U.S. Environmental Protection Agency Industrial Environmental Research EPA-600/7-76-016
Office of Research and Development Laboratory
Research Triangle Park. North Carolina 27711 Q CtObGf 1976
CONFERENCE ON
PARTICULATE COLLECTION
PROBLEMS IN
CONVERTING TO
LOW SULFUR COALS
Interagency
Energy-Environment
Research and Development
Program Report
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S.
Environmental Protection Agency, have been grouped into seven series.
These seven broad categories were established to facilitate further
development and application of environmental technology. Elimination
of traditional grouping was consciously planned to foster technology
transfer and a maximum interface in related fields. The seven series
are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
This report has been assigned to the INTERAGENCY ENERGY-ENVIRONMENT
RESEARCH AND DEVELOPMENT series. Reports in this series result from
the effort funded under the 17-agency Federal Energy/Environment
Research and Development Program. These studies relate to EPA's
mission to protect the public health and welfare from adverse effects
of pollutants associated with energy systems. The goal of the Program
is to assure the rapid development of domestic energy supplies in an
environmentally—compatible manner by providing the necessary
environmental data and control technology. Investigations include
analyses of the transport of energy-related pollutants and their health
and ecological effects; assessments of, and development of, control
technologies for energy systems; and integrated assessments of a wide
range of energy-related environmental issues.
REVIEW NOTICE
This report has been reviewed by the participating Federal
Agencies , and approved for publication. Approval does not
signify that the contents necessarily reflect the views and
policies of the Government, nor does mention of trade names
or commercial products constitute endorsement or recommen-
dation for use.
This document is available to the public through the National Technical
Information Service, Springfield, Virginia 22161.
-------
EPA-600/7-76-016
October 1976
CONFERENCE ON
PARTICULATE COLLECTION PROBLEMS
IN CONVERTING TO
LOW SULFUR COALS
G.B. Nichols, Compiler
Southern Research Institute
2000 Ninth Avenue South
Birmingham, Alabama 35205
Contract No. 68-02-2114
Program Element No. EHE624
EPA Project Officer: Dennis C. Drehmel
Industrial Environmental Research Laboratory
Office of Energy, Minerals, and Industry
Research Triangle Park, NC 27711
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Washington, DC 20460
-------
ABSTRACT
The papers in these proceedings discuss problems encountered
in burning low-sulfur coal in electrical utility power plant
boilers. Operating experience with electrostatic precipitators,
fabric filter baghouses, and wet scrubbers for control of fly
ash stack emissions and techniques for improving performance are
described. Increasing the collection efficiency of electrostatic
precipitators for high-resistivity fly ash by adding conditioning
agents (e.g_. , sulfur trioxide, sulfuric acid, ammonium sulfate,
sulfamic acid, and ammonia) to the flue gas is discussed, as are
possible mechanisms for their action. Experiments on the re-
entrainment of fly ash as the result of rapping precipitator col-
lection electrodes are described. A mathematical model that
calculates the collection efficiency of an electrostatic precipi-
tator as a function of particle size and operating conditions is
described. A mathematical model is also presented for fabric
filter action; it incorporates a model of collection efficiency
and a non-linear model of fabric drag. Fractional collection ef-
ficiency data for sub-micron fly ash particles are given for a
baghouse. Performance data are also presented for wet scrubbers
collecting fly ash and in one experiment include the chemical
composition (e_.g_. , content of trace elements) of the uncollected
fly ash. A multiple-stream process is described for cleaning
coal to produce fractions of low and medium sulfur contents by
specific gravity differences.
1X1
-------
ACKNOWLEDGEMENTS
Sidney R. Orem (Industrial Gas Cleaning Institute), Leslie
E. Sparks (Environmental Protection Agency), Richard S. Thorsell
(Edison Electric Institute), and James H. Turner (Environmental
Protection Agency) served as Session Chairmen for the Conference.
James H. Strickland, assisted by Marilyn Bailey and Patricia B.
Meekins (all at Southern Research Institute)., was in charge of
arrangements for the Conference. Charles E. Feazel (Southern
Research Institute) supervised the preparation of the Proceedings
xv
-------
CONTENTS
Abstract iii
Acknowledgements iv
Figures vii
Tables xii
Introduction
John K. Burchard 1
Paper 1. Operating Experience with ESP Conditioning
in Relation to an Electrostatic Precipitator Upgrading
Program
Scott H. Cragle 3
Paper 2. Sulfur Trioxide Injection at State Line Station
Ronald Cook and Michael Trykoski 20
Paper 3. Discussion and Comparison of Clinch River and
Glen Lyn Plant Electrostatic Precipitators
W. J. Buchanan 30
Paper 4. The MCCS Coal Cleaning Plant at Homer City
Station
Willard A. Crandall 36
Paper 5. Electrostatic Precipitator Options for Collection
of High Resistivity Fly Ash
L. E. Sparks 58
Paper 6. Mathematical Modelling of Fine Particle
Collection by Electrostatic Precipitation
John P. Gooch and Jack R. McDonald 68
Paper 7. Techniques for Conditioning Fly Ash
Edward B. Dismukes 107
Paper 8. Rapping Reentrainment Studies
Herbert W. Spencer, III 123
Paper 9. Fabric Filtration Performance Model
Douglas W. Cooper and Vladimir Hampl 149
-------
CONTENTS (Continued)
Paper 10. Operating Experience and Performance at
the Sunbury Baghouse
Harry Spagnola and James H. Turner 186
Paper 11. Scrubber Experience at Mohave
John M. Johnson, Dale G. Jones, Alexander Weir, Jr.,
W. Carl Martin, and Seymour Calvert 208
Paper 12. Evaluation of a Particulate Scrubber on a Full
Scale Utility Boiler
D. S. Ensor, L. E. Sparks, S. Calvert, D. V. Wallon,
and K. S. Campbell 225
Metric Conversion Factors 244
VI
-------
FIGURES
Figure Page
PAPER 1
1 Optical density vs. grain loading, Montour Unit 1 ... 6
2 Apollo injection locations (temperature full load -
temperature half load) i . . 9
3 Selected resistivity data with and without LPA-402
conditioning 13
PAPER 2
1 Sulfur burning flue gas conditioning system, State
Line Unit 3 22
2 Effect of conditioning level on collection efficiency,
State Line Unit 3 23
3 Corona power input as a function of SO3 addition ... 24
4 Flow diagram, flue gas conditioning unit 26
PAPER 4
1 New multi-stream coal cleaning process, Homer City
Electric Generating Station 41
2 Effect of size on quality of 1.3 sp. gr. float
product 42
3 Effect of specific gravity of separation on sulfur
content of cleaned coal 43
4 Homer City - MCCS simplified diagram 45
5 Homer City - MCCS simplified diagram, coal cleaning
circuits 46
6 MCCS coal sulfur balance 50
7 MCCS coal heat content balance 51
8 Photo of Homer City MCCS plant during construction . . 52
9 Photo of Homer City MCCS plant during construction . . 53
PAPER 6
1 "F" as a function of ideal efficiency and gas flow
standard deviation 78
2 Degradation from 99.9% efficiency with sneakage .... 80
3 Correction factor for by-pass sneakage when Ns=5 ... 81
4 Effect of reentrainment on the efficiency of a four-
section precipitator designed for a no-reentrainment
efficiency as indicated for a monodisperse
particulate 83
vii
-------
FIGURES (Continued)
Figure Page
PAPER 6 (Continued)
5 Schematic drawing of the laboratory precipitator . . 85
6 Inlet particle size distributions from Brink data - . 86
7 Experimentally measured and ideal calculated
fractional collection efficiencies in the
laboratory precipitator 88
8 Experimentally measured, ideal calculated, and
sneakage-corrected fractional collection efficiencies
for a current density of 53.8 nA/cm2 and a gas
velocity of 0.72 m/sec in the laboratory
precipitator 89
9 Experimentally measured, ideal calculated, and
sneakage-corrected fractional collection
efficiencies for a current density of 26.9 nA/cm2
and a gas velocity of 0.72 m/sec in the laboratory
precipitator 90
10 Experimental and theoretical data for a wire of radius
1.488 x 10~3m 91
11 Effective migration velocities for a full-scale
precipitator on a coal-fired boiler 93
12 Fractional collection efficiencies for a full-scale
precipitator on a coal-fired power boiler 94
13 Computed performance curves at 5 nA/cm2 94
14 Computed performance curves at 10 nA/cm2 95
15 Computed performance curves at 15 nA/cm2 96
16 Computed performance curves at 20 nA/cm2 97
17 Computed performance curves at 40 nA/cm2 98
18 Computed performance curves for "hot" precipitator . 99
19 Measured and theoretical fractional efficiency . . . 100
20 Effective migration velocity vs particle diameter . . 100
21 Computed and measured data from pilot precipitator . 102
22 Computed and measured data from pilot precipitator . 103
23 Comparison of measured and computed fractional
collection efficiencies 104
PAPER 7
1 Electrical resistivity of fly ash as a function of
temperature or water vapor concentration 110
2 Concentration of sulfur trioxide as a function of
sampling location or temperature Ill
3 Acidity and sulfate content of fly ash as functions
of temperature and water vapor concentration . . . 112
4 Resistivity as a function of the concentration of
injected sulfur trioxide 113
vixi
-------
FIGURES (Continued)
Figure Page
PAPER 7 (Continued)
5 Concentration of sulfur trioxide as a function of gas
temperature at the outlet of a precipitator .... 115
6 Reduction of rapping reentrainment by ammonia .... 119
7 Rapidity of the effect of ammonia on the voltage of
a precipitator 120
PAPER 8
1 Block diagram of experimental layout for a rapping
reentrainment study . 126
2 Schematic of diluter for real time particle sizing
system 129
3 Extractive sampling system for real time system . . . 129
4 Percent emissions due to rapping for particles with
diameters of 1 to 20 ym 131
5 Cumulative percent distribution for rapping puffs,
rapping intervals of 12, 32, and 52 minutes,
pilot test 131
6 Particle size distributions for inlet particulate,
outlet particulate with rapping, outlet particulate
without rapping, and average effective distribution
for rapping puffs . 133
7 Fractional efficiencies as function of particle
diameter for half normal current density test.
Test location Number 3 133
8 Average efficiencies for FluiDyne pilot precipitator
for various rapping intervals 134
9 Dust removal efficiency versus time interval between
raps . 134
10 Precipitator efficiency at Test location Number 1
as a function of the rapping interval 137
11 Percent of dust collected on precipitator plates
emitted from pilot precipitator due to rapping . . 137
12 Spatial distribution of particles in rapping puff . . 139
13 Rapping puffs at the exit plane of the pilot
precipitator, upstream and downstream raps .... 140
PAPER 9
1 Schematic of n-compartment baghouse 151
2 Theoretical and experimental curves of glass fiber
filter . 158
3 Theoretical and experimental curves of Dacron-type
filter 158
4 Baghouse simulation program flow chart 165
5 Fabric filter cake, resting on fabric 166
ix
-------
FIGURES (Continued)
Figure Page
PAPER 9 (Continued)
6 Schematic of flow model for analysis of impaction
and interception 171
7 Test run No. 5: 15-bag simulation pressure versus
time graph 174
8 Test run No. 5: baghouse simulation individual flow
rate graph . 174
9 Test run No. 13a baghouse simulation penetration . . 175
10 Test run No. 13b: baghouse simulation penetration
versus time graph; single bag, nonlinear drag/
with pinholes 179
11 Test run No. 15 baghouse simulation penetration
versus time graph; single bag, nonlinear drag,
without pinholes 180
12 Test run No. 16 baghouse simulation penetration
versus time graph; three bags, nonlinear drag,
with pinholes 180
PAPER 10
1 Location of dust removal equipment 188
2 Baghouse general arrangement 190
3 Gas flow through baghouse compartments during
normal operation and cleaning ..... 191
4 Differential pressure chart - indicating a damper
malfunction on 2A baghouse 198
5 Installing filter bag 200
6 Bolometer chart - indicating a bag failure in
compartment No. 14 of No. 2B baghouse 201
7 Baghouse performance at Sunbury Steam Electric
Station „ 206
PAPER 11
1 170 MW Vertical module (4-stage TCA) 210
2 170 MW Horizontal module (4-stage) 211
3 Inlet vs. outlet grain loading, 170 MW horizontal
module 213
4 Inlet vs. outlet grain loading, 170 MW vertical
TCA module 214
5 Inlet vs. outlet grain loading, 170 MW vertical
PPA module 214
6 Effect of turndown ratio on particulate removal . . . 215
7 Effect of liquid flow rate on vertical TCA module
particulate removal 216
8 Effect of liquid flow rate on horizontal module
particulate removal 217
9 Overall penetration as predicted by impaction
collection theory 219
x
-------
FIGURES (Continued)
Figure Page
PAPER 11 (Continued)
10 Comparison of power requirements to produce a given
"cut diameter" 221
PAPER 12
1 Scrubber system 227
2 Combined penetrations for diffusion battery and
cascade impactor (December 10, 1974) 231
3 Combined penetrations of diffusion battery and
cascade impactor (December 11, 1974) 232
4 Combined penetrations for diffusion battery and
cascade impactor (December 12, 1974) 232
5 Scrubber penetrations for selected elements 234
6 Cherokee No. 3 scrubber accumulative availability
from start-up to November 1974 240
7 Cherokee No. 3 scrubber accumulative availability
from start-up to November 1974 averaged over the
previous year on a running basis 242
XI
-------
TABLES
Table Pa8e
PAPER 1
1 Design Data on Precipitators Included in PL's
Particulate Compliance Program 5
2 Montour: LPA-40 Performance 7
3 Montour No. 2: LPA-402A Performance 11
PAPER 2
1 Typical Coal and Ash Analysis Arch Mineral No. 1 ... 21
2 State Line Station Generating Units 25
3 Summary of State Line Test Results 29
PAPER 3
1 Precipitator Design Parameters 31
2 Percent of Total Coal Tonnage Delivered in Two Sulfur
Ranges 33
3 Fly Ash Analysis 34
PAPER 4
1 Participants in Homer City MCCS Project 38
2 Typical Coal Analyses 48
3 Homer City Generating Station Alternative SO2 Control
Strategies(cost comparison) 48
4 Homer City Generating Station Alternative S02 Control
Strategies(estimated materials consumed) 49
5 Homer City Generating Station Alternative SO2 Control
Strategies(estimated effluent comparison) 49
6 Ambient Air Quality SO2 Concentrations (Forecast) ... 49
7 Concurrent Research & Development Programs 54
PAPER 5
1 Electrostatic Precipitator Specific Collector Area
Required for 99.5% Particle Collection for Various
Current Densities 61
2 Partial List of Conditioning Agents 63
3 Effect of Conditioning on Efficiency 64
Xll
-------
TABLES (Continued)
Table Page
PAPER 6
1 Typical Inlet and Outlet Particle Size Data for Wet
ESP Experiments 87
PAPER 7
1 Equilibrium Distribution of Oxidized Forms of Sulfur
at Various Gas Temperatures 108
2 Reduction of Rapping Reentrainment by Sulfur
Trioxide . 118
PAPER 8
1 Percentage Contribution of Rapping Reentrainment to
Total Emissions ....... 130
2 Effect of Rapping Reentrainment on Efficiency for
Rapping Losses Accounting for 20%, 30%, 40% and 50%
of Emissions for Various Overall Efficiencies . . . 142
3 Effect of Rapping Reentrainment on Performance for
Various Reentrainment Percentages as a Function of
Number of Sections and a Constant Collection
Efficiency Per Section . . . . 143
PAPER 9
1 Technical Data - Fabric Filters Selected for
Verification ........ 159
2 Experimental Data - Fabric Filters Selected for
Verification .......... 160
3 Test Run No. 13A Baghouse Simulation Input Data . . . 176
4 Approximate Sunbury Fly Ash Size Distribution .... 177
5 Parameters for Five Boilers Using Fabric Filtration . 183
PAPER 10
1 Typical Plant Coal Analysis 187
2 Menardi Filter Bag, Style No. 601T 192
3 Cleaning/Sweep Cycle Time'Sequence 193
4 Bag Failures 202
5 Installation, Operation and Maintenance Costs .... 203
PAPER 11
1 Operating Conditions at Design Gas Flowrate 212
PAPER 12
1 Scrubber Collection Efficiency . . 230
2 Penetration of Elements Through the Scrubber .... 233
3 Particle Penetration Summary. ... 235
4 Comparison of Mobile Bed Studies 236
xiii
-------
INTRQDUCTSOfM
John K. Burchard
Environmental Protection Agency
Industrial Environmental Research Laboratory
The purpose of this conference is to discuss the problems
involved in firing low sulfur coal and to explore the impact of
various control strategies. Federal standards for power plants
were established in December, 1971, and include sulfur dioxide
and particulate matter. At that time, switching to low sulfur
coal appeared to be the answer to the sulfur oxides problem. How-
ever, switching presented the problem of transportation of coal
from areas richest in low sulfur coal, the western United States,
to areas requiring the highest usage, the eastern United States.
An additional specific disadvantage in using low sulfur coal,
in utilities equipped with electrostatic precipitators for pollu-
tion control, is a corresponding decrease in collection efficiency
because of the increase in ash resistivity. At normal precipitator
temperatures, around 300°F (150°C), fly ash resistivity is rather
strongly inversely proportional to the sulfur trioxide concentra-
tion in the emissions. The S03 concentration will depend on the
sulfur content of the coal, the boiler operating conditions, and
the chemical composition of the fly ash.
In general, therefore, high sulfur coal gives higher S03 con-
centrations and lower resistivity, and a given electrostatic pre-
cipitator will perform better on high sulfur coal than on low sulfur
coal. Therefore, switching to low sulfur coal means reduced effi-
ciency and increased particulate emissions.
This is the problem to be discussed for the next two days.
The latest data on performance analysis of electrostatic precipi-
tators will be presented as well as alternative methods of partic-
ulate collection, such as bag houses and scrubbers.
-------
PAPER 1
OPERATING EXPERIENCE WITH ESP CONDITIONING IN RELATION
TO AN ELECTROSTATIC PRECIPITATOR UPGRADING PROGRAM
Scott H. Cragle
Pennsylvania Power and Light Co.
ABSTRACT
This report summarizes the major areas of an ongoing elec-
trostatic precipitator (ESP) upgrading program at Pennsylvania
Power and Light Co. Particular emphasis is placed on the area
of flue gas conditioning including eighteen months of operating
experience with Apollo Chemical Corporation products, a planned
SO3 injection trial, and other potential agents.
Results of ongoing trials of Apollo products at Montour SES
on two 750 MW bituminous coal fired units are presented. Per-
formance data, operating problems and cost information are cov-
ered. Eight chemical formulations have been injected into var-
ious combinations of four locations in efforts to improve pro-
duct performance, solve air heater plugging problems and improve
distribution of the product on the fly ash. A significant re-
duction of emissions on a difficult fly ash has been noted al-
though only half as great a reduction as originally expected
based on tests run with a good collecting fly ash. Air heater
plugging problems have been eliminated.
The report also covers some of the practical considerations
and potential problems that can occur in evaluating proprietary
additives.
Also, the report briefly summarizes projects on six of the
Company's ESPs including work in the following areas: fuel
quality effects; flue gas conditioning; rappers; voltage controls;
sectionalizing; gas flow distribution; operating and maintenance
and additional dust collector capacity.
INTRODUCTION
Pennsylvania Power and Light Company (PL) began a series of
efforts to upgrade its existing electrostatic precipitator (ESP)
capacity in mid-1973- At that time it had become apparent that
the particulate emissions from the two new boilers at PL's
-------
Montour station did not meet consistently acceptable levels.
The Montour station has two 750 MW units with pulverized bitu-
minous coal boilers. The ESPs were built in a chevron configu-
ration and designed for a 99.4% efficiency for coals with a 1.5%
to 3.0% sulfur and an efficiency of 99% for coals with a 1.0% to
1.5% sulfur content. More detailed information is shown in
Table 1.
INITIAL PERFORMANCE TESTS (9/73 - 11/73)
An initial series of efficiency tests on Unit 1 resulted in
efficiencies from 90.9% to 95.4%. A compliance test by the
Pennsylvania Department of Environmental Resources showed emissions
9 to 10 times the emission limit of 0.1 lb/106 Btu. Another series
of outlet tests was run to determine the effect of the sulfur con-
tent of the different fuels burned at the plant and also to deter-
mine the feasibility of using the two recently installed Lear
Siegler opacity monitors as indicators of ESP performance.
The tests showed a relation between optical density and dust
flow from the precipitator (Figure 1). As a result of this test,
the monitors have become an integral tool in monitorincr precipi-
tator performance. The tests showed that the ESPs performed
within regulation with certain high sulfur fuels, but performed
poorly with the low sulfur fuel from our Greenwich Mine, a major
fuel source for the plant. The test points are numbered on
Figure 1 in the order that the tests were performed. Tests 1
through 4 show that we didn't observe the full effect of a low
sulfur coal until it had been burned for several consecutive
days. This was due to the residual fly ash in the ESP. Tests
5 through 8 indicate that the good effects from a high sulfur
coal were not noticed on the first day after burning a low sulfur
coal, but only on the second and third day. Test 8 was a lower
sulfur coal, but high levels of SO3 were reported during the
test, possibly due to a high vanadium content in the ash.
EXPERIENCE WITH THE APOLLO FLY ASH CONDITIONING AGENT
Based on the test results, we viewed flue gas treatment to be
the major solution to the ESP performance problem at Montour.
Ongoing work done in other areas is covered briefly later in the
paper. After a short study (1/74) of conditioning alternatives,
we chose Apollo (2/74) on the basis of a low first cost and a
short three month lead time for installation. We thought that
Apollo would be a quick, low cost demonstration of the effective-
ness of fly ash treatment on our problem fuels. Other alterna-
tives required large capital investments and nearly a one year
lead time to set up any demonstration.
-------
U1
TABLE 1. DESIGN DATA ON PRECIPITATORS INCLUDED IN
PL'S PARTICULATE COMPLIANCE PROGRAM
Unit
Brunner Island 1
Brunner Island 3
Montour 1
Montour 2
Martins Creek 1
Martins Creek 2
Rating, Plow,
MW acfm x 10
350
750
750
750
160
160
Research
Cottrell (1961)
Buell (1965)
Western (1969)
Western (1972)
Western (1973)
Buell (1968)
Buell (1968)
550
550
2,300
2,300
2,300
600
600
SCA,
ft2/
3 10 'acfm
180
180
204
204
204
274
274
Face Treatment
Velocity, Time, Typical
ft, sec sec S,%
4
5
5
5
5
4
4
.2
.2
.3
.3
.3
.3
.3
4.5
3.5
4.6
4.6
4.6
6.2
6.2
1.
1.
1.
0.
0.
2.
2.
5-2.
5-2.
5-2.
8-2.
8-2.
0-2.
0-2.
5
5
5
5
5
5
5
Fuels
Ash,%
12-25
12-25
12-25
12-30
12-30
12-30
12-30
-------
0.8
0.7
t 0.6
CO
w 0.5
Q
-! 0.4
<
o
p 0.3
a.
°0.2
0.1
0.0
[ALLOWABLE EMISSION RATE
I
_ I
I
— I
8l
10
SULFUR
MEDIUM SULFUR
HIGH SULFUR
0.0 0.05 0.10 0.15 0.20 0.25
GRAIN LOADING, gr/ft3
0.30
0.35
Figure 1. Optical density vs. grain loading, Montour Unit 1, November 1973
-------
Our initial plans were for the trial to last three months on
one 750 MW unit. The cost projection was $150,000 (20«/kW) to
install the system and $90,000 for three months of chemical
supply. The system has now been expanded to both Montour units
for a cumulative capital cost of about $400,000 ($200,000 per
unit or 27C/kW)• The cost of chemicals to treat both units is
$650,000 per year (0.0067«/kWh).
Performance - LPA - 40® (9/74 - 1/75)
All conclusions about ESP performance in this report are
based on data taken by PL personnel.
The Apollo product used in the initial testing was LPA-40.
This was a sulfamic acid (PL analysis) based multicomponent pro-
duct that is injected into the flue gas upstream of the boiler
economizer. Through breakdown to NHa and SO3 it is intended to
affect fly ash resistivity, agglomeration, and space charge in
the flue gas.
Because of a 3 month strike at PL, initial operation of the
Apollo system was delayed until September, 1974. We quickly dis-
covered that we were limited to injecting less than 0.1 gal/ton
of coal (25 gal/hr) because of increased pressure drop across
the Ljungstrom air heaters. A series of emission tests were run
at this feed rate, and two opacity monitors made continuous ob-
servations. Table 2 shows the improvement that occurred with
this Apollo additive using our regulated emission limit (0.1
lb/106 Btu) as a base line (Ix). Emissions were reduced by 30-
45%.
TABLE 2. MONTOUR: LPA-40 PERFORMANCE
Fuel
Ash,%
Sulfur,%
LPA-40
Emissions
Greenwich
Greenwich
Oneida
Oneida
18
18
18
14
1.0
1.0
2.5
2.5
0.1 gal/ton
6x - 9x
4x - 5x
Ix - 2x
Ix (Base)
-------
Internal inspection of the air heater showed no adverse ef
fects related to the high pressure drop. Whenever treatment was
shut off for a few days the pressure drop would return to normal,
Most likely, the plugging was due to unvaporized product on the
air heater or recombination of vaporized product to ammonium
bisulfate.
During the next phase of the test program, our efforts were
centered on reducing the air heater plugging to allow higher
treatment rates. These five approaches were taken (in chrono-
logical order):
1. Product Atomization (10/74) - Air atomized nozzles were
added to the system to assist in faster breakdown of the
product. No major improvement resulted from this change,
2. Flow Distribution Tests (11/74) - We suspected uneven
treatment of the flue gases might be a significant pro-
blem because of the location of the injection ports
(Figure 2, Phase I).. Tests run at the inlet of the air
heater showed evidence that 80% of the gas stream was
getting uniform treatment and 20% of the gas stream was
receiving lower than normal concentrations of the pro-
duct. We are not certain if this situation improves up-
stream of the precipitator. Even if it is a significant
problem, no ready solution is available to us to allow
more even treatment of the flue gas.
3. LPA - 40, Neutral Formulation (11/74) - Apollo changed
the major component of their product from sulfamic acid
to ammonium sulfate (results of PL analysis) indicating
that the new product decomposed faster. No significant
improvement in performance resulted from the change of
product.
4. Flue Gas Temperature at the Point of Injection (1/75-
4/75) - Flue gas temperature at the point of injection
is apparently critical to the application of LPA - 40.
Injecting the product into the flue gas stream at too
high a temperature results in the chemical combination
of the products with the fly ash. Also, production of
SO 3 at high temperatures is not useful because the SO 3
will convert to SO2 by the nature of the S02 - S03 chem-
ical equilibrium. Injecting at too low a temperature,
the product does not decompose as completely and as
rapidly as necessary and air heater plugging results.
Our problem was low temperature at the injection point.
At the start of the program Apollo indicated that the
ideal temperature of injection was 800-1000° F. Follow-
ing their change to the neutral formulation, Apollo
8
-------
PHASE III
(1750 - 1250 °F)
PHASE II
(2300 - 2200 °F) —
PHASE I
(950 - 750 OF)
} TO
PRECIPITATOR
Figure 2. Apollo injection locations (temperature full load — temperature half load)
9
-------
indicated that they had seen good results at injection
temperatures above 1200° F and that temperatures as high
as 1500° F would be optimum for their product. Current-
ly they believe the optimum temperature is 1250-1300° F.
Figure 2, a boiler cross section, shows the locations we
have injected LPA-40 with the associated flue gas tem-
peratures for both full boiler load and half boiler load.
