.
1
2
3
4
5
6
7
74,
at
o '
9,
10*
lit
Gas Flow
(acfra)
570
570
560
560
520
530
560
~800
540
570
—
Grid
Voltage
(kV)
45
45
45
45
45
0
45
45
45
45
—
Inlet Dust
Loading
(gr/dscf @ 12% C02)
0.283
0.271
0.399
0.325
0.357
0.329
0.412
—
0.513
0.855
Outlet Dust
Loading
(gr/dscf
-------
APPLICATION SUMMARY NO. 4
APPLICATION:
TEST DATE:
STEEL SINTERING MACHINE TEST PROGRAM
FINE PARTICULATE COLLECTION (80% <1n
MAY 1980
For this application, the application's process parameters were very
closely duplicated on a 7000 acfm ELECTROSCRUBBER Filter stationary test
unit at CPC's test labs. In the actual sintering process, the customer had
developed a method of adding calcium chloride to the sinter mix to reduce
the alkali content of the sinter material for the blast furnace. The pro-
cess is successful, but the particulate released is almost entirely
submicron potassium chloride (KC1). As shown, a KCl fume generator produced
particulate as high as 90% in the submicron range with inlet loadings aver-
aging 0.26 gr/acf. Two of the data sets presented below clearly demonstrate
the effect of the electrostatic grid. It is interesting to note that the
highest power required for the grid occurred during the 75 kV test and was
less than 8 watts/1000 acfm.
Schematic of 7000 acfm ELECTROSCRUBBER Filter Stationary Test Unit
* ,00
>
u
z
uj
u 90
70
50
.060
r
; .050
i
( .040
i
| .030
i
i .020
l
! .010
.2 .3 .4 .6 7 1 2 345
PARTICLE AERODYNAMIC DIAMETER (MICRONS)
Fractional Particulate Collection
Efficiency vs. Grid Voltage.
10
30 40 SO 60 70 80 90 100
GRID VOLTAGE (kV)
Outlet Loading vs.. Grid Voltage
370
-------
APPLICATION SUMMARY NO. 5
APPLICATION: PULVERIZED COAL FIRED BOILER
In 1977, two DRY SCRUBBER filter units were installed on a pulverized
coal fired boiler. After startup, it was determined that the filter panel
area many times had inlet loadings higher than expected causing the filter
to "saturate" or to run with characteristics detrimental to the health of
the unit. A resultant effect was an outlet loading higher than allowable.
Similar to A.S. No. 1, electrostatic grids were retrofitted into these
units with the results as shown in the first table below. When the inlet
loading to the filter panel is below an allowable level, the filter pro-
vides below-code performance.
The second table below presents the performance of the improved
ELECTROSCRUBBER Filter as demonstrated on the 7000 acfm stationay test unit
described in A.S. No. 4. As has been suggested, there are many other
features in the ELECTROSCRUBBER in addition to the electrostatic grid and
the pneumatic media cleanup system which result in reduced outlet loading
for greatly increased inlet loadings.
SUMMARY OF PARTICULATE EMISSIONS TESTS ON 2 DS600C DRY SCRUBBERS
STARTED UP IN NOVEMBER 1977 AND RETROFITTED WITH
ELECTROSTATIC GRID IN AUGUST 1979
DRY SCRUBBERS
2 DS60UC
DRY SCRUBBERS
Gas Flow
Gas Temperature
Scrubber A~P
Inlet loading
Grid voltage
Actual outlet loading
Allowable outlet loading
Collection efficiency
before
Grid Installation
134,000 acfm
425 F
6-7 in. H20
0.45 gr/acf
0
0.134 gr/acf
0.067 gr/acf
70%
after Grid
Installation
124,000 acfm
425 F
6 in. HoO
0.46 gr/acf
20,000 volts
0.023 gr/acf
0.067 gr/acf
95%
SUMMARY OF PARTICULATE EMISSIONS TESTS ON PULVERIZED COAL FIRED DUST
WHEN TESTED IN 7000 ACFM ELECTROSCRUBBER FILTER STATIONARY TEST UNIT
Outlet
Gas Grid Grid Inlet Outlet Opacity
Flow Temp Voltage Current Loading Loading Efficiency Double Pass
(acfm) (F) (kV) (ma) (gr/acf) (gr/acf) (%) (%)
6450 290 20 0.4 0.530 0.008 98.5
6670 290 20 0.4 1.864 0.012 99.4
6600 300 20 0.4 3.387 0.024 99.3
3.7
3.0
3.5
371
-------
APPLICATION SUMMARY NO- 6
APPLICATION: LIME KILN
In 1977, a DRY SCRUBBER unit was started up on a small 200 ton per day
lime kiln. Although the unit regularly collected between 800 and 1000 pounds
per hour of particulate, the outlet loading was not consistently below
regulatory requirements. After making some other modifications, an
electrostatic grid was installed in the unit and activated in 1979. As
shown in the first table below, the performance improved dramatically.
Since the retrofit, the unit continues to consistently provide below-code
outlet loadings at power levels of less than 10 watts per 1000 acfm.
The second table below presents the performance of the improved
ELECTROSCRUBBER Filter as demonstrated on the 7000 acfm stationary test unit
described in A.S. No. 4. As has been suggested there are many other
features in the ELECTROSCRUBBER Filter in addition to the electrostatic
grid and pneumatic media cleanup system which produce very low outlet
loadings at high inlet loadings. Note that for lime dust, voltages much
in excess of 30 kV could be used which would result in even lower outlet
loadings.
SUMMARY OF PARTICULATE EMISSIONS TESTS ON 1DS400C DRY SCRUBBER
STARTED UP IN NOVEMBER 1977 AND RETROFITTED
WITH ELECTROSTATIC GRID IN NOVEMBER 1979
DRY SCRUBBER DRY SCRUBBER
1 DS400C (before Grid (after Grid
DRY SCRUBBER Installation) Installation)*
Grid voltage (kV)
Gas flow (acfin)
Gas Temperature (F)
DRY SCRUBBER AP (in. H20)
Inlet loading (gr/acf)
Actual outlet loading (gr/acf)
Allowable outlet loading (gr/acf)
DRY SCRUBBER collection efficiency (%)
N/A
36,000
375
7.5
2.6
.08
.10
96.9
10 kV
37,792
375*
7.6*
3.26
.030
.090
99.08
20 kV
41,182
375*
7.7*
3.04
.026
.092
99.15
30kV<$
37,724
375*
7.6*
3.18
.022
.090
99.31
Estimated
Grid current drain less than 10 ma in all cases
Higher voltages possible
SUMMARY OF PARTICULATE EMISSIONS TESTS ON LIME DUST
WHEN TESTED IN 7000 ACFM ELECTROSCRUBBER FILTER STATIONARY TEST UNIT
Gas Flow
(acfm)
6900
6400
6000
6000
6000
Grid Voltage
(Kv)
0
20
0
20
30
Inlet Dust
Loading
(gr/acf)
0.7
0.7
6.6
6.6
6.6
Outlet Dust
Loading
(gr/acf)
0.02
0.004
_
Efficiency
97.7
99.3
Opacity
Q
o
1
1
19
2
f i
372
-------
HIGH EFFICIENCY PARTICULATE REMOVAL WITH
SINTERED METAL FILTERS*
By: B. E. Kirstein, W. J. Paplawsky and D. T. Pence
Science Applications, Inc.
4030 Sorrento Valley Boulevard
San Diego, California 92121
T. G. Hedahl
EG&G Idaho, Inc.
P.O. Box 1625
Idaho Falls, Idaho 83415
ABSTRACT
Because of their particle removal efficiencies and durability, sintered
metal filters have been chosen for high efficiency particulate air (HEPA)
filter protection in the off-gas treatment system for the proposed Idaho
National Engineering Laboratory Transuranic Waste Treatment Facility. Proc-
ess evaluation of sintered metal filters indicated a lack of sufficient proc-
ess design data to ensure trouble-free operation. Subsequence pilot scale
testing was performed with flyash as the test particulate. The test results
showed that the sintered metal filters can have an efficiency greater than
0.9999999 for the specific test conditions used. Stable pressure drop char-
acteristics were observed in pulsed and reversed flow blowback modes of opera-
tion. Over 4900 hours of operation were obtained with operating conditions
ranging up to approximately 90°C and 24 vol% water vapor in the gas stream.
INTRODUCTION
Transuranic Waste Treatment Facility
During the past 26 years, thousands of tons of nuclear waste contaminated
with transuranic (TRU) elements have been stored or buried at the Idaho
National Engineering Laboratory (INEL). This waste, composed of both combus-
tible and noncombustible materials, may be retrieved, processed, and shipped
to a federal repository for permanent disposal. An attractive method for
processing this waste is to convert it by means of a high temperature slagging
pyrolysis incinerator (SPI) process into an inert, basalt-like solid that
encapsulates the TRU elements. The SPI has been selected as a processing
method to be incorporated into the INEL Transuranic Waste Treatment Facility
(TWTF) (1).
*Work performed under USDOE Contract EY-76-C-07-1570
373
-------
The major process operations in the TWTF will be waste receiving, waste
preparation, incineration, slag handling, and off-gas treatment. The Waste
Processing Building will incorporate the latest available technology in waste
handling, fissile material assaying, criticality control, instrumentation,
off-gas cleanup, remote operation/maintenance, and decontamination. Process
equipment generally will be operated remotely using both local and central
controls .
A key to ensuring safety and environmental acceptability for the TWTF is
an efficient off-gas treatment system. The off-gas treatment system for the
TWTF must establish effective control of radioactive and other potentially
harmful airborne materials. Various off-gas treatment technologies were
evaluated with respect to fine particle recovery for use in the TWTF. A
detailed evaluation of the material balance of the incinerator off-gas stream
indicated that high efficiency particulate air (HEPA) filters alone would
become overloaded and plug after approximately 1/2 hour of operation. There-
fore, some type of continuous particulate recovery equipment would be required
in the off-gas treatment system.
The removal efficiencies of venturi scrubbers fall off rapidly for par-
ticles with aerodynamic diameters below about 1 ym, which makes their use
impractical for this application. Electrostatic precipitators (ESP) also do
not appear to have sufficient fine particulate removal efficiencies, and the
design of an ESP requires far more information about the incinerator particu-
late than is available. Bag filters do not have a sufficiently fine particu-
late recovery efficiency for this application, and further, pose other safety
problems for a secure radioactive materials off-gas treatment system. A dry
off-gas treatment system using sintered metal filters (SMFs) was selected for
development and design, based on SMFs high particulate removal efficiency,
reduced maintenance (possibly in a remote environment), and absence of any
secondary contaminated liquid waste. However, the lack of adequate design
information for SMF's necessitated a pilot scale test program to verify the
reported fine particulate recovery efficiencies and operating conditions.
The description of this program is presented in the sections that follow.
Background of Sintered Metal Filters
The use of SMFs to remove particulate from a gas stream has been prac-
ticed for over 25 years (2). Yet, in this time, quantitative design equations
have not been developed nor has a satisfactory explanation been given of how
the filters operate without plugging over long operating times. These filters
have been used in a variety of nuclear applications that are only partially
documented with respect to operating conditions and particulate characteriza-
tion. Industrial experience with SMFs is virtually nonexistent in the litera-
ture. While it is known that there are industrial users, limited contact with
these users reveals their desire not to advertise certain aspects of their
manufacturing processes. In the presentation that follows, experimental
results are provided describing the operation of SMFs to remove flvash from
gas (air) stream where the primary objective was to obtain stable operation
with respect to pressure drop and to obtain high particle removal efficien-
cies. These results were then used in the conceptual design of an off-
8" a radl°aC thaf u^s a SPI
a
374
-------
EXPERIMENTAL PROGRAM TO TEST SINTERED METAL FILTERS
Purpose
The primary objective of the SMF test program was to verify the applica-
bility of the filters in the INEL TWTF off-gas treatment system. Two criteria
that were considered necessary for applying SMFs were particulate removal
efficiency and stable pressure drop. The particulate size distribution used
in the tests had to match or be finer than that expected from the SPI. The
pressure drop requirements were stable operation at as low a pressure drop as
practicable. A further objective was to test both blowback methods to deter-
mine any advantage of one over the other with respect to obtaining these
primary objectives.
Basic System Description
The pilot plant apparatus used for evaluating the performance of the SMFs
is shown in Figure 1. The major design considerations in the use of SMFs
involve controlling the gas flow through the filter and providing blowback to
clean the filter. A blower supplies air to the apparatus in which flyash is
redispersed by an air ejector. The air/flyash mixture flowrate is measured
and controlled prior to the filter vessel. Six cylindrical SMFs were instal-
led in the vessel, with filter exhausts manifolded to provide a single off-gas
stream for sampling. Total collection filters, or in-line filters, were used
to quantitatively collect all entrained particles for SMF efficiency
measurements.
An auger feeder gradually adds flyash to the ejector apparatus, which is
supplied by an external pressurized 310 kPa air supply. The ejector consists
of a stainless steel tube inserted in a tee to provide the aspirating action
necessary to finely disperse the flyash into the gas stream. Gas flow meas-
urements were accomplished using venturi and pitot tubes. The gas flow rate
was maintained at an essentially constant value of 1.25 normal cubic metres
per minute (Nm^/min).
The main air supply is provided by a regenerative blower. The total flow
from the blower is split into two streams: a main flow to the filters and a
vent flow used to control the main flow rate. As filter pressure drop
increases with time for a specified flowrate, the vent valve position provides
the control variable necessary to maintain the desired flow.
The tested SMFs were cylinders 0.91 m long by 6.8 cm OD with a 0.15-cm
wall thickness. The porosity of the filters was 0.5 absolute. One end of
each filter element is capped, and the other end is connected to exit gas
piping. Six filters were installed in a vessel 34.5 cm in diameter by 1.03 m
long. A 60-degree cone was welded to the bottom for particulate collection
following blowbacks.
The most important operating variable for an SMF appears to be the super-
ficial velocity. The recommended superficial velocity determined by experi-
ence is usually 1.2 to 2.1 m/min, or sometimes even less, even though pressure
drop versus superficial velocity for clean filters is advertised at
considerably greater velocities (3,4).
375
-------
Flyash
Air
supply
Flyash
ejector
J
•Auger
feeder
Blowback
valves (3)
Expansion
compensator
Valve •
Downstream
sample ports
Exit
Flow
measurement
venturi
Viewing
port
Sintered metal
filter
Q Sampling ports
INEL-A-14 138
Figure 1. General Schematic of Pilot-plant apparatus for testing sintered
metal filters.
-------
During operation of the pilot plant system, the gas and particles enter
the filter vessel such that the inlet stream does not impinge directly on a
filter surface. A baffle was installed to distribute the gas and particles
into the vessel. The gas and particles flow to the surface of the filter
where the particles are collected, and the gas flows through the filter ele-
ment. A pressure drop results due to the gas flow resistance of the porous
metal and filter cake. The pressure drop increases with time due to the
increasing cake thickness. When the pressure drop reaches a prescribed value
(depending on the test), blowback is initiated.
Two blowback techniques were tested and evaluated based on litera-
ture (5,6) and vendor information: pulsed and reverse flow. In the pulsed
blowback method, a tube or nozzle is directed to the filter exit. During
blowback, a pulse of approximately 414 kPa air was delivered to a pair of
filters for a fraction of a second through the blowback valves. Two filters
were blown at a time with a 20-sec pauses between pulses until all filters
were cleaned. During the series of SMF tests, the effect of blowback nozzle
tube diameter was also investigated.
The reverse flow method uses a secondary air system (1.4 m^/min regen-
erative blower) which operates continuously. During SMF loading, exhaust
valves remain open and the reverse flow air is vented; the blowback air supply
valves remain closed. When blowback is initiated (pressure drop controlled)
the exhaust valves close and the blowback valves are opened, thus diverting
the reverse flow blowback air through the filter set and into the vessel.
This flow configuration is applied to two SMFs at a time for approximately
3 sec. The other pairs of filters are then sequentially blown back at 30-sec
intervals.
Test Program (7,8)
Particulate Size Distribution and Loading
The particulate size distribution and loading of the redispersed flyash
reaching the SMFs were measured by sampling after the filter vessel when no
filters were installed. The reason for obtaining these measurements in this
manner was that the filter vessel had an efficiency for particulate removal
of about 60% for the flyash used. The filter efficiencies reported here are
based on these downstream measurements to ensure that artificially high
results are not reported.
The particulate size distribution expected from the SPI system is char-
acterized by 100 wt% less than 32 ym in aerodynamic diameter, 50 wt% less
than 10 ym, and 0.1 wt% less than 1 ym. The redispersed flyash that reached
the SMFs in the laboratory system was finer than the incinerator par- tides
below 10 ym as measured by a cascade impactor. The particulate load-
ings to the filters were in the range of 3 g/NnH with an initial flyash
injection rate of 10 g/min into a gas flowrate of 1.25 Nm^/min.
During the final stages of SMF testing, 5 wt% red iron oxide paint pig-
ment in the flyash was used as the test particulate. The iron oxide was added
to increase the percentage of fines, and cascade impactor measurements indica-
ted that the iron oxide increased the submicron fraction by approximately 80%.
377
-------
Sintered Metal Filter Efficiencies
The SMF filter efficiencies for both sets of filters were determined in
a total of 15 tests. Two of these tests in the pulsed blowback mode lasted
over 280 hours and yielded the highest estimates of particulate removal effi-
ciency (greater than 99.99999% for flyash removal on a mass basis). In each
long-term test, no mass change was observed on the filters. Since no mass
change could be observed on these filters, a mass of 0.1 mg collected was
assumed. The value of 0.1 mg was based on numerous experiments in the labora-
tory on filter handling procedures and repeated mass determinations. In other
tests with reverse flow blowback, some discoloration of the filter was
observed, but the mass gain was not measurable. This discoloration was deter-
mined to be due to the unfiltered blowback air. In the tests with iron oxide
plus flyash, no red discoloration of the sampling filters was observed.
The conditions of the two tests conducted for over 200 hours were a
superficial velocity of 1.16 m/min with a pulsed blowback initiation pressure
drop of 6.25 kPa. The gas was air at ambient temperature and humidity and
the temperature in the filter vessel was approximately 40°C. These condi-
tions are referred to as the base condition.
Thirteen additional SMF tests at other conditions were made with gas
particulate sampling times less than 200 hours. Since no mass gain on the
sampling filters was observed, with one exception, the 0.1 mg gain was
assumed. These tests yield efficiencies on the order of 99.999%. The one
exception, where a mass change was observed on the filter, was attributed to
the downstream piping being contaminated with flyash due to previous cascade
impactor measurements. Subsequent recleaning of the piping resulted in no
mass increase of the sampling filters. Pressure indications also showed that
no gross loading of particulate was occurring on the in-line filters.
It must be noted that a determination of the absolute particulate removal
efficiency of the SMFs was not possible with the hardware and techniques used.
Only bounded estimates could be obtained for the particulate source used.
Pressure Drop Characteristics
The porous SMF pressure drop immediately after blowback is called the
recovery pressure drop. For stable operation, this pressure drop must remain
well below the blowback initiation pressure drop. The recovery pressure drop
after every blowback in the tests described here was recorded. In the pulsed
blowback mode, the initiation pressure drop was 6.25 kPa. The recovery
pressure drop appears to stabilize at 3.75 to 4.25 kPa.
Over 100 blowback cycles were required to attain apparently stable opera-
tion. These test results do not indicate infinitely stable pressure drop
characteristics; however, the data for over 400 cycles on two different sets
of SMFs did not indicate an increasing trend in the recovery pressure drop.
Reverse flow blowback tests were conducted at the base condition.
Initially, the reverse flow superficial velocity was one-half the forward
velocity and resulted in a recovery pressure drop only slightly below the
378
-------
initiation pressure drop. Increasing the reverse flow superficial velocity
to 1.2 m/min resulted in a recovery pressure drop of 4.5 kPa for over
250 cycles.
Other tests conducted on pressure drop characterizations include varia-
tion of the blowback initiation pressure drop, increased superficial velocity,
particulate size distribution change, and temperature and humidity changes of
the process air stream. Changing the pulse blowback initiation pressure drop
had no observable effect on the recovery pressure drop at the base condition.
However, increasing the superficial velocity did result in an increase in the
recovery pressure drop. Returning to the base condition after increasing the
superficial velocity did not result in the prior recovery pressure drop. It
appears that the porosity and/or thickness of the permanent cake on the filter
are determined by the highest superficial velocity. Changing the particulate
size distribution with iron oxide, as previously noted, had no effect on the
recovery pressure drop compared to the base condition.
Probably the most significant tests with respect to pressure drop char-
acteristics were performed with conditions of high humidity and temperature.
By injecting steam into the air process stream, a gas composition containing
approximately 24 vol% water vapor was obtained. The process stream was heated
to approximately 90°C to maintain a dry system. These conditions resulted
in a rapid increase in the recovery pressure drop from 4.5 kPa at the base
condition to greater than 5.2 kPa. Temperature alone was the variable that
affected the recovery pressure drop. This was determined by allowing the
system to stabilize with steam injection, then stopping the steam injection,
and finally shutting down the process heater. There was no change in the
recovery pressure drop when the steam was stopped, but the recovery pressure
drop decreased to 3.9 kPa as the air stream cooled down after the process
heater was shut down. Repeated applications and removal of heat appeared to
indicate that the change in pressure drop is reversible. This pressure drop
change with respect to temperature appears to be due to a change in porosity
and viscosity, and has been quantified for a clean SMF.
CONCLUSIONS
Particulate Removal Efficiencies
The SMF particulate removal efficiencies reported here were obtained from
a specific experiment. The efficiency of the SMFs are absolute in the sense
that no measurable particulates were ever collected downstream from the filter
vessel. The estimated efficiencies are on the order of 99-99999%, based on
the minimum assumed mass gain of downstream collection filters. This estimate
is limited by the efficiency of the collection filters themselves, which were
HEPA filters. Other detection methods for particulate passing through the
SMFs could very well yield a different result, but the detection method used
in these experiments is similar to the intended application of HEPA filter
protection. The efficiencies reported here are only for the particulate size
distribution used and should not be interpreted or applied in any other sense.
379
-------
Pressure Drop Characteristics
The pressure drop across the SMFs appeared to stabilize after an initial
startup period. The pressure drop of interest is the recovery pressure drop
that occurs immediately after blowback. It is not possible to demonstrate a
nonplugging operation in an absolute sense, but only over the time frame of
an actual test. The pressure drop characteristics reported here are only for
the specific conditions and flyash tested. Over 4900 hours of operation with
over 5100 blowbacks were conducted on SMFs with no apparent plugging observed.
Other Applications of SMFs
A process design engineer is reluctant to use a unit operation that is
not well understood. In the case of SMFs, there are very few large applica-
tions of thousands of square feet of SMFs not only because of the cost, but
also because of the lack of design equations. On the other hand, there are
numerous applications involving hundreds of square feet of SMFs, where the
relative size involved presents a smaller risk to the user. At the present,
it appears that data and design information are emerging that will make rigid
barrier filters usable in such systems as incinerator flue gas filtration,
extreme temperature environments, and where extremely high particulate removal
efficiencies are required on a continuous use basis.
380
-------
ENDNOTES
1. Ralph M. Parsons Company, Conceptual Design Report: Design Description
for the Slagging Pyrolysis Incinerator Project, 5942-SPI-05, May 1980
(Prepared for U.S. Department of Energy, Idaho Operations Office).
2. Pall, D.B., Filtration of Fluid Catalyst Fines from Effluent Gases,
Ind. & E.G., 45, 6, p. 1197-1202, June 1953.
3. Mott Metallurgical Corporation, Farmington, Conn., Introduction to
Engineering Controlled Porosity Products, Catalog No. 1000.
4. Pall Trinity Micro Corporation, Cortland, New York, The Pall Porous
Metals Filter Guide, PSS-700a, September 1978.