Note how the temperature varies with load.
Phase I (950-750° F) - This is the original penthouse
location where we have had the best results so far. In-
jection here at the rate of 0.1 gal/ton resulted in air
heater plugging.
Phase II (2300-2200° F) - Injection above the firing
zone at high rates (0.4 gal/ton) had no effect. No
effect was seen on precipitator performance and no air
heater plugging occurred.
Phase III (1650 - 1250° F, upper ports; 1750 - 1350° F,
lower ports) - Injection into the side of the boiler
into the superheat area showed no effect at full load.
No effect was observed on precipitator performance and
no air heater plugging occurred.
Phase IV - The problem of finding an injection point with
the right temperature caused the next phase of testing.
Product was injected through the Phase I location during
periods of high boiler load and injected into the Phase
ill location during periods of low boiler load in an
attempt to keep the injection temperatures acceptable.
This approach was intended to reduce the chance of air
heater plugging while providing continuous treatment of
the fly ash at all boiler loads. No improvement in pre-
cipitator performance or in the plugging problem result-
ed from this approach.
5. Revised Apollo Chemical Formulations (5/75 - Present)
LPA 401, Catalytic Formulation - This is the LPA-40 for-
mulation with a metal added to cause faster breakdown.
This formulation caused more air heater plugging prob-
lems than the normal LPA-40.
LPA 402, Agglomerator Formulation - This formulation con-
tains an organic compound in trace quantities. Results
with this have been good. Air heater plugging has been
eliminated and ESP performance has been equivalent to
LPA-40.
10
-------
LPA 403, Combination of LPA 401 and 402 - No trial was
run due to the problems with 401.
LPA 404, Non-Ammoniated Formulation S03 Release Agent -
This was designed to produce S03 only. Results with
this chemical showed no improvements in S03 concentra-
tion in the flue gas or improvement of the in-situ
resistivity of the fly ash. ~~
LPA 405, Another Non-Ammoniated Formulation - This is
similar to 404, but has not been tried due to handling
concerns resulting from its low pH of 1.0.
LPA 402A, Agglomerator Formulation - This is similar to
402 only with a different organic compound. It is more
volatile than 402. Results with this chemical seem
identical to those with 402 and it is the formulation
presently being used.
Performance - LPA 402 and 402A
Table 3 shows results for Unit 2 with LPA 402A in the same
format as Table 2 shows results for LPA 40 on Unit 1. The Unit
2 precipitator in general performs better than Unit 1 due to
some variations in physical flow arrangement. The results in-
dicate that 402 and 402A are similar performers to LPA 40 in
the Montour boilers when injected at the penthouse location.
During April and May, 1976, a series of ESP efficiency tests
are being run on Montour 2. The tests are being run with high
sulfur coal, low sulfur coal, and low sulfur coal conditioned
with LPA-402A.
TABLE 3. MONTOUR NO. 2: LPA-402A PERFORMANCE
Fuel
Ash,%
Sulfur,%
LPA-402A
Emissions
Greenwich
Greenwich
O A
zo
1 O
0
A
0 . 1 gal/ton
1.
A
5x
- 12x
- 2x
n r -^
11
-------
Resistivity
Figure 3 plots in-situ resistivity taken by Apollo and
Southern Research Institute using a point to plane probe. These
were taken at various times during our test program. The scatter
of the data is evident in this graph and, as a result, it is
difficult to make conclusive statements about it. By looking
at selected points, we have noticed certain trends, however.
Increased sulfur content of the fuel has the expected effect
of lowerina the resistivity. Lower temperatures seem to produce
lower resistivities as expected. The school of thought at PL,
which has not been fully investigated, is that lower flue gas
temperatures may help the performance of the Apollo flue gas
treatment by lowering the resistivity. In general, for a low
sulfur fuel, it seems that the Apollo LPA 402 treatment lowers
the resistivity about one-half of an order of magnitude from
the low 1011 ohm-cm to the high 1010 ohm-cm.
At this point our data on available SO 3 in the flue gas is
incomplete. Apollo picks up most of the sulfur compounds in
hopper samples of fly ash.
It is difficult to draw conclusions from either resistivity
analysis or SO3 analysis because of the variable sulfur content
of fuel burned at Montour. Monitoring minute-to-minute changes
in the sulfur content of the coal is difficult.
GENERAL OBSERVATIONS
These are some suggestions that will be helpful to a company
considering a test program with a proprietary chemical flue gas
conditioning additive.
1. PL has found that trials planned to be quick and in-
expensive can turn out to be more involved than ini-
tially anticipated. If you are using expediency for
justification of a choice of a certain alternative,
be certain that your trial program will be expedient.
2. Prior to the start of a test program, define the ground
rules for the evaluation program.
a. What is the expected mechanism of action of the
additive?
b. What parameters will be used to measure the performance
of the additive (resistivity, S03 concentration, NH3
concentration, fly ash analysis, etc.)? Ideally you
will set up a program to measure the additive's effect
in more ways than just precipitator performance. In
this way you can determine if changes in ESP operation
are due to the additive or other parameters. Also it
will be easier to troubleshoot any lack of performance.
12
-------
1012
¥
HI
cc
1010
O —
°8o>
o®
©
260 270 280 290 300 310
TEMPERATURE, °F at measurement point
•- HIGH SULFUR FUEL - NO CONDITIONING
- LOW SULFUR FUEL - NO CONDITIONING
®— LOW SULFUR FUEL - WITH LPA 402
320
Figure 3, Selected resistivity data with and without LPA-402 conditioning
13
-------
3. Get a commitment from the vendor prior to the pro-
gram as to the "ideal" temperature and location for
injection of the product. If possible, investigate
the temperature (and its variation with boiler load)
at proposed injection locations before installing
ports and starting the trial.
4. Be sure your injection system is capable of adequate
distribution of the additive in the flue gas. Cutting
corners in this area may well limit the success of
your trial.
5. Use great care in installing the injection system and,
if possible, observe its operation (injecting water)
from inside the boiler during an outage. An unfor-
tunate error in our installation resulted in 16 econ-
omizer leaks and a six day unit outage.
6. To demonstrate the capabilities of the ESP, arrange to
burn some higher sulfur fuels and test the units under
ideal resistivity conditions.
7. If your system is installed during an annual outage,
allow the boiler to become dirty before testing emissions
and drawing conclusions on the performance of an addi-
tive product. PL's Chem Lab feels that a clean boiler
after an outage acts as a catalytic bed in the conver-
sion of SOz to SO3. During the period following an
outage the higher levels of SO 3 result in somewhat
better precipitator performance. Also the precipitator
is usually cleaner following an outage.
Also, note any modifications that are made during an
outage to the boiler such as addition of economizer
or other tube surface to the boiler. These may affect
the amount of S03 formed by catalytic action or they
may affect the temperature (thus resistivity) of the
fly ash.
8. Often, ESP performance improvements attributed to addi-
tives are due, in part, to changes in other parameters.
Data in this report was taken by PL under carefully
controlled conditions to minimize this problem.
a. During both baseline and additive testing, main-
tain uniform operating conditions in the boiler
(coal quality, excess air, gas temperature, mill
classifier setting, etc.). Variations in these
parameters can cloud your results.
14
-------
Continuous reading S02 and opacity monitors have
been a useful tool in quickly determining relative
changes in these parameters. Absolute values
from these instruments depend on monitor reliabil-
ity and maintenance.
b. Familiarize yourself with the operation of the ESP.
Use consistent set points for rappers and automatic
voltage controls (at sparking threshold) during all
testing. Observe the power levels during emission
testing to detect any abnormal conditions that may
invalidate results of that particular test.
9. At Montour we have found that changes in the type of coal
burned in the boiler do not have the full effect on the
precipitator for a number of days. Be careful not to
draw positive or negative conclusions too soon after
adjusting any operating parameter.
10. By following these guidelines and being open-minded to-
wards alternate chemical formulations and injection lo-
cations, recommended by the supplier, PL has benefited
with improved performance. Try to identify all possible
approaches before the start of the trial.
SUMMARY
The original plan for the Apollo trial was a quick, simple,
and inexpensive demonstration with hopes for impressive results.
Although the trial turned out to be long, involved, and costly
with less than spectacular results, day to day emissions at the
Montour plant have been reduced significantly through the use
of Apollo.
PL'S PARTICULATE COMPLIANCE PROGRAM (1/75 - Present)
PL's particulate compliance program is a series of trials
in 8 investigative areas on six of PL's precipitators.
Starting with the formation of a 5-man Air Quality Project
Team in January, 1975, PL started a two-year investigative pro-
gram to bring its units into full day-to-day compliance with
Pennsylvania air quality regulations. In the past, PL had
attempted to have ESPs comply by making modifications that
seemed most promising. Often, too many changes were made to a
unit at one time. This made it difficult to determine what, if
any, improvement was gained, and if the changes could be applied
to other units with similar problems. Previous work done on
Montour No. 1 is a prime example. This work included:
15
-------
1. Modification of rapper shafts and the rapping program.
2. Adjustment of electrode control set points.
3. Trials with -updated high voltage controls on sections of
the precipitator.
4. Installation of turning vanes in an attempt to improve
gas flow distribution.
5. Experimentation with adding weight to the electrode wire
tensioning weights.
6. Collecting plate realignment and wire support frame
stabilization.
7. Blending of high sulfur coal with low sulfur coal at the
mine in order to raise the overall sulfur level of the
coal.
8. Adjustments to operating parameters such as excess
oxygen and flue gas temperature.
9. Apollo fly ash conditioning trial.
The combination of concurrent changes made it extremely dif-
ficult, if not impossible, to reasonably evaluate the effect of
any single change. The intent of the present program is to pro-
duce as conclusive a result as possible about the effects of
various modifications that will be made. Table 1 lists the de-
sign parameters of the six precipitators.
This is a brief summary of the work in each investigative
area.
1. Fuel Quality Effect
The root of the problem with precipitators is often the
type of fuel burned. PL uses a wide variety of coals
ranging from 10% to 30% ash and from 0.8% to 2.5% sul-
fur. A study is being done on all of the six units to
determine the effects of the various fuels on the dust
collector performance. At the end of the study PL will
consider the economics and practicality of achieving
compliance by controlling the quality of fuel shipped
to our plants. This approach is possible by using the
capabilities of cleaning plants at our mines and coal
blending.
16
-------
2. Flue Gas Conditioning
a. Apollo - Use of the Apollo product will continue un-
less a more effective gas conditioning system is
demonstrated or additional collecting capacity is
installed at Montour. PL has an agreement with Penn-
sylvania regulatory authorities to continue any
reasonable activity that lowers emissions at Montour.
b. SO3 injection, Brunner Island Unit 1 - A Wahlco S03
injection utilizing a liquid SOz vaporizer system
will go into trial service on this unit during Au-
gust, 1976. A series of tests with the SO 3 injection
will be run to get data similar to that in Tables
2 and 3. With this data, we hope to assess how
effective such a system would be at Montour in com-
parison to the Apollo system. The total capital
cost of the trial system of this 350 MW boiler is
$445,000 ($1.28/kW) and the expected cost for liquid
SOa during the six month trial period will be about
$70,000 (.0058<:/kWh) • Any permanent installation of
this system will include a sulfur burner resulting
in $400,000 additional capital cost and a reduction
of chemical cost by a factor of 5 or 6. We expect
to have results available from this trial by October
1, 1976.
c. Trimex, Martins Creek 2 - PL's Research and Develop-
ment group is sponsoring this trial in an effort to
demonstrate the feasibility of reducing SOa emissions
from the boiler. An offshoot of this trial may be
conditioning of the fly ash. After almost five
months of testing our Chem Lab reports that the
Trimex chemical has shown no effect on the SOa emis-
sions or on the performance of the ESPs.
d. Other Additives - Although no additional trials are
planned, PL is investigating the use of triethylamine,
concentrated ammoniacal liquor, Betz proprietary
additives, and Nalco proprietary additives. A trial
with Nalco1s experimental catalytic additive was
cancelled after we had unsuccessful laboratory trials.
3. Rappers
Internal inspection of many of PL's precipitators
equipped with the vibrating rappers have shown heavy
dust buildups on the plates. On one of the two Montour
Unit 2 precipitators, the vibrating rappers have been
replaced with impact rappers. The new rappers' perfor-
mance is being evaluated.
17
-------
4. Voltage Controls
Because of the wide variety of fuels burned at PL's
plants, we have found that the automatic voltage con-
trols on many of the units are not capable of following
the varying conditions in the ESPs. Eight thyristor
controls manufactured by L. L. Little are being in-
stalled on one of the two precipitators at Brunner
Island Unit 3. We hope to evaluate any improvement in
the efficiency of the precipitator, any reduction in
burnout of wires, and the general ability of the control
to follow changes in operating conditions.
5. Sectionalizing
The precipitators at Martins Creek have a relatively
high degree of sectionalization and offer a certain
amount of flexibility in allowing quick modification of
their sectionalizing arrangement. Experiments are being
run with varying amounts of plate area per TR set to
determine what possible benefit may be gained by in-
creasing the sectionalization of the precipitators on
other PL units and future units. This work is in pro-
gress at present.
6. Gas Flow Distribution
The Brunner Island 3 precipitator and Montour 1 and
2 precipitators are built with a chevron inlet configu-
ration. There is evidence that the flow distribution
is not uniform over the face of the precipitator. Field
tests have been run to determine the flow distribution
and a model study is being conducted to determine if
improvement is possible. If modifications are made,
efficiency tests before and after the modifications
will be run to determine the effect of improvement in
flow distribution.
7. Operating and Maintenance
A complete review of operating and maintenance pro-
cedures is being conducted throughout the PL system.
Since the precipitators are not a megawatt producing
piece of equipment, historically they have received
less attention than other equipment. We are reviewing
the impact of operating parameters and maintenance on
the performance of the precipitators with our plant
personnel.
18
-------
8. Additional Collector Capacity
At the conclusion in early 1977, should the efforts
in the seven previously listed investigational areas
fail to bring Brunner Island 1 and 3 and Montour 1 and
2 into compliance (Martins Creek 1 and 2 are in compli-
ance) with air quality regulation, additional collector
capacity will be installed. The ground work in this
area is being done at present to allow for the shortest
lead time in purchasing and constructing any additional
capacity that may be necessary. The alternatives that
are being studied are hot precipitators, cold precipita-
tors, fabric filters, and particulate scrubbers.
19
-------
PAPER 2
SULFUR TRIOXIDE INJECTION AT STATE LINE STATION
Ronald Cook and Michael Trykoski
Commonwealth Edison Co.
INTRODUCTION
The passage of regulations in the late 1960's and early
70's regulating sulfur dioxide (S02) emission caused many util-
ities including Commonwealth Edison to switch to low sulfur fuel
as a means of compliance. The burning of low sulfur (approxi-
mately 0.2 - 0.5%) coal in boilers designed for medium sulfur
(approximately 3 - 4%%) coal created a host of problems, the most
severe ones being boiler slagging and fouling, carbon carryover
with its resultant ash hopper and air heater fire hazard, and
electrostatic precipitator degradation. This paper describes
Commonwealth Edison's efforts to overcome the last of these pro-
blem areas, electrostatic precipitator degradation, through the
use of sulfur trioxide (SO3) flue gas conditioning.
THE PROBLEM
Electrostatic precipitator operation depends on the flow of
an electric current between the electrodes of the precipitator.
This flow, in the form of ionized particles, must pass through
the layers of fly ash which are normally present on the elec-
trodes. For hiqh sulfur coal, the fly ash has a resistivity of
approximately 1010 ohm-cm, while for the low sulfur, low sodium
western coals presently being used by Edison, the value is 1013
ohm-cm. An analysis of a typical low sulfur coal being burned
in Edison boilers is shown in Table 1. Thus it can be seen that
low sulfur coal fly ash, with its higher resistivity, will re-
duce the current flow between the electrodes, and therefore the
precipitator performance, when compared to operation on high
sulfur coal.
20
-------
TABLE 1. TYPICAL COAL AND ASH ANALYSIS
ARCH MINERAL NO. 1
"AS RECEIVED" COAL ANALYSIS (% BY WEIGHT)
Moisture 10.8 - 12.3
Sulfur 0.52 - 0.86
Ash 10.94 - 14.43
Gram-cal/g 5,358 - 5,654
(Btu/lb) (9,644 - 10,176)
ASH ANALYSIS (% BY WEIGHT)
Silica
Alumina
Iron oxide
Titanium oxide
Calcium oxide
Magnesium oxide
Potassium oxide
Sodium oxide
31.6 -
16.9 -
10.6 -
0.5 -
15.3 -
3.0 -
1.0 -
0-64 -
39.7
19.0
18.8
0.7
18.8
3.6
1.6
0.74
Evaluation of the Edison generating system showed that a
total of 22 precipitators on ten generating units ranging in
size from 120 to 620 MW would suffer degradation while burning
low sulfur coal to the point where they could no longer meet
particulate emission standards while operating at full load. A
program was therefore begun to evaluate flue gas conditioning
as a means of restoring precipitator efficiency through the low-
ering of fly ash resistivity.
SELECTION OF APPROACH
Various flue gas conditioning agents have demonstrated the
ability to reduce the resistivity of fly ash. In the early
1970's the most promising appeared to be the use of S03 pro-
duced by one of the following methods:
1. Evaporation of liquid SO 3
2. Catalytic conversion of SOz to SOs
3. Vaporization of sulfuric acid
4. Burning of elemental sulfur followed by the catalytic
conversion of SOz to SO 3
After evaluating the available methods using criteria such
as installation and operating cost, reliability, safety, and
current generating station work practices, Edison elected to
install, for further investigation, a pilot sulfur burner at its
State Line generating station in Hammond, Indiana, and a full
scale liquid SOa system at its Waukegan Station in Waukegan,
Illinois.
21
-------
The State Line system went into service in May, 1973, an d
is shown schematically in Figure 1. The system consisted of a
li^iiiH Qui fnr feed to a an te burner where sulfur dioxiae is
.
sulated manifold to the probe and nozzle system loca ted inter
nally in the ducts leading to the electrostatic precipitator .
It was obvious soon after startup that the burner portion
of the system was undersized and did not have the capacity to
adequately condition the fly ash from the low sulfur Wyoming
coal which was currently being burned at State Line. Therefore,
when extensive testing of the system was performed in November,
1973, additional SO2 from an external source was added at a
point prior to the electric heater.
LIQUID SULFUR
STORAGE TANK
SO3 INJECTION PROBES
LLU U U
371 °C
+ CONVERTER
BY-PASS
ELECTRIC
HEATER
BLOWER
BURNER
BY-PASS
554°C
1ST STAGE
CONVERTER
2ND STAGE
CONVERTER
454°C
S03
COOLER
Figure 1. Sulfur burning flue gas conditioning. System State Line Unit 3
22
-------
Test results showed a dramatic improvement in precipitator
performance with flue gas conditioning. Precipitator efficiency,
shown in Figure 2, increased from approximately 85% to 96% with
an SO3 injection rate of about 40 ppm, while corona power input,
shown in Figure 3, increased ten-fold from about 25 watts/1000
acfm to about 250 watts/1000 acfm.
100rTTT
95
ss
>
75
i M 1 1 1 1 1 1 1 1 II ii • ii 1 1 1 1 1 1 1 1 1 1 1 1 i
FTT.
© 149.0 -154.4°C
0144.4- 145.0°C
-i i i i i i i i 11 I 11 i M I I I I I I I 11 I I 11 I 1111 11 I I 11 11 I I 11 11 I I I i 11
10 20 30 40
SO3 CONCENTRATION, ppm
50
Figure 2. Effect of conditioning level on collection efficiency, State Line Unit 3
23
-------
[ J_ J
POWER LIMIT FOR 230 MW (850,000 acfm)
BASIS-44 kV AND 6500mA
100
V)
UJ
O
-------
PROCUREMENT OF FULL SIZE SYSTEMS
Edison's Station Mechanical Engineering Department conduct-
ed an evaluation of available systems and vendor capability and
selected Wahlco, Inc., of California to build the first 11
systems required. These were installed at the State Line Sta-
tion, located in Hammond, Indiana, which had the earliest compli-
ance date. Table 2 lists the State Line generating units in-
volved and sizes.
TABLE 2. STATE LINE STATION GENERATING UNITS
Unit
1
2
3
4
MW (net)
171
140
190
318
No. Of Boilers/Precipitators
6
3
1
1
The Wahlco system, shown schematically in Figure 4, con-
sists of a sulfur burner followed by a single-pass catalytic
converter. Incoming combustion air is preheated by electric
heaters, which are cycled to maintain proper combustion tempera-
ture. From the converter, the S03 kis piped in insulated lines
to injection probes located in ductwork as far ahead of the pre-
cipitator as practical to allow maximum time for mixing. Un-
fortunately, with some of the older units, sufficient space was
barely available and considerable "shoe-horning" had to be done.
The system will follow boiler load and automatically modu-
late the liquid sulfur feed rate. The system is essentially
"fail-safe" with a malfunction indication tripping the sulfur
feed which results in system shut down as soon as residual sul-
fur already in the burner is consumed, which takes approximately
10 minutes, depending on what the burning rate had been. The
system is designed to operate without an operator except for
startup and normal shut down. Parallel sets of status lights,
temperature recorders, and trip buttons are installed in each
boiler control room and associated burner location.
The sulfur burner systems themselves were located through-
out the generating station in as close proximity to the point of
injection as possible to minimize heat loss in the piping and
to limit the exposure of SO3 piping from a safety standpoint.
One of the greatest problems with any SO3 injection system is
maintaining proper gas temperature to avoid corrosion problems;
as a precaution, thermocouples were attached to the probes
themselves.
25
-------
LIQUID SULFUR
250 - 300°F
CONTROLLED TO
800 - 825°F
AMBIENT
AIR IN
LIQUID
SULFUR
STORAGE
SULFUR
BURNER
BOILER FLUE
CONVERTER
AIR/S03
800°-1100°F-
INJECTION PROBES
TRANSFER PUMP
(OPTIONAL)
METERING
PUMP
CONDITIONED
FLUE GAS TO
PRECIPITATOR
Figure 4. Flow diagram, flue gas conditioning unit
The two liquid sulfur storage tanks of 190 tons capacity
for units 1 and 2 and 140 tons capacity for units 3 and 4 are
located at grade level outside the building. The tanks are
steam heated and the liquid sulfur lines to the various burner
locations are steam traced. The sulfur feed pumps are located
adjacent to the tanks with additional metering pumps installed
at each burner on the multiple boiler installations for units
1 and 2. The separation of the sulfur burner from the storage
tank presented somewhat of an engineering challenge in the case
of unit 4, where the burner skid was located on the roof approx-
imately 250 feet above grade.
STARTUP PROBLEMS
As with any new system, a certain number of problems can be
expected during startup and de-bugging. The SO 3 systems at
State Line were no exception; however, the supplier has made
every effort to solve and correct the problems associated with
his equipment.
Most of the initial problems were with the sulfur pumping
system. Liquid sulfur is a touchy thing; below 246° F it solid-
ifies and above 310° F it becomes extremely viscous. When thaw-
ing "frozen" sulfur lines, an allowance must be made for the
expansion of the material - this was discovered after several
pump diaphragms were ruptured while bringing the system into
26
-------
operation on several occasions. The unique temperature proper-
ties of sulfur also made it necessary to change from electri-
cally heated to steam jacketed pumps for better temperature con-
trol; to add additional steam tracing and insulation on pump
piping and valving; and to establish specific startup and shut
down procedures to insure that the sulfur was in the liquid
state before starting the pumps. Problems were also encountered
with the pump hydraulic oil breaking down under the 290° F
operating temperatures and with the air bleed system on the hy-
draulic circuit.
As the pumping system problems were being resolved and the
burners began accumulating operating hours, several new problems
appeared. Numerous failures were occurring in the electric air
preheater elements without apparent reason. It was eventually
determined by modeling studies that the U-shaped calrod type
heater elements were developing hot spots under certain air flow
conditions which resulted in premature failure. Turning vanes
and baffles were installed in the ductwork to keep temperatures
within limits. It was also necessary to install additional air
flow sensors and interlocks to insure that the heaters were de-
energized when a no air flow condition occurred. The need for
these was discovered after 300 kW of heaters were burned out by
a failure to de-energize following a unit shut down.
It was also necessary to revise the steam supply to the
sulfur tank heaters and steam tracing system. The initial de-
sign for the steam supply utilized a manifold arrangement so
that steam could be obtained from either one of the newer units,
3 and 4, or an older unit, 2. Unit 1 was not included because
it was generally shut down every night. An unexpected long term
outage of Unit 2 last fall, coincident with a planned overhaul
outage of Unit 4, placed the burden of steam supply on Unit 3.
Several tripouts of Unit 3 during this period therefore resulted
in a shutdown of the sulfur burners on Unit 1 which then requir-
ed a load reduction because of excessive particulate emissions.
Unit 1 has been subsequently interconnected to the steam supply
manifold.
It was recently discovered that the insulation on the in-
jection probes was being eroded by the fly ash particles in
certain locations. This was detected by changes in the readings
from the thermocouples on the probes, which were described ear-
lier. The insulation has been replaced and baffles installed
to protect the probes. Since this modification was only recent-
ly completed, we are unable to say whether it will resolve the
problem.
Generally, the sulfur burner and catalytic converter por-
tion of the system have proved to be very reliable. The systems
have followed boiler loads well and have demonstrated turn down
ratios of 10 to 1, except on the smaller Unit 1 burners, which
are still experiencing some sulfur metering pump problems.
27
-------
TESTING
To demonstrate to the Hammond, Indiana, Air Pollution Con-
trol Department that the flue gas conditioning systems were
effective, an extensive particulate emission testing program
was undertaken. First, it was agreed that testing was only
to be done while burning the low sulfur coal which appeared to
cause the most severe precipitator degradation of all the low
sulfur coals which were being shipped to the station. This re-
quired special coal shipments and stockpiling of the coal on
station property. It also caused some delay to the testing pro-
gram while supplies were being accumulated. It did, however,
demonstrate performance of the systems under worst case
conditions.
Second, the Hammond Agency required three tests at the
same load to demonstrate compliance at that point. Because of
the physical layout of the ductwork at State Line, it was not
feasible to use EPA Method 5 for particulate testing. Ap-
proval was given to use a modified ASME method utilizing a
0.3 ym Gelman in-stack filter as a substitute only after con-
currence was obtained from the Indiana Air Pollution Control
Department and the U. S. EPA Region V office. In addition, be-
cause velocities could not be measured while sampling, it was
necessary to conduct extensive pre-test velocity studies to
verify that flow rates and patterns did not change materially
with time.
After completing a satisfactory series of particulate emis-
sion tests on a given unit, the S03 system was then shut off
and the precipitator allowed to de-condition for about three
days. Particulate emission tests were again performed to
accurately assess the improvement in performance due to the con-
ditioning system and to determine what unit de-rating would be
necessary during outages of the conditioning system in order to
maintain compliance with particulate emission standards.
To date, seven of the 11 boilers at State Line have been
tested. It has not been possible to complete testing of the re-
mainder due to boiler-turbine related problems. For example,
Unit 2 has been out of service for turbine repairs since Septem-
ber, 1975. Of the seven boilers tested, only one, Unit 3, was
unable to meet emission standards and has had to be de-rated
from approximately 244 to 226 MW (gross). This is through no
fault of the SO3 system, but due primarily to the fact that
Unit 3's precipitator and control package were constructed in
1955 and were not designed to meet the current strict emission
limits. The test results to date are summarized in Table 3.
In addition, during the particulate emission testing, flue gas
analysis for S03 showed no significant difference in S03 levels
28
-------
between the precipitator outlet and a point prior to the SO 3
injection probes.
TABLE 3. SUMMARY OF STATE LINE TEST RESULTS
Boiler
1-1
1-2
1-3
1-5
1-6
3
4
T^mil Q G i on
Limit
(lb/106Btu)
0.24
0.24
0.24
0.24
0.24
0.26
0.24
Actual
"FTTTI o G -1 r\r\ a
At Full Load
With FGC (lb/106Btu)
0.09
0.10
0.14
0.13
0.10
0.26*
0.06
Efficiency, %
With
FGC
97.9
97.1
96.5
96.5
97.0
95.1*
97.0
Without
FGC
__
86.9
-
43.8
-
85.9
88.4
* Unit 3 had to be derated from 244 MW to 226 MW (gross) to
meet the emission limit.
CONCLUSIONS
The total cost of the 21 flue gas conditioning systems
being installed on the Edison system is approximately $11.5
million, of which the State Line installation represents $3.8
million. Installation of the flue gas conditioning systems
will be complete on all units by July 1, 1976. Testing has be-
gun on some of the other completed units and preliminary re-
sults show improvements in precipitator performance similar to
that obtained at State Line.
While the flue gas conditioning systems have not alleviated
the other operating problems associated with the burning of low
sulfur coal which still plague Edison's generating units, they
are enabling these units to meet particulate emission standards
with minimal de-rating at a cost far less than that of the only
other feasible alternatives - major precipitator rebuild or a
new precipitator.
29
-------
PAPER 3
DISCUSSION AND COMPARISON OF CLINCH RIVER AND
GLEN LYN PLANT ELECTROSTATIC PRECIPITATORS
W. J. Buchanan
American Electric Power Service Corporation
ABSTRACT
This paper reviews some of the factors that could affect
precipitator performance and discusses how several differences
may account for the ease on one hand and the difficulty on the
other in achieving satisfactory performance. To demonstrate
this, a comparison is made between the Clinch River and Glen
Lyn 6 precipitators.
INTRODUCTION
The principal point of interest in regard to these two Vir-
ginia precipitator installations is that in using low sulfur
coal at both plants, Glen Lyn 6 has performed well, with ease,
while at Clinch River considerable effort has been required in
order to meet performance requirements.
Glen Lyn 6 is a 225 MW pulverized-fuel fired dry-bottom
steam generator, while at Clinch River there are three compara--
ble 225 MW units. Gas flow and temperature conditions are com-
parable, i..e_., 815-860,000 acfm at about 250-300° F to the pre-
cipitators at Clinch River and 800-850,000 acfm at 270-305° P
for Glen Lyn.
Both the Glen Lyn and Clinch River precipitators were con-
servatively designed. They have a specific collecting area of
800 ft2/1000 Cfm and a design gas velocity of approximately four
feet per second at 900,000 acfm.
DESIGN-COMPARISON
A comparison of the Glen Lyn 6 and Clinch River precipitator
designs shows differences which favor the Clinch River installa-
tion (Table 1) .
30
-------
TABLE 1. PRECIPITATOR DESIGN PARAMETERS
1.
2.
3.
4.
5.
6.
7.
Design Efficiency*, %
Collecting Surface, ft2
Collecting Rappers
Collecting Surface
per Rapper, ft2
Corona Electrodes
Corona Electrode Rappers
Corona Wire Length
Glen Lyn
99.8
725,760
384
1,890
15,552
96
4,860
Clinch River
99.7
722,250
504
1,433
15,840
192
2,475
per Rapper, ft
Fields in Series
9. Transformer
Rectifiers
10. Collecting Surface
per TR Set, ft2
11. Collecting Electrode
Height, ft
8-9 ft 4 in.
48
15,120
2-6 ft, 6-9ft
Total - 8
48
1st two fields,
10,938
30
3rd through 8th fields,
16,413
30
* With any one field out of service
PERFORMANCE EXPERIENCE AND IMPROVEMENT PLAN
The Glen Lyn precipitator has required only normal atten-
tion since its initial start-up. After fine-tuning the power
supplies and rapping system, there has usually been no visible
emission.
On the other hand, the Clinch River precipitators experi-
enced marked deterioration in performance six weeks after a
promising start-up. Only one of the three units is on a stack
by itself—this unit had no visible emission when first started,
31
-------
In attempting to solve the performance problem at Clinch
River, which is related to the high resistivity ash, there has
been a continuing effort to increase the applied voltage and/or
corona current. Two approaches have been taken they are:
1. improve rapping,
2. optimize the electrical energization.
Reliability problems with some ash handling system compo-
nents are being corrected. Their effect on precipitator perfor-
mance is not known.
REVERSE IONIZATION, SYMPTOMS, ALLEVIATION
There were symptoms at Clinch River which indicated that re-
verse ionization, as associated with high resistivity ash and a
thin deposited layer, was an important factor in this case of
performance difficulty.
First of all, the corona electrodes were experiencing ab-
normal ash buildup, which is frequently a result of reverse
ionization. This symptom has been largely eliminated by replac-
ing all of the magnetic vibrators with heavier, more reliable,
pulsating pneumatic rappers. However, back corona continues to
have an adverse effect when the current density is above the
acceptable limit.
In addition, console indications suggested reverse ioniza-
tion. To confirm the presence of back corona, several trans-
former rectifier sets were monitored by making temporary pro-
vision for measuring the secondary voltage. Corona voltage-
current curves plotted for these sections revealed severe re-
verse ionization. By improving the match between transformer
rectifiers and load, the back corona has been reduced.
Although it was known that the Clinch River ash was of high
resistivity even before the precipitator was designed, its in-
situ resistivity was not determined until April, 1976. Southern
Research Institute was then asked to make the necessary tests.
As^suspected, the values were high, being 1012 ohm-cm at 290° F,
1011 ohm-cm at 235° F and having a peak resistivity of slightly
over 10 ohm-cm at 300° F. This resistivity data, together with
previous dew point test results, confirmed the ability to oper-
ate safely at temperatures as low as 250° F to the precipitator.
A reduction in flue gas temperature, according to the tempera-
ture-resistivity curve determined by Southern Research Institute,
should result in a reduction in ash resistivity as well as stack
heat loss.
According to Figure 18 in an EPA report on a mathematical
model of electrostatic precipitation1, it should be expected
32
-------
that average corona current density must be limited to approxi-
mately 10-20 yA/ft2, or less, to prevent back corona. To oper-
ate above this limit would most likely mean no practical gain in
performance. It seems very likely that average current densi-
ties of 30 to 40 yA/ft2, which occur at times in some fields,
are only indications of stable back corona which affects per-
formance adversely.
Based on our experience at Clinch River and other plants,
not including Glen Lyn, when high resistivity ash is likely to
be collected, it is of considerable value to install complete
secondary voltage and current monitoring and not simply current
indication. To be able to plot secondary voltage-current curves
for any field at any time is of great practical importance when
trying to analyze performance.
COAL SULFUR COMPARED
For the first three months of 1976, coal deliveries to
these two plants reflect that Glen Lyn normally received a high-
er sulfur content coal (Table 2).
TABLE 2. PERCENT OF TOTAL COAL TONNAGE
DELIVERED IN TWO SULFUR RANGES
Sulfur, %
0.60 - 0.80
0.90 - 1.0
Glen Lyn,%
42
58
Clinch River, %
82
13
Also, the largest deliveries of coal to Clinch River have
lower than the overall ^average of 0.71% sulfur coupled with an
ash content of 17,. 3%. Thus, the combination of low sulfur and
high ash at Clinch River aggravates the performance problem.
ASH CHEMISTRY COMPARED
At both Glen Lyn and Clinch River, the coal sulfur averages
less than 1% and the other elements, which affect resistivity
are shown in Table 3.
33
-------
TABLE 3. FLY ASH ANALYSIS
C1inch River, % Glen Lyn,%
Si
Al + Ti
Fe
Ca
Mg
K
Na
Free Lime
50.0
28.5
6.5
9.8
1.2
1.5
0.6
4.64
56.9
30.9
6.7
0.6
1.6
2 .0
0.3
0
Note the significant difference in values of calcium and free
lime.
FLY ASH CARBON CONTENT CONTRASTED
Although the exact effect is not known, there is probably
some advantage at Glen Lyn, as far as precipitator performance
is concerned, in having about 10% carbon uniformly distributed
and combined with the fly ash, while the ash at Clinch River
has a very low carbon content.
GAS VELOCITY DISTRIBUTION COMPARED
Another factor that may be of significance is the gas veloc-
ity distribution. The usual pre-operational tests were con-
ducted with the result that at Glen Lyn the RMS deviation was
13.67% and at Clinch River the RMS deviation was 16.
At Glen Lyn the velocity distribution standard used was
for 85% of the velocity readings to be within plus or minus 10%
of average velocity and no readings to be outside of plus or
minus 25%. At Clinch River the standard used was an earlier
one calling for 85% within plus or minus 25% of average and
none outside of plus or minus 40%.
INLET DUCT ASH FALLOUT
After a relatively short time in service, all of the Clinch
River inlet plenum turning vanes became filled with ash. This
plenum on each precipitator is a box approximately 70 ft wide,
28 ft high and 14 ft deep. It serves to turn the gas from a
downward vertical flow to horizontal while expanding about five
times.
34
-------
The turning vanes have vertical and horizontal sides with
an eight-inch radius corner between. When the inlet turning
vanes build up with ash they then present a flat surface which
slopes downward toward the single diffusion baffle at the pre-
cipitator entrance by 30 or more degrees.
The problem of ash fallout in the inlet turning vanes is
not unique to Clinch River. Glen Lyn is experiencing it, too,
but the ash has accumulated at a lower rate. The point being
made here is that this condition has a detrimental effect on
performance and could be expected to reduce any margin of
operating reserve.
EMISSION TEST AND SUMMARY
Even though all planned improvements have not been completed,
preliminary tests have confirmed that at Clinch River a level of
performance has been reached that satisfies the state's emission
regulation.
At these two installations, i_.e_. , Glen Lyn and Clinch River,
we see relatively minor differences in several areas which add
up to a difficult situation requiring an unusual effort to
correct.
REFERENCE
1. Gooch, J.P., J.R. McDonald, and S. Oglesby, Jr. A Mathematical
Model of Electrostatic Precipitation. EPA-650/2-75-037,
NTIS PB 246188/AS, U.S. Environmental Protection Agency,
Washington, D.C., 1975.
35
-------
PAPER 4
THE MCCS COAL CLEANING PLANT AT HOMER CITY STATION
Willard A. Crandall
GPU Service Corporation
ABSTRACT
The MCCS coal cleaning plant, now under construction at Homer
City Station, represents the implementation of several unique con-
cepts in coal cleaning technology - providing the means to utilize
available coal supplies in the generation of electric power in
compliance with currently applicable emission control regulations
for both S02 and particulate matter.
This paper discusses the process design and its application
to other installations where other fuel supply and environmental
considerations prevail. Concurrent with the basic process develop-
ment is the adoption and/or development of the latest state-of-
the-art technology in analytical instrumentation and process con-
trol—required for successful operation of the plant. Also con-
current are studies on the effect of coal cleaning, both the MCCS
and other advanced processes, on the coal properties in respect
to furnace combustion, electrostatic precipitation of fly ash,
etc. These studies are also discussed.
INTRODUCTION
The Homer City generating station, located near Indiana,
Pennsylvania, is operated by the Pennsylvania Electric Company
(PENELEC), which is part of the General Public Utilities System.
Ownership of the station is shared with the New York State Electric
and Gas Company.
The expansion of the Homer City generating plant by the addi-
tion of a new 650 MW coal fired unit necessitated an in-depth study
of the technology and feasibility of the various options available
for complying with the S02 and particulate matter emission control
regulations, both State and Federal, for the new Unit 3 and for
the existing Units 1 and 2 (also 650 MW each).
36
-------
An obvious option was flue gas desulfurization (FGD) by lime/
limestone scrubbing. While engineering design work was commenced
for such a system for Unit 3, concurrent studies were made to seek
a viable alternative to the adverse aspects of utilizing FGD.
Major emphasis was directed toward fuel beneficiation by employing
the best coal cleaning technology possible to the coal supply for
the Homer City Station. The results of this study indicated that
this route, utilizing various innovations developed by GPU and
associated organizations, was the most desirable on an economic,
operational, and environmental basis. The coal cleaning process
conceived (termed the "Multi-Stream Coal Cleaning Strategy") would
yield three coal product streams from treatment of the Homer City
mine-mouth coal supply. These would be:
— a low-sulfur, low-ash coal for use in Unit 3 to meet
Federal EPA "New Source" emission regulations of "less
than 1.2 Ib S02/106 Btu",
— a coal of medium ash and sulfur content for use in Units
1 and 2 to meet State emission regulations of "less than
4.0 Ib SC-2/106 Btu".
— and a small refuse stream of high-sulfur, high-ash mate-
rial which would be stored on site.
As a result of this study, work was stopped on the FGD pro-
gram and engineering design and construction was begun on the
MCCS plant which is now due to start up in late 1977.
This project involves the joint efforts of many organizations
and individuals — not only in respect to actual plant design and
construction, but also in the conducting of pertinent studies on the
technology involved and its ramifications. Table 1 lists the
major participants in the project.
PROCESS TECHNOLOGY
Let us take a brief look at the technology involved in the
process design for this coal cleaning concept.
Multi-stream coal cleaning, per se, is not a new idea. Nor
does the Homer City project involve new discoveries, special hard-
ware, exotic reagents, and "black boxes". The GPU-PENELEC MCCS
system, as the Homer City Station owners and operators are apply-
ing it, instead includes a number of innovative applications and
extensions of multi-stream coal cleaning technology which will be
pointed out as we proceed.
Let us consider a piece of coal—typical of the coal avail-
able at Homer City, mined from the upper and lower Freeport seams.
This piece of coal is not a homogeneous compound of carbon,
hydrogen, sulfur, mineral matter, etc. Instead, when it is
37
-------
TABLE 1. PARTICIPANTS IN HOMER CITY MCCS PROJECT
OWNERS AND OPERATOR OF HOMER CITY STATION
General Public Utilities Corporation
GPU Service Corporation
Pennsylvania Electric Co. (Operator, 50% Ownership)
New York State Electric & Gas Corp. (50% Ownership)
OPERATOR OF MCCS COAL CLEANING PLANT
Rochester & Pittsburgh Coal Co.
ENGINEERING DESIGN AND CONSTRUCTION
Heyl & Patterson, Inc.
CONSULTANTS
Australian Coal Industries Research Laboratories, Ltd,
Mr. J.W. Leonard Mr. A.C. Gilbert
Mr. E.A. Zawadzki Others
COOPERATING ORGANIZATIONS
Electric Power Research Institute
U.S. Environmental Protection Agency
Energy Research and Development Agency
U.S. Bureau of Mines
Babcock & Wilcox Co.
Research-Cottrell, Inc.
Southern Research Institute
Others
crushed down to small size particles, each individual particle
contains varying amounts of the volatile and "hydrocarbonaceous"
material (which provides the Btu "heating value") and the mineral
matter (which provides the "ash content").
38
-------
Chemical analysis of this coal has shown that the sulfur con-
tent totals about 2.9% and exists in three general forms:
- about 2.3% as pyritic sulfur (FeS2)
- about 0.6% as organic sulfur
- and less than 0.01% as sulfate (SCK~2)
The heating value of this coal was determined to be about 11,900
Btu/lb.
The specific gravity of the individual particles will vary
over a wide range, depending upon the size and composition of each
particle. For example, the following are the specific gravities
of a few pertinent mineral components of coal in their pure
crystalline form:
Silica (Si02), sp. gr. 2.3 Hematite (Fe203), sp- gr. 5.2
Alumina (A12O3), sp. gr. 4.0 Magnetite (FesCU), sp. gr. 5.2
Pyrite (FeS2), sp. gr. 4.9
The specific gravity of the combustible carbon-hydrocarbon portion
of coal is approximately 1.2.
- Conventional coal cleaning practice involves size reduction
to create heterogeneous particles, followed by separation
procedures based upon specific gravity. Over the years, a
large variety of hardware for this purpose has been devel-
oped to meet the then-existing needs of the industry.
- Specific gravity separations made on large particle size
material obviously limits the working range of the spe-
cific gravity cuts which can be applied and the composi-
tion of the material which can be removed.
- Specific gravity separation at high specific gravity would
remove only a portion of the ash and pyritic sulfur pre-
sent. Separation at low specific gravity would remove more
ash and pyritic sulfur, but also would remove combustible
matter. Imprecise separation procedures would reduce the
efficiency of the cleaning operation-
It can be said that coal cleaning technology, as generally
practiced, has been limited by the prevailing economic and fuel
quality criteria rather than by potentially available technology.
Now, however, the situation is different. The sulfur and ash
contents of available coal supplies are a matter of critical
concern as all fuels have become more costly, emission control
39
-------
requirements have become more severe, and emission control tech-
nology available today is both costly and uncertain in respect
to performance and reliability.
With this background, let us consider the composition of this
piece of coal and the S02 emission control requirements for the
Homer City Station.
The emission control requirements are as follows:
Units 1 and 2 - S02 less than 4.0 lb/106 Btu heat input.
Particulate matter less than 0.1 lb/10
Btu heat input.
Unit 3 - S02 less than 1.2 lb/106 Btu heat input.
Particulate matter less than 0.1 lb/106
Btu heat input.
This can be translated into the need of a coal for Units 1
and 2 having a sulfur content less than 2.0 lb/106 Btu and a coal
for Unit 3 having a sulfur content less than 0.6 lb/105 Btu.
Minimum ash content is obviously necessary to insure the attain-
ment of the low particulate matter emission levels by the use of
the electrostatic precipitators on each unit. It should be
noted that the electrostatic precipitators to be installed on
Unit 3 (furnished by Research-Cottrell) represent what we believe
to be the best state-of-the-art in conventional precipitator
design.
The analysis of the Homer City coal, as typified above, trans-
lates into approximately 2.4 Ib sulfur per 106 Btu.
The coal cleaning strategy evolved by GPU and PENELEC is
shown in the flowsheet of Figure 1. Multiple stream coal cleaning
circuitry is employed to both remove the maximum amount of sulfur
from the raw coal and to redistribute the remaining sulfur and
ash contents of the cleaned coal to produce the two product streams
of the desired quality.
As is well known, both particle size and specific gravity are
interrelated as factors affecting the efficiency of coal cleaning
operations. While crushing and size classification of the raw
coal, followed by a gravity separation at Sp. Gr. 1.80 can remove
a major portion of the sulfur and ash content, further treatment
is necessary to produce the desired quality coal. This consists
of gravity separations of selected particle size coal at a lower
specific gravity to distribute the sulfur and ash content of the
initially cleaned coal in the proper proportion between the two
desired product streams. Such gravity separations, both at
Sp. Gr. 1.80 and at a lower value, can be accomplished in heavy-
media cyclones, using suspensions of fine sized magnetite as the
"heavy-media" for the separations.
40
-------
INITIAL
COAL
CLEANING
PROCESS
EMISSION LIMIT
1.2 LBS. SO2/106 BTU
LOW SULFUR
COAL
EMISSION LIMIT
4.0LBS. SO2/106BTU
MEDIUM SULFUR
COAL ,
ADVANCED
COAL
CLEANING
AND
SEPARATING
PROCESS
PRESENT
UNITS NO. 1
AND NO. 2
Figure 1. New multi-stream coal cleaning process. Homer City Electric Generating Station
How is this accomplished? The theoretical basis for the
plant design is illustrated in Figures 2 and 3.
Figure 2 shows the effect of the particle size of the pro-
cessed coal on the sulfur content of the "float product" of a
float/sink specific gravity separation made at a Sp. Gr. of 1.30.
The two curves are for the coals from two different mines. As
can be seen, decreasing the particle size leads to a decrease in
sulfur content. However, the nature of actual gravity separation
process equipment limits the practical extent to which the parti-
cle size can be reduced to achieve lower product sulfur contents.
Figure 3 shows the effect of the specific gravity of separa-
tion on the sulfur content of the "float/sink" products when
making a gravity separation on a coal of a given particle size -
e.£., 1/4 x 100 mesh. As can be seen, in this particular case,
a gravity separation at Sp. Gr. 1.3 produces the desired sulfur
distribution.
41
-------
EMISSION STANDARD
3/8X1/4 1/8X10M 28MX65M
1/4X1/8 10MX28M 65 MX 100 M
SIZE FRACTION
Figure 2. Effect of size on quality of 1.3 sp. gr. float product
Now to tie this all together, the Homer City Multi-Stream
Coal Cleaning Strategy is based on the following assumptions
which have been implemented into the process design:
1. By removal of most of the pyritic sulfur and very
little of the combustible material from the coal
available for use at Homer City, the overall/com-
posite 3-0 Ib S02/106 Btu quality fuel required can
be created. Then, by redistributing the sulfur con-
tent of this coal, two product coal streams can be
created.
- one, < 0.6 Ib S/106 Btu, low ash, for Unit 3,
- the second, < 2.0 Ib S/106 Btu for Units 1 and 2.
2. With the proper selection of existing types of coal
cleaning hardware incorporated into a multi-stream
coal cleaning process design tailored for the avail-
able coal, the desired product coals could be obtained
if —
42
-------
1.32
3
m
-------
4. Development of a novel control system for tighter
control of the specific gravity of operation in the
nominal 1.3 specific gravity circuit operating on
fine size coal.
THE HOMER CITY MCCS COAL CLEANING PLANT
Figure 4 is a simplified process flow diagram for the Homer
City MCCS coal cleaning plant. It shows the basic circuits and
the major items of process equipment.
Figure 5 shows the overall process with boundary markings
added to segregate the coal cleaning circuits on the basis of
the coal particle size. Now, let's look at each circuit
separately.
Circuit A - Crushing and Classification
A portion of the cleaning plant is dedicated to crushing and
size classification. Provision is made to selectively crush all
major sources of supply to the cleaning plant in variable speed
cage crushers to control the production of plus 1/4 inch and minus
100 mesh material while optimizing the release of pyrite and ash.
Circuit B - Fine Coal, Deep Cleaning
The fine coal cleaning circuit cleans 2 mm x 100 mesh coal in
heavy media cyclones at 1.3 specific gravity. In this circuit,
underflow from the classifying and desliming screens is pumped .
to 14-inch classifying cyclones. Underflow, containing mostly
2 mm x 100 mesh coal, is cleaned at 1.3 specific gravity in heavy
media cyclones. Overflow product is partially dewatered and
washed of fines in spiral classifiers. Product makes up the
major portion of intensively cleaned coal for Unit 3.
Circuit C - Medium Coal Cleaning
The medium coal cleaning circuit cleans 1/4 x 2 mm coal in
two stages of heavy media cyclones. The first stage cleans at
1.3 specific gravity. The clean coal overflow product of this
circuit is available as intensively cleaned coal for use in
Unit 3. Surplus 1.3 float product from this stage is blended
into the moderately cleaned middling coal. Underflow from the
first stage is recleaned in the second stage of heavy media
cyclones at 1.8 specific gravity. Underflow from this stage is
refuse coal; overflow is 1.8 float -1.3 sink and is blended into
middling coal.
44
-------
HELVETIA COAL
TRUCK COAL
HELEN COAL
HYDROCYCLONE
DEISTER
REFUSE
1200 TPH 1-1/4 IN. XO
SCREEN 186 TPH -j.1/4 IM x 1/4 !N
TRASH SCREEN
REJECT
SCREEN
1/4 IN, X 2 MM
. 597 TPH 2 MM X 0
CLASSIFIER
VOR-SIV
CLASSIFIER
H.M.C. -1.8 SP. GR.
CENTRIFUGE
D&R SCREEN
DEEP CLEAN* COAL
111 TPH
MIDDLING tCOAL
98 TPH
MIDDLING
104 RPH
SPIRAL CLASSIFIER
H.M.C.
1.8SP. GR.
SEPARATORS
VACUUM FILTER
D&R SCREEN
REFUSE
CENTRIFUGES
THICKENER
NO. 2
DEEP CLEAN » COAL MIDDLING J COAL
234 TPH J*k- 203
MIDDLING 1 COAL
224 TPH
266 TPH 2 MM x 0
THICKENER NO. 1
Figure 4. Homer City—MCCS simplified diagram
-------
HELVETIA COAI
TRUCK COAL
HELEN COAL
1200 TPH 1-1/4 IN. XO
HYDROCYCLONE
186 TPH
1-1/4 IN. X 1/4 IN.
TRASH SCREEN!
REJECT I
417 TPH
SI/4 IN. X2MM
X
597 TPH 2 MM X 0
«__ f \
CLASSIFIER
D&R ^-^1.3 SP. GR.
SCREEN
VOR-SIV
CLASSIFIER
H.M.C. - 1.8 SP. GR. fti
D&RSCREEN
THICKENER
NO. 1
CENTRIFUGE
D&RSCREEN
DEEPCLEANtCOAL
111 TPH
MIDDLINGtCOAL
. - TPH
H.M.C.
DSR™*1
SCREEN
MIDDLINGtCOAL
104 TPH X
MAGNETIC SPIRAL
SEPARATORS CLASSIFIER
VACUUM
FILTER
D&R SCREEN
CENTRIFUGES
THICKENER
NO. 2
DEEP CLEAN?COAL
234 TPH
MIDDLING.? COAL
203 TPH _.
MIDDLINGtCOAL
224 TPH .
E )FINE COAL SCAVENGING
tod CRUSHING AND SIZE SEPARATION fr$ MEDIUM COAL - CLEANING
$gj} FINE COAL - DEEP CLEANING gjh COARSE COAL - CLEANING
Figure 5. Homer City-MCCS simplified diagram, coal cleaning circuits
-------
Circuit D - Coarse Coal Cleaning
The coarse coal cleaning circuit cleans plus 1/4 inch coal
in heavy media cyclones at 1.8 specific gravity. Cleaned plus 1/4
inch coal becomes part of the moderately cleaned middling coal
blend.
Circuit E - Fine Coal Scavenging
The fine coal scavenging circuit deals with 2 mm * 100 mesh,
1.3 specific gravity sink coal and minus 100 mesh coal. Coal pre-
pared in this circuit is used as moderately cleaned middling coal.
Coarser material is cleaned in hydro-cyclones. Hydro-cyclone
underflow is sent over deister tables for pyrite removal.
Now, after the local coal gets beaten, battered, screened,
swirled, and scrubbed in this coal cleaning plant, what is the
expected nature of the coals produced for use in our Homer City
Station?
Table 2 shows the expected typical analyses of the raw and
product coals. These data are based upon analyses made on coal
samples prepared by laboratory simulation of the coal cleaning
processes.
As the route of coal cleaning, rather than the route of flue
gas desulfurization, was chosen as the means to meet the SO2
emission control requirements for Homer City Station, a few com-
parisons of MCCS vs. FGD should be of interest.
Table 3 is an economic comparison of the two options consid-
ered for S02 emission control, based upon the most realistic
cost estimates available.
Table 4 is a comparison of the two options based upon process
material requirements.
Table 5 is a comparison of the effluents produced in operat-
ing each of these options.
Table 6 represents a comparison of the environmental effects
of each option, based upon predicted ambient air quality. In
addition to the ambient air quality, a major environmental aspect
to be considered is the solid and liquid waste situation associ-
ated with each process. The problem of disposing of vast quan-
tities of FGD sludge was a major factor in our choice between the
two options.
Figures 6 and 7 illustrate the predicted sulfur and energy
balances for the Homer City MCCS facility. As can be seen, the
energy loss is low in comparison with the reduction in sulfur and
ash.