5. Bjorklund, W.J., Development and Use of Sintered Metal Filters with
Fluidized Bed and Spray Calcination of Simulated High-Level Waste,
BNWL-2074, Battelle Pacific Northwest Laboratories, Richland,
Washington, July 1976.
6. Carls, E.L., and N.M. Levitz, Blowback of Sintered Metal Filters: A
Review of Tests and Operating Experience, ANL-7392, Argonne National
Laboratory, January 1968.
7. Kirstein, B.E., et al., R&D for an Off-Gas Treatment System for a
Slagging Pyrolysis Radioactive Waste Incinerator, Final Report for
Phase III, prepared for EG&G Idaho, Inc., Idaho Falls, Idaho, by
Science Applications, Inc., San Diego, California, June 1979.
8. Kirstein, B.E., et al., R&D for an Off-Gas Treatment System for a
Slagging Pyrolysis Radioactive Waste Incinerator, Final Report for
Phase IV, prepared for EG&G Idaho, Inc., Idaho Falls, Idaho, by
Science Applications, Inc., San Diego, California, February 1980.
381
-------
APPLICATION OF ELECTROSTATIC TECHNIQUES
TO THE REMOVAL OF
DUST AND FUME FROM THE INDUSTRIAL ENVIRONMENT
By: Stuart A. Hoenig, Professor
Department of Electrical Engineering
University of Arizona
Tucson, Arizona 85721
ABSTRACT
In earlier reports we discussed the application of charged fog to the
control of fugitive dust. Fogging units are now sold commercially but there
are many areas where it is impractical to use any water whatsoever.
For one application of this type we have designed and tested an electro-
static dust rejector that removes dust from the air being drawn into the tur-
ret of a military vehicle. Another system of this type is used to clean air
before it is drawn through the air filter of an internal combustion engine.
In both of these situations the dust is not "collected" but rather simply re-
jected back to the environment thereby simplifying the overall system.
Dust rejection can also be used to collect dust from air that is to be
passed from a source to a cyclone or baghouse. This reduces the load on the
regular control facilities while at the same time improving the overall sys-
tem operation by taking out the small (under 10 micrometer) particulates.
Other systems for control of water droplets have been developed and will
be discussed.
INTRODUCTION
In earlier publications [1,2] we reported on the development of charged
fog systems for control of fugitive dust. These units are now sold worldwide
y tne Kitu€ti Coirporfltion of Ardinoire. PPTTD*?vl iran•* a a-n/3 «-iun« • • * •
' ***x**uwtcj ••• cui.iay.LVciii.La, anc tnere is every indica—
tion that charged fog can be used to solve pollution problems in a number of
industrial areas. One study, under Environmental Protection Agency (?PA)
sponsorship, indicated that for a smelter application the cost reduction with
±^e t*0^*?™™^"^™ was * *«*or 0^ 28 rtuSTtiT
382
-------
SPINNING CUP FOG THROWERS
Conventional fog generators with nozzles are prone to clogging because of
dirt or dissolved salts that deposit when the units are turned off. In theo-
ry, filters and/or treated water can be used but we suggest that it is more
appropriate to design equipment that can withstand exposure to contaminated
water.
Spinning cup fog generators break up the water into fine drops by means
of a rotating element and an associated peripheral airflow that provides a
shearing effect [5]. Typical applications include throwing of limestone slur-
ries and drying of milk suggesting that clogging is not a problem. We have
developed a small 2 inch diameter spinning cup system with a blower to provide
the shearing airflow. A very satisfactory fog is produced; charging is accom-
plished by holding the cup at high (10 kV) voltage while isolating it from the
water line with a proprietary droplet generation system. In Figure 1, we show
a photograph of the unit in operation. Further development to produce a
"field useable" system is planned for the next calendar year.
PERMISSIBLE FOG GUNS
FOR EXPLOSIVE ENVIRONMENTS
At present the fog guns are not approved for use in flammable areas or
for underground operation in coal mining. There have been a number of tests
that indicated the foggers were very effective on coal dust, cotton linters,
and wood flour, suggesting that development of a permissible unit would be ap-
propriate. The problem here involves controlling the maximum energy release
in a way that precludes ignition of gas or dust/air mixtures. It is worth
noting that the problem involves not only controlling the amount of energy re-
leased, but the time or rate of release is important. A recent study [6] has
indicated that hot sparks of the type that might be produced by a momentary
short circuit in a fog gun are not nearly as effective for ignition as a long
term, lower temperature, heat source where a stuck roller on a conveyor belt
might be considered a typical example.
We have begun a study of the problems associated with meeting the "per-
missible standard" and hope to have an operational unit under test by the fall
of 1981.
ELECTROSTATIC DUST CONTROL TECHNOLOGY
There are many areas where it is not practical or possible to make use of
charged fog and we have been investigating the use of unconventional dust col-
lectors. One such system involves a sticky cloth collector material that can
be held at a positive voltage by a metallic backing material which is driven
by a small power supply. The cloth itself has a high resistance so that there
is no danger of shock on contact but the electrostatic field is large enough
to pull in dust in the vicinity of the cloth. The effect is enhanced if the
dust in the room is charged with a small negative ion generator. Figure 2
shows a system of this type in the off and on condition; in one case, the dust
simply "drifts" about while in the other, the dust is drawn in and held on the
cloth collector. We anticipate that devices of this type will be used in
383
-------
clean rooms or observatories where float dust is a severe problem.
We have developed a version of this device for use on an astronomical
telescope where falling dust can obscure the image from the primary mirror.
The system is shown schematically in Figure 3. In Figure 4, we show photo-
graphs taken on a model system some 24 inches in diameter. In one case, the
field was off and the dust simply floated about; in the other case, the unit
was "on" and the dust was drawn to the central collector. Calculations indi-
cate that systems of this type would be effective on mirrors with a diameter
as large as 2032 inches (8.0 meters) in an orbital telescope.
Another area of application for electrostatics in dust control relates to
the use of repulsion systems that can either "push" the dust, smoke, or fume
in a particular direction for collection or actually "hold back" the dust
while allowing clean air to move through the repulsion unit. The technology
is based on the system shown in Figure 5, where an array of needles is mounted
adjacent to a grounded mesh screen. When the needle array is driven at a high
(usually negative) voltage, with respect to the screen, the intense corona
discharge produces a significant electric wind (velocities as large as 200
m/min has been reported). At the same time there is a flow of oxygen 01 ions
that transfer their charge to any dust particles that they come in contact
with. These charged dust particles are then repelled by the electrostatic
field of the needle array. (It is important to appreciate that the electro-
static field from the needle array can penetrate the grounded screen thereby
providing a significant repulsion effect.)
In Figure 6, we show one of the dust repulsion systems in and out of ac-
tion; the smoke (ammonium chloride) was easily repelled when the system is
"on." It is worth noting that the design of these systems is not a simple
thing. Typical variables include needle to needle spacing, needle to grounded
screen spacing and screen opening. Figure 7 shows the results of one test
where only the needle to screen distance was changed; it is clear that this
distance is critical if best operation is to be achieved. We have found that
it is possible to optimize the electrostatic wind velocity or the current in
the space ahead of the screen but not both at the same time. This allows the
user to choose between a repulsion system that emphasizes the wind effects
versus one where the electrostatic effects dominate.
As an application of the pushing system, we show, in Figure 8, a diagram
of a system designed to remove the smoke and particulates from the chimney of
a home furnace burning wood or coal. (Reference 7 suggests that smoke from
wood combustion is a major source of air pollution in many parts of the United
States.) The moving belt system is part of the technology for removing the
collected material from the unit while at the same time allowing the chimney
to be cleaned by conventional sweeping techniques. (A proposal suggesting
further work in this area has been submitted to the Department of Energy.)
In Figures 9 and 10, we show the movement of smoke in a full scale model
of the dust-smoke collection system. In this case, the smoke was ammonium
chloride moving at 3 m/min and with the system on collection was quite effec-
tive. Similar results have been obtained with oil smoke but that material is
more difficult to photograph. In Figure 11, the same system is shown
384
-------
in action on participates (AC Fine) dropped through the system. Again, col-
lection was very effective and we see this as the prototype of a unit that
might be a significant step toward the improvement of urban environments even
with large scale combustion of fossil fuels for home heating.
Another version of the repulsion system has been developed for the turret
blower on the U.S. Army's M60-A1 Main Battle Tank. The blower is designed to
provide ventilation air at up to 1100 CFM and under normal conditions any dust
in the ambient is brought in with the air. Figure 12 is a schematic drawing
of a system designed to repel the dust before it can enter the blower while at
the same time providing the necessary ballistic characteristics to preclude
damage by light weapons in a combat environment. Figure 13 shows some of the
data with a system of this type; even at the high airflows significant repul-
sion of dust (AC Fine) occurred. The unit is even more effective at lower
airflows that might be used under nuclear, biological or chemical attack con-
ditions and some of these data are shown in Figure 14.
Another point of interest relates to the performance in "heavy dust" that
might be observed during desert operations. It is worth noting that here
"heavy dust" is a level far above that normally observed in industry. Typical
levels of military importance may range as high as 883 mg/cu m or even 8900
mg/cu (8.9 g/cu m). These conditions are far above 100% obscurity; you cannot
see your hand in front of your face. Typical results with a small scale re-
pulsion system are shown in Figure 15. It is clear that at dust levels that
might occur in industry there was very significant rejection. We see systems
of this type being applied in industry where it is important to keep dust from
escaping or entering a closed area while at the same time allowing the passage
of "clean air."
One very obvious application of this technology is in air cleaners for
internal combustion engines. Some modifications are necessary to permit ef-
fective operation in the confined area of an engine housing but in Figure 16,
we show a full scale unit designed for the air filter on a D-10 Caterpillar
Tractor. There was effective repulsion of the dust and some of the results
are shown in Figure 17. We see this as the first of a whole series of appli-
cations where the electrostatic technology is used to alleviate dust problems
without the complexities associated with conventional filters and collection
systems.
Another example of applied electrostatics exists in lead fume collection.
Lead fume is typically 0.1 micrometer in diameter and is therefore difficult
to collect with conventional systems. In many cases the fume source is a
smelting pot where crane movements preclude the use of conventional hoods. At
the same time the product (lead) is of some commercial value and it would be
advantageous to have a collection system that would allow the lead to be recy-
cled. In Figure 18, we show a schematic drawing of a lead collection unit in-
stalled above a smelting pot. The system is designed to swing out of the way
when the crane is in operation and to collect the lead fume on lead wool that
can be returned to the smelting pot after collection is complete. Figure 19
is a photograph of a prototype lead fume collector off and on. The lead fume
is drawn in by the lower electrostatic needle array that charges the fume and
forces it to deposit on the lead wool. We noted earlier that lead fume is
385
-------
difficult to charge and have therefore arranged for a second needle array to
charge and push any uncharged fume back toward the lead wool. If the two
needle arrays are properly arranged there will be a flow of "clean air"
through the system while the lead fume is collected on the lead wool. We see
this as the prototype of a system that might be used in a number of industrial
areas.
Large scale electrostatic wind arrays can be used to charge and push dust
from one area to another. One application of this type is shown in Figure 20,
where the objective was keeping dust out of an electrostatic precipitator in-
sulation box. At 30 kV and some 10 mA, the wind velocity, at a distance of
1.53 meters, was some 54.9 m/min.
One last application concerns the collection of diesel exhaust particu-
lates by a modified tube and wire electrostatic precipitator (ESP). ESP
units are well known to be effective on the fine (0.1 micrometer) carbon
flakes produced by a diesel engine but the problem has been removal of the
collected material and protection against insulator contamination by the con-
ductive carbon flakes. We have solved the cleaning problem by means of a ro-
tating liner and scraper-worm system that carries the collected material out
of the system. The insulators are protected by a small flow of clean air
that keeps the carbon from depositing. Some data taken with the cleaner on a
45 Hp engine supplied by the John Deere Company is shown in Figure 21. We an-
ticipate developing a cleaner-muffler unit that will combine both functions
into a single unit. This should allow the system to operate on a wide variety
of diesel powered systems.
CONCLUSIONS
We have discussed a number of applications for electrostatic technology
in the reduction of dust and fume. We see electrostatics as playing an impor-
tant part in low cost, low energy, pollution control systems.
ENDNOTES
1. Hoenig, S.A. Fugitive and Fine Particle Control Using Electrostatically
Charged Fog. EPA-600 7-79-078, March 1979. Available from NTIS, Spring-
field, Virginia 22161.
2. Hoenig, S.A. New Applications of Electrostatic Technology to Control of
Dust, Fumes, Smokes and Aerosols. IAS Conference Record. CH1575-0/80/
000-1032, September 1980. p. 1032-1037.
3. Daugherty, D.P., D.W. Coy. Assessment of the Use of Fugitive Emission
Control Devices. EPA-600/7-79-045, February 1979. Available from NTIS
Springfield, Virginia 22161.
4. Brookman, E.T. Demonstration of the Use of Charged Fog in Controlline
Fugitive Dust from Large-Scale Industrial Sources. (Presented at the
1980 Symposium on Iron and Steel Abatement Technology, Philadelphia
Pennsylvania, November 18-19, 1980.) '
386
-------
5. Hinze, J.O., H. Milborn. Atomization of Liquids by Means of a Rotating
Cup. Jnl. Appl. Mech. 17:145, 1950.
6. Eckhoff, R.K. Towards Absolute Minimum Ignition Energies for Dust Clouds ,
Comb, and Flame. 24: 53-64, 1975.
7. Cooper, J.A. Environmental Impact of Residential Wood Combustion Emis-
sions and Its Implications. Jnl. Air Poll. Cont. Assoc. 30:855,
August 1980.
Figure 3.
Figure 1. Spinning Cup Fog Thrower Figure 2. Electrostatic Dust Collec-
In Action. tor "Off" and "On."
9»nc4». **r*
( -rr
Til ^n
IUMCK
«Uff«ttMC
:•'
•A 1 «»»
IP
™.ir »—
ll
fl
Schematic Diagram Figure 4,
Electrostatic Dust Collec-
tor for 24" Astronomical
Telescope
Electrostatic Dust Collec-
tor "Off" and "On."
387
-------
Figure 5. Schematic Diagram of
Electrostatic Fence to
Reject Dust and Admit
Air.
Figure 6. Electrostatic Dust
Repulsion System "Off"
and "On".
APPIIEO
VOLTAGE
• V
25
2O
13
10
3
«OOC£OJ«0*rlVE
WIND (KM 1
a%
no
90
60
30
0
ar™
'OK
O.20
O.It
O 10
003
OOO
i
I
ft
n
J\
't
, '
ION CUNKNT
* WQUWXD SCMEM
\H*H WLTACC • • WMO SMCW*
\NCtBLi *«B«r
\
X
, .Tt
X
tH <
k
MEASUREMENT
^- POINT ON
•^ CENTERLINE
V~»
o a so
ALL DIMENSIONS IN MILLIMETERS
S 'J>
\ fl
mn
Figure 7. Effect of "d" Needle to
Screen Distance on
Electrostatic Wind
Parameters.
Figure 8. Schematic Drawing Cleaning
System for Wood Burning
Stoves.
388
-------
Figure 9. Electrostatic Smoke
Collector "Off."
Figure 10. Electrostatic Smoke
Collector "On."
Figure 11. Electrostatic Dust
Collector "On."
Figure 12. Schematic Drawing Inverted
Dust Rejection System for
M60-A1 Tank.
389
-------
OUST OCNSiTY
PAHTICtE OUMCTEH (MICROMEUflSI
Figure 13. Experimental Results
Electrostatic Dust
Rejection System.
OUST im.li WITH
ft PL*C£ '" \\ \ \
/ eUCTROSTAriC V\ V ^-^
' srsTEMorr NX N. \
MO OUST *l JCCTKW SYSTEM
CFM
'fffik. X'^^,!^C'"
fei:^ ^^>f-"
*' \ REDUCTION I uvjcfM ^*^»
,oocV5>vH tiSff" TOC'-
sazx'' \_ r M»».
nUXJCTICW ^.J I nOUCTIOK
PARTICLE OIAMETCR (MICROMETERS)
Figure 14. Experimental Results
Electrostatic Dust
Repulsion System.
SXTtBN41. CAN
, TO STIMULATS
TRACTOR OPMATION
I I I I I | AIRANOPUST
T T T T T f 31i (Kfl/oim
1 1-! i *;
'^iL\y.
171/i'g
•is'o
ir
SHOINE
AIR FILTER
fOR 0-10
CATERPILLAR
TRACTOR
IS'
\
OUST
OUTLET
HISM VOLTAbe
, HEEOtH ARRAT
\*
'0 SUCTION FAN
HIELD
Figure 15. Laboratory Data Figure 16. Schematic Drawing
Electrostatic Dust Repul- Electrostatic Test System
sion With the Bypass for Engine Air Filter.
Flow System.
390
-------
OUST REJECTION EFFICIENCY
100 r IN PER CENT
£L£AN HIR
(AT 00% f FFicieHCf Tire FIUER tire is CXTENOED »Y A
FACTOR OF FIV
1 T T T T T T
UP*tR NEEDLE AflSAT
LE40 WOOL
SUP'OHT
TTTTTTT]/ ^INOWVOT
Ufftfl
A
FR.TCR SURFACE AREA 3 SO F I. FLOW RATES
ISO.*50.900CF«
too IM too
FILTER F»CE V€LOCITT
Figure 17. Laboratory Results Tests Figure 18.
of Electrostatic System
for Dust Repulsion on
Engine Air Filter.
Schematic Drawing
Electrostatic Lead Fume
Collection System.
Figure 19. Electrostatic Lead Fume Figure 20. Large Scale Dust Pusher,
Collector "Off" and "On."
391
-------
200 rPM ( «O.S m/mln ) , CORONA VOUTAOS - 2O.OOO V.
2 J43S78S 10 :i
PARTICLE OIAM6TIR I HHCPOMETEaS i
Figure 21. Laboratory Experiments
Control of Diesel Exhaust
Participates by Means of
Electrostatic Techniques
392
-------
THE DRY VENTURI
BY: A.J. Teller, D.R.J. Roy
Teller Environmental Systems, Inc.
415 Boston Turnpike
Shrewsbury, Massachusetts 01545
ABSTRACT
The major problem in particulate emission recovery is the reliability
of collection of submicron particles. The use of electrostatic precipi-
tators has often resulted in time degradation of performance and that of
baghouses in blinding or low bag life. A dust agglomeration and capture
process (dry venturi) with a pressure drop of 0.2 KPa, providing removal
of the submicron particulates in the duct prior to the final collector
results in the increased effectiveness and reliability in the operation
of the final collector. The system has been successfully operated in the
fiberglass, municipal incineration, combustion, secondary aluminum, and
fertilizer industries.
INTRODUCTION
A process has been developed to effect reliable capture of submicron
particulates with low energy consumption. The mechanism of capture is
inertial impact in a "dry environment -I:
Inertial impact, an established method for collection of particu-
lates suspended in a gas stream, is normally conducted with the capture
targets as fixed members placed in the gas stream or liquid droplets
dispersed in the gas stream. The efficiency of capture has been modeled
by Ranz and Wong (1) and Langmuir and Blodgett (2).
The models both indicate that the efficiency of particle capture
is a function of equivalent particle size, target size and shape, and
relative velocity of the particle and target for a given fluid environ-
ment.
Simplistically, therefore for a tgiven particle to be captured,
Efficiency of capture = f (Vg Dpj = f (Ns)
The conventional inertial impact capture device, the wet venturi,
creates its own targets. The gas is accelerated to a high velocity and
impacted on liquid. The velocity energy is transferred to shearing action
thus creating the targets. The target droplets, formed at zero velocity,
have a relative velocity to the gas stream, equal to the gas velocity.
The drop size of the target is approximately inversely proportional to the
gas velocity [Nukiyama-Tanasawa] (3). Thus to achieve high efficiency,
significant energy must be expended to create the small droplet target.
393
-------
Typical anticipated behavior of a wet venturi based on established
relationships is as follows (provided an adequate number of targets is
available).
TABLE I
WET VENTURI CHARACTERISTICS
Eff.
%
Particle
Size (y)
Capture
1
0.5
0.3
1
0.5
0.3
Gas Vel
Venturi
m/s
52
52
52
91
91
Q1
Target
Droplet
Size (u)
140
140
140
75
75
75
AP Venturi
mm HaO
280
280
280
635
635
635
Power
Consumption
Gas
KW/1000 CMH
1.10
1.10
1.10
2.50
2.50
2.50
Ns
0.37
0. 18
0.11
1.21
0.61
0.36
97
90
80
99
96
92
The excessive power consumption prohibits the use of the venturi for
removal of submicron particulate when handling large volumes of gas.
The new system, dry venturi (fig. 1) , circumvents the limitation of
high energy consumption. The target material is supplied to the system as
a dry powder, in a crystalline form. Thus, the gas velocity can be re-
duced significantly.
For the same anticipated efficiency as a 635 mm H20 AP venturi, the
relative velocity in a dry venturi need be only 18 m/s, normal duct con-
veying velocity, if a 15y target is used.
Thus, the use of small solid targets, in the 5-25vi range permits the
achievement of high capture at duct velocities and with an energy consump-
tion approximately 4% of that required in a high energy liquid venturi,
Although dry powdered solids have in the past been introduced into
a duct (for various reasons), there has been little or no particulate
capture reported as a result. The reason is that solids have been intro-
duced into the gas stream in the conveying duct as a dense agglomerate.
By the time the powder was dispersed, a condition essential for capture,
the powder had attained the velocity of the gas since the dispersion
occurred due to the conveying capacity of the gas stream.
The dry venturi is designed to introduce the target material in a
dispersed state, at the maximum desired relative velocity to the gas
stream at minimal power consumption. The dry venturi permits the optimi-
zation of capture efficiency by selecting the target size, the target
population density, and velocities for a specific application. Typical
of presently operating systems the separation number function, Ns, obtained
for a 0.5 micron capture with a 20 micron target is 1.2.
394
-------
FIGURE
DRY VENTURI
co
vo
on
CONTACT
ZONE
t
CAPTURE
TARGETS
-------
The power required to effect this capture is 0.08 KW per 1000 cubic
meter per hour of gas treated, only 3-5% of that required in a convention-
al wet venturi. It is noted, however, that the power input for the capture
of the submicron particulate must be combined with that required to make
the final separation of captured particulates from the gas stream. Experi-
ence with commercial systems has shown that collection equipment in con-
junction with the dry venturi have demonstrated power consumption in the
order of 0.35 to 0.5 KW per 1000 cubic meter per hour of gas treated.
The dry venturi is operational in six industries and has been
piloted in the US, UK, and Japan. The final collectors have been the
electrostatic precipitator and the baghouse.
The effectiveness of the combined utilization of the dry venturi
and a baghouse final collector is unique.
Primarily, Ltpermits application of the baghouse to collection of
deliquescent, cohesive, adhesive submicron, and combustible particulates
where its use was previously restricted. It functions well with submicron
particulates and hydrophilic or cohesive particulates without blinding or
high pressure drop. It functions safely with combusting feed particulate.
But more significantly, the capture of the submicron particulates
in the duct by crystalline targets results in exceedingly low pressure
drop buildup in the baghouse. As a result, the shake cycle has been ex-
tended from the normally anticipated 5 min. to 30 min. to 4 hrs. to 36
hrs. Two characteristics are affected.
Bag life is extended because of the decrease in shaking requirement,
and the lack of permeation of submicron particulates into the interstices
in the cloth.
Lower and more constant pressure drop is obtained when the baghouse
is operated in conjunction with the dry venturi.
Residence time of the cake on the cloth is increased by a factor of
approximately 30. As a result, where the system provides the dual service
of particulate and acid gas collection, residual unreacted reagent is
accumulated and in residence on the bag surface. Thus "polishing" of acid
gases occurs in the long residence reaction time on the bag surface. The
advantage of this phenomenon becomes more significant when the acid gas
content is a variable.
Particulate emissions from the baghouse are lower than obtained in
conventional filtration.
No corrosion is evident after 5 years in acid gas service.
. _ Jhe performance of the dry venturi - baghouse system is indicated
in iabie II.
396
-------
TABLE II
PERFORMANCE CHARACTERISTICS OF THE
DRY VENTURI - BAGHOUSE SYSTEM
INDUSTRIAL
APPLICATION
Fiberglass
Glass
Municipal
Incinerator
Coke
Calciner
Secondary
Aluminum
AP MM
DRY VENTURI
25
30
20
25
20
H20
BAGHOUSE
80
100
90
125
150
SHAKE
CYCLE
HRS
24
16
8
8
INT
~48
PART
EMISSION
GR/DSCF
0.002-0.004
0.004-0.008
0.003-0.006
0.004-0.006
0.001-0.003
ACID GASES
COLLECTED
HF, SOX
SOX
HC£,HF,SOX
sox
HC£, HF
The dry venturi has operated commercially as an effective particu-
late collector with an energy requirement of less than 5% of an equivalent
performance wet venturi.
Because of removal of submicron particulates prior to the final
collector, the effectiveness of the final collector is enhanced.
Specifically, when used in conjunction with a baghouse, it permits
lower particulate emissions, decreases baghouse pressure drop, increases
the cycle interlude, extends bag life, and extends the reliable appli-
cation of the baghouse to recovery of materials previously considered
incompatible with baghouse collection.
ENDNOTES
(1) Ranz & Wong IEC, 44, 1371-1381, (1952)
(2) Langmuir & Blodgett U.S. Army Tech. Report 5418, (1946)
(3) Nuikyama & Tanasawa, Trans. Soc. Mech. Eng. (Japan)
5, 18, 68-75, (1939)
397
-------
FIBER BED FILTER SYSTEM CONTROL OF WELDING PARTICULATES
By: Judith Ann Bamberger and W. Kevin Winegardner
Battelle Memorial Institute
The Pacific Northwest Laboratories
ABSTRACT
The Electrostatic Fiber Bed Filter (EFBF), patented and developed at the
Pacific Northwest Laboratories of Battelle Memorial Institute, is a highly
efficient method for removal of submicron particles from gas streams. This
paper describes the EFBF system and the results of a parametric study to eval-
uate the EFBF's ability to control the fine particulates generated during
welding operations. The results of the parametric analysis show the depend-
ence of collection efficiency on electric field, face velocity, fiber bed
depth, and welding fume concentration.
System collection efficiencies above 95% were routinely measured at
velocities from 200 to 400 ft/min. For welding fume concentrations of 1 to
2 mg/m3 collection efficiencies were consistently 99.9%. Pressure drop
across the fiber bed was extremely low, ranging from 0.07 to 0.82 in. I^O.
INTRODUCTION
This report describes the investigation of the use of the Electrostatic
Fiber Bed Filter (EFBF) to reduce ambient concentrations of welding fume fine
particulates. Literature review suggests that exposure to the gases and par-
ticulates generated during welding operations is not a health hazard provided
adequate ventilation is maintained (1). However, welding fumes contain fine
or submicron particulates that can be inhaled deep into the lung. A growing
awareness of potential health effects is focusing attention on the need to
assure control of fume constituents below specified limits.
The EFBF is a patented, highly efficient system for removing fine,
highly resistive particles from gas streams (2). In addition to their small
size, welding fume particulates are electrically resistive. These two char-
acteristics make the EFBF an especially attractive particulate control device.
The parametric investigations discussed in this paper are part of a license
agreement with The Bahnson Company for application of the EFBF in the field of
building air cleaning.
THE ELECTROSTATIC FIBER BED FILTER
Basic components of the EFBF include a corona chamber and a highly
porous bed of dielectric fibers. Particles are first given an electrical
charge in the corona chamber and then collected in the downstream fiber bed.
Corona Chamber
A primary requirement of the particle removal process is the generation
of large quantities of gas ions for charging the particles of the aerosol
Ion generation is accomplished in a stable, self-maintaining discharge
398
-------
between two electrodes, termed a corona discharge. High voltage is applied
to one of the electrodes, a wire or other geometry with sharp radius. The
other electrode, usually adjacent parallel plates or a concentric cylinder,
is grounded. Such an electrode arrangement results in a highly non-uniform
electric field and ionization near the wire. The particles acquire an elec-
tric charge by passing through the dense, unipolar cloud of ions that fills
most of the space between the electrodes.
Fiber Bed
Highly efficient particulate removal in the fiber bed is attributed to
diversion of the charged particles from normal flow paths by the force of an
electric field. The field is believed to be the result of volumetric charge
buildup within the bed produced by particle deposition on the dielectric
fibers. Collection enhancement by the field is almost instantaneous. The
fact that particle collection is enhanced by a bed charged with the same sign
as incident particles is perhaps initially surprising. However, development
of an analytical model has revealed that the electric field resulting from
the deposited charges is significantly larger near a fiber than local Coulomb
repulsion and dramatically increases the capture cross section.
The fibers occupy only 5% to 8% of the total bed volume. The high fiber
bed void fraction and the inherently open corona chamber geometry result in
low pressure drop and energy requirements for routing an aerosol through the
system. The bed can be cleaned by liquid sprays, air jets or other methods,
depending upon the specific application.
EXPERIMENTAL APPARATUS
The modular EFBF bench-scale system used in the parametric study is
shown in Figure 1. A 3-ft-long diffuser section was provided upstream of the
approximately 10-ft-long, 10-in. x 10-in. main lucite duct. A wire-plate
electrode arrangement was used in the corona chamber. Plate-to-plate and
wire-to-plate spacings of 2 in. and 1 in., respectively, were used. One
FIGURE 1. SCHEMETIC DRAWING OF EFBF EXPERIMENTAL APPARATUS
ENTRANCE
DIFFUSER
PRECORONA
SAMPLING
SECTION
CORONA
CHAMBER
PREBED
SAMPLING
SECTION
FIBER
BED
POSTBED
SAMPLING
SECTION
BLOWER
399
-------
10-mil-diameter tungsten wire discharge electrode was centered between each
pair of the 1.5-in.-deep grounded aluminum plates.
The fiber bed, installed downstream from the corona chamber, consisted
of layers of uniformly packed polypropylene knit material. Particle leakage
around the bed edges was minimized by providing a small diversion baffle
(normal to the direction of flow) around the interior perimeter of the bed
holder. Sampling locations were provided upstream of the corona chamber and
fiber bed, and downstream of the fiber bed, permitting gravimetric determina-
tion of total system, fiber bed, and corona chamber particulate collection
efficiencies.
EXPERIMENTAL PARAMETERS/TECHNIQUES
The range of parameters used in the investigation is summarized in
Table 1. The fiber bed void fraction, the fraction of the bed not occupied
by fibers, was calculated from the weight of fiber per unit length, fiber
diameter, and the weight and volume of the bed. Fumes were generated by
shielded metal-arc welding.
Air Velocity
Welding Fume
Electrode
Concentration
Corona Chamber
Particle Residence Time
Applied Voltage
Fiber Bed
Material
Fiber Diameter
Bed Depth
Void Fraction
TABLE 1. SYSTEM TEST PARAMETERS
200, 300, 400 ft/min
(0.51, 0.76, 1.0 m/s)
E7024
1-35 tng/m3 (0.0004-0.015 gr/ft3)
0.019-0.038 s
0, -15, -18, -21 kV
Polypropylene
0.008, 0.012 in. (0.020, 0.030 cm)
1, 3 in. (2.5, 7.6 cm)
0.93
Total system, bed, and corona chamber particulate collection efficien-
cies were determined gravitnetrically from samples collected isokinetically
on fiber glass filters. As previously indicated, three sampling locations
were used. Further, the square, 10-in. x 10-in. duct was divided into quad-
rants for sampling. Samples were taken from each quadrant at each of the
three locations. Aerosol concentrations were determined from the quantities
collected on the filters, sampling times, and volumetric sampling flow rates.
Evaluation of a selected parameter, termed a test, consisted of a series
of efficiency measurements. Insofar as possible, variables were held con-
stant throughout the test. No attempt was made to clean the system until the
400
-------
test was complete. Results from a typical test, consisting of 11 efficiency
measurements, are shown in Figure 2.
FIGURE 2. COMPARISON OF SYSTEM, FIBER BED AND
CORONA CHAMBER COLLECTION EFFICIENCY
(200 ft/min velocity, 35 mg/m3 average
concentration, 3 in. thick fiber bed)
100
90
CJ>
i 80
o
o
;= 70
o
o
60
50
./•
A SYSTEM
• FIBER BED
• CORONA CHAMBER
2 4 6 8 10
MASS SAMPLE MEASUREMENT NUMBER
12
TEST RESULTS
Effect of Particle Charging
To confirm that the EFBF system is highly efficient only when the parti-
cles are electrically charged, efficiency measurements were made with and
without corona chamber applied voltage. Results in Figure 3 show that system
collection efficiencies of 88 to 99% were obtained with corona enhancement.
With particle collection only by the conventional mechanisms of impaction,
interception, and diffusion, system collection efficiencies dropped to
18 to 35%.
Concentration
Welding fume concentration ranged from 1 to 35 mg/m3 during the tests
(Table 1). The effect of concentration on system performance as a function
401
-------
FIGURE 3. EFFECT OF PARTICLE CHARGING ON EFBF SYSTEM,
FIBER BED AND CORONA CHAMBER PERFORMANCE
(300 ft/min velocity, 12 mg/m3 average
concentration, 3 in. thick fiber bed)
99.9
99 -
90
50
10
SYSTEM
COLLECTION EFFICIENCY
FIBER BED
COLLECTION EFFICIENCY
CORONA CHAMBER
COLLECTION EFFICIENCY
A WITH CORONA CHAMBER
APPLIED VOLTAGE
• WITHOUT CORONA CHAMBER
APPLIED VOLTAGE
FIGURE 4. THE EFFECT OF CONCENTRATION ON SYSTEM PERFORMANCE
(3 in. thick fiber bed, 0.008 in. fiber diameter)
>-
o
o
l_Lj
O
I—
o
o
o
2345
MASS SAMPLE MEASUREMENT NUMBER
402
-------
of velocity is shown in Figure 4. At low concentration (1-2 mg/m3) velocity
did not affect system performance (99.9%). At medium concentration
(5-10 mg/m3) system performance at 200 ft/min remained 99.9%, but dropped to
96% at 400 ft/min. At high concentration (15-25 mg/m3) system performance
varied from 96 to 98% at both velocities.
Bed Thickness
System performance of 1-in.-thick and 3-in.-thick fiber beds, both
having the same void fraction and fiber diameter, was investigated as a
function of velocity. Results presented in Figure 5 show that the average
collection efficiency of the 3-in. bed was 6% higher than the average
collection efficiency of the 1-in. bed.
FIGURE 5. EFFECT OF BED THICKNESS ON SYSTEM COLLECTION EFFICIENCY
99.9
99.5
98
t± 95
UJ '•"
z
2 90
o
o
80
60
A 3 INCH THICK BED
I . I
1 INCH THICK BED
I i I
200 400 200
VELOCITY, ft/min.
400
Fiber Diameter
Fiber beds of 0.008- and 0.012-in.-diameter polypropylene knit, with
identical void fraction and bed thickness, were tested at 200, 300, and
400 ft/min. Data presented in Figure 6 show the average performance of the
0.008-in.-diameter fiber to be 15% higher than the performance of the
0.012-in.-diameter fiber.
403
-------
FIGURE 6. EFFECT OF FIBER DIAMETER ON EFBF COLLECTION EFFICIENCY
99.9
99.5
>-"
^ 98
In )
^ 95
U_
LL.
2 90
o
S 80
i
o
60
/tn
A
-
L
i
\.
i
L
—
- A 0.008 in. DIAMETER
— —
: '
ill
•
_L
• 0.012 in. DIAMETER
I
200 400 200
VELOCITY, ft/min.
400
Pressure Drop
Pressure drop data for each fiber bed as a function of velocity are pre-
sented in Figure 7. The clean bed pressure drops ranged from 0.068 in. H20
for the 1-in.-thick bed at 200 ft/min to 0.540 in. H20 for the 3-in.-thick
bed, 0.008-in.-fiber diameter, at 400 ft/min. During operation, pressure
drop increases of 0.010 and 0.360 in. %() were measured for the 1-in. and
3 in. beds, respectively. Clean bed pressure drop as a function of velocity
appears to be linear up to 300 ft/min bed face velocity.
The pressure drop across the 1-in. bed was one-third that of the 3-in.
bed with the same diameter fiber. Pressure drop across the 3-in. bed with
0.008-in.-diameter fiber was consistently greater than that across the 3-in.
bed with 0.012-in.-diameter fiber.
404
-------
FIGURE 7. COMPARISON OF PRESSURE DROP ACROSS CLEAN BED AS A
FUNCTION OF BED FACE VELOCITY FOR THREE FIBER BEDS
0.600
o
CM
ca
z
<
LU
O
CO
GO
o
oi
o
D_
O
C£
O
to
LU
Qi
D-
0.500-
0.400
0.300 -
0.200 -
0.100 -
0.008 in. DIAMETER FIBER
A lin. THICK BED
• 3 in. THICK BED
0.012 in. DIAMETER FIBER
• 3 in. THICK BED
200 300 400
BED FACE VELOCITY, ft/min.
SUMMARY
Electrostatic Fiber Bed Filter System collection efficiencies of 95%
or greater were routinely obtained over the velocity range of this study,
200 to 400 ft/min. For welding fume concentrations of 1 to 2 mg/m3 collec-
tion efficiencies were consistently 99.9%. System pressure drop was
extremely low, ranging from 0.07 to 0.82 in. H20.
ENDNOTES
1. The Welding Environment. American Welding Society, Miami, 1973.
2. Reid, Donald L. Electrostatic Capture of Fine Particles in Fiber Beds.
In: Novel Concepts, Methods and Advanced Technology in Particulate-Gas
Separation, Ariman, Teoman (ed.). Univ. of Notre Dame, 1978. p. 305-316,
405
-------
OF GLASS CAPILLARY FILTERS TO CLASSIFY ACTINOLITE FIBERS
By:
: J. W. Gentry
Institute for
University of
College Park,
T. C. Chen, S
Department of
University of
College Park,
Physical
Maryland
Maryland
. W. Lin,
Chemical
Maryland
Maryland
Science &
207^2
P . Y . Yu
Technology
Engineering
207^2
ABSTRACT
Described in this paper are recent results in the development
of a GCAF (glass capillary array filter) inertial impactor. This
instrument was designed to separate mixtures of fibers and iso-
metric particles and to classify fibers according to their aero-
dynamic diameters.
Experimental evidence, suggesting that the more abundant
particles are removed by the GCAF while fibers penetrate, is
discussed. However, the principal thrust of the paper is the
description of developments in simulating instrument performance.
The simulations were designed to resolve two questions: l) why
do the GCAF stages preferentially remove isometric particles of
relatively small aerodynamic diameters; and 2) how can penetra-
tion measurements be used to obtain unambiguous estimates of
fiber length and diameter?
Because an exact simulation of the collection of particles
by the GCAF would require the solution of the non-linear Navier-
Stokes equations in three dimensions, analytical solutions are
not possible. However, analytical solutions are possible for
related problems which illustrate the collection mechanisms and
elucidate our experiments. These solutions are discussed.
A computer code showed that fiber length as well as diameter
could be characterized. The ideas developed here demonstrate
that the physical dimensions of the fibers could be determined
from penetration measurements if the fibers have a preferential
alignment, indicate the sensitivity of such measurements, and
indicate qualitatively the effect of polydi sper sity on the
measurements .
TEXT
The aerodynamic separation or classification of fibers poses
a number of interesting and, presently, unresolved problems In
a recent paper Spumy (l) suggested that an aerosol of asbestos
or actinolite fibers with a narrow size range could be prepared
406
-------
by first grinding a mineral sample and then producing an aerosol
with a vibrating bed generator. The fibers are then collected
and size classified by selective sedimentation in aqueous solu-
tions. This process is quite tedious requiring several months to
prepare a sample. A method based on either aerodynamic and/or
electrostatic classification of a fibrous aerosol presents the
possibility of rapid size classification.
In measuring fibers either in remote rural areas or in the
work place, a major difficulty is separating the more abundant
isometric particles from the fibers. Typically, the fiber to iso-
metric particle ratio is 1:3000 (2). Recently, we have developed
a GCAF (glass capillary array filter) inertial impactor which
shows promise both in preferentially removing isometric particles
and in classifying fibers according to their size.
The GCAF inertial impactor has been described previously
(3-5)- The principal properties of the instrument are as follows:
1. The impactor consists of six two-section stages.
2. The first section of each stage is a glass capillary
array filter (Galileo Electro-Optics Corporation) which hope-
fully will remove the isometric particles while allowing the
fibers to pass through.
3. The second section of each stage is a Mercer type
impactor (6).
k. The GCAF consist of uniform, equally spaced pores with
diameters of 10, 25, or 50 Urn. Their porosity is 50%, and their
pore length is 0.055 cm.
In this paper, results are presented which suggest that the
fibers preferentially penetrate the GCAF while the isometric
particles collect in the interstitial areas between pores. How-
ever, the bulk of this paper deals with the simulation of the
GCAF. Specifically, we consider the questions: l) why are iso-
metric particles collected in the interstitial areas between
pores; and 2) if the GCAF are mounted horizontally (the axial
flow is perpendicular to the gravitational field), can one deter-
mine fiber lengths as well as diameters?
Although in this paper there is insufficient space to des-
cribe the experimental measurements in detail, it is important to
indicate the key results. These are as follows:
1. Several investigators (7, 8) have shown that when poly-
styrene latex aerosols deposit on nuclepore filters, the deposi-
tion on the narrow bridges between pores is significantly greater
than around isolated pores.
407
-------
2. In our laboratory, the fractional penetrations of poly-
styrene latex aerosols as a function of flow rate through the
GGAF were measured (9). Typical results are shown in Figure 1.
The efficiency is plotted as a function of the Stokes number
(based on the jet diameter of the Mercer impactor) with the solid
and open symbols representing the absence and presence of the
impaction plate.
Several points are significant about these results. First,
there is little difference with or without the impaction plate
indicating that collection is on the GCAF. Secondly, the
efficiency is near its maximum value of 50% (the GCAF porosity).
In Figure 1, these experimental results were compared with the
Pich model for collection on capillary filters. Agreement is
good for the lower velocities. As the velocity (i.e. Stokes
number) increases, the collection efficiency decreases. This
result is typical of what would be expected if there were signif-
icant "bounce" and reentrainment. What is surprising is that
this apparent reentrainment occurs at velocities much lower than
predicted by theory (10).
In addition, there is qualitative evidence pointing to a
different mechanism than that suggested by Pich. From this
theory, one would expect that the particles would be collected at
the rims of the pores; whereas from scanning electron micrographs
the heaviest particle depositions were in the center of the
narrow regions between pores.
3. In our laboratories, the GCAF have been used to remove
agglomerates from test aerosols of actinolite. SEM micrographs
of crocidolite fibers passing through a 25 ym GCAF are shown in
Figure 2. What is significant from these micrographs is that the
fibers penetrate but that mo.st of the small isometric fragments
of fibers were removed. These results are especially significant
in view of recent theoretical and model simulations of fibers
penetrating through orifices (11).
Based on these experimental results, a prima facie case
exists for believing that the GCAF could be used to increase the
fraction of fibers.
The principal objective of the simulations was to determine
whether a filter with irregularly spaced pores would have the
same collection efficiency as a filter with regularly spaced
pores and the same porosity. As mentioned above, experiments
with nuclepore filters and with the GCAF showed a heavier deposit
in the interstitial areas. If the GCAF impactor is to operate
effectively, there must be significant collection of isometric
particles in the interstitial areas. Furthermore, there must be
relatively sharp velocity gradients in the vicinity of the pores
in order to impose a preferential alignment of the fibers It
should be stressed that the goal of the simulation is to deter-
408
-------
mine how the fluid velocity and its gradients are altered as the
distance between pores decreases. This differs significantly
from the objective of previous investigators (12-lU) who, in
their simulation of collection by nuclepore filters, have focused
on the collection efficiency of a centered pore within equally
spaced c ells .
In general, this simulation, because of the lack of angular
symmetry, would require a solution of the lavier-Stokes equation
including the inertial terms in three dimensions. Moreover, an
analytical solution is precluded because the Navier-St okes equa-
tion is non-linear, and one needs to reduce the dimensionality
to two in order to define a stream function i|>. In previous
numerical simulations, the dimensionality was two (the coordin-
ates were the radial distance from the center of the pore R and
the down stream distance Z) with the velocity field determined
from a finite difference algorithm.
Our approach was as follows:
1. The velocity field was determined analytically for two
related problems: the flow through a filter consisting of
identical cells, each cell containing two parallel slits; and
the flow through a filter consisting of identical cells, each
cell having a centered central pore surrounded by an annulus .
2. The filter was divided into equally spaced cells
assuming that laminar flow was established within the slit, pore
or annulus. By shrinking the distance between the slits (or the
pore and annulus), the closing of two pores together could be
approximated .
3. Since there is angular symmetry, a stream function can
be defined. Neglecting the inertial terms and eliminating the
pressure from the Navier-Stokes equations results in
where u is the vorticity and fy is the stream function.
h. For the "slit" model, the solution of the stream func-
tion is a Fourier series; while for the "pore and annulus" model,
the solution is an infinite series of Bessel functions. Manton
(15) has applied this method to simulate a "cell" model with a
single centered pore.
5. The three dimensional Navi er-St okes equation will be
solved with a self-adjusting finite element method (l6) with the
analytical solutions providing a check on the accuracy of the
numerical solution.
409
-------
6. Finally, the particle trajectories are determined by
numerical solutions for the particle trajectories (V)
dt
_ v)
^
where U is the dimensionles s fluid velocity, F is external forces
acting on the particle, and Stk is the^Stokes number. In the
simulations below, the fluid velocity U was determined from the
analytical solution for the slit model.
As pointed out by Manton, neglecting the inertial terms is
reasonable for nuclepore filters where the Reynolds number
(Re)<0.005. For the GCAF filters used in this study, Re~0. 1-1.0,
and the physical basis for neglecting the inertial terms is less
sound. Therefore, simulations for the analytical flow field can
best be regarded as asymptotic approximations. A "cell" with
two identical pores is simulated by requiring the flow through
the pore and annulus to be equal.
An important feature of these simulations is that very
accurate estimates of the velocity are required. If the velocity
field is to be used to determine the trajectory of fibers, not
only the velocity but the gradients and vorticity must be
specified. It is for this reason that a finite element rather
than a finite difference algorithm was deemed more useful.
For these simulations, the porosity was 0.10. The "cell"
was characterized by two dimensions: a coordinate perpendicular
to the slits and parallel to the face of the filter (Y), and a
coordinate perpendicular to the face of the filter (Z). The
slits were symmetrical around the centerline of the cell (Y=0)
with midpoints occurring at Y=+YO ) . The boundaries of the cell
were located at Y=+l and were characterized by VyEO and
3VZ/SYEO. The slits were assumed to be of sufficient length that
end effects could be neglected.