47
-------
TABLE 2. TYPICAL COAL ANALYSES
(Data Based On Laboratory Simulation of MCCS Process)
ANALYSIS MCCS PRODUCTS
Raw coal Deep-clean coal Middling coal Refuse
(Helen & Helvetia) (Unit 3) (Units 1&2)
Sulfur Content
Pyritic, %
Organic, %
Sulfate, %
Total %
Ash Content, %
Heating Value,
Btu/lb
Yield, %
Lb S02/106 Btu
2.28
0.58
0.007
2.87
22.02
11,923
4.81
0.27
0.57
0.018
0.86
3.06
15,299
27.1
1.12
1.48
0.50
0.006
1.99
16.35
12,880
.53.7
3.09
6.97
71.19
—
19.2
-
TABLE 3. HOMER CITY GENERATING STATION
ALTERNATIVE S02 CONTROL STRATEGIES
Cost Comparison - Coal Preparation vs FGD
CAPITAL INVESTMENT - $ x 106
Coal Preparation Facilities
Original Plant (for use with FGD)
MCCS Addition
FGD
Subtotal
FGD
18
59
77
MCCS
18
32
0
50
ANNUAL REVENUE REQUIREMENTS - $ x 106
Fixed Charges
Operating and Maintenance Expenses
FGD
Coal Preparation
Subtotal
11.6
7.5
0
7.6
15.1
48
-------
TABLE 4. HOMER CITY GENERATING STATION
ALTERNATIVE S02 CONTROL STRATEGIES
Estimated Materials Consumed - Coal Preparation vs FGD
Materials Consumed FGD MCCS
Coal, ton/yr @ 1200 Btu/lb
Boilers
Coal Dryers
Lime Kilns (Off Site)
Limestone, ton/yr
4,800,000
17,200
17,000
144,000
4,800,000
46,400
0
0
TABLE 5. HOMER CITY GENERATING STATION
ALTERNATIVE S02 CONTROL STRATEGIES
Estimated Effluent Comparison - Coal Preparation vs FGD
Effluents - Tons Per Year FGD MCCS
S02
Boilers
Coal Dryers
Lime Kilns (Off Site)
Subtotal
Sludge (50% Solids)
Fly Ash and Bottom Ash
Coal Refuse
176,000
600
700
177,300
334,000
1,057,000
552,000
176,000
2,100
0
178,100
0
577,000
1,032,000
TABLE 6. AMBIENT AIR QUALITY S02 CONCENTRATIONS (FORECAST)
FGD
MCCS
24 hr av
0.0771 ppm
0.0713 ppm
Homer City Generating Station Site
3 hr av
0.2475 ppm
0.2289 ppm
49
-------
REJECTS
HC 1 &2
2ND STEP
46%
HC3
EXISTING UNITS
NEW UNITS
Figure 6. MCCS coal sulfur balance
RAMIFICATIONS OF MCCS
This paper is being presented at an EPA-sponsored conference
which is concerned with the particulate collection problems
associated with conversion to low-sulfur coals. Therefore, it is
appropriate that we discuss the ramifications of the MCCS project
with respect to particulate emission control and other related
areas of concern.
First, this concept of a coal cleaning process is not re-
stricted to its application at Homer City. Multi-stream coal
cleaning processes, based upon Homer City MCCS design technology
and operating experience, provide a means for both reducing and
redistributing the sulfur content of coal - producing several
product streams tailored to the individual fuel requirements of
multiple-unit power generation complexes. This is illustrated
in Figures 8 and 9.
50
-------
COAL HEAT
CONTENT
REJECTS
—TITL
HC1&2
2ND STEP
62%
32%
l_
HC3
EXISTING UNITS
NEW UNITS
Figure 7. MCCS coal heat content balance
When the proper factors exist with respect to the nature of
the coal supply available and the S02 emission requirements of
existing and new generating units at some station or combination
of stations, this concept, with appropriate modifications, may be
a viable alternative to flue gas desulfurization.
Therefore, one must seriously consider what are the other
properties (beside lower sulfur content) of the coals produced
by such processes. Laboratory studies have verified the fact
that intensive coal cleaning operations alter the composition of
the fly ash which will be produced during combustion. Ash fusion
temperatures, slagging and caking characteristics, etc., may also
be altered.
51
-------
m
Figure 8. Photo of Homer City MCCS plant during construction
-------
I MJTW—•
* r
Figure 9. Photo of Homer City MCCS plant during construction
-------
Table 7 lists the various research activities being conducted
concurrently with the design, installation, and operation of the
MCCS facility at Homer City Station. Item 4 on this list is an
item of major concern to us.
TABLE 7. CONCURRENT RESEARCH & DEVELOPMENT PROGRAMS
1. Development of design and operating criteria for maximizing
the performance of coal cleaning equipment with the coal
available for use in the Homer City Generating Station.
2. Development and evaluation of process control instrumentation
systems for the precise control of heavy-media cyclone coal
washing operations,
3. Development of "on-line" analytical instrumentation for the
measurement of sulfur content, ash content, and heating
value of the coal being processed.
4. Prediction and evaluation of the combustion properties of
low-sulfur cleaned coal in respect to furnace firing, boiler
operation, and electrostatic precipitator performance and
collection efficiency.
5 . Development of assessment techniques for evaluating potential
coal supplies.
What kind of particulate matter, and how much, will be in the
combustion product gases when coals from our coal cleaning plant
and other intensive cleaning processes are burned in existing
and new utility boilers?
What will be the effects on furnace combustion and fly ash
collector performance as a result of changing the ash composition,
the coal particle size distribution, and the sulfur oxide concen-
trations in the flue gases?
Means for both measuring and predicting these phenomena with
reasonable accuracy must be developed now so that the combustion
and emission control equipment is compatibly designed and properly
operated to accommodate them.
We are currently working with the Australian Coal Industry
Research Laboratories, Southern Research Institute, Research-
Cottrell, Babcock and Wilcox, and others on this problem, as well
as conducting our own internal research program.
54
-------
This research is now taking the form of determining the
physical nature and the chemical composition, including complete
"ash analysis", of the Homer City MCCS product coals - then, using
both the classically accepted and the currently developing corre-
lation techniques, predicting their combustion characteristics
in respect to large utility boilers and their fly ash character-
istics in respect to electrostatic precipitator performance. This
requires the preparation of coal samples which resemble, as close-
ly as possible, the low sulfur (Unit 3) and medium sulfur (Units
1 and 2) coals which will be produced in the Homer City MCCS fa-
cility when completed. This was done by taking samples of the raw
Homer City coal that will be the input to the MCCS plant and sub-
jecting them to size reduction/classification and specific gravity
separation procedures in the laboratory in a manner which simulates
the various processing steps incorporated in the MCCS plant.
In addition to the analyses made in our laboratories, portions
of these samples were sent to other cooperating laboratories for
analyses and study. Samples of the low sulfur (Unit 3) coal have
been sent to Southern Research Institute and to the Australian
Coal Industry Research Laboratories. The tests in Australia in-
clude the combustion of the coal in a laboratory combustion train
under conditions simulating those for Unit 3 and the collection
of the ash/fly ash produced to determine furnace ash properties,
slagging characteristics, etc., and to determine the bulk elec-
trical resistivity and other properties of the fly ash which would
affect electrostatic precipitator performance.
Following the startup of the MCCS facility in 1977, these
test results and the predictions made therefrom will be compared
with analyses of the actual product coals and with the boiler/
precipitator performance data as these coals are burned in the
Homer City Station.
Items 2 and 3 in Table 7, concerning analytical and process
control instrumentation, also merit some discussion.
Item 3 is concerned with an anomalous situation in the current
state of development and availability of analytical instrumenta-
tion. While sophisticated instruments have been developed and
utilized to measure everything of interest contained in stack
gases, including quite possibly the "Hay Fever Index for Elderly
Asthma Sufferers", we are still forced to use crude sampling and
slow analytical procedures to determine the sulfur and ash con-
tents of the solid and liquid fuels we burn.
This can give us a good handle on what happened yesterday or
last week—why we had an opacity problem or our SOa emissions were
out of line. It can also tell us that we had received and burned
a coal that didn't meet specifications for sulfur and ash content.
55
-------
The present state of coal sampling and analysis technology
provides no on-stream capability for fuel quality assessment or
process control.
Such a capability, if developed, could be of infinite value
in the field of air pollution emission control. Therefore, we are
promoting and supporting various research efforts to develop prac-
tical analytical instrumentation systems of this type—
instrumentation which can be utilized for improved process control
in the Homer City MCCS facility and which can be used in the many
applications of coal processing and utilization where on-stream
evaluation of coal quality is of concern. It is hoped that through
the cooperative efforts of the Electric Power Research Institute,
the U.S. Environmental Protection Agency, and other organizations
working with us, we can achieve some measure of success in this
project.
Item 2 in Table 7 concerns another area in the field of pro-
cess control instrumentation where efforts conducted to satisfy
the needs of the MCCS program may result in technological advance-
ments which can be usefully applied elsewhere. The need for
accurately measuring and controlling the density of multi-phase
mixtures of such heterogeneous materials as water, pulverized coal
of varying size and consistency, finely-sized magnetite particles,
etc., requires the use of analytical and process control instru-
mentation which is currently not available "off-the-shelf" . The
further development of the design and application of currently
available instrumentation components is a major activity within
the MCCS program.
To further this effort, a major part of the MCCS process
development program is the installation and operation of a pilot
plant scale "test circuit" at the Bruceton Laboratory of the
Bureau of Mines where both the performance of process control
system concepts and component performance as well as the perfor-
mance characteristics of other items of process equipment will be
determined and evaluated.
CONCLUSIONS
The development and implementation of the MCCS coal cleaning
program for the Homer City Station is not merely an activity of
interest to those concerned with coal cleaning technology. The
successful completion of this program has many ramifications with
respect to the future utilization of our nation's available coal
resources. Among these are those which are of an internal'
nature—^concerned with the combustion and emission control problems
which may be involved with the utilization off the "cleaned coals"
produced in such facilities. This paper, while providing a brief
description of the MCCS process, is primarily intended to acquaint
those persons not intimately involved with coal cleaning technology
56
-------
with the implications and ramifications of this MCCS program in
the related areas of boiler design and operation, electrostatic
precipitator design and performance, etc. While this has not
been done in depth due to the complexities of the subject and the
length of the paper, it is hoped that an interest will be aroused
in those who hear or read this paper to pursue this matter fur-
ther in their various areas of expertise and concern.
The following references from which much of the material in
this paper was taken, provide more information on the background
and process design of the Homer City multi-stream coal cleaning
plant.
REFERENCES
1. McConnell, James F., and Charles W. Statler. Multi-stream
Coal Cleaning Strategy for the Control of Sulfur. Presented
at Fall Meeting, Society of Mining Engineers, Salt Lake City,
Utah, Sept. 10-12, 1975.
2. Statler, Charles W. Multi-stream Coal Washing—A Systems
Approach for the Control of Sulfur. Presented at Conference
on Air Quality Management in Electric Power Industry, Austin,
Texas, Jan. 28-30, 1976.
3. McConnell, James F. Multi-stream Coal Cleaning Strategy for
Control of Sulfur Oxide Emissions at the Homer City Generat-
ing Station. Presented at Coal Processing and Conversion
Symposium '76, West Virginia Geological and Economic Survey,
Morgantown, West Virginia, June 2-3, 1976.
57
-------
PAPER 5
ELECTROSTATIC PRECIP1TATOR OPTIONS FOR COLLECTION
OF HIGH RESISTIVITY FLY ASH
L. E. Sparks
Industrial Environmental Research Laboratory
Environmental Protection Agency
ABSTRACT
An analysis of the technical options for collecting high re-
sistivity fly ash is presented. The options are a large specific
collector area electrostatic precipitator operating at normal
temperatures (about 150°C), a wet electrostatic precipitator, a-
moderate to large specific collector area electrostatic precipi-
tator operating at either high (about 370°C) or low temperature
(about 110°C), or a moderate specific collector area at normal tem-
perature for an electrostatic precipitator plus fly ash condition-
ing. The advantages and disadvantages of each option are discussed.
The potential for adverse environmental impact of the conditioning
option is also discussed.
INTRODUCTION
Combustion of low sulfur coal generally produces fly ash
with higher resistivity at 150°C than does combustion of high sul-
fur coal. Thus, fly ash from low sulfur coal is more' difficult
to collect in electrostatic precipitators than is fly ash from
high sulfur coal.
There are several methods of overcoming the influence of
high resistivity caused by low sulfur coal. The options are:
1. Provide additional precipitator capacity; i^-e^ , add
collector plate area to overcome the poorer performance.
2. Use a wet electrostatic precipitator.
3. Change the flue gas temperature to achieve a more
favorable resistivity.
4. Add fly ash conditioning agents to change the resistivity
or otherwise modify electrical conditions in the
precipitator.
58
-------
In deciding which of these options is optimum for a given
situation, one must recognize that coals, even from the same
mine, vary. This means that the design coal will not be burned
all the time. The variation in coal composition, especially of
the sodium, iron and potassium content of the coal (Bickelhauptl)
will result in variations in fly ash resistivity with resultant
variations in electrostatic precipitator performance. Available
data indicate that coal composition variations can easily result
in day to day resistivity variations of plus or minus a factor of
two. The fact that coal composition and thus fly ash resistivity
will vary from day to day must be considered in selecting ESP
options. The method and costs of dealing with the variation
must be considered. Questions such as, "Will plate area be added
to handle the expected resistivity extremes?" or "Will the boiler
be operated at reduced load?", must be answered. Regardless of
which option is picked, it is essential that the precipitator be
well designed, built, operated and maintained. Gas flow distri-
bution must be good to excellent; sneakage and rapping reentrain-
ment losses must be minimized; wire alignment must be good; ash
handling must be properly designed; etc.
COLD SIDE ELECTROSTATIC PRECIPITATORS
Cold side electrostatic precipitators (operating at ^ 150°C)
can collect fly ash from low sulfur coal if sufficient collector
plate area is provided. The required collector plate area de-
pends on the resistivity of the fly ash and can be as high as
200 m2/am3/sec (^1015 ft2/akcfm). Specific collector areas in
the range of 80 m2/am3/sec to 120 m2/am3/sec (400 ftVakcfm to
600 ft2/akcfm are common) (am3 is actual cubic meters; akcfm is
actual thousand cubic feet per minute). High efficiency preci-
pitators with specific collector areas in this range have been
built for low sulfur applications.
If the required specific collector area is near the low
end of the range cited above, a normal cold side precipitator is
likely to be the most economical method of collecting high re-
sistivity fly ash. Some designers feel that in general cold side
units are less expensive than hot side units for collection of
fly ash from low sulfur coal (Matts2). Other designers believe the
opposite (Walker and Frisch3).
The disadvantages of cold side electrostatic precipitators
for collecting high resistivity fly ash all center around the
sensitivity of precipitator performance to small changes in
resistivity. This sensitivity to resistivity makes initial design
of the precipitator somewhat uncertain, since fly ash resistivity
is likely to vary somewhat during the life of the boiler. Once
the precipitator is built, normal variations in resistivity can
result in performance degradation such that violations of air
59
-------
pollution regulations occur. If the resistivity variation is of a
short term nature, emission limits can be met by reducing boiler
load. However, if the resistivity change is of long term, such
as might be expected if a new coal supply is used to fuel the
boiler, long term de-rating of the boiler or retrofit of additional
particulate collection equipment may be required.
WET ELECTROSTATIC PRECIPITATORS
Because the dust layer is continuously washed from the elec-
trodes and the gas is saturated with water vapor, dust resistivity
is not a factor in the performance of wet electrostatic precipi-
tators. Thus, wet electrostatic precipitators may be an economi-,
cally attractive method of collecting high resistivity fly ash.
Whether or not wet precipitators are economically feasible
for fly ash collection depends on the cost of purchasing and
operating the precipitators. Because wet precipitators must be
built of corrosion-resistant materials, they cost more per square
meter of plate area than do dry precipitators. Presently avail-
able information indicates that for utility boiler applications
wet precipitators are likely to cost 2% times or more per square
meter of collection plate area. If the total collection plate
area of the wet precipitator is small enough, the added cost per
unit of collector area can be overcome. Performance data (Gooch ,.
and McCain1*), on large plate-type wet electrostatic precipitators
indicate that a specific collection plate area of about
60 m2/am3/sec may be required for high efficiency fly ash collec-
tion.
Wet electrostatic precipitators have all the potential water
pollution problems that wet scrubbers have. The problems associ-
ated with meeting water pollution requirements are likely to in-
crease the cost of wet electrostatic precipitators.
RESISTIVITY MODIFICATION
As discussed in the Cold Side ESP section, precipitator per-
formance becomes extremely sensitive to slight changes in resis-
tivity when the resistivity is high. Under such conditions a
relatively small change in ash resistivity can lead to excessive
particulate emission. Compliance with emission standards can
require operation at reduced boiler load with consequent loss of
power generation. Therefore, at some point it becomes necessary
to consider steps to reduce the resistivity to a more acceptable
level.
Resistivity can be changed by operating the precipitator at
a higher or lower temperature or by adding chemical agents either
to the fuel or to the flue gas to condition the fly ash and reduce
its resistivity. Although resistivity modification may appear to
60
-------
be more expensive than living with the problem, the advantages of
improved confidence in design and lessened sensitivity to small
changes in resistivity must also be considered.
HIGH TEMPERATURE OPERATION
Location of the electrostatic precipitator before the air
heater allows operation at about 370°C. At this temperature
resistivity of most fly ash is in the range 2-4 x 1010 ohm-cm or
lower. Thus, except for rare problem ashes, hot side precipita-
tors can be designed for relatively high current densities and
with confidence that changes in fuel chemical composition are
unlikely to affect performance.
Calculations based on the electrostatic precipitator per-
formance model described by Gooch, et al5 and data reported by
Walker and Frisch3 indicate that a specific collector area of at
least 67 m2/am3/sec is required for 99.5% particle collection
efficiency in a hot side precipitator.
The primary disadvantage of hot side operation is the in-
crease in gas volume due to elevated temperatures. Thus the
reduction in SCA for high temperature operation is somewhat off-
set by the increased gas flow. Table 1 compares calculated col-
lector plate areas for 99.5% particulate removal for various
current densities at 150°C with collector plate area required at
370°C on the basis of m2/am3/sec and m2/nm3/sec. Note that the
hot side precipitator requires a specific collector area of
158 m2/mti3/sec.
TABLE 1. ELECTROSTATIC PRECIPITATOR SPECIFIC COLLECTOR
AREA REQUIRED FOR 99.5% PARTICLE COLLECTION
FOR VARIOUS CURRENT DENSITIES
Current Density, Temperature, Specific Collector Area
nA/cm2 °C m2/am3/sec m2/nm3/sec
5
20
40
30
150
150
150
370
122
53
39
67
.1
.4
189
82.
61
158
3
Normal conditions 0°C at 1 atmosphere pressure
Calculations based on Gooch, et al5
61
-------
Problems associated with thermal expansion must be given
special attention during the design and construction of hot side
precipitators. There have been cases where sufficient care was
not taken in dealing with thermal expansion which resulted in
poor electrostatic precipitator performance.
Hot side precipitators must be well insulated to reduce heat
losses. Insulation and unavoidable heat losses are likely to
increase the cost of hot side operation. Walker and Frisch3
present a lucid discussion of design, construction, and operation
of hot side electrostatic precipitators.
LOW TEMPERATURE ELECTROSTATIC PRECIPITATORS
For some cases a 20-50°C reduction in electrostatic precipi-
tator operating temperature may result in resistivities in the
acceptable range. The exact nature of the resistivity versus
temperature curve at low temperatures is variable and depends on
the particular situation. A tenfold or greater variation in low
temperature resistivity is common. Thus the option of reducing
gas temperature is subject to verification that the resistivity
does indeed decrease with decreasing temperature to a satisfactory
level at achievable temperatures. This option is also subject to
verification that the rate of corrosion is acceptable; this
verification may be possible if the corrosive effect of SOa is
minimized by the CaO content of the ash. Several precipitator
installations have performed satisfactorily at temperatures in
the 95-120°C range without undue problems, when sufficient atten-
tion was given to the ash-removal system to keep the collected
ash hot and to maintain the insulators free of condensed acid.
Achieving a flue gas temperature low enough to overcome the
resistivity problem presents problems that must be dealt with on
an individual plant basis. If there is sufficient capacity in
the air heater, the problem can be handled simply; however, this
is not often the case. Whitehead6 reported that 120°C is about
the minimum temperature that can be achieved even if the air
heater has sufficient capacity. An alternative approach to
reaching low flue gas temperatures in an existing boiler with in-
sufficient capacity in the air heater, or if temperatures below
120°C are required, is to increase the flow of secondary air
through the air heater and by-pass some of the excess air to the
stack. This approach requires additional fan capacity and im-
poses an energy penalty due to reduction in boiler efficiency,
and therefore in most cases is not too attractive.
Berube presents an excellent discussion of low temperature
operation at one plant7. For this plant a 0.8% reduction in
boiler efficiency was necessary to obtain the required tempera-
ture.
62
-------
In general, low temperature operation is sensitive to coal
composition. In a situation where coal composition varies over
a wide range, low temperature may not be practical.
CONDITIONING
The addition of a chemical agent to the fuel or to the flue
gas to reduce fly ash resistivity has been successful in some
cases. Whitehead6 and others report that flue gas conditioning
should be considered as an option for most high resistivity ap-
plications .
A partial list of presently available or proposed condition-
ing agents is shown in Table 2.
TABLE 2. PARTIAL LIST OF CONDITIONING AGENTS
Aluminum sulfate Organic amines
Ammonia Sodium carbonate
Ammonium bisulfate Sulfamic acid
Ammonium sulfate Sulfur trioxide
Hydrogen chloride Sulfuric acid
Iron oxide Vanadium oxide
Iron sulfate
The most commonly used conditioning agent is sulfur triox-
ide, S03. It is injected into the flue gas either as S03 or as
sulfuric acid, HaSOi*. Many of the other agents shown in Table 2
are sold as proprietary agents which may be blends of several of
the compounds listed. At the temperatures existing in the flue
gas many of the compounds in Table 2, for example, ammonia and
sulfur trioxide, dissociate into other compounds. Except for S03
there is little industrial experience with these conditioning
agents, some of which have not been demonstrated as being
effective.
With the exception of sodium and S03 conditioning, the
mechanisms by which many of the conditioning agents work are un-
clear. Both sodium conditioning and S03 conditioning reduce the
resistivity of the fly ash, as shown by Bickelhaupt8 and
Dismukes9. Some of the sulfate conditioning agents appear to
reduce resistivity probably because S03 is one of the products of
dissociation (Dismukes10). Vanadium and iron oxides are claimed
to catalyze the reaction of S02 and 02 to S03 and thus increase
the quantity of S03 present in the flue gas.
63
-------
The effectiveness of conditioning with some of the agents
listed in Table 2 is summarized in Table 3. Based on the avail-
able data it appears that properly engineered, operated, and
maintained conditioning systems using SO3 as the active agent can
improve resistivity-limited performance of electrostatic preci-
pitators. The data on most other agents are insufficient to
make a general judgment of the effectiveness possible.
TABLE 3. EFFECT OF CONDITIONING ON EFFICIENCY
ESP Efficiency for Particulate Emissions
Agent Without Injection With Injection
S03 94% 95.2% (Dismukes9)
S03 37.5% 51.4% (Dismukes9)
S03 67.3% 77.3% (Dismukes9)
S03 6?-7% 93,5% (Dismukes9)
Sulfamic acid No statistically significant difference,
(Dismukes1°)
Apollo LPA40 0.34g/10bJ 0.13g/106J, (McNinch12)
Apollo LPA40 90% 97% (Harris13)
S03 80-88% 95-96% (Klipstein11*)
S03 85% 94-95% (Cook15)
There are few data on the fate of conditioning agents after
they are injected into the flue gas. It is generally assumed
that all the conditioning agent is deposited on the fly ash and
then collected in the precipitator (Hall11). The available
data, although limited, cast doubt on this assumption.
There are more data on the effects of S03 than other con-
ditioning agents. Whitehead6 reported that in Great Britain
conditioning with up to 50 ppm S03 resulted in no increase in S03
emissions. Dismukes9 reported data for two plants that showed
that from 28 to 80% of the injected conditioning agent was emit-
ted. Published data on S03 emissions with and without condition-
ing for many plants with conditioning systems are lacking. Hall
attributes any increase in S03 to poor system design and opera-
tion1 x .
Because S03 emitted due to conditioning will form parti-
culate sulfuric acid or particulate sulfates essentially in-
stantaneously upon reaching the atmosphere, it seems reasonable
to count the excess S03 as particulate sulfuric acid. Since
1 ppm of sulfuric acid is 5.3 mg/nm3 (0.0023 gr/scf) of
particulate, it is clear that the amount of S03 that can escape
the stack, if S03 is counted as particulate, must be kept at a
minimum in order to reduce ambient particulate loading.
64
-------
Except for data taken by Harris13, there are no published
data on emissions from other agents. Harris reported that the
injection of Apollo Chemical's proprietary conditioning agent
LPA40 appeared to result in increased emissions of S03, ammonia,
surface active agent, and other organic compounds. Many of the
other proprietary conditioning agents and proposed conditioning
agents probably dissociate at flue gas temperatures to produce
S03 and/or NH3 — some of which may be emitted to the atmosphere.
The agents that are claimed to catalyze the oxidation of
S02 to SO3 are likely to catalyze S02 to S03 in the plume as well
as in the boiler. Thus, the emission of such an agent to the
ambient air may increase the ambient particulate concentration
above acceptable limits due to formation of sulfates.
It should be noted that many of the conditioning agents
listed in Table 2 are known to be toxic. Thus, an appreciable
increase in the emission of these substances may result in
adverse health effects — at least near the power plant.
As stated above there are few data to support or contradict
the discussion above. It is clear, however, that toxic sub-
stances are being used as conditioning agents and at least some
of these agents are emitted to the atmosphere. The impact of
these emissions on ambient air quality is unclear.
Conditioning does appear to be an acceptable option for
collecting high resistivity fly ash in electrostatic precipitators
under some conditions and with some agents. Whenever condition-
ing is used, careful measurements of all emissions are required
to ensure that emissions of toxic substances are not increased.
EPA is presently conducting research to determine the overall
environmental impact of conditioning agents.
Dismukes reviewed the capital and operating cost of S03
conditioning9, His data show that the cost of S03 conditioning
is sensitive to the method used to produce the S03 and the con-
centration of S03 injected.
McNinch reports data on costs of Apollo conditioning
agent12. Dismukes presents some data on costs of sulfamic acid
conditioning1°.
CONCLUSIONS
The ideal general solution to the problems associated with
collecting high resistivity fly ash from combustion of low sulfur
coals does not exist. Utilities that design plants initially
for operation using low sulfur coal have several options, one of
which is likely to be best for their application. In situations
65
-------
where fuel source and chemical composition are uncertain, hot
side electrostatic precipitators should certainly be considered.
In cases where coal supply is certain and coal chemical com-_
position is well defined, large cold side electrostatic precipi-
tators may be attractive.
Conditioning is likely to be the least expensive retrofit
solution. However, extreme care must be taken in all steps of
design, construction, and operation of conditioning systems to
avoid unacceptable emissions due to the conditioning agents.
REFERENCES
1. Bickelhaupt, R.E. Effect of Chemical Composition on Surface
Resistivity of Fly Ash. EPA-600/2-75-017, NTIS PB244885/AS,
U.S. Environmental Protection Agency, Washington, D.C.,
1975. 50 pp.
2. Matts, S. "Cold Side" Electric Precipitators for High-
Resistivity Fly Ash Require Different Design Philosophy.