With these assumptions, the stream function could be
expressed by:
00
llj
+ l a-, « (1 + A,Z) e J sin A.Y with
J=l J J J
12
, ^f-u .^3 t2 (sin ^,a - sin A.b)
+ A (b-a) (cos A.a + cos A.b)] t^\
From the stream function, the velocity components and vorticity
410
-------
can be expressed by:
- t
(«)
31 9Z
For a porosity of 0.1, the width of the slit is 0.1 with its mid-
point at Y0 . Even though the net flow through the slits is not
altered by the location Yo , the collection efficiency changes
significantly as Yo->0.05. In Figure 3, the collection efficiency
for a Stokes number of 0.1 is plotted as a function of YQ . The
upper curve represents the total collection efficiency, and the
dashed curve corresponds to the fractional collection efficiency
in the narrow region between the slits. The relative deposition
between the slits becomes much heavier as the distance between
the pores shrinks, and the overall collection efficiency in-
creases dramatically as the slits approach one another. This
result suggests an explanation as to why heavier depositions are
observed in the bridge between pores in nuclepore filters, and
it is a possible explanation as to why the experimental collec-
tion efficiency with nuclepore filters of very small particles
is larger than expected from theory (17)-
In Figures k and 5, a comparison is made of the trajectories
for a particle initially at a location of YQ-0.025 with the
Stokes number as a parameter. When Yo is equal to 0.15, the gas
stream lines (solid curve) are strongly distorted. The smaller
particles (small Stokes number) tend to follow the gas stream
lines and, consequently, are not collected. The particles of
very large size do not deviate significantly from the vertical
stream lines upstream from the filter. Intermediate particles
are collected as they follow the initial distortion of the
stream lines, but are unable to follow the sharp reverse bend
near the filter surface. Consequently, one has the interesting
result that for conditions where inertial impaction is the prin-
cipal collection mechanism and where there is no reentrainment
or bounce-off, the collection efficiency first increases and
then decreases as the particle size (or as the upstream velocity)
increases. This provides an explanation for the surprisingly
experimental results shown in Figure 1. When the slits are more
equally spaced (a value of Yo=0.5 would correspond to uniform
spacing), one does not have the distortion of the stream lines,
and this effect is not observed.
In considering particle collection by the glass capillary
array filters, account must be taken of several mechanisms: l)
diffusional collection within the pores; 2) gravitational
settling within the pores; and 3) inertial impaction on the
interstitial areas between pores at the filter face. Here the
collection of fibers within the GCAF is simulated. Specifically,
411
-------
we wish to examine whether it is possible to determine both
length and diameter from penetration measurements, and what
effect polydispersity of the fibers will have on the penetration.
The simulations were carried out using a computer code developed
in our laboratory and discussed elsewhere (l8). The principal
features of the code were:
1. Actinolite fibers were represented by prolate spheroids
of diameter Dp and mean aspect ratio Bo . The aspect ratio B was
distributed log normally with a standard deviation CT.
2. The fibers were aligned with their major axis parallel
to the direction of flow (an option of randomly distributed
fibers was provided).
3. The mechanical properties of the glass capillary filters
were completely determined by a porosity (e), a flow rate (Q), a
pore diameter (Df) and length (Lf), a cross-sectional area (Do)
and an angle of inclination (8).
k. The penetration where only gravitational settling was a
mechanism for particle removal was given by the expression of
Pich where the settling velocity was given by the expressions of
Oberbeck for spheroids.
5. When Brownian diffusion was predominant, the Gormley-
Kennedy expression was applied.
We first considered whether the length and diameter could
be determined independently. Experimental conditions were chosen
where diffusion could be neglected. The standard deviation a
was chosen as 0.1 corresponding to a distribution of almost
constant diameters and lengths. In Figure 6, the ratio of the
apparent aerodynamic diameter at an angle Q to the apparent
diameter at 0° is plotted as a function of angle 9. The aero-
dynamic diameter is the diameter of a hypothetical sphere of unit
density that has the same settling velocity as the fiber.
An "apparent" diameter corresponds to the diameter of a
monodisperse particle distribution having the same penetration
as the distribution. For o< 0 . 1 , the "apparent" aerodynamic
diameter is simply the aerodynamic diameter.
For spheres or randomly aligned fibers of a monodisperse
distribution, the aerodynamic diameter does not vary with angle.
This follows because the penetration Pt is a function of a
variable x where:
V
X ~ 5- cos 9
(7)
with Vs the settling velocity. If the particle were spherical
412
-------
or if it were randomly aligned, Vs could "be constant. On the
other hand, if the fibers have a preferential orientation, the
values of Vs (and the aerodynamic diameters) would vary with
orientation. In Figure 6, the ratio is plotted as a function of
angle for different values of mean aspect ratio. For B =1
(spheres), as expected, it is observed that the aerodynamic dia-
meter remains constant, whereas for values of B= 10 and B= 100
the aerodynamic diameter decreases with respect to B. This
would suggest that it is feasible to distinguish between aero-
dynamic diameters and hence fiber length depending on the angle
of orientation.
However, if the method is to be useful, one must be able to
detect meaningful differences in the experimental penetration at
different orientations. In Figure 7, the penetration is plotted
as a function of Q/cos 6 for two different aspect ratios. For
each value of mean aspect ratio, three angles of orientation are
used: 0°, ^5°, and 60°. Since the penetration varies by more
than 5%, greater than the uncertainty in measurements, penetra-
tion measurements have the capability of discriminating between
fibers of different lengths when the fibers align with the flow
field.
In this paper, it was demonstrated that one could determine
fiber length from fractional penetration measurements if the
angle of the elutriator or settling chamber were varied and if
the fibers acquire a preferential alignment. These results
suggest a method for classifying fibers according to their
length as well as diameter.
Numerical simulations indicate why much heavier depositions
are observed in the interstitial areas between pores. These
results and preliminary experimental measurements suggest that
the GCAF may be used to remove isometric particles and to
concentrate fibers.
ACKNOWLEDGEMENT
This research was partially supported by the Environmental
Protection Agency (Grant # R 8065 180 l) and by the National
Science Foundation (Grant # CPE 80 11269).
413
-------
STK'/Z
Fig. 1 Collection Efficiency
for 0.62 ym PSL Aerosols
(25 ym GCAF)
Fig. 2 Chrysotile Fibers After
a 50 ym GCAF
.3
.2
.2
Yo
.4
.14
1.44
.1
.2
.4
.5
3 Collec t ion
Efficiency at
Stk=0.1 for
"Slit" Model as
Function of Slit
Loc at ion
it Particle
Trajectory as
a Function of
Location and
Stokes Number
5 Particle
Trajectories as
a Function of
Location and
Stokes Number
Fie. 6
THETA
414
-------
ENDNOTES
1. Spumy, K., ¥. StBber, H. Opiela, and G. ¥eiss. Am. Ind.
Hyg. Assoc. J. k: 198-203, March 1980.
2. Spumy, K. , J. Gentry, and W. St8ber. In: Fundamentals of
Aerosol Science, Shaw, D. (ed.). New York, John Wiley &
Sons, Inc., 1978. p. 257-32U.
3. Kellert, R., R. Preston, C. Johnson, D. Safferman, M.
Shapiro, and J. Gentry. I&EC Prod. Res. and Develop. 19:
502-506, December 1980.
k. Lin, S., L. Colcord, M. Triantafillou, C. Field, and J.
Gentry. (Presented at the Fine Particle Society Meeting,
College Park, Md., Sept. l6-l8, 1980).
5. Lin, S., R. Preston, and J. Gentry. In: Atmospheric Pollu-
tion, Benarie, M. (ed.). Amsterdam, Elsevier Scientific
Publishing Company, 1980.
6. Mercer, T., and R. Stafford. Ann. of Occup. Hyg. 12: itl-U8,
1969.
T. Fan, K., C. Leaseburge, Y. Hyun, and J. Gentry. Atmospheric
Environment. 12: 1797-1802, No. 8, 1978.
8. John ¥., G. Reischl, D. Bettencourt, S. Goren, and D.Plotkin.
(Presented at the 10th Aerosol Technology Meeting, N.M., 1977).
9. Lin, S. M.Sc. Paper, University of Maryland, 1980.
10. Dahneke, B. J. of Coll. & Inter. Sci . , 51: 58-65, Apr 11,1975-
11. Cohen, A., J. Gallily, A. Schiby, W. Hollander, P. Mause, D.
Schiless, and ¥. StSber. (Presented at the GAeF Conference,
Schmallenberg, ¥est Germany, 1980).
12. Smith, T., C. Phillips, and 0. Melo. Environ. Sci. Technol.
10: 27^-277, March, 1976.
13. Parker, R. Ph.D. Diss. Duke University, N.C. 1975-
Ik. Parker, R., and G. Buzzard, J. Aerosol Sci. 9- 7-l6, No. 1,
1978.
15- Manton, M. Atm. Environ. 12: l669-l675, 1978.
16. Babuska, I. In: The Mathematics of Finite Elements and
Applications, ¥hiteman, J. R. (ed.). London and New York,
Academic Press, 1976. p. 125-1^2.
415
-------
IT- Gentry, J., K. Spurny, and J. Schoermann. (To be published
in Atm. Environ. 198l).
18. Colcord, L., Y. Park, P. Anderson, and J. Gentry. (Presented
at the GAeF Conference, Schmallenberg, ¥. Germany, 1980).
416
-------
ULTRA-HIGH EFFICIENCY FILTRATION SYSTEMS
(AIR RECIRCULATION)
Robert W. Potokar
Staff Development Engineer
General Motors Corporation
Manufacturing Staff
Warren, Michigan
ABSTRACT
The metal casting industry is becoming increasingly aware of the role
energy availability and its efficient use has on productivity and pro-
fitability. One of the major uses of energy in a foundry is for heating
and distributing make-up air. Efficient filtration and recirculation of
presently exhausted air could be a viable technique for energy conser-
vation in some cases. This paper will review new hardware and test data
from pilot testing sponsored by the American Foundrymen's Society (AFS)
which indicates that particulate removal efficiencies in excess of
99.96% are feasible from casting cleaning operations. In addition,
guidelines for a complete system design will be presented.
TEXT
One of the major uses of energy by all foundries is for heating and
distributing make-up air. On a yearly basis, heating, ventilation, and
air conditioning (HVAC) energy usage averages over 15% of the total
energy used by an iron foundry. During a winter month, this percentage
may rise significantly. The amount of energy going to HVAC has accel-
erated in recent years because of increasing pressure from OSHA, unions,
and employes to meet contaminant standards and improve worker comfort.
The recirculation of exhaust air can provide for significant reductions
in consumption of energy for tempering make-up air. Filtered and
cleaned exhaust air can help improve the in-plant environment by al-
lowing the ventilation engineer to use more air volume to capture air
contaminants without overtaxing make-up air systems and disrupting the
static pressure balance within the plant.
Why air recirculation as an energy conservation means? The majority of
the air exhausted from foundries is near ambient temperature. The volume
may be in the order of 10 million ft /minute in some of the large auto-
motive casting facilities. The magnitude of the air handling problem is
exemplified by the fact that some iron foundries use up to 40 air changes
per hour resulting in 112 tons of air being handled per one ton of iron
poured. The fact that we are dealing with large air volumes at rela-
tively low temperatures makes the cost effectiveness of heat recovery
marginal. Also, heat recovery equipment (heat wheels, heat pipes, air-
to-air heat exchangers, etc.) only offer a 40% to 70% BTU recovery po-
tential, whereby air recirculation approaches 100% recovery of the
energy in an air stream.
417
-------
Present day foundry process exhaust systems have, as an integral part of
particulate removal systems, baghouses, electrostatic precipitators, or
wet scrubbers. However, the efficiencies to meet EPA emission require-
ments do not approach the efficiencies needed to meet recirculation
guidelines. What we find in many foundries is the make-up air system
inlet near an exhaust duct resulting in the worst system possible. The
heat is lost from the exhaust, but the contaminants are partially recir-
culated. Some states do have prescribed limits for the type and con-
centration of contaminants allowed in recirculated air, but many do not.
A general rule is that the recirculated air should contain no contami-
nant in excess of 10% of the permissible exposure levels (PEL) for each
substance. These PEL are set by OSHA.
To help coordinate the metal casting industry's efforts on air recircu-
lation and stimulate supplier R&D, the American Foundrymen's Society
organized the Air Recirculation for Energy Conservation Committee in
1977. The Committee has focused its efforts on filtration and rein-
troduction of exhaust air into the work space to conserve energy. Not
considered is the utilization of dirty exhaust in other processes or the
use of unheated make-up air around ventilation hoods to provide the
needed draft and exhaust volume. These types of conservation techniques
are best handled on an individual plant basis. Also, because of the poor
efficiency of heat recovery and the fact that suppliers, engineers and
engineering societies are addressing indirect heat recovery, these areas
would not be considered.
From the Committee's inception up to today, significant progress has
been made where filtration hardware is available which we believe can
meet the requirements for ultra-high filtration efficiencies on heavily
loaded particulate streams such as casting cleaning process exhaust.
Before a detailed review of where the technology stands today, a brief
outline of the Recirculation Committee's activities to date is warranted.
Foundries, as all complex multi-process facilities, have numerous ex-
haust streams, all with different temperatures and contaminants.
Exhaust emanates from the mold room, core room, shakeout, sand handling
systems, cleaning room, and other processes. To obtain a background on
present recirculation installations and past attempts, a questionnaire
was sent to 3500 industry executives. Based on this survey, and on the
fact that the exhaust from the cleaning room should have no products of
combustion, it was decided that the first process to be fully investi-
gated for recirculation would be the cleaning room. The cleaning room
operations are a conglomerate of processes such as grinding and debug-
ging. They employ the largest percentage of people and run production
during periods when other heat-producing operations of the foundry would
not be able to supply heat necessitating the use of outside energy or
make-up to replenish exhaust. Generally, the cleaning room exhaust
volume is about 20% of the total foundry exhaust. This may add up to
1,500,000 cfm or more in a large automotive facility.
418
-------
From the survey, we found no installation that represented the best
engineering practices, which includes properly specified air filtration
equipment followed by a return air monitoring system and a bypass to
atmosphere in case of failure or efficiency problems. Some of the newer
recirculation installations from blast cleaning in the cleaning room are
using opacity meters to monitor the cleanliness of the exhaust before
return to the work space. However, the sensitivity of opacity meters is
quite limited. It would require a path length of 10 feet between the
light source and sensor to enable a 1% opacity change to relate to a
rise in particulate concentration from 1 mg/m to 2 mg/m . Because
opacity is an optical measurement, it is dependent on particle size
distribution and reflectivity characteristics of the particulates in the
gas stream. So, to use an opacity meter, the instrument has to be
calibrated versus in-stack mass measurements and the particle charac-
teristics must remain constant over all conditions.
Getting back to the results of the survey, it was found that some
foundries have attempted to recirculate from different operations
without success. In one case, the recirculation was in operation for a
year before it was discovered that the concentration of free silica in
the breathing zone was too high due to the recirculation. From this
background, it was decided that a demonstration recirculation system be
installed on the cleaning room exhaust following these steps:
(1) emission characterization, (2) filtration equipment specification,
(3) monitoring equipment specification, (4) integration of a complete
system, (5) build and installation, and (6) performance testing.
To determine the type of filtration required to recirculate cleaning
room process exhaust emissions, characterizations were done on a select
blast cleaning process at five different foundries. AFS research funds
were used to accomplish this task, and the five participating facilities
provided the required test ports, access scaffolding, and site coordi-
nation. The five facilities ranged from the highly automated automotive
casting plants to small batch operations, all with a variety of venti-
lation and collection systems. The tests were conducted in the follow-
ing manner. The data will not be identified as to the source to insure
confidentiality.
Simultaneous inlet and outlet process emission samples from a cleaning
room shot blast exhaust stack and existing collector discharge were
performed in duplicate at each of the five foundries. Each emission
characterization included all the parameters listed in Table I and was
conducted in the prescribed manner.
Both inlet and outlet from existing collectors is important in the
development of design criteria for either a complete new process fil-
tration system or for an add-on polishing filtration system.
For uniformity in having equipment suppliers bid on systems to meet
recirculation requirements, a "typical exhaust" was formulated in
Table II. The particle size distribution of the inlet and outlet
419
-------
mass is shown in Figures 1 and 2, respectively. Because of the large
amount of submicron material, with the inlet having approximately 700
mg/m of particulate < Jy , fabric filtration is a good candidate treat-
ment system type.
Using these data, over 50 equipment vendors were requested to specify a
5000 cfm collection system to act as either a process collector or a
finishing collector on top of the existing collector. Manufacturers of
all generic types of air pollution control equipment were included on
the bid list. Of the over 50 suppliers contacted, four generic types
appeared, on paper, to have the potential for high efficiency and
reasonable maintenance.
All four systems were tested at the Central Foundry Division of General
Motors Corporation, Pontiac plant. The pilot units were tested on a
slip stream of direct process exhaust from an iron casting blast clean-
ing cabinet. The unit's performance was tested using the same sampling
procedures previously described. All of these filtration systems are
now commercially available. The four systems included an electrostati-
cally augmented baghouse, a two-stage baghouse system, a single baghouse
using polyester felt bags with a teflon lamination, and a two-stage
cartridge filtration system.
The operation of the electrostatically augmented baghouse uses an
ionizing section directly upstream from a baghouse. As particulate
matter passes through, it is charged to saturation. The charged par-
ticles create a polarization effect between the particles and the fabric
surface. Our pilot testing confirmed that significantly higher filtra-
tion efficiencies occur with the use of the high voltage charge. Also,
a reduced pressure drop was witnessed when the electrostatic charging
section was used. With the charge, the particulate removal efficiencies
were in excess of 99.98%. During the course of over two months of on-
stream running, it was determined that if the electrostatic section was
turned off, the pressure drop across the system would build up six times
faster. This is due to the polarized particles which form a "fluffier"
dust cake on the fabric and to precipitation within the ionization
section on the cylinder walls supporting the corona wires.
The unit tested had the ionizing hardware integrated into the baghouse
system where the reverse jet air was also used to purge dust from the
corona wires, grounded tubes, and supports. Preliminary investigations
have been made by the supplier to determine if a separate ionization
section could be used to achieve the same effects. It is envisioned
that the ionizer could be located in ductwork upstream from an exist-
ing baghouse, thus enabling retrofit of conventional equipment now in
the field. A drawback of this approach is that the ionized particles
tend to precipitate out on the grounded duct between the ionizer and
the entrance to the baghouse. Also, compressed air or another cleaning
mechanism is required in two locations to remove dust build up in the
corona section and in the baghouse.
420
-------
Before I discuss the remaining work, I want to point out that the AFS
and Committee policy is not to mention suppliers by name, but merely
refer to generic system type.
The second pilot system tested was a two-stage baghouse. The first
stage was a reverse jet system with a 7:1 air-to-cloth ratio using
dacron felt bags. The second stage was a shaker-type unit with a 6:1
air-to-cloth ratio, with cotton bags and a cellulose pre-coat filter
aid. The filter aid effectively forms a high efficiency filter in
place. The first stage-is designed to reduce dust loading from the
approximately 1000 mg/m down to approximately 20 mg/m . The second
stage, ultra-filtration step, reduces dust loading to less than 0.3
mg/m . The test data shows that upon start up, the system efficiency
was 99.96-99.97% and after one month of continuous on-stream operation,
the efficiency rose slightly to 99.98%. Before concluding all tests
with this system, a breakdown in the first stage baghouse was stimu-
lated. Two of the 36 bags in the first stage were intentionally torn.
The subsequent emission tests documented that the efficiency was still
in excess of 99.96%. A second stage baghouse using a pre-coat filter
aid can be used to retrofit an existing system presently exhausting to
increase efficiency for possible recirculation. However, the efficiency
in the first stage must be adequate; that is, the outlet from the first
stage should be in the magnitude of 20 mg/m to avoid overload of the
second stage.
The third pilot system tested utilized only the first stage baghouse
from the aforementioned system. The dacron felt bags were replaced with
a new type of bag. The new bags have a microporous teflon membrane
laminated to a felt filter media. The manufacturer claims that unlike
typical fabric or felt filter media, which depends upon particle build
up or a pre-coat for filtration, the filtration takes place both on the
surface of the membrane and on the collected particles. Thus, a steady
high efficiency can be achieved without the need for a pre-coat. Our
tests with these membraned bags resulted in a particulte removal effi-
ciency of 99.59% on the blast cabinet exhaust. Only limited testing was
done with these bags due to time restraints. No estimate can be made on
bag life or if the efficiency would increase or decrease with use. The
emission testing was done the same day the new bags were installed.
Only 45 minutes of on-stream running occurred before emission testing
commenced. The efficiency of the pilot one-stage system with conven-
tional bags was 98.6%, which means a significant increase in efficiency
did occur. The penetration of particles decreased with the new bags
from 1.4% to 0.41% which is over a three-fold decrease in penetration.
The final pilot filtration system evaluated was an aspirated cartridge
dust collector with a second stage filter for monitoring the entire air
stream. The primary pleated cartridges instead of bags are used for
particulate collection and compressed air is used to back flush the
cartridges for cleaning. The cartridges are made of a non-woven,
synthetic and cellulose material, and the back-up monitoring filter
is also of cellulose. During the first tests on the unit, the total
421
-------
3 3
particulate load was dropped from a 925 mg/m inlet to 0.20 mg/m on the
outlet. After this test, a team from the manufacturer checked the
system for leaks and one cartridge was changed because of a suspected
leak. This leak may have occurred in shipment or during manufacture.
Another test was performed and during this test, the efficiency in-
creased to 99.997%. The inlet concentration was 1248 mg/m and the
outlet 0.033 mg/m . This system was on-stream for four months with no
apparent problems.
The particulate removal efficiencies quoted and summarized in Table III
are based on total gross particulate load. A complete breakdown of
composition of particulates and gases will be made part of our committee
report due for publication this year, if our committee completes the
laborious task of writing in a timely manner.
Our goal as an American Foundrymen's Society committee was not to
certify or test every piece of filtration equipment, but to stimulate
suppliers to develop equipment to meet the needs of high efficiency
filtration and the unique needs of our industry. We feel this goal has
been met. The pilot demonstration and test program also showed that
this type of equipment can operate in a foundry environment without
oberburdening maintenance.
Before any of these or any other system should be considered for an
application, a complete specification should be developed. The criteria
of loading, particle size, particulate and gas composition, along with
all the other parameters outlined in Table I, must be measured. The
data should include the extremes that the exhaust stream may vary to.
Any control system must be able to not only meet the normal conditions,
but also perform under the abnormal conditions.
The allowable contamination in the recirculated air is not a definitive
value. Some states and foreign countries use the value of 1/10 the TLV
for any contaminant in the recirculated air. NIOSH, the National Insti-
tute for Occupational Safety and Health, has published "A Recommended
Approach to Recirculation of Exhaust Air" in an attempt to address this
question. This 184 page document presents complex mathematical means
for back calculating the effect of recirculated air on the air that
workers will breathe. The AFS Committee on Exhaust Recirculation has
taken the stance that (1) the recirculated air should not contain any
contaminate in excess of four times the Primary Ambient Air Standard.
(If a particular contaminant is not covered under the Ambient Air
Standard, the value 1/10 the established PEL'should be used.)
The quality of the recirculated air should be monitored on a regular
basis. The definition of regular is subject to a great variance in
interpretation. As a committee, we recommend that a method be used
to continuously monitor the recirculated air return and a bypass to
atmosphere in case of filter failure be tied into the system. Also
periodic measurements should be taken, at least upon system installa-
tion and at a minimum of once a year thereafter in workers breathing
422
-------
zones, to determine the effect of the recirculation system. There are
a variety of methods which can be used to continuously on-line monitor
the particulate content in recirculated air. It must be re-emphasized
that our efforts to date have centered on air streams which do not
contain toxic or detrimental gaseous constituents. If gaseous contami-
nants are to be continuously monitored, the specific component may be
determined by infrared or another method. A number of such continuous
readout monitors are on the market today.
For a determination of available equipment for particulate monitors at
the low concentration levels required, we as a committee sent a survey
to over 40 suppliers in the particulate sampling and monitoring busi-
ness. Again, this effort was not only to determine the present state of
the art, but also to stimulate effort for this market. The response has
been gratifying. There are various light scattering instruments, both
side scattering and forward scattering instruments, which are now coming
on the commercial market. We are now working on a plan to evaluate a
group of these instruments. All of these instruments cost between
$2500-5000 per sensor point, and the sensitivity is approximately
+ 0.005 mg/m in the 0-0.5 mg/m range.
Another method to continuously monitor the particulate level in recir-
culated air is by an after filter. This filter can either be for the
entire air stream or for part of the air stream. The pressure drop
across the after filter is sensed, and if the pressure drop exceeds a
pre-determined level or rate of rise, an alarm can be activated along
with a bypass to atmosphere. We plan to continue to gather data and
information on feasible monitoring methods. This information will be
made available through the American Foundrymen's Society.
The economics involved with air recirculation are continually changing.
Today in the geographic climate area of Chicago or Detroit, the energy
cost to heat and distribute one cfm of make-up air is about 50c/year.
The direct heat energy dollar value is not the only dollar savings.
Other factors should be included such as the filter and maintenance cost
for an existing conventional make-up air system and exhaust cleaning
system. With a properly operating air recirculation system, both jobs
can be done in one unit. Other benefits include: (1) tendency to
neglect dust collectors is minimized because recirculation systems can
substantially reduce the plant's energy bill; consequently, the equip-
ment gets the inspection and maintenance it requires; (2) heat added by
the dust collector fan (approximately 1°F for every two in. w.g.) is
distributed along with the recirculated air; (3) tendency to starve
dust collector fans because of inadequate make-up air is minimized;
(4) a plant can afford to handle larger air volumes, increasing the
capture efficiency of hoods, since make-up air heating costs are not
a consideration.
423
-------
As more plants become air tempered or conditioned for summer, the
dollar savings for energy will approximately double. In addition, if a
facility is in a non-attainment region and there is a need for a parti-
culate emission reduction or tradeoff, high efficiency particulate
removal systems may be dictated. If so, it would make good sense to
consider recirculation. The estimated cost of a nominal 40,000 cfm
process filtration system for a typical casting cleaning room exhaust
capable of recirculation is in the range of $2-4 per cfm (not installed).
In the decade of the 1980's, the metal casting industry, as all American
industry, will have difficulty in meeting the demands of increasing
productivity and profitability while still meeting ever tighter govern-
ment regulations and employe expectations. The AFS is committed to
coordinating increased information and stimulating research, develop-
ment, idea and experience exchange to help meet this challenge. On the
subject of air recirculation, the AFS 10H Committee needs to continue to
receive your comments and cooperation.
As I said before, a detailed report on this subject will be available
soon. However, feel free to contact the AFS or me if we can be of
service.
424
-------
TABLE I
TEST METHODS
TABLE II
TXPICAL EXHAUST
Parana ter
Total particulate concentration
and stack conditions, tempera-
ture, velocity, etc,
Particle size distribution
Carbon monoxide
Phenols
Formaldehyde
Farticulate chemical analysis
on fraction 25v and <5y
«, Free silica
b. Cristobalite
c. Quartz
d. Tridymite
c. Cu
f. Zn
g. Pe
h. Organics
i. Total inorganics
Method
E?A Methods 1-5
In-stack cascade
itipaction
net chemistry
Met impingement,
distillation ar.d
gas chroraatog.rajrhy
Wet chemistry
X-ray diffraction
Atomic absorption
•pectrophotometry
Hexane extraction
Difference
Inlet
Outlet
Emission par-ciculates, avg/ai* 3500 60
Size distribution
B?50' Dicr°ns 3-0 1.7
03 4.3 3.0
Temperature, °F 70° 70°
Hoistura content, 5 4.0 4.0
Particulate breakdown
Fe total, mg/p,} 595 9
Fe <5y, mg/a* 525 9
Si total, mg/ms 4.2 0.0
Hi <5n, sncj/m5 2.8 0.0
SiOj total, mg/ms 80S 10.2
SiOi <5v, tng/ia' 630 7.2
Benzene extractable total, ing/m3 56 1.1
Benzene extractable <5y, rng/m' 4£ 0.8
TABLE III
SYSTEM EFFICIENCIES
JTOTAL PARTICULATE)
UMIT TYPE
Electrostatic Baghouse
Two-Stage Baghouse
Two-Stage Baghouse
(after 1 month)
Two—Stage Baghouse
(2 bag failures first stags)
Single-stage Baghouae
(conventional baja)
Single-stage Baghous«
(iBecbrane lined bags)
Cartridge Filter
iT_t
1
2
1
2
1
1
1
1
1
2
INLET
1054
743
1040
848
656
860
1804
1300
925
1248
OUTLET
(mg/m')
0.19
0.11
0.26
0.29
0.13
0.29
25.17
5.37
0.20
0.033
EFFICIENCY
99.962
99.985
99.975
99.966
99.980
99.966
98.605
99,586
99.978
99.997
425
-------
10 IS 20 30 «0 40 to 70 80 §S 90 95
.2
3X1
Figure 1 - Typical Inlet Particle Size Distribution
426
-------
I .
•f.
98%
SO 60 70 SO BS 90 K
Figure 2 - Typical Outlet Particle Size Distribution
427
-------
THE HET UALL ELECTROSTATIC PRECIPITATOR
By: Dr.-Ing. J. Starke
Phys. Forschungsinstitut (AP)
Bayer AG
Leverkusen, W. Germany
Dip!.-Ing. J. Kautz, Ober.-Ing. K-R. Hegemann
Gottfried Bischoff GmbH & Co., KG
Essen, W. Germany
ABSTRACT
The operating principles of electrostatic precipitators, particularly
those of conventional wet electrostatic precipitators, are well known. The
incentive to develop a new variation of the wet ESP - the wet wall electro-
static precipitator (WWP) - derived from investigations on the removal of
participates from waste gas from a plant designed to incinerate organically
loaded wastewaters having high salt contents. Conventional ESP's, both wet
and dry, as well as venturi scrubbers and fabric filters were tested and
found to be uneconomical.
The most efficient and economical solution proved to be a wet ESP of
new design with vertical, water-cooled ground electrodes. The water film
resulting from vapor condensation on the cool electrodes instantaneously
dissolves the soluble particulates and suspends the insoluble dusts, flushing
both from the discharge zone. Thus the efficiency of such a precipitator
remains constant since it is free of dust deposits and uninterrupted by
flushing cycles.
The wet wall electrostatic precipitator has been thoroughly tested. It
has a simple structural design permitting reasonable capital and operating
costs. In the laboratory typical separation efficiency curves were recorded.
Of several different applications tested in the field the most promising were
those dealing with submicron particulates and aggressive gases.
INTRODUCTION
The development of this new wet-type electrostatic precipitator resulted
from experiments on the precipitation of particulates originating from a
plant for the incineration of organically contaminated, strongly saline efflu-
ent. These particulates form during cooling of the flue gas and have diameters
in the order of 0.1 pm.
Conventional dust collection processes were initially studied in attempts
to deal with particulates having these characteristics. With dry type ESP's
the extreme fineness of the particulates led to very low migration rates In
addition the particulates had a very low bulk density and were therefore'
difficult to remove from the collector plates by rapping.
428
-------
The acidic pollutants present in the flue gas and the Tatters high
dew point dictated that only fabric filters such as glass, with working
temperatures above 250°C5 were applicable. It is known that it is difficult
to remove sub-micron particulates from fabrics without damaging the fabric
and lessening the service life. Reasonable service lives could only be
obtained by setting specific loading per unit of surface area at uneconomical
levels.
A conventional low pressure scrubber was tested at an efficiency of only
50%. By extrapolation, a venturi scrubber capable of adequate removal effi-
ciency would require a pressure differential of 2,000 mm W.G. and unaccept-
able energy costs.
Experiments with vertical tubular ESP's of the wet type were more
encouraging. There were; however, two remaining disadvantages. In wet ESP's
designed for corrosion protection the collecting electrodes are fabricated
from non conductive plastics. Plastics formulated to be conductive are not
sufficiently resistant to damp corrosive flue gases. The surface of plastics
which are used becomes conductive through deposition of moisture from the
water separated; however, these electrodes become progressively drier and
less conductive in the cleaned gas direction. Thus the contribution of this
section of the ESP to cleaning is minimal.
The second disadvantage is that the electrodes must be rinsed down at
regular intervals. During this period the collection efficiency drops consid-
erably, a characteristic which is overcome in actual operating practice by
providing more expensive dual zone precipitators.
Design and Operating Principle of the Wet Hall Precipitator
The Wet Wall Electrostatic Precipitator (WWP) was designed to incorporate
the useful features of conventional wet type ESP's and overcome the disadvan-
tages referred to above. As seen in Figure 1, the novel feature is the
cooled collecting electrode design which facilitates the condensation of
water vapor from the gas being treated. The condensed water film then pro-
vides the desired surface conductivity and rinsing characteristic. The
thermal conductivity of the plastic tubes is an important consideration.
These concepts were initially tested in a wet wall ESP which was retro-
fitted to permit collector electrode cooling. A steady flow of cooling water
caused water to condense on the collecting surface at a dew point of 60°C in
the gas being cleaned. The resulting thin film of water was sufficient to
dissolve and constantly flush away the saline deposits. It was possible to
operate the precipitator continuously at high collection efficiency for
extended periods of time with gas velocities of about 1 m/s and particulate
concentrations of 20 g/m3.
As a result of these encouraging findings the full scale installation
shown in Figure 2 was designed and built. The most important component is
the tubesheet, which is made entirely of plastic and enclosed in a rubber-
lined steel casing. To compensate for the differences in thermal expansion
of steel and plastic, special packing was provided between the respective
429
-------
. Clean gas
Cooling
water
Cooling
water
Raw gas ^ -
Water
dram
Figure 1.
Bischoff Wet Wall Electrostatic Precipitator With
Indirectly Cooled Collecting Electrodes
Figure 2.
Precipitation of Extremely Fine Salt Aerosols
from an Effluent Incinerator
The water inlet and outlet were carefully designed to ensure their
structural rigidity and to provide uniform cooling. In other respects the
design does not deviate markedly from that of conventional vertical, wet type
electrostatic precipitators.
Nozzles for spraying the collecting surfaces were provided but were only
used to develop surface conductivity during static (non-operational) tests.
No provisions were made for periodic flushing.
The success of this small commercial unit in treating flue gas from
saline liquor incineration led to a decision to investigate the concept
further, both in the laboratory and in the field.
The program consisted of:
* Laboratory comparison tests with standard particulates in which
test runs would be compared with results already know for high
performance scrubbers.
430
-------
* Application oriented field tests with small experimental WWP's.
* Analysis of experience gained from operating commercial units.
Experimental Program
Laboratory Experiments
Initial tests demonstrated that even dusts insoluble in water were easily
flushed off. Experimental conditions could therefore be selected which were
identical to those used during experiments previously carried out with wet
type precipitators.
The pilot unit was designed for a throughput of 1,500 m3/h. The test
setup is shown in Figure 3.
The raw gas was saturated with water and raised to the desired tempera-
ture by injecting steam into the input air. Measurements have shown that
this technique has no influence on the precipitator's collecting efficiency.
Particulates were fed into this gas stream using a compressed air injector
and a manually controlled vibrating chute. All raw gas ducts were insulated
and heated. Upstream from the particulate feed there is a mist eliminator to
prevent any agglomeration of dust due to carryover of drops of condensate.
The pilot precipitator is of vertical tubular design. The precipitation
unit consists of nine internally rubber lined steel tubes of 200 mm diameter.
Tube clusters varying in length from 1 to 3 meters were evaluated. Cooling
water flows continuously around the outside of the tubes. The remaining
elements of the equipment are of conventional design.
On the clean gas side the residual dust content is measured in a long
straight section of the tube by the standard method of isokinetic partial flow
aspiration (1,2). In addition, the gas velocities, pressures, and temperatures
are measured at various locations in the experimental setup, cf. Figure 3.
In the experiments to date the following parameters have been varied
within the indicated limits:
* Length; of the precipitation unit*(l to 3m)
* Raw gas temperature, T (50° or 80°C)
* Gas velocity in the precipitation tube, v (1 to 2.5 m/s)
* Raw gas particulate loading, 6, (1 to 5 g/m)
* Electrode voltage, ug (30 to 70 kV)
* Type of particulate (grain diameter, electrical resistance,
wettability)
Data evaluation in respect to the total collection efficiency, n, and
the size fraction collection efficiency, nf, is made in accordance with the
details given by Holzer (2).
431
-------
1200
Static pressure meos.
Q) Dynamic pressure meas.
Temperature meas.
IT) tsokinetic dust sampling
Figure 3. Experimental Apparatus
Oust loading . <5< = 1g/m* ot WWP inlet
25 35 45 55 65 u. tkV] 75
Figure 4. Collecting Efficiency n of the Test WWP as a Function of Electrode Voltage, U
100
\
as temperature . T=~~8Q*C I atWWP
ust loading g; = Ig/m1 [ tnle_t
\
Tube length L = 2 m
Gas velocity v ; 2m/s
Coitecting efficiency ij =98%
Figure 5. Retained Fraction Curves, Rj and R,,, and Fractional Collecting Efficiency, nf.
432
-------
Some typical results from the experimental program are given in Figure 4,
where total collection efficiency, n, is plotted as a function of electrode
voltage, u . Parameters varied included the length of the collection tubes
and the velocity of the gas in the tube. The experiments were carried out
using "Sillitin", a fine quartz powder,as the test particulate at a raw gas
loadiig of 1 g/m3 and a temperature of 50°C.
When the length of collection tubes and gas velocity ware varied the plots
relating collection efficiency to electrode voltage are seen to be a family of
similar curves. With increasing voltage these curves rise sharply at first,
then progressively more gradually, and finally coincide. At constant voltage,
lengthening the tube and reducing the gas velocity produced, as expected, an
improvement in collection efficiency.
It is noteworthy that with a collection tube length L=lm and a gas
velocity of v=lm/s better collection efficiency values are achieved at all
voltages then when L=2m and v=2m/s. The generally accepted correlation
between collection efficiency and specific area of precipitation A/Q (with a
constant tube radius A/Q is identical to the retention period) does not yield
conclusive results in this instance, cf also (3). Clarification of this
matter is still under study.
The migration velocities are about 25 cm/s, a normal rate for experimental
precipitators. For practical applications these discrepancies are of less
importance.
The most important result, which can be seen in Figure 4, is that start-
ing from a relatively low voltage, i.e. about 50 kV, the length of the col-
lection tube, the gas velocity, and the voltage itself no longer are control-
ling factors and that a collection efficiency of 99.9% is attained. Experi-
ments at higher gas temperatures and other particulate loadings also did not
provide data significantly different from that plotted in Figure 4.
Figure 5 shows the size fraction collection efficiency, nF, of the
condensation electrostatic precipitator as a function of the grain diameter,
d. The size fraction collection efficiency is calculated from the residue
curves Rj and RH for raw and cleaned gas respectively (2).
Unlike the case with wet scrubbers, a minimum is noted at a grain dia-
meter of about 0.6 ym. Overall, the curve is close to the total collection
efficiency of 98% (see also (3), where similar curves are shown from both
theoretical and experimental research ). The minimum is typical of electro-
static precipitators since with grain sizes in the range of 0.5 ym charging
of the particle is less intense because of ion deposition and without ion
diffusion (3,4).
The evaluation of the size fraction collection efficiency could not be
carried out for grain diameters below 0.1 ym since the grain analysis method
used (centrifugal sedimentation) does not yield reliable results in this
range. The limiting grain diameter as normally defined cannot therefore be
determined for the condensation precipitator from the available data. It can
only be concluded that the limiting grain diameter is less than 0.1 ym.
433
-------
Further research is in progress to determine the extent to which the
principle of wet wall condensation contributes to the level of collection
efficiency.
Field Tests
Glass Trough
The condensation precipitator concept was tested on waste gas from a
molten glass trough. The gas contained a mixture of solid dust particles
and P205mist.
At the time of these tests the plant had a venturi scrubber installation
which did not meet the collection efficiency required under the latest emis-
sion control regulations - approximately 20 mg/m3 NTP.
A pilot WWP with one pipe, but retaining the original geometry of the
precipitation field, was installed in parallel with the scrubber plant. The
waste gas was cooled from about 150°C by spraying with water in a pipe imme-
diately upstream of the filter, passed through the precipitator, and fed into
the gas line again upstream of the existing blower.
The following residual contents were achieved at gas flow ratesof
1.2 - 1.3 m/s:
P20s approx. 1 mg/m3 NTP
As203 0.1 mg/m3 NTP
Total Collection Efficiency >99.8%
Commercial Units
The condensation precipitator concept has been employed with great
success in the full scale plants built to date, particularly where full
corrosion protection has been employed. Two of the plants already built
provide for the precipitation of pigments from flue gas and the precipitation
of sulfuric acid mists, mixed with particulates, from an acid waste incinerator.
Comparison of Costs and Performance
Performance Compared With Wet Scrubbers
In Figure 6 the power consumption of various types of scrubbers is
plotted against the limiting particle size of the test dust, i.e., against
the particle diameter at which just 50% is cleaned. Below the typical curves
drawn for each scrubber type is a lower boundary line below which no scrubber
dropped.
The performance characteristic of the WWP is designated by a cross-hatched
area rather than a linear plot. There are two reasons for this. As discus-
sed above, the limiting particle size for the condensation precipitator can-
not be defined since it lies beyond the range of available metrological equip-
ment. The lack of precision in particle size data also contributes to a lack
434
-------
of precision in power consumption figures. The performance band given for
the WWP in Figure 6 does; however, clearly show that power consumption lays
one or two powers to ten below the level for conventional scrubbers and the
particle diameters which can be handled lie outside the range that scrubbers
can normally cope with. In other words, the condensation precipitator can
operate beyond the capability of scrubbers.
As part of the energy consumption study, the power required for ionization,
that for producing the small pressure drop across the electrostatic precipi-
tator, and that for insulator heating, was considered.
The low water consumption is also worthy of note. During the series of
tests it was not necessary to spray additional water into the precipitator
inlet beyond that needed for gas saturation. The condensate produced was
more than enough for satisfactory operation. Pollutants were separated in
optimal concentrations of 25 g/1.
Comparable results can only be obtained from scrubbers by multiple
recycling which is generally not possible in practice because of difficulties
such as increased wear on pumps and nozzles.
Costs and Performance Comparison With Conventional Precipitators
Cost comparisons between conventional electrostatic precipitators and
the wet wall electrostatic precipitator were developed on a percentage basis
with assumptions set out below.
The WWP operates with essentially constant collection efficiency and can
therefore, other conditions permitting, be safetly operated with just one
collection zone.
By contrast, two zones must be provided in a conventional wet electro-
static precipitator since the collection efficiency drops appreciably as the
solid precipitated particles are rinsed away. It was assumed for the purposes
of this comparison that the migration velocity would be 50% of the normal
value when rinsing was being carried out in one of the two zones. This means
that if the precipitation area of the WWP is taken as 100%, that of the con-
ventional precipitator will have to be 133%.
The wet wall precipitator involves more structural expense due to the
cooled tubes and the dead space for water cooling. The precipitator shell
must also be larger. In the conventional precipitator; however, the required
significant amount of dead space was included in the second zone. Two spray
zones and rectifier systems were allowed for the conventional precipitator,
reduced size not playing a significant role here.
The relationships can be seen in Figure 7 which shows diagramatically a
wet wall precipitator and its functionally equivalent conventional precipi-
tator. Costs for the five major elements of the WWP are given the value 1.0;
corresponding costs for the same elements in the two stage conventional
electrostatic precipitator are assigned proportional values.
435
-------
Casing
Collecting
electrode
Discharge
electrode
Flushing
system
Size
1.0
1.0
1.0
0.5
Costs
1.0
1.0
1.0
1.0
1.0
0.375
0,375
0.12
0.10
0.03
1.0
ES
'kWh"
1000m'
Conventional Preclpitator
Size
1.26
1.33
2x 0.67
2x 0.67
2x 0.67
Costs
1.35
0.95
Ratio
0.51
0.36
0.22
0.18
0.06
1.33
Figure 7. Size and Cost Comparisions Wet Wall Precipitator
and Conventional Wet Precipitator
Ul-fca
am
2 3456 789Q1
56)89is*[Mm2]
Mr
1
3
4
5
6
7
8
9
10
11
12
13
U
15
16
17
11
20
21
22
23
type
wet cyclon
washing tower
vortex
jet- scrubber
washing tower
vortex
LP-Venturi
rotation scrubb
LP-Ventun
impinger
LP- v?nt * imp.
vwt cyclonette
LP-Vmturi
Mixer-type
1
^rotation scrubber
}
] Mixer type
rotation scrubber
} HP Venturi
I
WWP
mark
KBVT
6HY
TAS
X KO
^GLK
TCR
0 HS
0 CS
a sc
• PB
nPB
KBT
a KV
• vo
o ST
o pz
°CN
• PD
• RD
o KN
D LR
O BG
'////,
gas
volume
450
250
3000
500
1000
2500
2000
17500
3 000
1 000
1 000
500
3500
1000
300
600
2000
2800
1500
500
1500
pressure
drop
m bar
19
6-20
20
1*21
6-18
15-22
15-20
10
15-40
18
22-37
18
20
20
a
18-70
1+15!
20-50
MS 1
8-10
25-100
40-100
2
water
•SS/fflf"
0.2 -O.S
0,5-1
3-10
0,5-1
2-3
0,5
0,5-1
0,6-1
0,5-2
0,1-0,4
0,5
0,5-0,8
1-2
0,8
1-3
0,04*
Figure 6. Power Consumption Versus Diameter
For Various Wet Gas Cleaning Installations
The actual calculations were too involved to be fully described in this
paper. The size of the precipitator shell was worked out on the basis of
surface area; costs were for realistic designs. The actual or equivalent cost
of electricity required to overcome the head loss can also be expected to
involved extra operating costs of about 35%.
These figures demonstrate that it is worth considering the installation
of a wet wall precipitator for all possible applications.
436
-------
Summary - Applications - Future Work
The first precipitators using the wet wall principle have all been
designed for the wet separation of corrosive waste gases and were therefore
fully protected against corrosion. Despite these expensive provisions for
corrosion protection the plants offered overall savings as compared to the
costs for simple wet precipitation.
The condensation precipitator principle permits the use of highly
insulating, corrosion-resistant plastics as electrodes and collecting surfaces.
In addition to the special applications described above it is believed
that the basic principle has application for all waste gases containing fine
particulates and for which a given water vapor dew point can be maintained.
In addition to low inital costs, the virtual absence of maintenance expense
is especially notable.
Further investigations are needed before more definitive application
prediction's can be made. The correlation between gas dust content and amount
of condensate must be studied. Dust characteristics must be investigated to
learn why materials such as soot resist washdown.
Random observations suggest that condensation effects significantly
promote precipitation. More detailed study of the influence of condensation
near the precipitator walls on the precipitation process is being made.
References
1. VDI - Richtlinie 2066. Leistungsmessungen an Entstaubern.
2. Holzer, K. Nassabscheidung von Feinstaeuben und Aerosolen.
VDI-Berichte Nr. 349, 1979. p. 102-109.
3. Maartmann, S. Moeglichkeiten zur Feinstaubabscheidung mit
trockenarbeitenden Electroentstaubern. Staub-Reinhaltung Luft
34 Nr 10, 1974. p. 353-380.
4. White, H. I. Industrial Electrostatic Precipitation. Reading, Mass.
Addison - Wesley Publishing, Inc. 1963.