In: Proceedings of Symposium on Electrostatic Precipitators
for Control of Fine Particles. EPA-650/2-75-016, NTIS
PB 240440/AS, U.S. Environmental Protection Agency, Washing-
ton, D.C., 1975. pp 229-236.
3. Walker, A.B., and N.W. Frisch. U.S. Design and Operating
Experience with Hot Electrostatic Precipitators on Coal-
Fired Utility Boilers. Paper presented at Joint US/USSR
Symposium on Control of Particulate Emissions and Related
Problems, 1976.
4. Gooch, J.P., and J.D. McCain. Particulate Collection
Efficiency Measurements on a Wet Electrostatic Precipitator.
EPA-650/2-75-033, NTIS PB 244173/AS, U.S. Environmental
Protection Agency, Washington, D.C., 1975. 60 pp.
5. Gooch, J.P., J.R. McDonald, and S. Oglesby. A Mathematical
Model of Electrostatic Precipitation. EPA-650/2-75-037,
NTIS PB 246188/AS, U.S. Environmental Protection Agency,
Washington, D.C., 1975. 162 pp.
6. Whitehead, C. Gas Conditioning. In: Proceedings of the
Electrostatic Precipitator Symposium. Birmingham, Alabama,
1971. pp 177-222.
7. Berube, D.T. Low Gas Temperature Solution to High Resis-
tivity Ash Problems. In: Proceedings of the Electrostatic
Precipitator Symposium. Birmingham, Alabama, 1971.
pp 223-280.
66
-------
8. Bickelhaupt, R.E. Sodium Conditioning to Reduce Ply Ash
Resistivity. EPA-650/2-74-092, NTIS PB 236922/AS, U.S.
Environmental Protection Agency, Washington, D.C., 1974.
9 PP-
9. Dismukes, E.B. Conditioning of Fly Ash with Sulfur Triox-
ide and Ammonia. EPA-600/2-75-015, NTIS PB 247231/AS,
U.S. Environmental Protection Agency, Washington, D.C.,
1974. 155 pp.
10. Dismukes, E.B. Conditioning of Ply Ash with Sulfamic Acid,
Ammonium Sulfate, and Ammonium Bisulfate. EPA-650/2-74-114,
NTIS PB 238922/AS, U.S. Environmental Protection Agency,
Washington, D.C., 1974. 50 pp.
11. Hall, H.J. SOa/HaSOi* Gas Conditioning for Control of High
Resistivity Ash Problems in Electrostatic Precipitators.
Technical Report HAR 73-35, 1973.
12. McNinch, M.O. Chemotherapy Tames Fly Ash from Low Sulfur
Coal. Electrical World; 32, January 15, 1976.
13. Harris, D.B. Personal communication, 1976.
14. Klipstein, D.H. Improved Precipitator Performance by Gas
Conditioning. Combust ion; 43, October, 1975.
15. Cook, R.E. Sulfur Trioxide Conditioning. In: Proceedings
of Symposium on Electrostatic Precipitators for Control of
Fine Particles. EPA-650/2-75-016, NTIS PB 240440/AS, U.S.
Environmental Protection Agency, Washington, D.C., 1975.
pp 315-326.
67
-------
PAPER 6
MATHEMATICAL MODELLING OF FINE PARTICLE COLLECTION
BY ELECTROSTATIC PRECIPITATION
John P. Gooch and Jack R. McDonald
Southern Research Institute
ABSTRACT
A mathematical model is described which calculates collection
efficiency in an electrostatic precipitator as a function of par-
ticle size and operating conditions. The model takes into account
the dynamics of particle collection by determining the electric
field, particle charge, and removal efficiency as functions of
position along the length of the precipitator. The model con-
tains procedures for estimating collection efficiency losses
caused by non-uniform gas velocity distributions, gas bypassing
the electrified regions, and particle reentrainment. Comparisons
of the model predictions with experimental data obtained under
essentially idealized conditions in a laboratory scale precipi-
tator indicate that the model provides a reasonable basis for
describing the precipitation process. Model predictions are
compared with field data obtained from precipitators used to
collect fly ash and the results are discussed.
INTRODUCTION
The techniques used in the sizing of electrostatic precipi-
tators for various applications differ with the manufacturer and
the application area. These techniques may be classified into
two general categories:
1. The development of statistically-based relationships
between precipitator performance and the properties of the dust
or fuel, or the operating parameters of the precipitator. Usually
this approach involves a correlation of a precipitation rate para-
meter which is used as an empirical factor in the Deutsch-Anderson
equation.
2. The use of a pilot plant and prior experience for deter-
mining the collection area required for the design efficiency at
a given application.
Although these techniques can be effective, they can also
result in serious design errors if improperly used. Furthermore,
68
-------
in order to design a precipitator for a new installation, it is
necessary either to have extensive data for similar applications,
or to conduct a test program with a pilot unit. A preferable
approach is to develop theoretical relationships to the point
that precipitator design can be performed by calculation from
theoretical principles and the properties of the dust to be
collected. This approach offers the potential for increased
confidence in design and in cost savings by preventing under-
sizing on the one hand and over-sizing on the other. The re-
liability of predictions obtained from such an approach is sub-
ject to the extent to which certain fundamental parameters are
known, the degree to which the theoretical relationships describe
precipitator operation, and the accuracy with which the factors
that correct for non-ideal conditions can be modelled and deter-
mined. At present, efficiency losses due to non-ideal conditions
are accounted for only by estimation procedures in which assumed
values of the descriptive parameters are normally used.
In order to model particulate collection by the electro-
static precipitation process from first principles, it is nec-
essary to (1) calculate electric fields as a function of position
and operating conditions, (2) calculate particle charge as a
function of time, electrical conditions, and particle size, and
(3) mathematically represent the process by which charged par-
ticles are transported from the bulk gas stream to the collection
electrode. The mathematical model described in this paper in-
cludes these operations and predicts collection efficiency as a
function of particle size and operating conditions in electro-
static precipitators.
Comparisons of the model predictions with experimental
fractional collection efficiency data obtained in a laboratory
precipitator indicate that the model provides a reasonable basis
for describing the precipitation process. Application of the
model to simulate conditions existing in full-scale and pilot-
scale precipitators used to collect fly ash leads to results
which are in reasonable agreement with experimental field
measurements. In certain cases, however, part of the ideally
calculated fractional collection efficiency curve lies below the
measured curve. This type of discrepancy can be attributed in
part to certain unmodelled and approximately modelled processes
which are important under certain conditions in precipitators.
DESCRIPTION OF THE MATHEMATICAL MODEL
Ideal Calculation of Particle Collection Efficiency
The details of a mathematical model for electrostatic pre-
cipitation and a computer program which performs the various
operations are described elsewhere in the literature.1'2'3 In
the following, a brief description of the physical processes is
69
-------
given, and the mathematical framework of the model is developed.
The first fundamental step in the model is to determine the
electric field existing in the precipitator. An accurate calcu-
lation of the electric field is important because of its influence
on particle charging and particle collection. Although the elec-
tric field in a precipitator is a function of position, the value
of the electric field used in the particle charging calculations
is approximated by the average value between the discharge and
collection electrodes. In order to calculate the velocity of
charged particles near the collecting electrode, however, it is
necessary to compute the local electric field values in this
region of space.
In the model the electric field is determined as a function
of position by the numerical relaxation technique suggested by
Leutert and Bohlen. "* Using this technique, Poisson's equation
and the current continuity equation are solved simultaneously
and subject to the existing boundary conditions. These two
equations can be written in two dimensions in discrete form as:
and
2 _ /AV Ap , AV ApN
P ~ e° \Ax Ax A^ Ay/
where
V = potential, volts
p = space charge, coul/m3
x = distance perpendicular to gas flow from wire to plate, m
y = distance parallel to gas flow from wire to wire, m.
The numerical solutions to Equations (1) and (2) are obtained
by iterating on a grid of potential and space charge density until
convergence is obtained. In the iteration procedure, the space
charge density at the wire is adjusted until solutions are found
that satisfy experimentally measured boundary conditions of
applied voltage and average current density at the plate. The
major approximation in the method is the assumption that the
motion of all charge carriers can, on the average, be described
by a single "effective mobility". in addition, the "effective
mobility" is assumed to be independent of position in formulating
Equation (2). In previous work,3 it has been shown that this
technique yields theoretical electric field distributions which
agree well with existing experimental data in the literature.
70
-------
The introduction of a significant number of fine dust par-
ticles or a heavy concentration of large particles into an electro-
static precipitator significantly influences the voltage-current
characteristics of the interelectrode space. Qualitatively, the
effect is seen by an increased voltage for a given current com-
pared to the dust-free situation. The increased voltage results
from the lowered "effective mobility" of the charge carriers
which occurs as the highly mobile gas ions are bound to the
relatively slow dust particles, thus creating a "space charge".
It is desirable to determine the space charge resulting from
dust particles because this quantity influences the electric
field distribution, especially near the collecting plate where,
for the same current, it is raised above the dust free situation.
Also, the "space charge" is a function of position along the
length of the precipitator.
It has been shown3 that the effect of "space charge" on the
electrical conditions can be estimated by using an "effective
mobility" given by:
- bi (
200
jj
(3)
where
b_ =
b. =
3m =
effective mobility, m2/(volt-sec)
ionic mobility, m2/(volt-sec)
total current density at the plate, amp/m2
= particulate current density at the plate, amp/m2
The derivation of Equation (3) is based on the assumptions
that (1) free electron current can be ignored, (2) the current
carried by the particulate is insignificant, (3) particle mobility
can be treated as independent of size and on the average, bi =
200 bp, and (4) the particles instantaneously acquire a limiting
charge given by the "saturation charge" from field charging theory.
The quantity jrp is measured experimentally and jp can be deter-
mined for a given incremental length of the precipitator from an
estimated or calculated overall mass efficiency.
In a given incremental length of the precipitator, the
effect of "space charge" on the electrical conditions can be
determined by solving Equations (1) and (2) subject to the con-
dition that:
JT = beppEp
(4)
where PpEp is the product of the space charge density and electric
71
-------
field at the plate averaged over the number of grid points used
in the direction of gas flow.
The second fundamental step in the model involves the cal-
culation of particle charge as a function of particle size, re-
sidence time, and precipitator operating conditions. Accurate
values of particle charge are necessary because the electrical
force which causes the particle to migrate towards the collection
electrode is directly proportional to the charge on the particle.
In the model particle charge is calculated using the charging
rate equation developed by Smith and McDonald5 and given by:
dt
where
and
^s
80
q =
<3s =
p
e0 =
NO =
e =
EO =
Eo =
N o ebq /
s I
4eo ^
[3ar02-r03
)9 •) 'V
Tra^evN
+ 2
(K+2)+a3 (K-l)] \
kTr02 (K+2) /
o O) / \
tra evNo / qe \
2 ex"
= 4Tre0E0a2
\ Tr£°a /
L+2 ^ }
I11 -K+2 / '
/ \
qe(r0-a)
sin 6d6
= arccos
(^)
instantaneous charge on the particle, coul
saturation charge, coul
azimuthal angle in a spherical coordinate system with
origin at the center of the particle, radians
maximum azimuthal angle for which electric field lines
enter the particle, radians
free ion density, number/m3
electronic charge, coul
permittivity of free space, cou!2/(Nm2)
average electric field between the electrodes, volt/m
72
-------
b = ion mobility, m2/(volt-sec)
v = mean thermal speed of ions, m/sec
a = particle radius, m
k = Boltzmann's constant, J/°K
T = absolute temperature, °K
t = time, sec
K = dielectric constant of the particle
ro = radial distance along 6 at which the radial component of
the total electric field is zero, m.
Equation (5) represents the sum of three charging rates cor-
responding to three different charging regions on the particle
surface. The first term corresponds to a region of the particle
surface where field charging occurs; the second, to a region
where field-enhanced diffusion occurs; and the third, to a region
where diffusion occurs with the effects of the applied electric
field neglected as an approximation. The charging rate given by
Equation (5) approaches the results obtained from the classical
field charging equation for large particles and high electric
fields and reduces to the classical diffusion charging equation
in the absence of an applied electric field.
The agreement between the results predicted by Equation (5)
and Hewitt's6 experimental data for dioctyl phthalate droplets
is within 25% over the entire range of data that is available
and is within 15% for practical charging times in precipitators.
The agreement between the theory and the experiment over the
particle size range 0.1-1.3 ym and a wide range of electric
field strengths indicates that Equation (5) should predict
closely particle charge as a function of time in the precipi-
tator model.
The third fundamental step is the calculation of particle
collection efficiency. Once the particle charge and the electric
field adjacent to the collecting electrode are calculated, the
electrical drift velocity, or migration velocity, resulting from
the coulomb and viscous drag forces acting upon a suspended par-
ticle can be obtained. For particle sizes in the size range of
interest, the time required for the particle to achieve the
steady-state value of velocity is negligible, and the migration
velocity is given by:
qE C
73
-------
where
w = migration velocity of a particle of radius a, m/sec
E = electric field adjacent to the collecting electrode,
P volt/m
C = Cunningham correction factor
y = gas viscosity, kg/(m-sec).
Gas flow velocities in most cases of practical interest are
between 0.60 and 1.8 m/sec, while theoretical migration velocities
for particles smaller than 6.0 ym are usually less than 0.3 m/sec.
The path of these smaller particles therefore tends to be domi-
nated by the turbulent motion of the gas stream in the inter-
electrode region. The classical equation for describing particle
collection in electrostatic precipitators under turbulent flow
conditions was derived by Deutsclv and gives collection effi-
ciency as a function of gas volume flow, collection area, and
migration velocity in the form:
n = 100 [1 - exp (-A w/Q) ] , (7)
where
n = collection efficiency of a particle of radius a, %
A = collecting area, m2
Q = gas volume flow, m3/sec.
The assumptions on which the derivation of Equation (7) is
based are discussed in detail elsewhere in the literature and
will not be restated here.1'8 Of greatest concern in the model-
ling of the precipitation process is the assumption that gas
turbulence provides sufficient mixing to establish a uniform
particle concentration at any cross section of the precipitator.
The validity of this assumption is most important in determining
theoretical collection efficiencies for fine particles.
The mathematical model uses the Deutsch equation to predict
the collection fraction, m j , for the i-th particle size in the
j-th incremental length of the precipitator. Thus, the Deutsch
equation is applied in the form:
-W.,j A./Q
where Wi . (m/sec) is the migration velocity of the i-th particle
74
-------
size in the j-th increment, and Aj (m2) is the collection plate
area in the j-th increment. Since the Deutsch equation is based
on the assumption that the migration velocity is constant over
the collection area of the precipitator, it is necessary to make
the incremental lengths sufficiently small that the electric
field at the plate and the charge accumulated by a given particle
size remain essentially constant over the increment.
The collection fraction (fractional efficiency) r\j_ for a
given particle size over the entire length of the precipitator is
determined from:
En. . N. .
i/D 1/3
H' J--- '
where Nj_f j is the number of particles of the i-th particle size
per cubic meter of gas entering the j-th increment. The quantity
N. . can be written in the form:
1 / j
where Nj^ i = Nj_ Q, the number of particles of the i-th particle
size per 'cubic meter of gas in the inlet size distribution.
The overall mass collection efficiency n for the entire
polydisperse aerosol is obtained from:
n = E n^ , (ID
i
where P^ is the percentage by mass of the i-th particle size in
the inlet size distribution.
Methods for Representing Non-Ideal Effects
In the preceding section, a basis for calculating ideal col-
lection efficiencies has been developed. This section will dis-
cuss the non-idealities which exist in full-scale electrostatic
precipitators and describe calculational procedures for estimating
the effects on predicted collection efficiencies. The factors of
major importance are: (1) gas velocity distribution, (2) gas
sneakage, and (3) particle reentrainment due to rapping.
Non-idealities will reduce the collection efficiency that
may be achieved for a precipitator operating with a given specific
75
-------
collecting area. Since the model is structured around the Deutsch
equation for individual particle sizes, it is convenient to re-
present the effect of the non-idealities in the model as correc-
tion factors which apply to the exponential argument of the
Deutsch equation. In the subsequent discussions, these correction
factors will be used as divisors for the theoretical migration
velocities. The resulting "apparent" migration velocities are
empirical quantities only and should not be thought of as an
actual reduction in the migration velocity in the region of
space adjacent to the collecting electrode.
Although it is widely known that a poor velocity distribution
gives a lower than anticipated efficiency, it is difficult to
apply a numerical description for gas flow quality. White8 dis-
cusses non-uniform gas flow and suggests corrective actions.
Preszler and Lajos9 assign a figure-of-merit based upon the rela-
tive kinetic energy of the actual velocity distribution compared
to the kinetic energy of a uniform velocity. This figure-of-merit
will be a measure of how difficult it may be to rectify the
velocity distribution but not necessarily a measure of how much
the precipitator performance would be degraded.
It is possible to develop an approach to estimating the de-
gradation of performance due to a non-uniform velocity distribu-
tion based upon the velocity distribution, the ideal collection
efficiencies, and the Deutsch equation.1'2 It will be assumed
that the Deutsch equation applies to each particle size with a
known migration velocity and that the specific collection area
and size of the precipitator are fixed. A constant k can be de-
termined from the Deutsch equation such that:
A W
k = -T2— = u In ( T±- ) , (12)
where A-j_ is the total inlet cross-sectional area (m2) and ua is
the average inlet velocity (m/sec). Now, the mean corrected pene-
tration p for a given particle size can be written as:
N k
1 V Ui
P = NU~ LJI ui e • (13)
a i=l
where N is the number of points in a velocity traverse and u. are
the point values of velocity (m/sec). x
For any practical velocity distribution and efficiency, the
mean penetration obtained by summation over the velocity traverse
will be higher than the calculated penetration based on an average
velocity. If an apparent migration velocity for a given particle
size is computed based upon the mean penetration and the Deutsch
76
-------
equation, the result will be a value lower than the value used
for calculation of the single point values of penetration. The
ratio of the original migration velocity to the reduced migration
velocity is a numerical measure of the performance degradation
caused by a non-uniform velocity distribution. An expression for
this ratio may be obtained by setting the penetration based on the
average velocity equal to the corrected penetration obtained from
a summation of the point values of penetration, and solving for
the required correction factor, which will be a divisor for the
migration velocity.
The correction factor "F" may be obtained from:
exp
Therefore,
(14)
u,(ln Pi '
Ci.
Whether the quantity F correlates reasonably well with statistical
measures of velocity non-uniformity is yet to be established. A
limited number of traverse calculations seem to indicate a cor-
relation between the factor F and the normalized standard devia-
tion of the velocity traverse. Figure 1 shows F as a function of
the ideal efficiency for several values of gas velocity standard
deviation. These curves were obtained by computer evaluation of
Equation 15, and the data on which the calculations are based
were obtained from Preszler and Lajos.9 The standard deviations
have been normalized to represent a fraction of the mean. The
overlapping of the curves for standard deviations of 1.01 and
0.98 indicates that the standard deviation alone does not com-
pletely determine the relationship between F and collection
efficiency.
The data in Figure 1 were used to obtain the following
empirical relationship between F, the normalized standard de-
viation of the gas velocity distribution, and the ideal collec-
tion predicted for the particle size under consideration:
1.786
F = 1 + 0.766 ncr + 0.0755 a In
where
77
-------
.0.42
0.59 0.68 1.01 0.98 1.18
1.58
2 3
CORRECTION FACTOR F
Figure 1. "F" as a function of ideal efficiency and gas flow standard deviation
78
-------
;
N
1
N Z~l (u -u.
_i a i
ua
(17)
This relationship is based on a pilot plant study, and should
be regarded as an estimating technique only. If it is desirable
to simulate the performance of a specific precipitator , the pre-
ferred procedure would be to obtain the relationship between F,
n, and 0g for the conditions to be simulated from a velocity tra-
verse at the entrance to the unit.
Gas sneakage occurs when gas bypasses the electrified areas
of an electrostatic precipitator by flowing through the hoppers
or through the high voltage insulation space. Sneakage can be
reduced by frequent baffles which force the gas to return to the
main gas passages between the collection plates. If there were
no baffles, the percent sneakage would establish the minimum
possible penetration because it would be the percent volume
having zero collection efficiency. With baffles, the sneakage
re-mixes with part of the main flow and then re-bypasses in the
next unbaffled area. The limiting penetration due to sneakage
will therefore depend on the amount of sneakage gas per section,
the degree of re-mixing, and the number of sections.
If the simplifying assumption is made that perfect mixing
occurs following each baffled section, an expression for the
effect of gas sneakage may be derived. Let:
S = fractional amount of gas sneakage per section,
n = collection fraction of a given size particle obtained
with no sneakage for total collection area,
n . = collection fraction per section of a given particle size
3 1/Ns
= 1 - (1 - n)
N = number of baffled sections, and
s
p . = penetration from section j .
Then the penetration from section one is given by:
pi = S + (1 - rij) (1 ~ S)
and from section two,
79
-------
P2 =
+ (i -
(i - s)p
= Pl [S + (1 - Tij) (1 - S)]
= [s + (i - n •) (i - s) ]2
and from section N (the last section),
s
pM = [S + (1 - n.) (1 - S)]Ns
ru ~1
O "*
= [s + (i - s) (i - n)1/Ns]Ns
(18)
Figure 2 shows a plot of the degradation of efficiency from
99.9% design efficiency versus percent sneakage with number of
baffled sections as a parameter. For high efficiencies, the
number of baffled sections should be at least four and the amount
of sneakage should be held to a low percentage. With a high
percentage of sneakage, even a large number of baffled sections
fails to help significantly.
99.9
99.5 —
U
z
III
o
LL
LL
UJ
O
UJ
Q
UJ
O
N = NUMBER OF
BAFFLED SECTIONS
0.001
1/10%
0.01 0.1
1% 10%
S, % SNEAKAGE PER SECTION
Figure 2. Degradation from 99.9% efficiency with sneakage
80
-------
A gas bypass or sneakage factor B can be defined in the form
of a divisor for the effective, or length averaged, migration
velocity in the exponential argument of the Deutsch equation.
The factor B is obtained by taking the ratio of the effective
migration velocity we under ideal conditions to the apparent
value of the effective migration velocity we' under conditions
of gas sneakage so that
B =
In (1-n)
We = In(l-n) =
V ln PNs Ng In [S + (1-S) (1 -n)
(19)
Figure 3 shows a plot of the factor B versus sneakage for a family
of ideal efficiency curves for five baffled sections. Similar
curves can easily be constructed for different numbers of sections.
0 ~TO""~ 20 30 40
S//o SNEAKAGE PER SECTION FOR A MONODISPERSE PARTICULATE
Figure 3. Correction factor for by-pass sneakage when Ns = 5
The foregoing estimation of the effects of sneakage is a
simplification in that the sneakage air passing the baffles will
not necessarily mix perfectly with the main gas flow, and the
flow pattern of the air in the bypass zone will not be uniform
and constant. The formula is derived to help in designing and
analyzing precipitators by establishing the order of magnitude
of the problem. Considerable experimental data will be required
to evaluate the method and establish numerical values of actual
sneakage rates.
81
-------
Rapping reentrainment is defined as the amount of material
tnat is recaptured by the gas stream after being knocked from the
collection plates by rapping or vibration. With perfect rapping,
the sheet of collected material would not be reentrained, but
would migrate down the collection plate in a stick-slip mode,
sticking by the electrical holding forces and slipping when re-
leased by the rapping forces. However, the rapping forces are
necessarily large to overcome adhesion forces, and much of the
material is released into the gas stream as sheets, agglomerates,
and individual particles. Most of the material is recharged and
recollected at a later stage in the precipitator.
Based upon the simplifying assumptions that a fixed fraction
of the collected material of a given particle size is reentrained,
and that the fraction does not vary with length through the pre-
cipitator, an expression can be derived identical in form to that
obtained for gas sneakage:1'2
PR = [R + (1-R) (l-n)1/NR]NR (20)
where
p = penetration corrected for reentrainment
i\
R = fraction of material reentrained
N = number of stages over which reentrainment is assumed to
occur
n = collection fraction of a given particle size obtained
with no reentrainment.
Since Equations (18) and (20) are of the same form, the
effect of reentrainment can be expected to be similar to the
effect of sneakage, provided that a constant fraction of the
material is always reentrained. It is doubtful that such a con-
dition exists, since precipitators frequently use different
rapping programs on different sections, agglomeration occurs dur-
ing collection, and different holding forces exist in different
sections. However, until sufficient data on rapping losses per
section as a function of particle size can be accumulated, the
relationship may be used to estimate the effect of rapping re-
entrainment on precipitator performance.
Figure 4 shows the effect on resultant efficiency for a
given size particle of various degrees of reentrainment for a
four-section precipitator with the indicated values of no-
reentrainment efficiency.
Since reentrainment and sneakage effects are estimated with
identical mathematical expressions, a combined correction factor
B' is used in the mathematical model. From input values of the
82
-------
99.9
REENTRAINMENT PER SECTION -
20 30 40 50 60 70 80 90
'0 OF COLLECTED DUST REACHING HOPPER
100
Figure 4. Effect of reentrainment on the efficiency of a four-section precipitator
designed for a no-reentrainment efficiency as indicated for a monodisperse
particu I ate
83
-------
fraction of material assumed to be lost by reentrainment and
sneakage, and the number of stages over which losses are assumed
to occur, B1 is determined from the ideal collection fraction for
each particle size.
In summary, the mathematical model takes into account the
non-ideal effects of non-uniform gas velocity distribution, gas
sneakage, and rapping reentrainment by reducing the ideally cal-
culated migration velocities we by the correction factors F and
B1. An "apparent" migration velocity we' is determined for each
particle size from:
w
w
e . (21)
e F-B1
Using we', the corrected fractional collection efficiencies are
calculated.
EXPERIMENTAL LABORATORY MEASUREMENTS AND COMPARISON
WITH MODEL PREDICTIONS
Laboratory-Scale Precipitator
A laboratory-scale precipitator was constructed for the pur-
pose of studying collection of fine particulate under idealized
conditions. The test unit has an electrode irrigation capability
and provisions for humidifying flue gas produced from a gas-fired
burner. For the experiments described here, however, the carrier
gas was ambient air, and the particulate source was an atomizer
which produced a polydisperse aerosol of dioctyl phthalate (DOP).
Electrode irrigation and humidification were not used, and the
collected oil droplets drained from the collection electrodes by
gravity.
Figure 5 is a schematic drawing of the laboratory precipi-
tator assembly. Wire-plate geometry is employed, with plate-to-
plate spacing and wire-to-wire spacing of 12.7 cm (5 in.). Dis-
charge electrodes consist of 0.269 cm (0.106 in.) diameter wires.
The collecting plate area is 2.32 m2 (25 ft2) which results in a
specific collecting area of 35.0 rn2/(m3/sec) (178 ft2/1000 cfm)
at a gas velocity of 1.37 m/sec (4.5 ft/sec). The precipitator
is divided into four equal-length sections separated by baffles
to limit bypassage, or sneakage, through the hoppers and the
region above the collecting electrode. Three power supplies are
used, the first two of which each energize a 76.2 cm (30 in.)
length section. The third power supply is connected to the last
two sections.
Gas bypassage was estimated by measuring gas velocity with a
thermal anemometer in the hoppers and in the region above the
collection electrodes. The estimated amount of bypassage, ex-
pressed as a fraction of the total gas flow through the unit, was
84
-------
POWER SUPPLIES
MIXING CHAMBER
SPRAY
TOWERS
BLOWER
Figure 5. Schematic drawing of the laboratory precipitator
between 8% and 10% for the two values of gas flow reported here.