437
-------
TROUBLESHOOTING MULTIPLE CYCLONES ON FUEL-OIL-FIRED BOILERS
By: F. Crowson and R. L. Gibbs
Naval Surface Weapons Center
Dahlgren, Virginia 22448
ABSTRACT
The body of information presented in this paper is directed to the
particulate control of boilers fired with No. 6 fuel, a problem when stringent
standards must be met. The success of meeting regulatory standards is highly
dependent on the availability of data for emission characterization, as well
as the availability of space and funds for control equipment. Early efforts
toward controlling emissions from boilers fired by No- 6 fuel at two Navy
shore facilities resulted in the selection of multiple cyclone collectors.
Since their installation, both facilities have failed to meet compliance with
local standards. A study was made to troubleshoot the boilers and their
associated pollution control equipment. The problems associated with the
multiple cyclone collector design, installation, and operation are examined.
INTRODUCTION
The U. S. Navy is required by executive order to assume the initiative
in complying with state and federal regulations governing discharge of
emissions from stationary sources (1). Initial efforts to meet the require-
ments were directed toward providing the quickest, least-cost, and best
control technology available to meet existing regulations. In the mid-197J's,
heavy-fuel-fired boilers at two Navy installations were equipped with multiple
cyclone collectors to meet the emission regulations for the District of
Columbia and the state of Maryland. Initially, these boilers were fired with
No. 6 fuel, and dust collector design was based upon this factor. Unfortu-
nately, limited funds precluded obtaining particle sizing data and flue gas
characteristics to allow the evaluation of other control devices. The major
factors governing the selection of cyclones were space availability, design
simplicity, and cost.
Since the multiple cyclones were installed, emissions from the units
under full-load operation at both facilities have not met the requirements set
by the regulating districts. Source emission tests (SETs) were performed at
various boiler settings in an attempt to verify at what loads compliance could
be achieved. These tests attempted to confirm acceptable operation until
modifications or necessary design changes could be made to ensure a more
efficient system. To date, no systematic study has been conducted to
determine emission characteristics at various loads and to quantify necessary
design parameters. The lack of stack gas data and incomplete operation
records further complicated effecting a ready solution. The purpose of this
investigation was to troubleshoot the boilers with the present multiple
cyclone collectors and to provide an engineering assessment for further action
based on existing information.
438
-------
PROCESS EVALUATION
Certain criteria were used for evaluating each process at both facilities.
The factors for the evaluation of the boilers and burners included (1) instal-
lation, (2) maintenance, (3) internal deterioration (corrosion/erosion), (4)
flame impingement, (5) flame pattern and length, (6) stack outlet conditions,
(7) fuel flow, temperature, and pressure, (8) fuel analysis, and (9) instru-
mentation for boiler operation and stack gas analysis. Assessment of the
pollution control equipment included surveying (1) design, installation, and
maintenance, (2) dimensional parameters, (3) wall roughness of the cyclone
tubes, (4) plugging and bridging in vanes, cones, or hoppers, (5) leaks or
infiltration at gaskets and valves, (6) erosion and corrosion, (7) hopper
loading, (8) pressure drop and inlet flow distribution, (9) insulation, (10)
auxiliary equipment such as fans and expansion joints, and (11) instruments
for opacity measurement.
Washington DC Facility
The Washington DC Facility consists of twin, No. 6 fuel boilers each
rated at 100 million British thermal units (MBTUs) and utilized for heating
buildings. Each boiler is fired by a steam-atomized burner capable of a
maximum flow rate of 700 gallons per hour (GPH) . The average firing rate for
these boilers is 84 MBTUs. An evaluation was first made of the boilers and
burners. Historical data were obtained by reviewing previous operation and
maintenance records, Orsat analyses, SETs, and installation drawings. A
visual inspection was then made of the equipment and boiler room areas.
Minimum deposits on the boiler tubes confirmed that the boiler internals
had received proper lancing during the scheduled weekly maintenance period.
No corrosion was evident in either of the units. Surface temperatures were
low and no hot spots could be found indicating that the boilers were properly
insulated. Flame patterns were observed to fill the firebox with only minor
impingement occurring in one boiler. Boiler controls appeared to be adequate
and properly calibrated. The fuel oil being used during the survey was
reported to be a blend of No. 6 and No. 4 which was necessary to lower the
sulfur content to an acceptable value. During the initial inspection, the
fuel oil pressure was maintained at 157 psi for a flow rate of 240 GPH. The
fuel oil temperature was 130°F which appeared to be low for the blend of oil
reported. No viscosity measurement could be made during the allotted survey
time to verify the temperature required for good combustion.
Next, the multiple cyclone collectors were evaluated. Each collector
consists of 128 high-efficiency cyclones each having a 6-in. inside diameter
and a 33-in. total length. The vanes for turning the gas flow within the
cyclone are mounted on the 4-in.-diameter exit tube external of the cyclone
body. This arrangement allows the stack gas to enter the cyclone radially,
as well as axially. Exit angles of the vanes are approximately 45°. The
cyclones were designed for a total gas flow rate of 35,000 actual cubic feet
per minute (ACFM) at a maximum pressure drop of 4.5 in. H20 across the entire
collector. A design efficiency of 80% at these conditions is based on a
particle size distribution with a mass median diameter less than 13y.
439
-------
The first efforts in evaluating the dust collectors were to measure
dimensions for the individual cyclones and to monitor pressure drops across
the collector. With the data obtained and SET report results, a relative
evaluation could be made of the adequacy of design for these units to meet
the required emission levels. This approach was taken in an attempt to
estimate cyclone performance prior to making a decision to conduct extensive
emission tests.
The first problem encountered in design verification was the unavailabil-
ity of particle size distribution data for these particular boilers. A brief
search was made of air pollution literature to locate particle sizing for
similar boiler operations. Figure 1 shows the particle sizing data for No. 6
steam-atomized burners together with data on the outlet emissions of a
multiple cyclone installed on a similar No. 6 fuel boiler (2,3). The mass
median particle size can be seen to decrease from about 13. Oy to a value of
0.38y. The data presented for No. 2 fuel oil shows the particle sizes are
similar to those emitted from an efficient multiple cyclone.
Using the data in Figure 1 and the dimensions obtained during the field
inspection, calculations for vane-axial cyclones were made to confirm the
reported fractional efficiencies and the total mass efficiencies. Entering
the data of the upper curve into vane-axial cyclone design equations, overall
mass efficiency was calculated to be 69% which is low compared with the
specified efficiency of 80%. If the efficiency of the collector can be
increased to the 80% design value, a mass emission rate of 0.050 Ib/MBTU
could be achieved, which is well below the DC standard of 0.061 Ib/MBTU.
Observations made during the field visits uncovered more significant
problems with the dust collectors. Initial installation of the collector
at the Washington DC Facility resulted in a poor flow distribution to the
inlet of the multiple cyclone. The contractor installed a series of flow
distribution vanes, and adjustments were made based on pressure measurements
before the vanes were permanently fixed in position. The overall pressure
drop was measured as 0.4 in. H20 maximum instead of the 4.5 in. H~0 design
value. Inspection of the interior of this collector showed a hardened
buildup on the cyclone tubes and bridging of material between the cyclone
turning vanes. The material that was impacted in certain areas appeared to
be collected as agglomerates of soot. All of these factors suggest low gas
flows and periods of operation below the dew point of the flue gas constit-
uents. However, no severe corrosion or erosion was observed. Since no dust
was being collected, the hopper and its method of emptying were inspected.
The dust could be emptied through a rotary valve into a 10-ft length of 10-in.
diameter duct. The duct can then be emptied into a 32-gal container through
a slide gate valve. The rotary valve had never been checked for wear on the
rotor blades and maintenance had been limited to lubrication of the electric
motor bearings. When the slide gate was opened, dust particles were observed
to fall from the outlet and then to be pulled back up into the duct work.
Additionally, inleakage could be felt around the slide gate flange. Gaskets
on the hopper access doors were inspected and found to be in good condition.
No inleakage could be determined at any of the hopper flanges.
440
-------
Upon completion of the survey, the following observations and recommend-
ations were made using the process evaluation criteria discussed earlier. The
boilers appeared to be installed and maintained properly resulting in minimal
or no internal deterioration. The method of operation apparently contributed
very little to the emissions. The burner flame exhibited minor impingement,
but flame pattern and length was good in general. Measurements taken at the
outlet stack conditions showed moderate to high boiler efficiencies. The fuel
oil temperature appeared to be low; however, a viscosity measurement would be
required to confirm an optimum temperature setting. The boiler instrumenta-
tion appeared adequate and properly calibrated. An opacity monitor was in use,
but its location prevented representative sampling of a sufficient path length.
A camera was being used to visually monitor smoke from the stack.
The individual cyclones within the collector were determined to be
correctly designed for the assumed particle sizes; however, the low pressure
drops measured across the collector indicated the total unit was oversized
for the particular system. Additionally, the original installation and current
operation and maintenance of the pollution control equipment contributed to the
pollution problem. This conclusion was based on the following observations:
(1) scale was evident on the cyclone walls, (2) some bridging of the inlet
turning vanes was obvious, (3) inleakage was occurring through improperly
maintained valves and airlocks, and (4) the exhaust gas was being diluted
through deteriorated expansion joints at the induced draft fan inlet.
Since some problem areas would require additional time to correct, minor
burner adjustments (primarily nozzle replacement) and increased fuel oil
temperatures were tried to optimize combustion and minimize emissions. A
private contractor was used to refurbish the the burners and to conduct stack
sampling. Fuel oil temperatures were increased to 180°F. All efforts to
meet standards using this approach have been unsuccessful because optimized
boiler operations did not significantly reduce emission levels. To comply
with the District of Columbia regulations, the boilers are being operated at
a maximum load of 80%. Further work on the multiple cyclone collectors has
been postponed until installation of preheaters (for energy conservation
purposes) is completed. The suggested method for blocking the individual
cyclones is presented in Figure 2. Blocking the gas flow consists of
capping the cyclone outlet tube and the cyclone discharge tube using a
threaded rod to hold tension at both ends. This method is preferred for
this particular collector primarily because of access to individual cyclones.
Maryland Facility
The Maryland Facility has four boilers that are used for general heating
and intermittently for operating a steam injector at a research area. Three
of these were originally rated at 40 MBTU and the fourth at 20 MBTU. The
exhaust for the small unit is common with the exhaust of one of the larger
boilers. The other two boilers have separate exhausts. All units are fired
by pressure-atomized burners at flow rates of 260 GPH and 120 GPH respective-
ly. The boilers and burners were evaluated using the same approach and
techniques discussed earlier in this section.
441
-------
Review of previous SETs and boiler plant records produced a list of
several problems with the entire system. The boilers, which are 1945 vintage,
had operated fairly efficiently until the dust collectors were installed in
1977. Routine Orsat analyses had been discontinued after oxygen analyzers
were installed at the same time as the collectors. Unfortunately, the
oxygen analyzers had never operated properly and no plant records of analyses
were available regarding combustion efficiencies. SETs conducted immediately
after the collector installation and at two-year intervals reported excess
air values in the stack greater than 150%. Efficiencies ranged from 78% to
82% for boiler loads through 30 MBTU. These low efficiencies apparently
contributed to the emission levels measured during the SETs.
Review of records prior to the current effort revealed the following
facts leading to the existing conditions, all of which affect the variables
to be evaluated in order to eliminate the pollution problem. The induced
draft fans were original equipment. After installation of the cyclone
collectors, the 10-hp blower motors had to be replaced with 30-hp motors and
the blower speeds had to be increased to overcome the pressure drop of the
collectors. The increased speed resulted in excessive noise from the stack
which required installation of mufflers and a reduction in the speed to attain
acceptable noise levels. This loss in induced draft fan capacity lowered the
maximum firing rate to 30 MBTU for the larger boilers. After carefully
reviewing the existing data and taking the above facts into consideration, an
on-site inspection was made of the boiler plant using the criteria listed
initially under PROCESS EVALUATION.
Inspection of the interior of each boiler confirmed that the brickwork
and refractory were in good condition. Further inspection of the exteriors
revealed areas of severe inleakage. Inleakage was attributed to missing or
deteriorated gaskets and packing around inspection and cleanout ports,
blowout doors, steam drums, and the outlet steam manifolds due to poor
maintenance. In some instances, leakage was determined by locating cold spots
on the boiler casing. Accumulation of extraneous material was observed on
the forced draft fan blades and a considerable amount of rubbish was removed
from the tunnel supplying forced draft air to one boiler allowing higher flows
to be attained. Flame patterns were pulsating and smoke and sparks were often
observed. Impingement was observed on both walls. Orsat analyses were made
for one boiler operating at 60% boiler load. Excess air was about 190% and
the combustion efficiency was determined to be about 80%. Fuel temperatures
were between 145°F and 150°F which was considered low for the No. 6 fuel being
used. The instrumentation used for boiler control was outdated and in need
of calibration. Oxygen analyzers, opacity monitors, and the viscosity monitor
were all inoperative.
The pollution control equipment at the Maryland Facility consists of four
multiple cyclone units. The collectors for the three large boilers contain 36
high-efficiency cyclones and the collector for the small boiler has 18 cyclones.
The cyclones have an inside diameter of 9-in. and are 56- to 80-in. in length.
The turning vanes are of the vane-axial type with exit angles of approximately
45°. The cyclones were designed for a total gas flow of 22,000 ACFM at a
pressure drop of 3-in. H20 across the total collector. The design efficiency
442
-------
at these conditions was reported to be 50% for mass median particle sizes less
than 10-y; however, particle siz:e/efficiency design data could not be located
for this particular installation.
During the initial survey, the following observations were made regarding
the collectors. Internal inspection showed few deposits, no bridging or
scaling, and no corrosion or erosion. Impacted material was white and very
little soot was observed. The hopper was emptied by two slide gate valves
into a sealed 55-gal barrel. No evidence of inleakage through the hopper was
observed. Calculations for overall mass efficiencies using the No. 6 fuel
particle size data in Figure 1 resulted in a predicted collection efficiency
of 50% at existing flow conditions.
Since several problem areas required addressing, a full analysis of the
Maryland Facility was undertaken. The objective of the analysis was to reduce
emissions by modification or adaptation of existing equipment as a short term
solution. To prevent interference with plant operations, one boiler was taken
off line for modifications. The approach used included the following steps:
(1) repair monitoring equipment, (2) seal the boiler, (3) calibrate the boiler
controls, (4) modify the burners, and (5) optimize the collector efficiency.
Steps (1) and (2) were accomplished by refurbishing all instruments and
installing new gasket materials around all leak areas. In order to calibrate
the viscosity monitor, a fuel viscosity determination was made. To improve
burner operation, several modifications were made to the existing burners.
These included installing new burner tips, tightening the air register for
better adjustments, and modifying the diffusers. The boiler controls
manufacturer was contracted to optimize the boiler control instrumentation.
A check was made of the collector pressure drop. A pressure drop of 1.8 in.
H20 was measured at 22,000 ACFM instead of the 3.0 in H20 design value. This
confirmed that the collector was oversized and that the required pressure drop
would be achieved only by blocking off a certain number of the individual
cyclones. The method used for blocking individual cyclones for this specific
design is shown in Figure 3- Custom made collars were made for the cyclone
inlets. The cyclone discharge tube was capped by merely fitting a flat plate
into the existing ring. Blocking the cyclone tubes increased the collector
pressure drop significantly enough to further reduce the capacity of the
induced draft fan resulting in even lower firing rates. The compromise was
to seal only the number of tubes required to achieve the design pressure drops
and velocities while maintaining an adequate boiler firing rate. Twelve of
the thirty-six tubes were blocked and preliminary testing was conducted at a
new firing rate of 23 MBTU. A pressure drop of 2.4 in. H20 was obtained at a
flow rate of 15,000 ACFM which is near the design values predicted for the
specific cyclones. Higher flow rates were unattainable due to the limitation
of the induced draft fan.
Improvements in combustion during initial tests were attributed to lower
fuel viscosities, improved flame patterns, lower excess air values, and higher
combustion efficiencies. Some flame impingement occurred during optimization
tests; however, excess air was reduced to 33% at the boiler outlet and
efficiencies were on the order of 86%. Preliminary SET data indicated that
emissions are near or below the 0.03 gr/dscf requirement. Preliminary particle
443
-------
size data indicate that inlet and outlet sizes are approximately l.Oy mass
median diameters and collector efficiencies are very low because of these
small particles. If the preliminary particle size distribution data is
correct, low efficiencies would be expected. However, since mass loadings
are near the applicable standard, only small increases in efficiency would be
required. Installation of an induced draft fan having sufficient capacity
to return the firing rate to 37 MBTU would increase the velocity in the cyclone
to its design value and provide a small efficiency increase.
CONCLUSIONS AND RECOMMENDATIONS
Selection of air pollution control equipment requires a good knowledge of
the process gas stream to be controlled as well as a basic awareness of how
the control device operates. When an off-the-shelf device is chosen, care
must be taken to ensure that the collector is compatible with the process and
can effectively remove the pollutant. After installation, the collector
should be thoroughly inspected for proper flows, pressures, temperature
limitations, construction, and maintenance required. Guarantees for the
level of emissions should be obtained and tests should be required to confirm
the collector's efficiency prior to acceptance of the device. Care and logic
should be used in selecting the types and location of the instrumentation for
monitoring outlet stack conditions.
In view of future decisions to install economizers and possibly replace
the existing control devices, two approaches were recommended as intermediate
actions to improve the collection efficiencies for the two cases discussed in
this paper. For the Washington DC Facility, inleakage through the collector
should be stopped by ensuring that valves and fittings are adequately main-
tained and sealed. Velocities through the cyclones should be increased by
decreasing the number of cyclones. This decrease can be achieved by blocking
the inlets and outlets of certain cyclones in a manner that will not affect
the flow distribution through the overall device that could result in
channeling through adjacent cyclones. The collector should be preheated and
remain heated during startup and lancing operations to prevent scale buildup.
Additionally, the induced draft fan size should be increased commensurate with
the increased collector pressure drop. Efforts to reduce pollution at the
Maryland Facility will require increasing the velocity through the collector
to its design value. Additional induced draft fan capacity is required to
accomplish this end and to return the firing rate to its original value of
37 MBTU.
In summary, the evaluation of multiple cyclone collectors for heavy-fuel-
oil boilers should consider the following recommendations. The collector
design and installation should be commensurate with the boiler loads encounter-
ed in routine operation. Maintenance of particulate collectors is often
neglected and adequate procedures, such as maintaining temperatures during
downtimes to prevent scaling, are required. Prior to acceptance of a
mechanical collector, dimensional parameters should be double-checked to ensure
proper sizing for a given process. After installation and initial operation,
collectors should be inspected for erosion, plugging, and bridging of internal
components. Inspection for inleakage should be routine and procedures should
444
-------
be exercised to prevent excessive inleakage. Low hopper loadings are a sure
indication of inleakage through the collector. Pressure drops across the
collector should be routinely monitored to provide first indications of
collector deterioration. The inlet flow distribution to any mechanical
collector should be consistent to prevent irregular velocity patterns and
inefficient collector operation. Finally, ancillary items such as collector
insulation, instrumentation and stsck gas monitoring equipment should be
properly installed and operated to insure that the collector operation is
adequate at all times.
ENDNOTES
1. Executive Order 12088, Federal Compliance with Pollution Control Standards,
13 October 1978.
2. McGarry, F. J. and C. J. Gregory. A Comparison of the Size Distribution
of Particulates Emitted from Air, Mechanical, and Steam-Atomized Oil
Fired Burners. J. Air Poll. Control Assoc. 22:636, 1972.
3. Cheng, R. J. et. al. Characterization of Particulates from Power Plants.
J. Air Poll. Control Assoc. 26:787, 1976.
4. Navy Environmental Support Office. Source Emission Tests of Boilers
Alpha and Beta at the Naval Training Center Great Lakes, Illinois.
NESO Document No. 2-093- Port Hueneme, California, July 1978.
445
-------
100
1 10
GO
B #6 STEAM ATOMIZED
A #6 MC EMISSIONS
O #2 STEAM ATOMIZED
2 102030405060708090 98
CUMULATIVE PERCENT LESS THAN PARTICLE SIZE
Figure 1. Particle Size Distributions
446
-------
PLATE v
PLATE
Figure 2. Proposed Cyclone Sealing Method at the Washington DC Facility
447
-------
COLLAR
PLATE
56-80
Figure 3. Cyclone Sealing Method Used at the Maryland Facility
448
-------
COLLECTION EFFICIENCIES OF CYCLONE SEPARATORS
By: P.W. Dietz
General Electric Company
Corporate Research & Development
Schenectady, New York 12301
ABSTRACT
An improved model for particle collection in cyclone separators is pre-
sented. The model is predicated on the identification of three distinct re-
gions: the inlet region, the downflow region, and the core region. In each
of the three regions, turbulent mixing is assumed ro'result in a uniform radial
concentration profile. The effects of cyclone geometry on the gas flow pattern
are included so that the model incorporates the distribution of gas residence
times.
Experimental results from several sources are compared to the model pre-
dictions and favorable agreement is obtained.
INTRODUCTION
With the growing concern for the environmental effects of particulate
pollution, it is becoming increasingly important to be able to design opti-
mized pollution control devices. Unfortunately, existing models for many
control devices are inadequate for this task. For example, although measure-
ments of the fractional efficiency of cyclone separators can be used to pre-
dict the performance of a given design in new applications, the efficiency of
new cyclone designs cannot be predicted based on current theories. In an
effort to address this situation, an improved cyclone model has been developed!
In the present paper, an overview of this new cyclone model is presented and a
preliminary qualification of this new design tool is achieved through a com-
parison of the model with the measured efficiencies of cyclones tested at the
Exxon Pressurized Fluidized Bed (PFB) miniplant and at the National Coal
Board's (NCB's) PFB facility.
Historically, cyclones were characterized by a cut size (d ) which was
defined to be the particle size for which the cyclone efficiency was 50 per-
cent. This cut size can be approximately computed by equating the centrifugal
force on a particle (due to its angular velocity) to the drag on the particle
due to the radial gas velocity.(2,3) Unfortunately, this procedure does not
allow prediction of the shape of the grade-efficiency curve. Nonetheless, this
simple approach can be used to scale experimental data for geometrically sim-
ilar cyclone.(3,4) Deviations from this model have been attributed to re-
entrainment, particle bounce, and gas by-passing (sneakage). Unfortunately,
new cyclone designs must be tested before this technique can be employed.
Recent papers by Leith and Licht (5) and Dietz (1) have proposed improved
models which recognize the turbulent nature of the flows within cyclones. Con-
sequently, both models, which incorporate turbulent back-mixing of the sus-
pended particles, provide reasonable agreement with data. The Dietz model is
preferred for two reasons. First, it incorporates the distribution in gas
449
-------
residence times. Consequently, the model describes the particle by-pass
caused by short gas residence times. Second, the Dietz model more accurately
describes conventional reverse-flow cyclones because it includes the exchange
of particles between the core and annular regions. Consequently, particles
which enter the core region can be collected.
Theory
An analytic model (1) has been developed for predicting the collection
efficiency of a conventional, reverse-flow cyclone (see Figure 1). This model
is predicated on the identification of three distinct flow regions: the en-
trance region, the downflow (or annular) region, and the upflow (or core) re-
gion. Within each region, turbulent mixing is assumed to be effective at
maintaining uniform radial concentration profiles. In addition, both particle-
and gas-exchange are allowed along the entire interface between the annular
and core regions. Consequently, the model includes the effects of particle by-
pass due to short gas-residence times.
In developing the model, certain assumptions were incorporated with the
goal of simplifying the analysis while retaining the critical elements of the
cyclone system. These assumptions are as follows:
• the effective engagement length of the cyclone, D, is equal to the
actual engagement length minus half of the axial extent of the inlet,
D = S - a/2
• the effective length of the cyclone, JL, is the length of the cyclone
below the outlet duct,
jfc = H - S
Of course, if the length is greater than the natural turning length
of the vortex, (6) then
/R 2\ 1/3
A = 7.3 R t -~- )
t \ab /
• the radial velocity does not vary axially
• the tangential velocity does not vary axially
• the radial dependence of the tangental velocity is given by a modified
form of the equation for a free vortex in an inviscid fluid
where m is between 0.5 and 1.0 and is typically taken as 0.7.
particles which impact the cyclone wall are collected (consequently,
re-entrainment, particle bounce, and base pick-up are neglected).
the radius of the boundary between the core region and the annular
region is equal to the radius of the exit tube. Thus
R = R_
v t
450
-------
(This boundary can also be taken as the location of the maximum separ-
ative force (2) with little effect on the predictions of the model.)