Gas flow uniformity was determined by conducting a traverse with
the thermal anemometer at the precipitator inlet. The standard
deviation of the velocity distribution, expressed as a percentage
of the average velocity, was 11.2% for an average gas velocity of
1.09 m/sec.
Measurement Technique
Fractional efficiency measurements were conducted using a
five-stage Brink impactor at the precipitator inlet and outlet.
Aluminum foil substrates were used to facilitate weighing of the
small masses collected on the individual stages. The mass de-
terminations were made with a Cahn Electrobalance. For the
experiments conducted at the lower gas velocity, a sampling time
of six hours was required at the precipitator outlet to obtain
weighable quantities on the impactor substrates. Six and four
point traverses were conducted with the impactor at the inlet
and outlet, respectively, in order to insure that a representative
size distribution was obtained at each location.
85
-------
Since the particulate exiting from an electrostatic pre
cipitator will have an electrical charge, it was necessary to
employ a charge neutralizer at the inlet of the sampling line
for all measurements at the precipitator outlet. Measurements
with an optical particle counter indicated that, if metal sub-
strates were used in the impactor, the charge acquired by the
particulate substantially influenced the size distribution ob-
tained with the impactor. Since the impactor was calibrated with
an uncharged aerosol, the use of the charge neutralizer minimized
sizing errors due to electrostatic attraction.
Table 1 gives a typical set of inlet and outlet data ob-
tained from duplicate experiments at the indicated conditions.
Good reproducibility was obtained for all stages at the inlet,
and for stages 2 through 5 at the outlet. Isokinetic sampling
was not employed, since the effect of anisokinetic conditions
on particles with diameters less than 2 ym is expected to be
negligible. The effect of changing the gas flow rate through
the impactor on the indicated size distribution was checked by
reducing the flow rate from 56.6 cm3/sec to 28.3 cm3/sec. The
higher flow rate, which was used for all of the fractional
efficiency determinations, resulted in a gas velocity of 0.76
m/sec at the sample point. Figure 6 shows the inlet size dis-
tributions from Table 1 and the distribution obtained at the
10.0
DIAMETER, Mm
o .->
'_» 0
A
• a
Ab
*c
• d
• e
FROM TABLE 1
REDUCED FLOW RATE
*
4
A'
M
•
*B
»
**
A
»
*¥
•
0.01 0.1 1 10
% SMALLER THAN INDICATED SIZE
50
80 90
Figure 6. Inlet particle size distributions from Brink data
86
-------
TABLE 1.. TYPICAL INLET AND OUTLET PARTICLE SIZE DATA FOR WET ESP EXPERIMENTS
(OOP Sprayer, Gas Velocity = 1.37 m/sec, Current Density = 55 nA/cm2)
CO
Mass Loading, rag/ am3
Test No.
Stage
1
2
3
4
5
Lower Size
Limit, urn
3.04
1.80
1.24
0.65
0.46
2ai
26.341
18.659
7.249
7.448
2.954
2bi
28.956
13.526
6.285
6.984
1.775
Inlet
2ci
24.596
14.128
7.238
6.743
2.514
2di
24.088
15.226
8.226
7.419
2.654
Average
25.
15.
7.
7.
2.
995
385
250
148
474
Outlet
2ao 2bo
0.0414 -^0
0.1494 0.1103
0.2575 0.2115
0.5724 0.4782
0.3862 0.3770
Geometric
Mean
Diameter.
Average pm
0.
0.
0.
0.
0.
207
1299 2.34
2345 1.49
5253 0.90
3816 0.55
Collection
Efficiency,
-
99.16
96.77
92.65
84.58
-------
same location with the reduced impactor flow rate. These data
indicate that the same size distribution is obtained at either
flow rate. The higher flow rate is more desirable because of
the smaller cut points and the increased amount of gas that may
be sampled per unit time.
Experimental and Theoretical Results
Fractional efficiency measurements were performed with the
Brink impactor and DOP aerosol at current densities of 26.9,
53.8, and 107.5 nA/cm2 at a gas velocity of 1.37 m/sec, and at
current densities of 26.9 and 53.8 nA/cm2 at a gas velocity of
0.72 m/sec. The results obtained from these experiments, ex-
pressed as collection efficiency as a function of particle size,
are presented in Figures 7, 8, and 9. Computed collection
efficiencies obtained with the mathematical model are also shown.
In view of the difficulties in making these types of measurements,
the agreement between measured and calculated efficiencies shown
99.9
I I I
j = 26.9 nA/cm2, v = 1.37 m/sec
— A j = 53.8 nA/cm2, f = 1.37 m/sec
j = 107.5 nA/cm2, v = 1.37 m/sec
*, A, • — EXPERIMENTAL
DATA
PARTICLE DIAMETER, Mm
10.0
Figure 7. Experimentally measured and ideal calculated fractional collection efficiencies
in the laboratory precipitator
88
-------
99.99
99.98
99.95
99.9
35
O
g 99.8
o
Ul
S 99-5
o
ai
J 99.0
O
O
98.0
95.0
90.0
1 1 1 I I I I |
1 IJINII
THEORETICAL / /
'CORRECTED
/ FOR 8%
t SNEAKAGE
j = 53.8 nA/cm2
v = 0.72 m/sec
0.1
EXPERIMENTAL
i i i
1.0
PARTICLE DIAMETER,]
10.0
Figure 8. Experimentally measured, ideal calculated, and sneakage-corrected fractional
collection efficiencies for a current density of 53.8 nA/cm2 and a gas velocity
of 0.72 m/sec in the laboratory precipitator
89
-------
*
0
HI
o
u_
u.
LU
.LECTION
•wl
O
O
99.99
99.98
99.95
99.9
99.8
99.5
99.0
98.0
95.0
90.0
0
1 1 1 | 1 1 II 1 1 1 1 I 1 Ml
„,-„ A/ 2 THEORETICAL
j = 26.9 nA/cm^ .
v = 0.72 m/sec 1
r // ~
/ /
/ •
/ '
//
/ /
/ <*'
/ / CORRECTED FOR
• / 8%SNEAKAGE
/ / O
1 t
/ /* Q EXPERIMENTAL
- //o —
m/ '/*'
.1 1.0 10.
PARTICLE DIAMETER, Wm
Figure 9. Experimentally measured, ideal calculated, and sneakage-corrected fractional
collection efficiencies for a current density of 26,9 nA/cm2 and a gas velocity
of 0.72 m/sec in the laboratory precipitator
90
-------
in Figure 7 is considered good. Figures 8 and 9 suggest that the
agreement obtained between computed and measured results decreases
with^decreasing gas velocity and current density. The assumption
of 8% gas sneakage over four stages improves agreement between
computed and measured results at a gas velocity of 0.72 m/sec.
However, the data obtained at this lower gas velocity with 26.9
nA/cm are considerably below the computer projections. Possible
causes of this lack of agreement are unmodelled effects such as
non-uniform current density and electric field, and particle con-
centration gradients in the inter-electrode space.
Another series of experiments is being conducted at the
present time in which fractional efficiencies are being measured
for different wire-to-plate spacings and wire sizes. Figure 10
shows experimental and theoretical fractional collection effi-
ciencies and migration velocities for a plate-to-plate spacing
of 19.05 cm (3.75 in.), wire radius of 0.1588 cm (0.125 in.), and
gas velocity of 1.46 m/sec (4.79 ft/sec). These results indicate
that the model predictions give good agreement with experimental
data for current densities of 26.9 nA/cm2 or less. However, for
current densities greater than 26.9 nA/cm2 it appears that the
agreement worsens and, in fact, the ideal calculated fractional
99.9
99.8
o
HI
o
ul
LL
0!
O
ai
O
O
99
2 95
90
80 —
60 —
30 —
1 ' ' I 'A
NEGATIVE CORONA T
GAS VELOCITY = 1.46 m/sec |
O EXPERIMENTAL
j = 107.5/^A/m2
O EXPERIMENTAL,
j = 268.8/jA/m2
A EXPERIMENTAL
j = 430jzA/m2
— THEORETICAL
•-THEORETICAL
8%SNEAKAGE
j I I
I
I I I
28.0
24.0
20.0
o
1C.O
12.0
8.0
4.0
h-
§
LU
z
o
cc
(2
0.1 1.0 10.0
GEOMETRIC MEAN DIAMETER, am
Figure 10. Experimental and theoretical data for a wire of radius 1.488 x 10~3m
91
-------
efficiency curve lies below the measured curve.
The disagreement in Figure 10 between theory and experiment
for current densities greater than 26.9 nA/cm2 occurs at a plate
spacing and gas velocity which are not too different from those
encountered in the field. Since space charge effects are not
significant in these experiments, it might be suspected that
particle concentration gradients exist under certain conditions
and that in these cases the collection mechanism in the model is
not strictly applicable.
Measurements have been made at the outlet of the laboratory
scale precipitator under several operating conditions to deter-
mine whether or not concentration gradients existed. A "Climet"
optical particle counter was used to count 1.0, 0.5, and 0.3 ym
diameter particle sizes at several positions located between the
collecting plate and the wire. The measurements at the 12.7-cm
plate spacing under all conditions presented here showed no
build-up of particles near the collecting electrode. These
results are consistent with the fact that the model yields good
agreement with experiment under all conditions at the 12.7-cm
plate spacing. The measurements at the 19.05-cm plate spacing
showed definite concentration gradients existing for all three
particle sizes under several conditions. In these cases, the
concentration was greatest near the collecting plate and de-
creased towards the discharge electrode. This type of behavior
might be expected due to the directed coulomb motion of a
particle which is superimposed on the turbulent motion of the
gas stream. The ratios of the concentration near the collecting
plate to that near the wire for 1.0, 0.5 and 0.3 ym diameter
particles were 1.7, 1.3, and 1.3, respectively, at approximately
43.0 nA/cm2. The causes for such gradients, the conditions under
which they would exist, and their effects on predicted collection
efficiency of fine particles need further investigation.
FIELD MEASUREMENTS AND COMPARISON WITH MODEL PREDICTIONS
In this section the predictions of the mathematical model
are compared with experimental field data for several different
precipitators. These precipitators were involved in the col-
lection of coal fly ash under various operating conditions. Also,
"typical" performance curves are generated using the model for
"cold" precipitators with operating current densities from 5 to
40 nA/cm2 and for a "hot" precipitator operating at a current
density of 30 nA/cm2.
Figure 11 gives a comparison of fractional effective migra-
tion velocities obtained from the model with those obtained from
inertially-determined fractional efficiency measurements on a
coal-fired power boiler. This "cold" full-scale precipitator
has a low specific collection area (SCA) and moderate current
density. The particulate under collection had a mass median
92
-------
o 28.0
0)
i 24.0
g 20.0
_i
LU
> 16.0
< 12.0
cc
8.0
O
4.0
•CALCULATED FROM IIMERTIALLY
DETERMINED FRACTIONAL
EFFICIENCY MEASUREMENTS
COMPUTED AT
20 nA/cm2
S = 0, ag = 0.25
J L
0.1 0.2 0.4 1.0 2.0
PARTICLE DIAMETER, w
4.0
10.0
Figure 11. Effective migration velocities for a full-scale precipitator on a coal-fired
boiler, SCA = 55.7 m2/(m3/sec)
diameter (MMD) of approximately 25 ym and a low value of re-
sistivity. In addition to the ideal predictions, Figure 11
shows the effect of correcting for a gas velocity standard de-
viation (ag) of 0.25. A comparison of the we values in Figure
11 indicates that the theoretical predictions are low for the
particle diameters below 2.0 ym and high for larger particles.
Due to limitations in the measurement technique, no data are
available for particle diameters larger than 5.0 ym. Figure
12 shows the fractional efficiency data from which the we values
of Figure 11 were calculated. The overall mass efficiency pre-
dicted by the model was 99.8%, as compared to a measured value
of 99.6%.
Figures 13 through 18 give results predicted by the model
in terms of overall mass efficiency as a function of SCA for
current densities of 5, 10, 15, 20, and 40 nA/cm2 at 150°C, and
for a current density of 30 nA/cm2 at 370°C. The relationships
presented in these figures are based on a 22.86-cm (9-in.) plate
spacing, and secondary volt-amp curves and particle size dis-
tribution (MMD = 25 ym) considered to be typical for a coal-fired
power boiler. Also given are test results obtained under con-
ditions approximating the electrical conditions represented by
the given values of current density. A comparison of the limited
amount of applicable test data with the computed results indicates
that the ideal predicted overall mass efficiencies are generally
higher than those obtained from the field measurements. The mea-
surements were taken with sampling techniques which insured that
essentially all of the mass for particles larger than 0.3 ym in
diameter was captured by the sampling device.
93
-------
5?
o
UJ
U
ul
u.
UJ
z
O
O
LU
O
O
99.99i
99.90
99.8
99
98
95
90
80
60
0.1
• EXPERIMENTAL
DATA
COMPUTE DAT
20 nA/cm2
S = 0, CT = 0.25
_L
J L
1.0
PARTICLE DIAMETER, Mm
10.0
Figure 12.
Fractional collection efficiencies for a full-scale precipitator on a coal-fired
power boiler
100 200 300 400 500 600 700 800
SPECIFIC COLLECTING AREA, ft2/(1000 ft3/min)
Figure 13. Computed performance curves at 5 nA/cm2
94
-------
99.99]
99.98
99.95
* 99.9
o
Z
UJ
o
99.8
S 99.5
£ 99.0
o
HI
IJ 98.0
8
95.0
90.0
80.0
I
I
I I
100 200 300 400 500 600 700 800
SPECIFIC COLLECTING AREA, ft2/(1000 ft3/min)
Figure 14. Computed performance curves at 10 nA/cm2
95
-------
90.0
100 200 300 400 500 600 700 800
SPECIFIC COLLECTING AREA, ft2/(1000 ft3/min>
Figure 15. Computed performance curves at 15 nA/cm2
96
-------
99.99
100 200 300 400 500 600
SPECIFIC COLLECTION AREA, ft2/(1000 ft3/min)
Figure 16. Computed performance curves at 20 nA/cm2
97
-------
99.99
95
100 200 300 400 500 600
SPECIFIC COLLECTING AREA, ft2/(1000 ft3/min)
Figure 17. Computed performance curves at 40 nA/cm2
98
-------
99.99
90 __
100 200 300 400 500 600
SPECIFIC COLLECTING AREA, ft2/(1000 ft3/min)
Figure 18. Computed performance curves for "hot" precipitator
The use of the empirical correction factors reduces the
computed values of overall mass collection efficiency to the
range of values obtained from the field measurements. Reentrain-
ment and sneakage effects were combined by assuming that the
indicated fractional losses per stage occurred over four
effective stages. Note that high efficiency (>99.5%) precipi-
tators operating on the low-temperature side of the air heater
fall reasonably close to the computed line obtained with 0g =
0.25 and S = 0.1. A gas velocity distribution with a standard
deviation of 0.25 is generally considered to be a good distri-
bution for a full-scale unit. The computed results show that
a poor distribution (erg = 0.5) seriously degrades performance.
The detrimental effects of sneakage and reentrainment are also
indicated in the decrease of computed performance resulting
from variations in the parameter S.
Figures 19 and 20 show results obtained from the analysis
of a "cold" full-scale precipitator having a large SCA [560 ft2/
(1000 ftVmin)] and low current density (12 nA/cm2) . The ex-
perimental results were obtained from inlet and outlet size
distribution measurements obtained using inertial, optical, and
electrical sizing techniques. The precipitator is preceded by
99
-------
0.01
0.05
o o
<
cc
10
LU
Z 30
LU
60
90l
0.01
THEORETICAL
"
• IMPACTORS (W/O RAPPING PUFFS)
A OPTICAL \ MO RAP DATA
OEAA f
i
0.1 1
PARTICLE DIAMETER, urn
99.99
99.9
99
90
o
LU
O
LL
LL
LU
z
60 2
o
LU
40
__J10
10
O
o
Figure 19. Measured and theoretical fractional efficiency
20i
u
-s
u
a
o
2 10
z
o
I-
ol—
0.01
I
o o o
• IMPACTORS (W/O RAPPING PUFFS)
AOPTICAL
OEAA
NO RAP DATA
THEORETICAL.
I
I
0.1 1.0
PARTICLE DIAMETER, urn
10.0
Figure 20. Effective migration velocity vs particle diameter
100
-------
a mechanical collector so that the inlet size distribution con-
sisted of finer particles with an HMD of approximately 3 ym. The
fly ash under collection had a high resistivity resulting in
limitations on the current density. The measurements were
performed with no rapping so that reentrainment should be mini-
mized. The ideally predicted fractional efficiency and migration
velocity curves lie below the experimental data for particle
sizes in the 0.06-2.5 ym range. Over this size range the theo-
retical curve has the same shape as the experimental data with
the minimum occurring at about 0.4 ym in both cases. The over-
all mass efficiency predicted by the model is 99.7% as compared
to the measured value of 99.9+%.
Figures 21 and 22 show results obtained from a study of a
"cold" pilot precipitator operating at various SCA values and
current densities ranging from 7 to 40 nA/ft2. The experimental
results were obtained from inlet and outlet measurements with
mass trains using in-stack filters. Particle size distribution
measurements made at the pilot precipitator inlet with inertial
impactors indicated that a typical fly ash distribution was ob-
tained and that for particle diameters below about 5 ym, the
cumulative mass loadings obtained in the main duct were within
the range of those obtained at the pilot plant inlet.
Both Figures 21 and 22 indicate that in order to match the
test data with the predicted model results at comparable current
densities, it is necessary to either assume a poor gas velocity
distribution (a = 0.50) or relatively large losses due to re-
entrainment and gas bypassage. The bypassage and reentrainment
losses are assumed to occur over three effective stages, since
the pilot unit consisted of three sections. Gas flow measure-
ments with a thermal anemometer indicated that at this installa-
tion gas flow quality was poor (a > 0.46), and that gas sneakage
was estimated to be no smaller than 8% and no greater than 30%
of the total gas flow. Resistivity data from in-situ measure-
ments and voltage-current curves from the pilot unit power
supplies indicated that the current densities of 11 and 32 nA/
cm2 could be sustained without experiencing electrical break-
down of the dust layer deposited on the collection electrodes
under the test conditions.
Figure 23 shows results obtained from the analysis of a
"hot" full-scale precipitator having an SCA of 430 ft2/(1000 ft3/
min) and current density of 35 nA/cm2. The experimental results
were obtained from inlet and outlet size distribution measure-
ments obtained using inertial, optical, and diffusional sizing
techniques. For this installation, the fractional efficiency
curve obtained from, the measurements is lower than the theoretical
curve obtained from the mathematical model. Although the average
electrical operating conditions for this unit are used in the
model, some of the electrical readings for individual sets were
characterized by anomalously low current and voltage levels,
101
-------
COMPUTED
CD = 30 MA/ft2
• EXPERIMENTAL o
200 500 800
SPECIFIC COLLECTING AREA, ft2/1000 cfm
Figure 21. Computed and measured data from pilot precipitator
102
-------
> 99.!
Z
LU
o
LL
U-
LJJ
99.5
O
LU
-I
O
o
99
95
90
I
COMPUTED
CD = 10 MA/ft2
• EXPERIMENTAL
DATA
I
I
100 200 500 700
SPECIFIC COLLECTING AREA, ft2/1000 cfm
Figure 22. Computed and measured data from pilot precipitator
103
-------
LL
LU
O
LU
O
o
99.98
99.9
99.5
98
95
90
60
A
301—
0.05
0 C£ °
MEASUREMENT METHOD:
ACASCADEIMPACTORS
O OPTICAL PARTICLE COUNTERS
• DIFFUSIONAL
PRECIPITATOR CHARACTERISTICS:
TEMPERATURE - 335°C
SCA - 85 m2/(m3/sec)
CURRENT DENSITY - 35 nA/cm2
I L
0.1
0.5
1.0
5.0
10.0
PARTICLE DIAMETER, Mm
Figure 23. Comparison of measured and computed fractional collection efficiencies
suggesting the possibility of electrode misalignment. This is a
contributing factor in the disagreement between theoretical and
measured results, since the average electrical conditions will
not adequately represent the effects of low voltages and current
in individual sets on collection efficiency. The theoretically
predicted overall mass efficiency was 99.9% at the indicated
condition, and the average measured collection efficiency was
99.3%. If a gas velocity distribution standard deviation of
0.25 is assumed, sneakage and reentrainment losses of 10 to 20%
over three stages are required to reconcile the computed and
measured mass efficiency.
CONCLUSIONS
Comparisons of theoretically calculated collection effici-
encies with laboratory data obtained under essentially idealized
conditions indicate that the mathematical model gives good agree-
ment with experiments for particle sizes from 0.55 to 2.5 ym in
diameter over a wide range of current densities at gas velocities
of 0.72 and 1.37 m/sec, with a 12.7-cm plate spacing. For a
19.05-cm plate spacing and gas velocity of 1.46 m/sec, the model
predicts ideal fractional efficiency curves which lie below the
measured data for current densities greater than 25 nA/cm2.
This discrepancy may at the present be attributed to unmodelled
concentration gradients which were measured for 0.3, 0.5, and
1.0 ym diameter particles under these conditions.
104
-------
Comparisons of model predictions with field data obtained
from full-scale and pilot precipitators indicate that the theory
generally gives a good prediction of overall mass collection
efficiency although, in certain cases, the fine portion of the
fractional efficiency curve lies below the measured data. The
laboratory measurements obtained to date indicate that the
cases in which the model underpredicts fine particle collection
may be due to concentration gradients which exist under certain
conditions.
Considering the broad range of laboratory and field data
with which the theory has been compared, it can be concluded
that the mathematical model provides a basis for indicating
performance trends caused by changes in specific collecting
area, electrical conditions, and particle size distribution, pro-
vided that back corona does not exist. Current density, applied
voltage, and the particle size distribution are the most im-
portant variables in the calculation of overall mass collection
efficiency for a given specific collection area. The theoretical
calculation of ideal overall collection efficiency of polydis-
perse particulate in an electrostatic precipitator generally
gives results higher than those obtained from performance mea-
surements on coal-fired power boilers. Corrections to the
idealized or theoretical collection efficiency to estimate the
effects of non-uniform gas flow, rapping reentrainment, and
gas bypassing the electrified sections reduce the overall values
of calculated efficiency to the range of values obtained from
field measurements. These calculations suggest that the theo-
retical model may be used as a basis for quantifying performance
under field conditions if sufficient data on the major non-
idealities become available.
105
-------
ACKNOWLEDGEMENTS
The work described in this paper was supported under con-
tracts with the Environmental Protection Agency, the Electric
Power Research Institute, and the Tennessee Valley Authority.
REFERENCES
1. Gooch, J. P., and N. L. Francis. A Theoretically Based
Mathematical Model for Calculation of Electrostatic Pre-
cipitator Performance. J. Air Pollution Control Assoc.
25_ (2) :108-113, 1975.
2. Gooch, J. P., J. R. McDonald, and S. Oglesby, Jr. A Mathe-
matical Model of Electrostatic Precipitation. EPA-650/2-75-
037. NTIS PB 246188/AS, U.S. Environmental Protection
Agency, Washington, B.C., 1975. 162 pp.
3. Gooch, J. P., and J. R. McDonald. Mathematical Modelling of
Fine Particle Collection by Electrostatic Precipitation.
1975 Air Symposium (in press) Amer. Inst. Chem. Eng.
4. Leutert, G., and B. Bohlen. The Spatial Trend of Electric
Field Strength and Space Charge Density in Plate-Type
Electrostatic Precipitators. Staub 32 (7):27, 1972.
5. Smith, W. B., and J. R. McDonald. Calculation of the Charging
Rate of Fine Particles by Unipolar Ions. J. Air Pollution
Control Assoc. 25 (2):168-172, 1975.
6. Hewitt, G. H. The Charging of Small Particles for Electro-
static Precipitation. Trans. Amer. Inst. Elec. Eng. 76,
Part 1:300-306. 1957. ~~
7. Deutsch, W. Ann. Phys. (Leipzig) 6_8:335, 1922.
8. White, H. J. Industrial Electrostatic Precipitation.
Addison-Wesley, Reading, Mass., 1963.
9. Preszler, L., and T. Lajos. Uniformity of the Velocity Dis-
tribution upon Entry into an Electrostatic Precipitator of
a Flowing Gas. Staub 32 (ll):l-7, 1972.
106
-------
PAPER 7
TECHNIQUES FOR CONDITIONING FLY ASH
Edward B. Dismukes
Southern Research Institute
ABSTRACT
Conditioning of fly ash is discussed in 'connection with elec-
trostatic precipitators from several points of view: the purposes
served, the types of chemicals used, and their mechanisms of action.
Major emphasis is given to resistivity modification with sulfur
trioxide and other agents. Comments are also made on alternative
conditioning mechanisms: increasing the cohesiveness of deposited
fly ash to minimize reentrainment and improving the electrical pro-
perties of flue gas to increase the efficiency of collection.
INTRODUCTION
Conditioning of fly ash usually involves the injection of a
chemical (most often sulfur trioxide) into flue gas to control the
electrical resistivity of the ash and improve its collection in an
electrostatic precipitator. Currently, however, the term is as-
suming a much broader meaning, for conditioning involves processes
other than resistivity modification. This paper discusses condi-
tioning to achieve resistivity modification, but it also deals with
two other purposes of conditioning: to increase the cohesiveness
of fly ash particles deposited in a precipitator and to improve
the electrical properties of the gas stream flowing between dis-
charge wires and collection electrodes.
RESISTIVITY MODIFICATION: TREATMENT OF ASH FROM LOW - SULFUR COAL
Conditioning By Sulfur Trioxide
Origin of High Resistivity. The production of fly ash with a high
electrical resistivity is the usual—but not inevitable—conse-
quence of the combustion of a low-sulfur coal. The problem of
high electrical resistivity is clearly associated with the com-
bustion of Western coals containing around 0.5% of sulfur. It is
widely assumed to occur when Western coals of relatively low-sulfur
content (often around 1%) are substituted for the more prevalent
Eastern coals containing 3 to 4% sulfur. However, in terms of the
107
-------
frequently quoted upper limit of acceptable resistivity values
(about 1 x 10 * ° ohm-cm), ash from Eastern coals presents less of
a difficulty than ash from Western coals. There are at least two
factors that control resistivity: one is the sulfur content of the
coal, and the other is the overall elemental composition of the fly
ash.
Sulfur occurs in coal as organic compounds and as inorganic
compounds—notably pyrite and sulfate salts. Regardless of the dis-
tribution of the sulfur in these different forms, however, the pre-
dominant product of combustion in the boiler is sulfur dioxide.
From a thermodynamic point of view, sulfur dioxide would be the only
product found in the boiler, and sulfur trioxide (or, more exactly,
sulfuric acid) would be the only product present when the flue gas
reaches a temperature around 150° C where electrostatic precipita-
tors normally operate,1 as shown in Table 1. Thermodynamics evi-
dently is the process-limiting factor in the boiler but not toward
the end of the flue-gas train, for the kinetics of oxidation does
not allow more than a small fraction of the sulfur oxides to appear
as sulfur trioxide.
TABLE 1. EQUILIBRIUM DISTRIBUTION OF OXIDIZED FORMS
OF SULFUR AT VARIOUS GAS TEMPERATURES S
Temperature Relative concentrations, %, of sulfur compounds
°C ' S02 S03
1400
1000
800
600
500
400
300
200
100
99.7
97-6
88.7
42.2
12.3
1.6
0.1
0.0
0.0
0.3
2.4
11.3
57.7
87.3
94.8
59.3
6.5
0.0
0.0
0.0
0.0
0.1
0.4
3.6
40.6
93.5
100.0
a. Calculated from the data in JANAF Tables,lassuming con-
centrations of oxygen and water vapor equal to 4% and 10?
by volume, respectively.
b. The maximum absolute concentration of each compound is
sharply limited below 300° C as a result of the condensa-
tion of the predominant compound, HaSOit, in a binary
liquid mixture.
108
-------
Even so, the small fraction of the total of sulfur oxides oc-
curring as sulfur trioxide at 150° C can be sufficient to lower
the resistivity of ash to an acceptable range. Obvious factors of
importance are the available concentration of sulfur dioxide
(roughly proportional to the amount of sulfur in the coal) and the
extent of oxidation to sulfur trioxide. A less obvious factor is
the nature of the interaction of sulfur trioxide with the fly ash.
We are only beginning to gain answers to this last question, but
this much seems clear: sulfur trioxide and water vapor are jointly
adsorbed or condensed on the surfaces of fly ash particles, and
either vapor assists in the deposition of the other. The result-
ing acid may remain essentially intact in a chemical sense, pro-
ducing a conductive surface layer on an otherwise poorly conducting
substrate. The acid may, however, react with basic constituents
of the ash and thus undergo conversion to a nonconducting layer of
sulfate salts. Calcium oxide, a fairly abundant component of ash
from Western coals, is a probable cause of acid neutralization and
its nullification as a conductor. Thus, two coals similar in sul-
fur content may ultimately produce similar concentrations of sulfur
trioxide, but the one producing a more alkaline ash is likely to
have a substantially higher resistivity.
Conditioning by Naturally Occurring Sulfur Trioxide. Some of the concepts
outlined above were verified2 in a study of the natural condition-
ing of fly ash in a plant burning an Eastern coal containing
about 2% of sulfur and producing an ash in which the total per-
centage of alkaline oxides was low, about 7% by weight (4.2% as
the alkali metal oxides Li2O, NaaO, and KaO, and 2.4% as the alka-
line earth oxides MgO and CaO). Experimental data showing fly ash
resistivity in a range of temperatures at the precipitator inlet
are shown in Figure 1. The upper curve shows the expected
maximum—in this instance, around 175° C—that occurs as the re-
sult of the transition between volume conduction at high tempera-
tures and surface conduction at low temperatures. The lower seg-
ment of a different curve indicates the effect produced by injecting
water vapor in an amount sufficient to double the normal level—i.-6..
increasing the concentration of water vapor from 7% to 14% by volume
Other experimental data of value in explaining the observed
changes in resistivity are given in Figures 2 and 3. Measured con-
centrations of sulfur trioxide in gas samples taken upstream and
downstream from the air preheater indicate the occurrence of two
effects: (1) the thermodynamically predicted increase in the oxi-
dation of sulfur dioxide to sulfur trioxide as the gas temperature
falls across the air heater (but short of the extent predicted for
equilibrium) and (2) the loss of sulfur trioxide from the gas phase,
a phenomenon aided by either the lowering of temperature or the in-
jection of water vapor. Measurements of slurries of fly ash in
water show corresponding increases in the acidity of the ash (de-
creases in pH) and in the sulfate content of the ash.
109
-------
1013
u
E
10
12
10
11
WITHOUT H2O INJECTION
WITH H2O INJECTION
e
140 150 160 170 180
TEMPERATURE, °C
190
200
Figure 1. Electrical resistivity of fly ash as a function of temperature or water vapor
concentration
Conditioning by Injected Sulfur Trioxide. In plants where the naturally
available concentration of sulfur trioxide is too low—in an
absolute sense, as the result of a low sulfur concentration in
the coal or, in a relative sense, as the result of reaction with
a highly alkaline ash—the conditioning agent can be introduced
by an artificial process. Sources of sulfur trioxide are vapor
generated by vaporizing the liquid form of this compound, by
vaporizing sulfuric acid, or by oxidizing sulfur dioxide gas on
a vanadium pentoxide catalyst. In the last-mentioned type of
process, sulfur dioxide may be purchased as the liquid compound
under pressure and evaporated prior to conversion, or it may be
generated at the site in a burner for elemental sulfur. The mo-
lecular composition of vapors injected in the various systems de-
pends not only on the source material but on the temperature and
water vapor concentration in the dilution air. The vapor may be
sulfur trioxide per se, or it may be sulfuric acid. It is of
little consequence which substance is injected for, once injected,
either compound will occur only as sulfuric acid vapor in the gas
stream. This statement stems from the rapid rate of the gas phase
reaction S03 + H20—*-H2SCH under typical flue-gas conditions (tem-
peratures around 150° C and water-vapor concentrations around 7%
to 10% by volume), to approach equilibrium conditions as indicated
in Table 1.
110
-------
20
15
a
a
A
O
CO
o
h-
QC
H
Z
LLJ
u
o
o
10
UPSTREAM FROM-
PRECIPITATOR
UPSTREAM
FROM AIR
PREHEATER
O WITHOUT H2O INJECTION
• WITH H2O INJECTION
140 150
160 170 180
TEMPERATURE, °C
190
330
340
Figure 2. Concentration of sulfur trioxide as a function of sampling location or
temperature
111
-------
a
140 150 160 170 180 190
TEMPERATURE, °C
0.7
S? 0.6
H
Z
LU
I-
1 0.5
ui
LL
D
CO
0.4
0.3
I I
140 150 160 170 180 190
TEMPERATURE, °C
Figure 3. Acidity and sulfate content of fly ash as functions of temperature and water
vapor concentration
112
-------
Investigators at the Institute have studied the results
achieved by the injection of sulfuric trioxide or sulfuric acid in
a number of full-scale power plants.2 We have considered both
practical and theoretical questions and obtained answers that, to
us, seem generally satisfactory, subject to the limitations im-
posed by the characteristics of plants available for study.
One question considered is whether there is any high
resistivity ash that cannot be suitably conditioned with sulfur
trioxide. We believe that the answer is no, but we do recognize
that the quantity of conditioning agent required varies with such
factors as the chemical composition of the ash and the gas tem-
perature. The importance of these factors is illustrated in
Figure 4, which gives the results obtained with ashes character-
ized broadly as acidic, neutral, or alkaline at temperatures vary-
ing from 110 to 160° C. If a resistivity of about 1 x 1010 ohm-
cm is required, more sulfur trioxide is required if the ash is
alkaline or if the gas temperature is high.
• PLANT 2
A PLANT 3
B PLANT 5
^ PLANT 6
V PLANT 7
u
o
1-
C/3
CO
LJJ
a.
PLANT 5 (135°C, BASIC ASH)
- PLANT 3 (110°C, BASIC ASH)
^
*<
**tij
107 _
0 5 10 15 20 25 30
CONCENTRATION OF SO3 INJECTED, ppm
Figure 4. Resistivity as a function of the concentration of injected sulfur trioxide
113
-------
Another question considered is whether the efficiency of con-
ditioning is markedly affected by the choice of location in _ the
flue-gas train where the agent is added. Our studies have in-
cluded plants with and without mechanical collectors to aid pre-
cipitators in the collection of fly ash; they have included plants
with injection ahead of the mechanical collectors, between the
mechanical and electrostatic collectors, and ahead of precipitators
operating without mechanical collectors. No important change in
the efficiency of conditioning can be attributed to these varia-
tions in plant operation. We have not had experience with injec-
tion ahead of the air preheater, which would be attractive from
the point of view of ensuring uniform introduction of sulfur tri-
oxide or sulfuric acid as a vapor.
Another question was whether a choice is to be made among the
different types of injection systems (based on injection of sulfur
trioxide directly, as catalytically oxidized sulfur dioxide, or as
sulfuric acid). We have not recognized any change in the efficiency
of conditioning that we believe to be fundamentally attributable
to the type of system. We did have experience with one acid-
injection system that was inefficient as the apparent result of
excessive condensation near the point of injection.
A fourth matter of practical interest was whether a signifi-
cant fraction of the injected sulfur trioxide is lost through the
stack. A close examination of this point was made in only one
plant where conditions favored a stack loss, as the result of the
acidic character of the ash and incomplete removal of the condi-
tioning agent. Figure 5 compares the measured concentrations of
sulfur trioxide with and without injection as a function of dis-
tance across the outlet duct at the precipitator or, in other words,
as a function of the gas temperature (a variable across the duct
as a result of the influence of the Ljungstrom air heater) . The
difference in the areas under the two concentration curves corre-
sponds to an average concentration of 5 ppm, about 30% of the 14
ppm injected. An increase in the sulfur trioxide concentration
by 5 ppm would produce a concentration of condensed sulfuric acid
of 20 yg/m3 (about 0.01 gr/ft3) in the relatively cool plume from
the stack—a significant level in terms of "clear stack" concen-
tration levels for fly ash. We recognize that, under typical con-
ditions with an alkaline ash from low-sulfur Western coals, the
stack loss of sulfur trioxide will be lower and perhaps insignifi-
cant. But it is not a matter that can be ignored.
A theoretical question of some importance is: What is the
mechanism of conditioning? One facet of this question is whether
deposition of the agent into ash occurs by adsorption or condensa-
tion. Our conclusion is that adsorption of vapor above the acid
dew point can certainly occur, although condensation may also
occur if the gas is below the dew point. This conclusion is
based on a prediction of dew points from thermodynamic data that
114
-------
are applicable only if the condensed vapor occurs as a binary
Izquid mixture of sulfuric acid and water. The possibili?y lhat
reaction of vapors with alkaline fly ash constituents may take
?£?S Hen?e °Ver Jonden?ati°n makes prediction of dew points on
this basis uncertain; if such a reaction occurs, however, the
deposition of vapors would be even more likely to occur as an
adsorption process.
o
o
cc
D
H
<
t£
LLJ
Q.
150
140
130
120
110
i—i—I—r
I- x
A
I
SO3 WITHOUT
INJECTION
10
LU
O
O
O
CO
O
V)
0
0
1.0
Figure 5.
0.2 0.4 0.6 0.8
FRACTION OF DISTANCE ACROSS GAS DUCT
Concentration of sulfur trioxide as a function of gas temperature at the outlet
of a precipitator
Conditioning By Other Agents
No discussion of conditioning of high-resistivity fly ash
would be complete without mention of the role of ammonia, although
the subject must be approached with caution. It appears that this
agent sometimes is effective by other mechanisms of conditioning,
as discussed subsequently, and that its ability to alter resis-
tivity is not clear-cut.
The value of ammonia as a conditioning agent for one type of
suspended particulate matter—catalyst dust in a petroleum refinery,
rather than fly ash from a coal-burning boiler—has been known for
many years. J. F. Chittum of the former Western Precipitation
Company was one of the pioneers in the use of ammonia. He believed
that ammonia was of greater value than sulfur trioxide in treating
an acidic dust, such as the catalyst material in a refinery.3
We may very well ask: What can be expected of ammonia as a
conditioning agent for fly ash? Suppose we accept Chittum1s pre-
mise that the use of ammonia and not sulfur trioxide is indicated
115
-------
if the ash is acidic. We would then have to imagine a fly ash
that is at once acidic and high in resistivity. The experience
and interpretation of ash chemistry leads to the viewpoint that
these may be mutually exclusive factors. None of the constituents
of fly ash except adsorbed sulfur trioxide is sufficiently acidic
to have a strong affinity for ammonia, which is a rather weakly
basic compound. But, if adsorbed sulfur trioxide is present, how
can resistivity be high?
These are ideas that will have to change if enough clear-cut
evidence of a resistivity effect by ammonia becomes available.
There are data available either directly indicating that ammonia
lowers resistivity or indirectly indicating that ammonia has this
effect, such as changes in precipitator voltages and currents that
are most easily explained as a result of lowered resistivity. It
does not appear, however, that conclusive evidence of lowered
resistivity with ammonia conditioning is now in existence.
What about other agents? Sodium is an example that we believe
to have clear-cut significance in lowering resistivity. The im-
portance of sodium first became evident in a comparison of resis-
tivity values of ashes varying naturally in percentages of sodium
oxide. It was later confirmed in trials of various sodium com-
pounds (notably the carbonate) as additives to coal being fed to
the boiler of a power plant. Our experience on the use of sodium
is entirely restricted to boiler additions (not flue gas treat-
ment) , ** but others have reported the use of sodium as a flue gas
additive.5
What about some of the proprietary agents now on the market?
For some of these agents, the claim is made that they function as
resistivity modifiers, but the data upon which such claims are
made are not often, if ever, given. From what is known about the
compounds present in the proprietary agents and what has been pub-
lished from laboratory studies of the compounds, their functioning
as resistivity modifiers is not an unreasonable assumption.6 One
of the compounds identified is sulfamic acid; another is ammonium
sulfate. It is perhaps noteworthy that each of these compounds is,
in theory at least, derived from sulfur trioxide and ammonia.
From a constitutional point of view, sulfamic acid is equivalent
to these constituents in a 1:1 mole ratio (HO-S02-NH2 = S03 + NH3).
Similarly, ammonium sulfate or (NBU^SOij is equivalent to sulfur
trioxide, ammonia, and water in the molar proportions
S03:NH3:H20 = 1:2:1. Thermal decomposition of either compound can
thus produce both sulfur trioxide and ammonia as gases along with
further breakdown products, such as S02 and N2
116
-------
RESISTIVITY MODIFICATION OF ANOTHER TYPE: TREATMENT OF
ASH FROM HIGH-SULFUR COAL MIMCIMI ur
It is important to realize that the problem of excessively
low resistivity occasionally exists, although it is less widely
recognized than the problem of high resistivity. It occurs in
the burning of a high-sulfur coal, and it conceivably may occur
if ash from a low-sulfur coal is overtreated with sulfur trioxide.
If resistivity is low—below 1 x 107 ohm-cm—the practical conse-
quence may be excessive reentrainment, owing to the inadequacy of
the electrical force across the ash deposit for maintaining the
physical integrity of the deposit under the influence of scouring
or electrode rapping.
In view of the likelihood that low resistivity can be attri-
buted to an excessive amount of sulfur trioxide, it is reasonable
to try ammonia as an agent for chemically neutralizing the acid.
From a theoretical point of view, one must be troubled over the
prospects of increasing resistivity sufficiently by simply con-
verting sulfuric acid to ammonium sulfate. The conductivity of
ammonium sulfate in water is lower than that of sulfuric acid,
but the ratio of conductivities is roughly 1:4 and thus not as
great as one order of magnitude. A study of the mechanism of
ammonia conditioning under circumstances where it apparently gave
the desired resistivity change failed to confirm the mechanism
assumed.
REENTRAINMENT CONTROL: INCREASING THE COHESIVENESS OF
PRECIPITATED ASH
If, as just discussed, the resistivity of ash from a high-
sulfur coal can be increased, one has a means for minimizing re-
entrainment. If one is unable to make any significant change in
the resistivity and thus in the electrical force needed to main-
tain physical integrity of deposited ash, he has the option of
increasing mechanical forces that bind individual particles
together.
Dalmon and Tidy7 recognized the value of sulfur trioxide as
an agent for increasing the binding forces between particles of
fly ash and unburned carbon. Carbon is a highly conductive mate-
rial (relatively speaking) and is especially difficult to retain
after it is electrostatically precipitated, except by means of
cohesive forces. The method used by Dalmon and Tidy for demon-
strating the reported activity of sulfur trioxide involved pack-
ing a bed of ash and carbon in a bed with one end detachable from
the other and determining the mechanical force needed to rupture
the bed of particles.
Investigators at the Institute2 obtained evidence in one
plant for a similar effect of sulfur trioxide, although in this
instance there was no indication of excessive carbon being the
117
-------
cause of excessive reentrainment. In this plant, conditioning by
sulfur trioxide had been adopted as a measure for restoring the
original efficiency of a precipitator after a change had been made
to burn coal with a lower sulfur content. Our data failed to in-
dicate that resistivity was a problem; the resistivity of untreated
ash was found to be about 1 x 1010 ohm-cm, and the value with 14
ppm of sulfur trioxide injected was about 1 x 109 ohm-cm. One
might have expected increased reentrainment as a result of the
change in resistivity. In actuality, however, we found that in-
jection of sulfur trioxide increased the precipitator efficiency
markedly. Data in Table 2 show a parallel between the effects
achieved by injecting sulfur trioxide and by discontinuing elec-
trode rapping. Our conclusion, therefore, is that the principal
mechanism of conditioning in the circumstances investigated was
increased cohesiveness of the fly ash and decreased rapping re-
entrainment.
TABLE 2. REDUCTION OF RAPPING REENTRAINMENT
BY SULFUR TRIOXIDE
80s injected, Precipitator
ppm Rapping efficiency, %
0 Normal 69.7
Off 82.0
14 Normal 93.5
Off 96.8
We also found evidence that ammonia injection suppressed
rapping losses in plants burning high-sulfur coal.2 A reproduc-
tion of a chart recording the signal from a transmissometer at
the precipitator outlet in one of these plants is given in Figure
6. After ammonia injection was started, the intensity of rapping
puffs was slowly suppressed. After ammonia injection was discon-
tinued, the intensity of rapping puffs gradually increased again.
The findings at the plant in question would not be complete with-
out mention of other experimental results: no measurable effect
of ammonia on the resistivity of the ash (about 1 x 109 ohm-cm)
was evident, but significant effects of ammonia on the electrical
properties of the gas stream were discerned, as discussed later.
It appears that triethy1amine, which has recently been found
a promising agent in Australia,8 may act in the manner described
for ammonia but even more effectively. Both agents are basic,
but the basic strength of triethylamine is greater and the re-
ported stronger effect of triethylamine seems logical. It must
be recognized that the Australian workers have concluded that
triethylamine acts as an "agglomerating" agent, which we interpret
118
-------
to mean an agent that increases particle size before precipitation
occurs. However, if this mechanism of conditioning does occur,
specific evidence for the process does not appear to have been
obtained.
1200
1300 1400
1500
1100
1600
1000
1700
0900
0800
1800
1900
100
20
60
100
RELATIVE VALUE OF
LIGHT OBSCURATION
Figure 6. Reduction of rapping reentrainment by ammonia
ALTERATION OF ELECTRICAL PROPERTIES OF THE GAS STREAM: THE SPACE-
CHARGE EFFECT
Conditioning through this mechanism has been observed in
several power plants where ammonia was used for conditioning fly
ash.2'9 Two of the plants investigated burned high-sulfur coals
and produced fly ash with resistivities around 1 x 108 to
1 x 109 ohm-cm. Another two plants burned so-called low-sulfur
coals from Eastern mines, but in these instances the sulfur level
was around 1.0% and measured resistivities were about 1 x 10ll
ohm-cm. Moreover, significant concentrations of sulfur trioxide
were found in the flue gas—around 2 to 5 ppm, in contrast to
usually undetectable levels (less than 1 ppm) in plants burning
0.5%-sulfur Western coals.
No measurable effect of ammonia on resistivity could be de-
tected. However, the gradual change in the intensity of rapping
puffs, indicating an increase in the cohesiveness of deposited
ash, was detected clearly in one plant. Moreover, there were
other changes that were virtually instantaneous. One effect was
a change in the precipitator electrical conditions in the direc-
tion consistent with a decrease in the mobility of charge carriers
119
-------
in the gas phase, as shown in Figure 7. Another phenomenon that
occurred rapidly was a marked enhancement in the concentration of
suspended fine particles at the precipitator inlet. Coupled with
these effects were pronounced decreases in the concentration of
sulfur trioxide as a constituent of the gas stream.
50
> 40
LLJ
C3
O
>
20
NH3 ON
(20ppm)
NH3OFF
1000
1100
HOUR
1200
Figure 7. Rapidity of the effect of ammonia on the voltage of a precipitator
Taken together, the observations suggested a chemical reac-
tion between the injected ammonia and the normally present sulfur
trioxide (actually, the vapor of sulfuric acid) to produce a fume
of fine particles of ammonium sulfate or perhaps the bisulfate,
as shown by the following equations:
(NEU) 2SCH (solid)
(solid or liquid)
2NH3(gas) + H2SCMgas)
NH3(gas) + H2SCMgas) — > N
The bisulfate in the second equation is shown as either solid or
liquid; the transition point is about 144° C.
What is the likelihood of a similar space-charge effect in a
plant burning a typical low-sulfur Western coal, and what would
be the value of such an effect? The reaction of injected ammonia
with normally occurring sulfur trioxide would generally be less
likely to occur in such a plant because of the insignificant
quantity of sulfur trioxide present. However, much higher
120
-------
concentrations of ammonia than are generally used for conditioning-
say, 100 to 200 ppm—could react with trace concentrations of
sulfur trioxide (less than 1 ppm). Concentrations of ammonia in
this range, it may be noted, are used to advantage with high
resistivity Australian ashes. Such concentrations of ammonia may
also react with sulfur dioxide to produce fine particles of ammo-
nium sulfite, although this does not appear likely from the ther-
modynamic data now available.
What about the possible value of the space-charge effect if
it does occur? Possibly this effect could suppress the abnormal
currents that are encountered in instances of back corona. Data
published by Watson and Blecher in Australia some years ago10
gave evidence of back corona suppression by ammonia, at a rate
rapid enough to be attributed to a space-charge effect.
PERSPECTIVE ON THE FUTURE OF CONDITIONING
Chemical conditioning is only one of the alternatives open
to the power industry in dealing with difficult fly ash. The tech-
nology of conditioning has been tried far more often on an exper-
imental basis than it has been adopted for full-time continuous
use. When permanent installations for conditioning have been made,
usually they have been intended to upgrade the performance of an
existing precipitator, not to ensure adequate performance by a
new collector.
A major factor leading to this policy has been the unpopular-
ity of conditioning with sulfur trioxide, mainly as a result of
operating difficulties with injection systems for this agent and
hazards associated with the compound. Another factor has been the
frequent lack of success with ammonia and other flue gas addi-
tives as alternative agents and the absence of an understanding
of the mechanisms by which they operate.
Some of the needs that must be met if conditioning is to
have a continuing or expanding application are: (1) development
of new agents that are attractive on the basis of cost and con-
venience and (2) research on the mechanisms of action by various
agents that will give a more dependable basis for predicting per-
formance. Coupled with these developments, there must be a care-
ful diagnosis of the specific problems in individual power plants
that cause inefficient precipitation and a recognition of the
fact that conditioning cannot rectify all difficulties. In other-
words, chemical companies, equipment vendors, and utility companies
must all accept their responsibilities if, in the long range, con-
ditioning is to serve the immediate interests of all concerned.
121
-------
One final point to be made is that research must be done to
ensure that conditioning has a favorable overall impact on the
environment. Regulatory agencies will not, in the long run,
tolerate conditioning as a means of reducing fly ash emissions
if it leads to other undesirable effects that can be avoided
through the adoption of some other corrective measure.
REFERENCES
1. Stull, D.R., and H. Prophet (ed.). JANAF Thermochemical
Tables. Washington, National Bureau of Standards, 1971.
Unnumbered pages listed in this alphabetical order: H20,
H2CHS, 02S, and 03S.
2. Dismukes, E. B. Conditioning of Fly Ash with Sulfur Trioxide
and Ammonia. EPA-600/2-75-015, NTIS PB 238922/AS, U. S.
Environmental Protection Agency, Washington, D.C., 1975.
50 pp.
3. Chittum, J.F. Western Precipitation Corporation, Los Angeles,
California. Unpublished data from studies in 1942-1945.
4. Bickelhaupt, R.E. Electrical Volume Conduction in Fly Ash.
J. Air Pollution Control Assoc. 2£ (3): 251-255, 1974.
5. Selle, S.J., and L.L. Hess. Factors Affecting ESP Perfor-
mance on Western Coals and Experience with North Dakota
Lignites. Symposium on Particulate Control in Energy Pro-
cesses. San Francisco, May 11-13, 1976.
6. Dismukes, E.B. Conditioning of Fly Ash with Sulfamic Acid,
Ammonium Sulfate, and Ammonium Bisulfate. EPA-650/2-74-114,
U. S. Environmental Protection Agency, Washington, D.C., 1974-
51 pp.
7. Dalmon, J., and D. Tidy. The Cohesive Properties of Fly Ash
in Electrostatic Precipitation. Atmos. Environ. (Oxford,
England) 6_ (2): 81-92, 1972,
8. Potter, E.G., and C.A.J. Paulson. Improvement of Electrostatic
Precipitator Performance by Carrier Gas Additives. Chem.
Ind. (London) 1974; 532-533, July 6, 1974.
9. Dismukes, E.B. Conditioning of Fly Ash with Ammonia. J. Air
Pollution Control Assoc. 25 (2): 152-156, 1975.
10. Watson, K.S., and K.J. Blecher. Further Investigation of
Electrostatic Precipitators for Large Pulverized Fuel-Fired
Boilers. Air Water Pollution Int. J. (Oxford, England) 10
(9): 573-583, 1966.—
122
-------
PAPERS
RAPPING REENTRAINMENT STUDIES
Herbert W. Spencer, III
Southern Research Institute
ABSTRACT
The results of experimental investigations of rapping re-
entrainment in four electrostatic precipitators connected to
coal-fired boilers and in a large pilot electrostatic precipita-
tor are reported. The objective of the studies was quantifica-
tion of rapping reentrainment in terms of the percentage and
particle size distribution of the reentrained particulate.
Impactors and mass trains were used to measure separately
rap and nonrap emissions. An extraction real time optical parti-
cle sizing system was also used to determine the percentage of
rapping emissions in five particle size bands. This system also
supplied data on the temporal variations of the emissions due to
rapping. During the pilot-scale test and during one of the field
tests, the rapping emissions were observed to depend on the time
interval between raps. Rapping emissions decreased from 53% to
18% of total emissions for the pilot test as the time interval
between raps was increased. The percentage contribution of rap-
ping reentrainment to total emissions for all tests ranged from a
high of 53% to a low of 6.5%. Data on the particle-size distribu-
tion of the rapping puffs indicated that the mass median diameters
of the particles were on the order of 10-20 ym. The large parti-
cles emitted during the rapping puffs appeared to be agglomerates.
Rapping emissions were observed during the pilot-scale test to
have a significantly higher concentration in the lower half of
the precipitator. One of the major causes for this appeared to
be hopper "boil-up". It was photographically observed that most
of the particulate dropped into the hoppers; then a portion re-
bounded and slowly escaped over the hopper baffles to exit the
precipitator.
INTRODUCTION
Rapping reentrainment accounts for significant losses from
many electrostatic precipitators. The losses occur when the
collected dust is removed from precipitator collection plates by
123
-------
rapping or shaking. The dust is lost by direct reentrainment in
the gas stream and by dispersion when the dust falls into the
collection hoppers.
The rapping losses are in addition to losses from uncollected
material and to losses from other sources of reentrainment. Other
sources of reentrainment can include the direct scouring action
of the gas on the collected dust, scouring during sparking, and
sweepage of dust directly from the hoppers, caused by poor gas-
flow conditions or by air leakage into the hoppers.1
This paper discusses the results of experimental investiga-
tions of rapping reentrainment of fly ash in four electrostatic
precipitators connected to coal-fired boilers,2'3 and the results
of an experimental investigation of rapping reentrainment of fly
ash in a large pilot precipitator at FluiDyne Engineering Corpora-
tion's Rosemount Laboratory.1*
The main objective of the studies was quantification of
rapping reentrainment in terms of the percentage of total emis-
sions and in terms of the particle size distribution of the
reentrained particulate with the intent of providing data for
inclusion of losses due to rapping reentrainment in a systems
model of electrostatic precipitation. In addition, during the
studies using the pilot precipitator, the basic mechanics of
removal of dry dust by rapping and the variations in the removal
mechanisms with changes in dust properties were investigated.