• the drag on the particles is given by Stokes1 drag (the result can be
easily modified to incorporate other drag laws).
From this model, a set of coupled, nonconstant-coefficient, ordinary dif-
ferential equations is obtained for the particle concentration. An analytic
solution to these equations has been developed and an expression for the frac-
tional efficiency of the cyclone is obtained (the symbols are defined in the
nomenclature and in Figure 1).
where
K -
o
exp
-2 R U D
c pw
%0
RU +RU+RU
c pw v r v pv
2R U
v pv
and
where
and
RU -RU-RU
_ v pv v r c pw
1 2R U
v pv
U
9 9
2 P R;U:
p P tw
pw
U = U 1^
pv pw\ R
Ur 27TR £,
The tangential velocity at the cyclone wall, Ufcw, is obtained by match-
ing the angular momentum of the flow in the inlet to the angular momentum of
the flow in the entrance region. Because many inlet geometries are employed,
this balance will be illustrated for a scroll inlet (see Figure 2).
R
1
R
U rdr = 27T
o
vo
r2dr
where U =
vo
o a(R,-R !
1 c
451
-------
Thus,
tw
(3-m)Q
__
vo
V
(For tangential inlets, the tangential velocity at the outer wall is approx-
imately the same as the inlet velocity for many cyclones.)
Comparison With Experiment
Particulate captive efficiencies have been reported for various cyclones
operating on the Exxon Miniplant PFB Combustor and on the Natural Coal Board's
Coal Utilization Research Laboratory PFB Combustor. The dimensions of these
cyclones are reported in Table 1. The data are plotted in Figures 3 through 7
and the theoretical efficiencies are presented in the same Figures. During the
testing, the Exxon facility was operated over a range of conditions. However,
since the range was small, the data from all of the runs are presented on the
same Figures. The data have been scaled to an assumed operating condition
through the conventional, dimensionless scaling parameter
p Ud 2
The use of this scaling factor affect the data only slightly since over
three quarters of the data was corrected by 5% or less.
Discussion
In the present paper, the three-region model for cyclone performance has
been compared with cyclone efficiencies from PFB applications at the Exxon mini-
plant and the NCB/CURL facility. Based on the excellent agreement between the
predictions of the model and the experimental data, a preliminary qualification
of this new design tool has been achieved for cyclones in which the assumptions
outlined in this paper are applicable.
Notation
a
A,C
b
D
axial extent of inlet (inlet length)
intermediate variables for computing |3
radial extent of inlet (inlet width)
effective engagement length
K ,K,,K intermediate variables in computing efficiency
I
m
n
Q.,
effective cyclone length
vortex strength (0.5 < m < 1.0)
number density
volume flow rate
452
-------
R radius of cyclone
R particle radius
Rfc radius of exit tube
R radius of vortex
S engagement length of cyclone
ut tangential gas velocity (U , wall; U , vortex)
U particle velocity (U , wall; U . vortex)
P pw pv
/? exponent in solution
T? cyclone efficiency
p particle mass density
\i gas viscosity
F particle flux
^ dimensionless cyclone scaling parameter
Acknowledgment
This work was performed under DOE Contract No. EX-76-01-2357 to the
General Electric Energy Systems Programs Department by Corporate Research
and Development. Data analysis were performed by Dr. J.T. Kuo.
References
1. P.W. Dietz. Collection Efficiency of Cyclone Separators. Accepted AIChE
J., 1981.
2. C.J. Stairmand. The Design and Performance of Cyclone Separators. Trans.
Inst. Chem. Engrs, 29, 1951. p. 356.
3. A.C. Stern, K.J. Caplan and P.O. Bush. Cyclone Dust Collectors. Am.
Petrol Inst., New York, 1955-
4. C.E. Lapple. Process Uses Many Collector Types. Chem. Eng., 58, 144,
1951.
5. D. Leith and W. Licht. The Collection Efficiency of Cyclone Type Particle
Collectors - A New Theoretical Approach. AIChE Symp. Series. Series,
No. 126, Vol. 68, 1972. p. 196.
6. R. Mck. Alexander. Fundamentals of Cyclone Design and Operation. Proc.
Aust. Inst. Mining and Met., 152-153, 1949. p. 203.
7. Advanced Cleanup Device Performance Design Report (Task 4.3), Volume A
- Cyclone Theory and Data Correlation. General Electric Co., Schenectady,
NY, Report No. FE-2357-70, DOE Contract DE-AC21-76ET10377, October 1980.
453
-------
Exxon 1200-1540 F
900 KPA
TABLE 1. CYCLONE DIMENSIONS AND OPERATING PARAMETERS
NCB/CURL Cyclones
Diameter (2R)
Length (H)
Barrel Length (h)
Engagement Length (S)
Inlet Height (a)
Inlet Width (b)
Exit Tube Diameter
(2Pt)
Inlet Type
Inlet Velocity
Pressure
Temperature
Primary
Stream 1
50.8
193
66.5
+
23.4
10.7
15.2
Scroll
18.6 /sec
m
600-620kPa
840 °C
Secondary Secondary
Stream 1 Stream 2
39.1 25.4
148 102
51.6 50.8
+ 19.1
18.0 12.7
8.2 6.4
7.4 12.7
Scroll Tangential
31.4 /sec 30.5 /sec
m m
Exxon Cyclones
Cyclone
B
17.8
68.0
23.5
7.62
8.18
3.66
5.25
Scroll
45.7 /sec*
m
920kPa*
840 °C
Cyclone
C
17.8
71.2
26.7
7.62
8.18
3.66
3.51
Scroll
All lengths in centimeters
k
Typical
Van Tongern proprietary
454
-------
Figure 1. Conventional Reverse-Flow Cyclone
{Mat velocity U0>
Figure 2. Top View of Typical Scroll Inlet
455
-------
99.5
99.8
93.5
99
98
95
90
i 30
o 70
t SO
u 50
o
«: 40
a:
^ 30
«=• 20
10
5
2
1
0.5
0.2
0.1
0
Figure 3.
99.8
99.5
99
98
91
90
>•
£ 80
Ul
5 70
i eo
" 50
o
•i 40
CE
•2 3°
V 20
10
5
2
0.5
0.2
0.!
0
1 1 — r 1 1 i Ml 1 i — ! ! 1 i I II * 1 1 r • r i i i i
1 « /
- * °/
• * A /
t i /o
e /a
' 7
V
A -
/a
' *
/ LEGEND
/ TEST 1 - *
.x TEST 2 - XACO
X TEST 3 - •
DIETZ THEORY '
1- MODIFIED FOR -
SCROU. INLET _|
ACCELERATION j
1 I l 1 1 l 1 i 1 I ' 1 1 I i I M 1 1 1 1 1 1 1 i -
.1 i 10 100
PARTICLE SIZE (MICRON)
Fractional Efficiency of the Primary Cyclone on Curl Stream 1
1 ! I 1 1 1 ! 1 i 1 1 1 : ! 1 II 1 i i 1 1 I : ,
f
/
/
•
f, / % 8 - o •
i f* s * « •
* V * . * »
1*2* / **
5 «• f o i A
g 6 '
/
/ LEGEND
/ TEST 1 - *
/ TEST 2 - XADO
x TEST 3 •
— DIETZ THEORY
MODIFIED FOR ~
SCROLL INLET
ACCELERATION
1 1 1 i 1 1 1 P i 1 1 l l i l i l 1 I 1 i l i i "
-1 1 10 100
PARTICLE SIZE (MICRON)
Figure 4. Fractional Efficiency of tne Secondary Cyclone on Curl Stream 1
456
-------
999
S9.8
9?.5
99
98
95
90
>•
£ 80
UJ
o 70
It 60
" 50
I
530
e- 20
0.'
0.1
t i f 1 1 1 1 1 1
OPERATING PARAMETERS
0 5"
T I550°F
f 9.lotm
- "^
- /<
: * . • *
/
/
/ D •
/
/
o
-
-
D
1 1 1 1 i ' i I'
I i : i i i i 1 1 i i i i i i i !
/
JC •/
-------
"•9I 1 1—I—I I I I I I— 1 1—I I I I I I 1 1 1—I—MM
99.8
995i_ OPERATING PARAMETERS x
99
98
T I550°F
95
90
2
I
0.5
0.2
O.I
D 6"
150 fps
P = 9.1 aim - *
»XVoo
O x v x v v
80
70
60
50
V x
x
LEGEND
x 109 C
• III C
* O II2C
O V II5C
20
10
5
DIETZ THEORY
J I I I I I I I I I I I I I I I I | I I
O.I I 10 100
PARTICLE SIZE (MICRON)
Figure 7. Fractional Efficiency of the Exxon Cyclone C
458
-------
ELECTROSTATICALLY AUGMENTED COLLECTION IN VORTICAL FLOWS
By: P.W. Dietz
General Electric Company
Corporate Research & Development
Schenectady, New York 12301
ABSTRACT
Electrostatic forces have been demonstrated to provide significant en-
hancement in the efficiency of cyclone separators. To study these effects,
experiments have been performed in a two-dimensional vortical flow. In these
experiments, the effect of an applied electric field on collection efficiencies
has been measured for a range of flow rates for 2.02 jjm particles. These meas-
ured efficiencies are compared with an analytic model and excellent agreement
is demonstrated.
INTRODUCTION
Electrostatic augmentation of the inertial forces in conventional cyclones
has been proposed as a means for improving cyclone efficiencies. Historically,
various methods for electrostatically augmenting axial and reverse-flow cyclones
have been investigated by Molyneux, (1) Petroll and Langhammer, (2) Reif (3—5)
and Giles. (6) Although encouraging results have been claimed for these "elec-
trocyclones," this enhancement technique has yet to find significant applica-
tion. However, as specific applications are identified which are not amenable
to conventional cyclone technology, the potential of electrostatic augmentation
will be investigated. To this end, recent work (7—9) has again demonstrated
that cyclone efficiencies can be significantly enhanced through the application
of an electric field within the cyclone. This work has also demonstrated po-
tential advantages of the electrocyclone — including improved filtration during
flow transients and turn-down capability. Theoretical models (10,11) have been
developed which indicate that electrocyclones may have radically different
scaling laws than conventional cyclones. In particular, large electrocyclones
can be as efficient as small ones!
The ability to significantly enhance the collection efficiency of cyclone
separators may have an additional benefit: it may be possible to enhance the
collection of particles in the respirable size range. Because these particles
are preferentially deposited in the lungs, particles between 0.3 and 3.0pm
(the respirable size range) are of special concern. Since potentially toxic
and/or carcinogenic compounds may be concentrated in these fine particles,
significant accumulations of these particles in the respiratory tract can have
a detrimental effect on human health.
To develop an advanced cyclone which employs these electrostatic effects,
a research program has been established under DOE funding. The program has
two immediate goals: to demonstrate significant electrostatic enhancement in
reverse flow cyclones and to develop an understanding of the physical phe-
nomena which lead to the enhancement. The first goal has been achieved (7-9)
and work is proceeding to develop a practical electrocyclone for use in the
high-temperature, high-pressure environment prior to the gas turbine in a coal
fired combined cycle power plant.
-------
To address the second task, a simplified cyclone geometry is developed
(see Figure 1). The flow within a cyclone is modeled in a two-dimensional
geometry. The resulting vortical flow retains the essential features of both
the inertial and electrostatic forces on the particles without the compli-
cations of the three-dimensional flow patterns. As this system lends itself
to analysis, a collection model has been developed.(7) Thus, what is reported
here is the data from the vortex system and a comparison between the theory
and experiment.
Experimental Apparatus and Results
To study the combination of electrostatic and inertial forces in vortical
flows, a simple two-dimensional test cell was designed (see Figure 1). Particle
laden gas is introduced tangentially into a cylindrical test volume defined by
two coaxial cylinders. The outer cylinder contains eight tangential slots for
introducing the gas. The inner cylinder is constructed of sintered metal to
provide sufficient pressure drop that axial velocity components can be neglected.
The end pieces are fabricated from clear plastic so that a potential difference
can be applied between the two cylinders.
Sampling/instrumentation ports on the end piece are provided. Measure-
ments with a hot-wire anemometer indicate that the flow is two-dimensional.
In addition, the flow measurements indicate that the tangential velocity is
described by
/R
Ut = Uto
where U is the tangential velocity at the cyclone wall, R is the radius of
the cyclone and the power m = 0.7.
A Climet Particle Generator is used to disperse 2.02/^m Dow microspheres
into the inlet gas stream (see Figure 2). These particles can be electrically
charged prior to introduction into the electrovortex inlet plenum with the
main gas flow. Particle concentrations within the electrovortex are measured
with a Climet Particle Counter.
Sampling probes are designed to be nearly isokinetic. Measurements of
particle concentration indicate that 2.02 pirn particles do not require isokinetic
sampling. Similarly, the sampling of charged particles is only slightly af-
fected by probe voltage (provided the probe is not isolated).(12) Nonetheless,
sampling conditions were maintained nearly isokinetic and "iso-electric."
Particle charging is accomplished by a corona charger. The mobility
analyzer (11) indicates that the electrical charge acquired by the particles
is quite uniform. In particular, for the charger parameters employed here
the electrical charge is
q = 9.6 x 10 coul (2)
Because the particle generator displayed a slight drift in its output concen-
tration, the order in which the data was taken is also reported.
460
-------
Theory
In the laminar, two-dimensional flow of the electrovortex, the collection
efficiency is characterized by a single cut-size, d , above which all par-
ticles are collected. This cut-size can be computed from a force balance on
the particle at the position of maximum separative force, (10) R..
/iTi2 / „ \ 2m-1
R.
i
where b is the electrical mobility of the particle in an electric field at the
(2 \
13 = p Pp J , and U is the radial gas
18M /
velocity at r = R.. For typical corona chargers.
b = ° P ° (4)
-12
where 6 = 8.854 x 10 F/m is the permittivity of free space, d is the par-
ticle diameter, E is the charging field, and |U is the gas viscosity.
For the particle sizes considered here, the inertial force is negligible
compared to the electrostatic force, In this case, an explicit expression for
the cut-size can be developed.
v - T^¥ <5>
* r* f\ r
o o c
Below this cut-size, additional collection results from the inlet flow. Be-
cause this gas is introduced over a finite thickness (see Figure 7), the radial
velocity of the gas goes to zero at the outer wall. Consequently, even ex-
tremely fine particles which enter near the wall can be collected. Integration
over this entrance layer (10) gives (again assuming negligible inertial effects)
bE
n = i - B_ = _2 for d < d (6)
n U - pc
oo
This theoretical result is presented as the solid lines shown in Figures 3
to 6. (The zero voltage intercept is taken to be the average of the data.)
Discussion
The theoretical model developed by Dietz (10) predicts that electro-
statically enhanced cyclones can be substantially more efficient than conven-
tional cyclones. In addition, the model anticipates two important features of
electrocyclone performance for PFB CFCC applications. First, a large electro-
cyclone can be as efficient as several smaller cyclones. Consequently, a
single, large electrocyclone may be able to replace many smaller cyclones
(multiclone) and eliminate the associated problems of gas distribution and
461
-------
dipleg plugging. And second, the efficiency of an electrocyclone does not
necessarily degrade with flow reductions. Consequently, cyclone efficiency
will not be lost during power plant turndown.
In the present article, experimental data has been presented for electro-
statically enhanced collection of 2-micron particles in a vortical flow. Com-
plete collection of these fine (respirable) particles has been demonstrated at
moderate applied voltages. Comparison of the results with the theoretical
model yields satisfactory agreement, and this agreement is viewed as a pre-
liminary validation of the model. Consequently, the unique scaling of the
electrocyclone is substantiated.
Acknowledgment
This work was supported by DOE (Dr. Fred E. Witmer of the Environmental
and Safety Engineering Division) under Contract DE-AC02-79-EV10143.
References
1. F. Molyneux. Electrostatic Cyclone Separator. Chem and Process Eng.,
517, 1963.
2. J. Petroll and K. Langhammer. Vergleichsversuche an Zyklonabscheidern.
Freiberger Forschunsheft A-220, 175, 1962.
3. R.B. Reif et al. Final Report on Highly Efficient Dust Filtration
System for Use in Cotton Textile Mills. USDA Contract 12-14-7001-365,
1977.
4. D.P. Thibodeaux, A. Baril and R.B. Reif. A Wet-Wall Electroinertial
Precipitator: A Highly Efficient Air Cleaner for Cotton Dust. Annual
Meeting IEEE Ind. Appl. Soc., 1976. p. 333-339.
5. R.B. Reif. Electro-Inertial Air Cleaner. US Patent No. 4,010,011, 1977.
6. W.B. Giles. Electrostatic Separation in Cyclones. Proc. of the First
Symposium on the Transfer and Utilization of Particulate Control Tech-
nology, Vol. 3, 1978. p. 291.
7. W.B. Giles. Submitted for the Third Symposium on the Transfer and
Utilization of Particulate Control Technology, Orlando, Florida, 1981.
8. R.R. Boericke, J.T. Kuo, P.W. Dietz and W.B. Giles. Electrocyclone for
High-Temperature, High-Pressure Dust Removal. AIAA 19th Aerospace
Sciences Meeting, St. Louis, Missouri, January 1981. p. 12-15.
9. Novel Methods for Respirable Dust Control. General Electric Company
Report No. EV-01043-1, DOE Contract DE-AC02-79EV10143, September 1980.
10. P.W. Dietz. Electrostatic Effects in Vortical Flows. Session Ql,
Second Symposium on the Transfer and Utilization of Particulate Control
Technology, Denver, July 1979.
462
-------
11. P.W. Dietz. Electrostatically Enhanced Cyclone Separators. Submitted
Powder Technology, 1980.
12. W.B. Giles and P.W. Dietz. Electrostatic Effects on Sampling Through
Ungrounded Probes. Second Symposium on the Transfer and Utilization
of Particulate Control Technology. Volume IV Special Applications
for Air Pollution Measurement and Control, EPA-&00/9-80-039d, Sep-
tember 1980.
13. J.R. Melcher and K.S. Sachar. Charged Droplet Scrubbing of Submicron
Particulate. Environmental Protection Technology Series, No. EPA-
650/2-07, 1974.
463
-------
Vlfi at £.i;T««OTtl
IMLET
'UMM
•us^
'- i
, i
V«*E«9
MET*.
!PO*OUS,
ELEtTftbe*
wnr*
>, ?u»«»
\ '••
\ r
L-|
(—• •
1
L
-own ;-CCT»OM
»,T^ 0,ST»l»l-£5
TtME^T'Jtl. iHLETS
Figure 1 Electrovortex Apparatus (Top and Side Views)
GAS
MOBILITY
ANALYSER
SAMPLE
LINE
ELECTROVORTEX
APPARATUS
VACUUM
Figure 2 Schematic of Experimental System
464
-------
Qu « 8.t?ft!/min
Ec = 13,0 kV, lc * 3.5 x 10-'A
PARTICLE DIAMETER: iff^m
4.0 6.0
VORTEX VOLTAGE (kV)
Figure 3 Particle Concentration in Electrovortex (8.17 ft /rain)
i.Or
OL
Q, « 6.18 tp/min
Ec = 13.0 kV. lc = 3.5 x 10-!A
PARTICLE DIAMETER. 2.02,-m
2.0 4.0 6.0
VORTEX VOLTAGE (dV)
3, -
Figure 4 Particle Concentration in Electrovortex (6.18 ft /min)
A65
-------
iooo r
G, = 4.23 ftVmin
6C « 130kV, ic - 3.5 x 10'A
PARTICLE DIAMETER: 2.02nm
2.0 3.0
VORTEX VOLTAGE (kV)
Figure 5 Particle Concentration in Electrovortex (4.23 ft'Vmin)
1.8 r
O.; » 2.5 tt'/min
E,. = 13.0 kV, I. = 3.5 x IO-*A
PARTICLE DIAMETER: 2.02ym
1.0 2.0 3.0
VORTEX VOLTAGE (kV)
4.0
Figure 6 Particle Concentration in Electrovortex (2.5 ft /min)
466
-------
AIR
Figure 7 Close-up View of Distributed, Tangential Inlets
467
-------
HIGH PERFORMANCE CYCLONE DEVELOPMENT
Bv W.B. Giles,
Corporate Research and Development
General Electric Company
Schenectady, New York
Abs tract
The results of cold flow experiments at atmospheric conditions
of an air-shielded 18 inch diameter electrocyclone with a central
cusped electrode are reported using fine test dusts of both fly-
ash and nickel powder. These results are found to confirm ex-
pectations of enhanced performance, similar to earlier work on a
12 inch diameter model.
An analysis of the combined inertial-electrostatic force
field is also presented which identifies general design goals
and scaling laws. From this, it is found that electrostatic en-
hancement will be particularly beneficial for fine dusts in
large cyclones.
Recommendations for further improvement in cyclone collection
efficiency are proposed.
I n t r o d u_c t i o n
Earlier experiments found (1) a marked influence of natural
electrostatic forces in enhancing cyclone collection efficiencies,
particularly at low velocities. This naturally-occurring phen-
omena, if present, is evident as a relatively constant collection
efficiency with throughput. Evidence of this anomolous behavior
is present in the literature (2,3,4,5,6) without explanation.
Also Siemens' experience found weak influence due to both velo-
city and cyclone size (7). Experiments using a Faraday cage to
sense air-borne particle charge levels show that triboelectrie
charges are induced by particle-wall collisions. Certain dusts
are found to have a much higher propensity for this charge gen-
eration than others. For example, Exxon flyash have been observ-
ed to generate levels of 100 fold greater than CURL flyash. With-
in the cyclone, these charged particles are mutually repulsive
and the resultant space charge augments inertial separation. In
the present work, applied electrostatics are used to enhance
performance. Similar effort is found in the literature (8,9,10).
One study (2), in fact, concludes that the benefit does not jus-
tify the complication. However, for hot gas cleaning in coal-
fired power generation systems large cyclones offer an economic-
ally attractive option. Small, multicyclones pose a substantial
risk of fouling, whereas large conventional cyclones have poor
performance for fine particle collection. Thus, the objective
is to attempt to obtain, in large cyclones, the equivalent per-
formance of small, inertial cyclones through the application of
468
-------
electrostatics. In addition, the cyclonic action provides a
mechanism of dust removal from the collecting electrode surface
that precludes the problem of dust conductivity at high tempera-
ture which inhibits collection with conventional electrostatic
precipitators.
Preliminary Experiment
The general characteristics of using applied electrostatics
are shown in Figure 1. Here, a central cusped electrode was
supported within the exhaust duct to protrude down into the cyc-
lone body. When charged, this electrode provides an electric
field from the electrode to the grounded cyclone body with a cor
ona source at the four cusped edges of the electrode. The data
indicates that the application of a charge results in a signifi-
cant improvement in collection at the lower test velocities.
ISLET VELOCITY (ft/sec)
80
70
60
SO
40
30
20
10
5
2
1
0.5
0.2
0.1
10
20
40 60 SO 100
—1 1 I I I I I
Test Dust:
COM, 3rd Stage Flyash
"50
1.5 win
I
I . I
200 400 600 BOO 1200 2000
FLOW BATE (acfm)
12-INCH AIR SHIELD MODEL WITH CUSPED ELECTRODE
General Theory
In a. reverse flow cyclone, a
swirl flow is induced, and the
flow moves radially inward to the
exhaust. A centrifugal, or iner-
tia force, F£, is produced on the
air-borne particles acting against
the inward drag force, F
-------
The maximum g-field occurs at the edge of the core flow reg-
ion which is assumed here as equal to the exit duct radius or
r=D /2. The maximum E-field, however, is greater at smaller
radius,e.g. particles convected to smaller radius can become el-
ectrostatically dominated. For present purpose, the E-field will
be evaluated also at r=D /2.
e
Since the radial inflow velocity is approximately uniform,
2
The spin-up, via conservation of angular momentum, is
(1 + 22
Then equation 1 becomes
l+2x')D V. = D V , x' = x/D
!i U-^f
D (V./l
el- i
(2)
where ,
2 p d D 3
KP P e
Equation 2 indicates a minimum exists when V^/De = /k„, and
this minimum separative parameter is
V
'ii i • n r\ r\ i / i H- / -v • \
S
_ _ _ _
min 1.5U 3 D ,2 / In D/Del
The design of the electrocyclone should be such as to make
Smin as large as possible. This indicates that the desired
features are :
1. Long cyclone length, L/D, consistent with vortex stability
2. Small inlet area, x', consistent with good inlet flow swirl
turning, e.g. flow acceleration into the annular passage
3. Small exit diameter, De
4. High applied voltage, Vo , consistent with arc-over con-
s tr aint s , and
5. Large electrode diameter, Del, consistent with the internal
corona production.
In addition, cyclone operation should be at maximum allowable
velocity (subject to constraints of erosion, pressure loss and
particle bouncing). This allows minimized cyclone size, D , to
maximize performance.
*Some independent evidence suggests that in the presence of a
strong corona source the E-field may be relatively constant.
470
-------
The influence of electrostatic augmentation, relative to pure
inertial separation, SQ, is indicated by the ratio,
, V /D \
6 o j
p-p-i N- / J.+2xV\ln D/Del/
thus, there is a greater enhancement with large electrostatic
cyclones, than with small cyclones.
Test Model Design
(4)
at 3u
"1
\ Exhaust
The general design configuration was derived from Stairmand's
High Efficiency design with major modifications. An outline of
the design is shown in Figure 2. Similarity is found in the
cylindrical, conical and exhaust length-to-diameter ratios, plus
use of a small inlet area (Ai=0.lD2). In addition, the air shield
feature (12) is employed using a double scroll inlet with the
clean air inlet sized for 80% of the total flow, and a conical
section added to the exhaust inlet to increase gas spin-up.
The electrostatic features included
the electrical isolation of the lower
end of the exhaust duct with high vol-
tage supplied to a central electrode.
The original configuration, shown in
Figure 3, consisted of bundled wire and
is similar to the electrode Petroll and
Langhammer (2). It was found to result
in excessive vibration, singing, and
poor cyclone performance. The next de-
sign used a central cusped electrode
supported by crossed non-conductive rod.
Figure 4 shows the\original installa-
tion in the 12-inch diameter air shield
cyclone as used in the experiment in
Figure 1. This electrode was then in-
corporated in the 18-inch diameter el-
ectrocyclone testing program.
electrical
insulation
.2230
7
2.5D
I
FIGURE 2: 18" D ELECTROCYCLONE MODEL
The completed cold flow electrocyclone
installation is shown in Figure 5.
General Experimental Technique
The experimental procedure consisted
of supplying metered clean and, dirty flows (80%
-! IT ^i 1 TT \ 4- /-» t- Vi a T- T.T r* r» -
-------
FIGURE 3: BUNDLED WIRE
ELECTRODE
FIGURE 4: CENTRAL CUSPED
ELECTRODE IN AIR SHIELD
EXHAUST
FIGURE 5: EIGHTEEN INCH DIAMETER
ELECTROCYCLONE TEST INSTALLATION
472
-------
and outlet using two PILLS V Mass Concentration Monitors. The
other measured size distribution at both inlet and outlet using
two Royco Airborne Particle Counter Systems. The latter used
isokinetic probe sampling, followed by dilution to avoid coinci-
dene e er rors .
Results and Discuss i o n
Pressure Loss
The flow impedance of the cyclone was determined by measuring
pressure loss, Ap, versus input volume flow. The correlation
based on inlet kinetic heads to give
Ap/—pV.2 = 8.2 for D /D = —
2M i e' 3
This may be compared to the reported (11) High Efficiency
Stairmand design giving a value
Ap/ipV.2 = 6.0 for D /D = -i-
i i e 2
Scoping Experiments
Preliminary experiments were first tried with an external
corona particle charger, and separately, with an internal voltage
field. Neither were found to yield significant enhancement. How-
ever, tests with an internal corona source to produce both par-
ticle charging and an applied voltage field did show promise as
found in Figure 1. The results of a series of scoping tests
using the 18 in. diameter electrocyclone are summarized in Figure
6. These results indicated that an upstream corona source was
not significant and a positive corona was slightly beneficial at
atmospheric conditions, relative to a negative corona source.
The data shown
is plotted in terms of the overall efficiency vs.
volumetric flow times the square
of input mass mean particle size
as an indicator of the inertial
similarity parameter. CURL fly-
ash is used as the test dust and
the exhaust duct is insulated
from ground.
Inertial Performance with
Flyash and Nickel
Figure 7 shows the ex-
perimentally determined frac-
tional efficiency of the elec-
trocyclone operated in an un-
charged state with flyash; and
with nickel shown in Figure 8.
99
96
95
90
80
70
60
40
20
10
5
2
i
• 1 — ' | ' 1 ' 1 Mill ' 1 ' 1 ' 1 ' 1 1
— — -A^ -A -fifl A?
J^g- ^^
/^"^
ft®
O No Charge
— D Upstream Corona (+Bkv)
A Upstream Corona/+65Xv Electrode
^7 Upstream Corona/-65kv Electrode
TTT
~
-
-
-
-
_
-
~
~
^) No Upstream Corona/+65kv Electrode
"\) No Upstream Corona/-65kv Electrode
_
1 . . .liiiil .^ 1 i — Li 1 . 1 1-
-
-U.J-
10"
"
INERTIAL SIMILARITY
SCOPING EXPERIMENTS KITH 18" D ELECTROCYCLONE MODEL
473
-------
Correlation with the inertial separative parameter is found to
be excellant for the case of flyash. Good repeatability is noted
with replicated runs. Also overall efficiency versus mass aver-
age separative parameter (solid symbols)
with fractional efficiency data.
99.9
99.8
' 1 — ' 1 ' 1 ' 1 Mill ' T
r
' 1 ' 1 ' 1 INN ' T-
D D
D
D 0 0
' 1 ' 1 ' 1 1 1 1 1
O
O
CURL 2nd Stage Dust
Inlet Cyclone
Velocity Plow
ft/sec ft/sec
"e 0 97
D 5
O «
V 6
A i
*4 4
1319
1013
740
490
289
195
Note: Solid Symbols Mass
Effi
Aver
cienc
SEPARATIVE PARAMETER
FIGURE 7: UNCHARGED ELDCTROCYCLONE (1>"D)
is in good agreement
The data using nickel,
in Figure 8, (with
particle density taken
as 8, versus 2 for fly-
ash) shows the same
approximate agreement,
particularly for data in
the range of 21 to 75
ft/sec inlet velocity.
The mass average effic-
iency curve, however, is
at significant variance
from the fractional eff-
iciency curves. The be-
havior of the former is
suggestive of "coarse
particle bouncing." The
data suggests that this
effect is primarily de-
pendent on particle
kinetics, PpV-^2 , rather
than particle size,dp,
since the effect is not
evident in the fraction-
al efficiency data. This
is taken to infer that
performance degradation
might become important
for flyash at velocities
greater than 100 ft/sec;
FIGURE Si UNCHARGED ELECTROCYCtONE (18"D)
Electrostatic Performance
however
erences
to play
r ole .
, material diff-
may be expected
an important
The inferred fraction-
al efficiency is found
to fit the approximate
empirical expression;
1/2
(5)
Figure 9
ciency with
shows the relative influence on overall cyclone effi-
a charged central cusped electrode. The exhaust duct
474
-------
was electrically insulated from ground and hence, could float at
some intermediate voltage level. Typical current flux was meas-
ured at 0.35 ma. It is noted that performance is substantially
independent of cyclone inlet velocity and significantly superior
" operation. Figure 10 shows the same data after minor
to inertial
correction for particle size errors, associated with
sensor versus the
but ions. The
the PILLS
mass average separative parameter. Also shown
are typical measurements of inlet and outlet flyash size distri-
very close similarity of these distributions can
lead to experimental errors in
deducing fractional efficiency.
Figure 11 shows the deduced
fractional efficiency with a
charged central electrode.
Generally, it is found that
efficiency is substantially in-
dependent of inlet velocity.
Using nickel as a test dust,
as shown in Figure 12, there
is a more distinct difference
between inlet and outlet dust
distributions, and thus mark-
edly less ambiguity in measur-
ing fractional efficiency.
Again, there appears to be a
pronounced performance degra-
FIGURE 9: OVERALL EFFICIENCY AT 18" D ELECTROCYCLONE WITH CENTRAL datlOH dU6 tO Particle kinBtlC
CUSPED ELECTRODE AND FLOATING EXHAUST r
energy. This is clearly evident
in Figure 13 which shows the overall efficiency as measured by
the PILLS instrumentation versus cyclone velocity.
Theoretical Correlation
The performance data with a charge and using flyash may be
achieved by replacing the inertial term in equation 5 with the
complete separative parameter, or
1000
CYCLONE FLOW Ccfm)
ln(l-n) = -1-75 k' (l
0.5
f-)10
(6)
where
__
18y
e
D
An analytic fit may be taken from the experimental data at a
median velocity of 40 ft/sec using Tlf = 0.83 at dp = 2y and nf =
0.935 at d = 4y to give,
P
(7)
475
-------
Outlet Mass
Distribution
r
20 j-
10 r~
SI-
st Dust: CURL 2 Stage Plyash
o.s
0.2
MASS AVERAGE SEPARATIVE PARAMETER
PARTICLE SIZE 111)
wherein the particle
size is in microns and
the cyclone inlet velo-
city is in ft/sec.
The trends of this
theoretical correlation
are shown in Figure 14
over the general range
of experimentation. It
is noted that a loss of
performance is antici-
pated at increased vel-
ocity due to a weakening
of the relative influ-
ence of electrostatics
and the predicted per-
formance increases with
particle size. Equation
FIGURE 10: OVERALL EFFICIENCY 6 SA1IPLE DISTRIBUTION AT 18"D ELECTROCYCLONE TEST
7, however, predicts a higher performance level for coarse par-
ticles than found experimentally.
The indicated enhancement for electrostatic enhancement is
ln(l-nf)
14245. »
e D
o
26
V /D
o e
p d V.
p p i
D
l+2x' In D/Del
Thus, for a fixed voltage grad-
ient, geometric similarity, and
fixed inlet velocity, performance
is improved with electrostatic
augmentation at increased cyclone
scale.
For application in the PFB-CFCC
system, special interest is dir-
ected to controlling erosive
particles of the order of 5 mic-
rons and larger. Assuming cyc-
lone inlet velocities of the or-
der of 100 ft/sec, the indicated
enhancement in separative effect-
iveness (from equation 8) is in-
creased by 28% for an 18 inch diameter cyclone, or byl!2% for a 6
ft dia. cyclone. This empirically-deduced theoretical correla-
tion finds that electrostatic augmentation should be highly de-
sirable for turbine erosion control using large cyclones. How-
ever, the apparent variation of coarse particles is critically
important.
PARTICLE DIAMETER
FRACTIONALLY EFFICIENCY OF
CUSPED ELECTRODE AT «70kv I
w,TH CENTRAL
476
-------
PARTICLE DIAMETER (U)
PARTICLE DIAMETER (w)
99.9
99. B
99.5
99
98
95
Test Dust: Nickel Powder
INLET VELOCITY ft/Bee
10 20 30 50 70 90
I I . I I . ll . I I ll I ill I
FIGURE 12: FRACTIONAL EFFICIENCY 6 SAMPLE DISTRIBUTION AT 18"0 ELECTROCYCIONE TEST
100 1000
CYCLONE FLOW (cflu)
13: OVERALL EFFICIENCY OF ELECTROCYCLONE WITH
CENTRAL CUSPED ELECTRODE § FLOATING EXHAUST
Flyash
I.I I I I I
.6 .a i
PARTICLE DIAMETER
-------
Particular attention may be drawn to the ratio of flow area
at the cyclone annulus versus inlet. It is noted that excessive
flow diffusion exists with the High Efficiency configuration.
This would be expected to result in excessive flow separation
and turbulent mixing at the cyclone inlet. This situation is
also evident in the High Flow design, but to a much lesser ex-
tent. A preferred design would provide for an accelerating
inlet flow turn (or the use of axial swirl vanes) as indicated
for a recommended design shown in Figure 16. The ideal design
is intended to suggest preferred trends.
g eo
y 70
stairmand High Efficiency
Non-Shielded Cyclone
0/0 - 1/2
Uncharged,
Shielded
Electocyclone
16"D, 0 /D = 1/3
V Stairmand High Flow {11J
Son-Shielded Cyclone
D /D - 3/4
_LJ_L 1.1 LL.. ... L .-_!-.
1-U
10
T
-.45D-4
T
• 3/4D -
J_
.250
T
SEPARATIVE PARAMETER S. = •
FIGURE IS: COMPARATIVE CYCLONE PERFORMANCE DATA
Summary
-IGURE 16: BASIC RECOMMENDED DESIGN
These cold flow investigations show an electrostatic enhance-
ment particularly for fine dusts which are projected to be es-
pecially beneficial for large cyclones. Present data suggests,
however, that performance is inhibited for dense dusts and at
high velocity. It is hypothesized that the main problem is due
to an inlet flow maldistribution associated with the use of
small inlets, typical of high performance cyclones.
A c k n o w 1 e d g em e n t
This work was performed under sponsorship of the U.S.
Department of Energy, Contract No. DE-AC01-80ET17091 .
Reference s
1. Giles,W.B. "Electrostatic Separation in Cyclones" Symposium
on the Transfer and Utilization of Particulate Control Tech-
nology, Vol. Ill, Sect. B., p. 291-302, Feb., 1979.
478
-------
Inlet Type
Inlet, x/D
Overall
Length, L/D
Dust Exit,
De/D
Inlet Area/D2
rt
Annulus Area/D
Outlet, De/D
?
Outlet Area/D
Inlet Vol . /D3
Body Vol. /D3
Spin-up Ratio
Length/D
TABLE 1
BASIC CYCLONE DESIGN PARAMETER COMPARISON
(Body Diameter = D)
Stairmand
High
Ef f ic iency
Tangential
0.5 x 0.2
0
Stairmand
High Flow
Scroll
0.75 x 0.375
Basic
Air Shield
Scroll
0.75 x 0.375
Recommended
Scroll
0.9 x 0.45
4.15
1/8
1/8
1/4
Vortex Shield
Diffuser
*** Axial Length Between Inlet and Exhaust
R&D
Ideal
Annular
Turn
0.9x0.45
3/8, Dump
0.10
0.589
0.5
0.1963
0.2944
1.776
2.8
3/8, Dump
0.2813
0.3434
0.75
0.4416
0.783
1.482
2.33
3/8, V.S.
0. 2813
0.3434
0. 67
0. 3489
0.783
1.085
2.63
3/8, V.S.
0.405
0.3434
0.5(Dif . )**
0.1963
0.663
1 .772
3.90
3 / 8 , V . S .'
0.405
<0.3434
<0 . 5(Dif.)
<0.1963
>0.663
*3. 225
>3.90
-------
2. Petroll, J. & Langhammer, K., "Comparative Tests on Cyclone
Precipitators," Breiberger Forschungsheft, Vol. A220, pp. 175-
196, 1962.
3. Ludewig, H., "Cyclone Model Experiments Regarding the Effect
of the Dip Pipe Depth on Separating Efficiency and Pressure
Drop," Vol. 7, No. 8, pp. 416-421, 1958,Maschinenbautechnik.
4. ter Linden, A.J., "Investigations in Cyclone Separators," VDI
Seminar, Vol. 3, 1954, VDI Verlag.
5. Berth, W. & Trunzk, "Model Test with Water Stream Cyclone
Separator for Predetermining Removal Efficiency," Z.F. Angew ,
Mat and Mech., Vol. 30, 1950, No. 8/9.
6. Rammler, E. and Breitling, K., "Comparative Tests with Centri-
fugal Separators," Freiberger Forschungsheft, A56, 1957.
7. Verbal communication with Weber and Klein of SiemensKraftwork
Union.
8. Reif, R., U.S. Patent 4,010,011, March 1, 1977.
9. Hodson, P., U.S. Patent 2,748,880, June 5, 1956.
10. Rommel, W.R., U.S. Patent 2,594,805, June 26, 1945.
11. Stairmand, C.J. "Design and Performance of Cyclone Separators"
Trans. Instn. Chem. Engr's., Vol. 29, 1951, p. 356-383.
12. Anomy. "PBG-CFCC Development Program-Adv. Cleanup Device
Performance Design Report (Task 4.3) Vol. B-Air Shield
Cyclone Evaluation", Prep, for U.S. DoE Contract No. EX-76-C-
01-2357, Dist. Category UC-90e, FE-2357-70.
13. Bekofske, K.L."Air Shield Cyclone in Non-Shielded Configura-
tion", Letter report dated 1/24/80.
Nomenclature
D
D<
D
el
L
Ap
P
Q
qP
r
S
cyclone inlet area
particle diameter
cyclone diameter
exhaust diameter
electrode diameter
electric field
drag force
electrostatic force
inertial force
length
differential pressure
penetration = l-ri£
cyclone flow
particle charge
radius
separative parameter
Si =
Ur "
vi =
Vt =
Vo =
x =
Greek
n
11 f
P
Pr
V
e
inertial separating parameter
radial velocity at r
cyclone inlet velocity
cyclone tangential velocity
voltage differential
inlet scroll width
overall cyclone efficiency
fractional efficiency
gas density
particle density
absolute gas viscosity
permittivity of air
480
-------
AUTHOR INDEX
AUTHOR NAME
Albrecht, P.R.
Anderson, M.H.
Arce-Medina, E.
Ariman, T.
Armstrong, J.A.
Bakke, E.
Balfour, W.D.
Bamberger, J.A.
Bergmann, L.
Berlant, M.J.
Bernstein, S.
Beutner, H.P.
Bickelhaupt, R.E.
Boericke, R.R.
Bohn, R.
Borenstein, M.
Brookman, E.T.
Bump, R.L-
Bush, P.V.
Calvert, S.
Carnes, D.
Carr, R.C.
Chamberlain, H.L.
PAGE
IV-116
11-334
11-76
III-290
IV-188, IV-252
1-236
III-119
III-398
1-323
11-218
11-405
111-71, III-228
1-165
III-353
IV-344
111-90
IV-125
11-425
1-157
III-l, 111-10, IV-156
IV-135
1-118
IV-406
481
-------
AUTHOR INDEX (cont.)
AUTHOR NAME PAGE
Chambers, R. I'45
Chiang, T. III-250, III-261
Chou, K.H. IV-73
Cowen, S.J. IV-264
Crippen, L.K. 1-148
Crowson, F. Ill-438
Crynack, R.R. H-242
Czuchra, P.A. IV-55
Dalmon, J. 11-390
Demski, R.J. 1-341
Dennis, R. 1-1, III-140
Dietz, P.W. III-449, III-459
Donovan, R.P. 1-11
Drehmel, D.C. III-341, IV-210
DuBard, J.L. IV-383
Durham, M. 11-54, IV-285
Ensor, D.S. 1-176, IV-242
Eskinazi, D. III-238
Faulkner, M.G. 11-199, IV-395
Feldman, P.L. IV-3
Ferrigan III, JJ. 1-197
Finney, W.C. 11-358
Fjeld, R.A. 11-179
Fortune, O.F. 1-82
482
-------
AUTHOR INDEX (cont.) '
AUTHOR NAME PAGE
Frazier, W.F. III-171
Gardner, R.P. III-128
Gaunt, R.H. , !_216
Gehri, B.C. 1-333
Gentry, J.W. III-406
Giles, W.B. III-468
Hardison, L.C. 111-33
Harmon, D.L. IV-317
Hawks, R.L. III-221
Helfritch, D. 1-75
Henry, F. III-301
Henry, R.F. IV-63
Hesketh, H.E. IV-222
Hoenig, S.A. III-382
Hovis, L.S. 1-23
Hyde, R.C. 1-129
lionya, K. III-181, III-321
Jaworowski, R.J. 1-185
Jensen, R.M. 1-138
Joergensen, H.J. 11-370
Johnson, C.A. 1-352
Kalinowski, T.W. III-311
Kanaoka, C. III-280
Kirstein, B.E. III-373
483
-------
AUTHOR INDEX (cont.)
AUTHOR NAME
Kolnsberg, H.J. IV-179
Krishnamurthy, N. IV-232
Ladd, K. !-55' !-65
Lagarias, J.S. I"272
Landham, Jr., E.G. I"237
Langan, W.T. III-211
Lawless, P.A. H-25, 11-35, 11-44
Leith, D. I"-270
Leonard, G.L. n"120
Maartmann, S. 11-130
Mahoney, D.F. I'206
Mappes, T.E. Ill-ISO
Martin, D. IV-145
Masuda, S. "-189, 11-380
Mathai, C.V. IV-200
Mazumder, M.K. 11-160, 11-169
McCrillis, R.C. IV-306
McElroy, M.W. 1-94
McLean, KJ. I-2S5, 11-304
Menegozzi, L. 11-404
Menoher, C. Ill-Ill
Mitchner, M. H-97
Moore, W.E. IV-105
Mormile, D. IV-363
484
-------
AUTHOR INDEX (cont.)
AUTHOR NAME PAGE
Moslehi, G.B. 11-109
Mosley, R.B. n_1; n_13
Musgrove, J.G. III-193, III-201
Noonan, P.M. IV-326
Oglesby, H.S. HI-80
Ostop, R.L. 1-107
Parker, R. 111-51, IV-2
Parquet, D. III-363
Parsons, Jr., E.L. 1-303
Patton, J.D. Ill-160
Pearson, G.L. 1-120
Pedersen, G.C. 111-60
Petersen, H.H. 1-291
Piulle, W. 1-253
Potokar, R.W. III-417
Prem, A. 111-21
Presser, A.M. IV-26
Pyle, B.E. II-66
Raemhild, G.A. II-349
Reardon, F.X. III-102
Rimberg, D.B. H-262
Rinaldi, G.M. IV-95
Rinard, G. H-283, 11-295
Rubow, L.N. IV-83
485
-------
AUTHOR INDEX (cont.)
AUTHOR NAME
Rugg, D. 11-273
Samuel, E.A. H-149
Schliesser, S.P. 11-252
Semrau, K.T. HI-43
Shilling, N.Z. H-230
Smith, W.B. 1-96
Snaddon, R.W.L. IV-74
Sparks, L.E. H-314, 11-326
Spawn, P.O. IV-335
Starke, J. III-428
Stevens, N.J. 1-313
Sullivan, K.M. H-141
Tatsch, C.E. IV-353
Teller, AJ. III-393
Thompson, C.R. 11-415
Urone, P. IV-275
VanOsdell, D.W. 1-35
Viner, A.S. IV-168
Wakabayashi, A. III-332
Wang, H.H. IV-36
Wang, J.C.F. IV-373
Wegrzyn, J. IV-46
Weyers, L.L. 1-226
Wilks, W.H. IV-15
486
-------
AUTHOR INDEX (cont.)
AUTHOR NAME PAGE
Williamson, A.D. IV-297
Yamamoto, T. H-87
Yung, S. IV-1, IV-155
Zarfoss, J.R. H-208
487
^USGPO: 1982—559-092/0431
------- |