BACKGROUND
The removal of particulate from a gas stream by a dry
electrostatic precipitator involves four separate processes. The
first is the precipitation of the particulate from the gas stream
onto the collection plate. A Southern Research Institute report
discusses the physical mechanisms involved and describes a com-
puter model for the dust collection from the gas stream.5 The
second process is the removal of the particulate from the collec-
tion plate, and the third is the transfer of this particulate to
a hopper. The fourth process is the removal of the collected
particulate from the hopper to a disposal area.
In present-day precipitator applications, two approaches are
prevalent with regard to the removal and transfer of the particu-
late from the collecting plates. One approach is to rap often
and to provide maximum rapping acceleration to these plates dur-
ing each rap in an attempt to minimize the thickness of the re-
sidual dust layer. The other approach is to vary the intensity
and frequency of rapping in an attempt to minimize the quantity
of material reentrained. A determination of the best rapping
technique for a specific application depends on an understanding
of the method by which dust is actually removed and transferred
124
-------
from the collection plates during a rapping sequence and of the
effects of residual dust layers.
The mechanics of the dust removal process vary with the
properties of the dust, precipitator operating conditions, and
rapping parameters. Dust properties and precipitator operating
conditions affect the adhesion and cohesion of the dust layer.
The adhesion and cohesion of dust layers depend upon particle-to-
particle forces. According to Tassicker,6 the component forces
are: London- van der Waals, triboelectric, capillary, surface
dipole, and electric- field corona forces. These component forces
are influenced by the following: particle diameter, porosity and
compaction of the layer, complex dielectric constant, humidity
in the gas, adsorbed surface dipolar molecules, work-function
interfaces on the material, and the electric field and current
density in the dust layer.
Theory of the Mechanics of Dust Removal. An elementary theory of dust
removal which considers only the tensile strength (P) of the dust
layer and the acceleration (a) normal to the plate has been devel-
oped by Tassicker.7 The theory predicts that the dust layer is
removed only when
ST M/A
where & is the bulk density, £ is the dust layer thickness, and
M/A the mass per unit area. According to this equation, for a
given dust thickness, the rapping intensity must be of sufficient
magnitude to produce an acceleration greater than the ratio of
the tensile strength of the ash layer and the mass per unit area.
For a given acceleration, dust is removed only when
M/A > P/a; (2)
that is, when the mass per unit area (dust surface density) is
greater than the ratio of dust layer tensile strength to the
normal plate acceleration. Since the mass per unit area depends
on the dust layer thickness, which in turn is related to collec-
tion time between raps, the time interval between the raps is
directly related to the efficiency of dust removal from the plates
Experimental data obtained by Sproull8 and by Penney and
Klingler9 show that the requirements for removal of a precipi-
tated dust layer are in basic agreement with Tassicker 's elemen-
tary theory for dust removal.
Emissions Due to Rapping. Emissions due to rapping and their de-
pendence on rapping parameters have been reported previously by
Sproull,10 Plato,11 Sanayev and Reshidov,12 Schwartz and Lieber-
stein,13 and Nichols, Spencer, and McCain.2 Sproull found that
125
-------
reducing the intensities of the raps lead to a reduction in rapping
emissions. He also observed a vertical stratification of the
emissions during rapping, with higher concentrations in the lower
portion of the precipitator. The work by the others showed im-
provements in performance of full-scale precipitators when the
time intervals between raps were increased.
Although these previous studies have added to our understand-
ing of rapping reentrainment, they have not provided quantitative
data on the amounts of emissions due to rapping reentrainment.
EXPERIMENTAL PROCEDURES
A complete characterization of rapping reentrainment requires
the measurement of a large variety of variables. A block dia-
gram of an experimental layout for our pilot rapping reentrain-
ment study is shown in Figure 1. In addition to the data that is
ELECTRICAL
CHARACTERISTICS
RAPPING
VARIABLES
DUST LOAD
ON PLATES
PLATE ACCELERA-
TION
GAS ANALYSIS
SO3, SO2. H2O
TEMPERATURE
TIME INTEGRATED
PARTICLE SIZE
MEASUREMENTS
MASS LOADING
VELOCITY
DISTRIBUTION
• TWO SETS AT 3 LOCATIONS: ONE TO MEASURE DURING
RAPS AND ONE TO MEASURE BETWEEN RAPS,
" TWO SEPARATE UNITS: ONE TO LOOK AT LOWER HALF OF THE
PRECIPITATOR OUTLET AND ONE TO LOOK AT UPPER HALF OF
THE PRECIPITATOR OUTLET.
PRECIPITATOR
OBSCURATION
METER
TIME INTEGRATED
PARTICLE SIZE
MEASUREMENTS
CAMERA AND
LIGHTING
MASS LOADING
UPPER HALF
MASS LOADING
LOWER HALF
HOPPER
SAMPLES
" REAL TIME
PARTICLE SIZE
MEASUREMENTS
Figure 1. Block diagram of experimental layout for a rapping reentrainment study
obtained with this arrangement, a complete characterization
utilizes the precipitator design data.
The field experiments included a similar set of measurements
to those made during the pilot studies. However, sampling view
ports for photographing rapping emissions and for determining the
vertical stratification of the rapping emissions were not avail-
able in the full scale units nor were load cells for measuring
126
-------
the quantity of fly ash collected on the collection plates.
Hence these measurements were not included in the field tests.
The adhesion and cohesion properties of the collected dust
layers were also not determined during each test since suitable
field test equipment was not available. However, some informa-
tion about the tensile strength of the collected ash layers was
obtained during the pilot studies by using the elementary theory
of adhesion given in the introduction, the load cell data, and
plate acceleration data.
Quantification of rapping reentrainment in terms of the
percentage of total emissions and in terms of the particle size
distribution of the reentrained particulate required determining
the amount of emissions with and without rapping reentrainment.
To define the mechanisms by which the rapping losses occurred re-
quired time-resolved data on the particulate concentrations and
size distributions across typical portions of the precipitator
exit plane.
During the investigations the above requirements were met by
using real-time systems and integrating systems. A real-time
system based on an optical single-particle counter was developed
which included a five-channel analog ratemeter, which provided
parallel monitoring of the instantaneous concentrations of
particles in five pre-selected size intervals ranging from 0.6 to
20 ym.
To conform to instrumental limitations on the total concen-
tration of aerosol particles in the sample gas stream arriving
at the sensor, the aerosol sample from the flue was diluted be-
fore measurement. Because of the very steep gradient in the size
distribution, on a number basis, at the exit of a precipitator
on a power boiler, the diluter was made as a size-selective device
which, under ideal conditions, dilutes the concentration of small
particles in the sample gas stream by fairly large factors while
passing a relatively confined and undiluted stream of the lower
concentrations of large particles directly to the particle sensor.
Figure 2 illustrates the operating system for the particle dilu-
tion train. Because of the potentially large probe losses that
can result from impaction and settling in the probe, the system
was intended to be used only from below a duct with a minimum
number of bends (one) between the sampling point and the particle
sensing volume. The probe was oriented in the vertical direction
so that settling was in the direction of the probe axis and con-
sequently did not result in a loss of particles.
For those circumstances in which it was not possible to
sample from below the duct, a second sample extraction system
was constructed. These samples were removed at high flow rates,
0.002-0.02 m3/sec (5-40 cfm), through a large bore probe (4 cm
diameter) and conveyed to a suitable location beside or on the
127
-------
top of the duct, at which point a secondary sample was extracted
into the diluter and counter as illustrated in Figure 3. This
sampling method provided information on relative concentrations
of particles of various sizes during and between puffs, but did
not provide quantitative concentration data because of the un-
certainties in the probe losses and in the degree to which the
secondary sample represents the average concentration in the high
flow rate probe.
Integrated measurements of rapping emissions were made with
separate sets of inertial impactors and mass trains during the
pilot studies to independently quantify rapping reentrainment
losses and non-rapping reentrainment losses. One set of impac-
tors and mass trains measured losses between raps and the other
set measured rapping losses. The dust feed was turned off dur-
ing rapping. During two of the field tests, losses were measured
with separate sets of impactors and mass trains during alternat-
ing periods with rapping and without rapping. The weighted time
average of the measurements was compared with data obtained dur-
ing normal operation to determine if losses were affected by use
of the non-rapping intervals. During the other two field tests,
measurements were made during non-rapping periods and compared
with measurements during either normal operation or operation with
a modified rapping frequency.
RESULTS OF EXPERIMENTAL STUDIES
In the tests on the five electrostatic precipitators, the
following data were obtained: the percent of total emissions due
to rapping reentrainment, the percent of emissions for a given
particle size due to rapping, the particle size distribution of
the particulate emitted due to rapping, and the effects of rapping
reentrainment on fractional collection efficiencies. Selected
examples of the test results are contained in this paper.
The percentage contributions of rapping reentrainment to
total emissions at each of the test locations and for various
test conditions are tabulated in Table 1. The percentage con-
tributions ranged from a high of 53% to a low of 6.5%.
The percentage of emissions for a specific particle size
that are due to rapping are given in Figure 4. This data indi-
cates that rapping reentrainment accounts for only 2-5% of the
emissions of 1 ym diameter particles, and that the percentage of
emissions due to rapping increases significantly with increasing
particle size, ranging from 24% to 48% for 4 ym diameter particles
and from 50% to 88% for 12 ym diameter particles.
Averaged particle size distributions for rapping puffs are
plotted in Figures 5 and 6. The particle size distribution data
shows that the rapping puffs consisted of relatively large parti-
cles when compared with non-rapping emissions.
128
-------
SAMPLE
FLOWRATE
MANOMETER
//
<-
u
GAS FLOW
1
J
PROBE
HEATER
PROCESS EXHAUST
LINE
0.02 urn
FILTER
BLEED
VALVE
DILUTION AIR
MANOMETER
AND ORIFICE
VERTICAL
ELECTRICAL
LEADS, ETC.
LARGE PARTICLE
COUNTER
MAIN FRAME
Figure 2. Schematic of diluter for real time particle sizing system
BLOWER
FLOW
REGULATOR
DILUTER
/AND COUNTER
(X)
£
/
EXHAUST
DUCT TOP
PROBE
.GAS
"FLOW
Figure 3. Extractive sampling system for real time system
129
-------
Table 1. PERCENTAGE CONTRIBUTION OP HAPPING REENTRAINMENT TO TOTAL EMISSIONS
u>
Location Comments
1
SO 3
injection
S03
injection
pilot test
pilot test
2 pilot test
pilot test
pilot test
3
4
5
Type
Rapper
Vibrator
Vibrator
Vibrator
no rap
no rap
Drop hammer
Drop hammer
Drop hammer
Drop hammer
no rap
Drop hammer
two plates
Drop hammer
two plates
Rotating
Drop hammer
Avg. Plate
Plate Rap Gas Current
Acceleration Intervals Velocity Densities**
G's Min. m/sec mA /cm2
x,y,x axis
3.7, 1.6, 2.9 4, 6.5, 6.5 7.6, 10.4,
15.3, 13.9
3.5, 1.8, 2.6 8, 13, 13 11.2, 11.2,
16.7, 12.5
4, 6.5, 6.5 11.2, 11.2,
16.7, 12.5
11, 16, 15 12 0.87 23.3
32
52
150
10, 10, 8.4, 9.1,
20, 20 1.52 13.2
60, 60 15.1, 12.3
10, 20, 1.25 11.1, 17.6,
60 22.7
4.6, 7.8,
11.3
6, 6, 12, 1.83 14, 24.6,
12 33.0, 47.7
Gas SCA* , Total Penetration
Temp mz Penetration Due to Rapping
°C m'/sec % Reentrainment , %
137 32.1 46
137 25.8 33
132 6.2 48
17.4
3.2
122 33 11.4 53
7.6 32
6.1 18
6.9 25
5.2
154 110 0.85 30
157 48 0.40 35
157 0.95 38
162 52 0.19 6.5
*SCA - specific collection area
**for each electrical field
-------
100
a.
a.
cc
O
LLJ
a
oo
O
GO
GO
LOCATION 4
NORMAL CURRENT
20
8 12 16
PARTICLE DIAMETER, «m
4. Percent emissions due to rapping for particles with diameters of 1 to 20 jjm
0.01 0.1 1 10 20 40 60 80
PERCENT LESS THAN INDICATED SIZE, by mass
Figure 5. Cumulative percent distribution for rapping puffs, rapping intervals of 12, 32,
and 52 minutes, pilot test
131
-------
A selected example of the effect of rapping reentrainment on
precipitator fractional efficiency is shown in Figure 7. Frac-
tional efficiencies for alternating periods with the rappers off
and with the rappers on are given along with fractional effici-
encies determined during normal rapper operation. These data
show that collection efficiency is relatively independent of
rapper operation for particles with diameters from 0.1 ym to
2 ym with the possible exception in the region of 0.3 to 8 ym in
which there is some ambiguity in the data. Above 2 ym, rapping
significantly reduced collection efficiency.
Discussion of Experimental Results
The percentage contribution of rapping reentrainment to
total emissions varied significantly for changes in test condi-
tions at the same location and for the different precipitators.
The lowest rapping emissions were apparently obtained for a
precipitator with high plate accelerations, i_.e_., accelerations
greater than 150 Gs (1.5 x 105 cm/sec2) and "with a rap. interval
of 30 minutes. There are several possibilities for the apparent
low rapping emissions at this location that are of interest and
which merit further investigation. One is the use of large and
relatively deep hoppers (depth 6 meters). It is suspected that
these aided in reducing hopper "boil-up", which is discussed in
the section on reentrainment mechanisms. Another possibility is
the effect of having a relatively fine inlet dust. There also is
the possibility that other sources of reentrainment may have con-
tributed to non-rap emissions, resulting in the apparently low
rapping emissions.
The highest percentage rapping emissions were obtained dur-
ing the pilot studies with a rapping interval of 12 minutes and
with normal plate accelerations on the order of 11 Gs (1 x 1011
cm/sec2). The rapping emissions during the pilot study (test
location number 2) decreased with increasing time between raps.
The effect on overall efficiency is shown in Figure 8. The per-
centage of the collected dust removed from the precipitator plates
also increased with increased time between raps, as shown in
Figure 9.
The collecting of a larger mass per unit area (dust surface
density) produces the two effects mentioned above. According to
the elementary theory of dust removal, the product of the normal
plate acceleration and dust surface density has to be greater
than the tensile strength of the layer, typically 0.5 x io3 to
3 x 103 dynes/cm2, if the dust layer is to be removed. This
theory indicates that dust removal efficiency should increase with
increased time between raps, as illustrated by the data shown in
Figure 9. Laboratory measurements by Sproull"^ produced similar
data. Figure 9 also illustrates the build-up of a residual dust
layer that was not removed with normal plate accelerations on the
order of 11 Gs. There are several possible causes for the
132
-------
102
CO
•o
10°
10
1-1
INLET. DIVIDED BY 102 _
10° 101
GEOMETRIC MEAN DIAMETER, jjm
102
Figure 6. Particle size distributions for inlet paniculate, outlet paniculate with rapping,
outlet paniculate without rapping, and average effective distribution for
rapping puffs
u.u
0.1
0.2
1
a""1 ^
0
ffi 10
UJ
g 20
°~ 30
40
60
80
I i i i | • i i | | i ill
« •"•""
:§ • • • _
— Q S g A So0
0 B B a 0 o 0
moo ^^ ^^™
•» g 9 8
_
— 0 ° —
O RAPPERS ON TEST (8)
— WRAPPERS OFF TEST (8) —
ONORMAL RAP INTERVAL TEST (6,7,9)
_ —
« .iii | 1 1 tJ J 1 1 L_
99.9
99.8
99
98 ^
95 >
t-
90 £
o
mum
80 t
HI
70
60
40
20
Figure
GEOMETRIC MEAN DIAMETER,Mm
7. Fractional efficiencies as function of particle diameter for half normal current
density test. Test location Number 3
133
-------
100
5?
>
U
UJ
o
ul
90
WITHOUT RAPPING
20 60 100 140
TIME INTERVAL BETWEEN RAPS, min
180
Figure 8. Average efficiencies for Flu/Dyne pilot precipitator for various rapping intervals
OF PARTICULATE
COLLECTED BETWEEN RAPS
% OF TOTAL PARTICULATE
COLLECTED ON THE PLATES
140
160
40 60 80 100 120
TIME INTERVAL BETWEEN RAPS, min
0.26 0.78 1.3 1.8 2.3 2.9 3.4
MASS/AREA GAINED BETWEEN RAPS, kg/m2
Figure 9. Dust removal efficiency versus time interval between raps
134
-------
development of the residual layer. For one, the dust layer
directly in contact with the collection plates has a much higher
tensile strength than the remainder of the layer. Estimates for
removal of the layer called for accelerations greater than 103Gs
(9.8 x 10 cm/sec ). Consolidation of the dust that remains on
the plate after a rap also aids in producing residual layers.
The vibrations during a rap can have the effect of compacting the
dust layer if it is not removed making it more difficult to re-
move. A third possible cause of the residual layer is the removal
of patches of dust only from selected locations on the collection
plates where the removal criteria are met. Dust can be removed
from one location during one rap and from another location on the
next rap due to changes in distribution of the dust surface den-
sity. This results in a nonuniform dust layer and the presence
of a residual layer. This is often the result of nonuniform
plate accelerations. At one location where plate accelerations
are on the order of only 4 to 5 Gs (3.4-4.9 x 103 cm/sec2),
residual dust layers as thick as 2 cm were observed in the vicin-
ity of plate baffles where the plate accelerations are dampened.
Between the baffles, the residual layers were only 1 to 2 mm
thick.
The decrease in rapping emissions observed during some of the
tests with an increase in the time between raps or, corresponding-
ly, an increase in dust surface density appears to be the result
of a shift in the size distribution of the reentrained particulate
to larger particle sizes. Thicker dust layers appear to produce
larger particles. In this case the large particles that are
emitted are not the primary particles that entered the precipita-
tor but agglomerates consisting of smaller particles. Inspection
of impactor substrates after collecting dust at the outlet of
test locations number 2 and 3 indicated that the majority of the
large particles in the rapping puffs were agglomerates. The
shift to larger particles in the rapping puffs observed during
the pilot studies as the time between raps was increased is
illustrated by the shift of the cumulative percent particle size
distribution of the rapping puffs shown in Figure 5.
This effect was previously observed by Plato,11 who found
that with a dust layer surface density greater than 1 kg/m2,
dust was removed in cakes and for values less than this it
formed clouds. Producing agglomerates instead of individual
particles when the plates are rapped reduces the quantity of
material that escapes from the precipitator due to rapping. This
is because the large agglomerates are recollected faster by
gravitational settling and by electrostatic precipitation than
discrete single particles or smaller agglomerates.
The above provides an explanation for the results that are
given in Figures 8 and 10. Sanayev and Reshidov have also used
the above explanation for the decrease in the percentage of dust
collected on the plates that is emitted due to rapping with an
135
-------
increase in dust surface density. A comparison of results ob-
tained during the pilot studies^ and by Sanayev and Reshidov1
is shown in Figure 11.
Both sets of data indicate that for the conditions used dur-
ing the tests, minimum reentrainment loss occurred with a dust
surface density on the order of 1 kg/m2 or larger. The Sanayev
and Reshidov data also show the expected dependence on gas veloc-
ity. The results given in Figure 11 do not include the effects of
collecting thicker layers on overall performance.
The 50 to 88% emissions due to rapping reentrainment observed
for large particles (>10 ym) which are shown in Figure 4 can be
explained either by a continuing transfer of collected large
particles from the inlet to the outlet by rapping or by the above
mentioned agglomeration of the collected particles. Both effects
probably occur, with agglomeration being the dominant effect,
as indicated by the agglomerates observed on the upper stage of
inertial impactors.
During collection by an electrostatic precipitator, the mass
median diameter of the fly ash from a coal-fired boiler typically
decreases from the inlet of the precipitator to the outlet. This
is illustrated by a shift in the particle size distributions of
the inlet and outlet dust concentrations. In Figure 6, the
dm/dlogD particle size distributions are plotted for the inlet
dust at test location number 4 and for the outlet dust with the
rappers off and with the rappers on. The expected shift in
particle size which can be theoretically predicted for the inlet
and outlet dust concentrations with the rappers off is clearly
evident. When the plate rappers are on, the mass median diameter
of the particle size distribution shifts back towards larger
size particles as shown by the dm/dlogD plot in Figure 6 with the
rappers on and by the effective average dm/dlogD for the rapping
puffs. The large particle size of the particulate in the rapping
puffs is also illustrated by the data given in Figure 5. The
rapping emission percentages in Figure 4 reflect the shift in the
outlet particle size distribution produced by plate rapping. At
locations such as at test location number 3, where the precipita-
tor is very large and preceded by a mechanical collector, almost
all of the large particles (5lO ym) are collected before reaching
the outlet. Thus the percentage of emission due to rapping can
approach 100% for the large particles.
Rapping reentrainment has the effect of transferring parti-
cles from particle size intervals consisting of small single
particles to particle size intervals consisting of larger
particles.
At test location number 4, data were obtained at normal
operating current densities (see Table 1) and at one-half the
normal operating current density. This resulted not only in a
136
-------
100
95
90
O 85
Z
UJ
o
u.
80
75
70
65
NO RAP
NO RAP
N
2N
TIME INTERVAL BETWEEN RAPS
Figure 10. Precipitator efficiency at Test location Number 1 as a function of the rapping
interval
20
O
Z
5!
o.
<
<= 15
UJ Q
10
00
UJ
• SPENCER4 GAS VELOCITY 1 m/sec
• SANAYEV & RESHIDOV15 GAS VELOCITY 2 m/sec
• SANAYEV & RESHIDOV15 GAS VELOCITY 2 m/sec
0.5 1.0 1.5
DUST SURFACE DENSITY, kg/m2
7 7. Percent of dust collected on precipitator plates emitted from pilot precipitator
due to rapping
137
-------
decrease in efficiency between raps but also, as shown by the
data given in Figure 4, in an increase in the percentage contri-
bution of rapping reentrainment for large particles. This im-
plies that a change in particulate properties such as an increase
in resistivity that reduces operating current density can result
not only in a decrease in efficiency between raps but also in an
increase in the percentage emissions due to rapping reentrainment.
However, since the forces holding the dust layer together theo-
retically increase with resistivity, the effects may not be the
same as those obtained during the test at location number 4,
where dust resistivity was held constant.
REENTRAINMENT MECHANISMS
The amount of rapping reentrainment that occurs depends on
the processes by which the dust is reentrained into the gas
stream. Sproull10 has suggested that the rapping processes
occur in a sequence in which the collected dust is removed by
what is termed a slip-stick process. The concept is that the
dust layer builds up to a sufficient thickness that the
acceleration of the plate causes a shearing force between the
dust layer and the plate. The dust then slides along the surface
of the plate until it is recollected on the plate. The process
is repeated with each successive rap until the dust finally falls
into the hopper.
Another concept of rapping is that the collected dust is
dislodged in large agglomerates or clumps and falls directly into
the hopper without being recollected.
Neither concept adequately explains the rapping process as
observed in the FluiDyne pilot precipitator or in the Southern
Research Institute small-scale precipitator. Motion pictures
(32 frames/sec) of the removal of a dust layer (2-3 mm thick) by
rapping in the SRI pilot unit show the dust layer fracturing
along lines of discontinuity in the dust surface. The resulting
fractured sheet of dust starts to fall as separate sheets which
break up as they encounter other falling sheets and patches of
unremoved dust. The dust appears to fall without being recollect-
ed and to become turbulently mixed as it falls. The motion pic-
tures show the majority of the dust dropping into the hoppers
from which a portion boils up and becomes reentrained into the
gas stream. Motion pictures taken in the large pilot precipita-
tor at the FluiDyne Engineering Rosemount Laboratory showed
similar behavior.
Measurements of the vertical distribution of the rapping
loss at the exit of the FluiDyne unit indicated that 82% of the
rapping emissions occurred in the lower 32% of the precipitator.
This effect, evidently due to hopper "boil-up" combined with
138
-------
gravitational settling of the reentrained particulate, produced
a significant vertical stratification of the reentrained emis-
sions from the large pilot precipitator as shown in Figure 12.
10°
z
o
LU
z
1* 105
2
O
O
UI
o
z
DC
£z 104
— o
GO —
li
o °-
rf °°
D: < 103
U- I—
Z <
— GO
i__ ii
i ^
o ?•
O (3
0 <
1- DC
ioi
i i i i 'Ml \ I —
* • • 1.5 -3.0 urn
•P
• 3.0 - 6.0 jjm
^6.0-1 2.0 wm
^, * • • • 12.0 -24.0 wm
"*" • • A > 24Mm
^
— • • * ^
* -^ *"
•
— "A —
A
A A
A ^
I I I I ,', I I I I
B 20 40 60 80 10o' ' 330 350 370 390 41
A
C DISTANCE FROM BOTTOM BAFFLE, cm
F
L
E
Figure 12. Spatial distribution of particles in rapping puff
Rapping puffs observed in the lower portion of the precipi-
tator occurred in two bursts for both upstream and downstream
raps as shown in Figure 13. The first burst lasted 2-4 seconds.
This burst was interpreted as being the result of particulate
reentrained directly in the gas stream and being carried out of
the precipitator at the velocity of the gas through the unit.
The longer lasting second burst, which for the larger particles
was a series of puffs, can be interpreted as resulting from
hopper "boil-up". These data indicate that hopper "boil-up" con-
tributes significantly to rapping reentrainment emissions.
139
-------
24 wm DIAMETER PARTICLES
ou
u
j* 40
0
= 30
CO
Z 20
8 10
0 500
0)
"5 400
c
£ 300
1 200
O
0 100
5x103
u
8 4
o
= 3
CO
Z 2
0 1
0 '
5x1 03
u
IA 4
O
1 3
CO
Z 2
8 i
5x103
| 4
° 3
COUNTi
-» M
1 1 1
—
—
—
. 1 1 1 1 1 1 1 1 1
II —
1
LLI U ~
12 - 24 urn DIAMETER PARTICLES
—
—
—
—
—
-
..j-
6- 12 «m DIAMETER PARTICLES
T _-„-,„„
_,_„ |||
_ I _
p L
i
3 - 6,um DIAMETER PARTICLES
_ UPSTREAM _
—
—
L
' I DOWNSTREAM
1.5 - 3 Mm DIAMETER PARTICLES
1 ft :
- 11
: I .:
1 1 1 Lf 1 1 H 1 1 1 / 1 Kf
TIME, 12sec/div
Figure 13. Flapping puffs at the exit plane of the pilot precipitator. upstream and
downstream raps
140
-------
EFFECT OF RAPPING REENTRAINMENT ON COLLECTION EFFICIENCY
PENETRATION, AND SIZING OF ELECTROSTATIC PRECIPITATORS
Rapping reentrainment can have a considerable effect on
collection efficiency, penetration, and sizing of electrostatic
precipitators. In this section two approaches to estimating the
effects of rapping reentrainment are discussed. The examples are
based on the results of the FluiDyne pilot tests and computer
calculations by Gooch et al.5 The results of the example calcu-
lations are tabulated in Tables 2 and 3.
These examples show that the percentage increase in collec-
tion efficiency obtained with no rapping reentrainment decreases
with increasing efficiency of the precipitator. However, due to
the exponential dependence of precipitator efficiency on particle
migration velocity, plate collection area, and gas volume flow
rate, the effect of rapping reentrainment on the size of a pre-
cipitator required for a given collection efficiency is substan-
tial. Estimates of the effect of rapping reentrainment on pre-
cipitator size are tabulated in Tables 2 and 3 for various con-
ditions. Two approaches were used to generate the tabulated
values. One approach was to assume that the percentage penetra-
tion due to rapping was independent of the size of the unit.
This assumption, although not strictly true, may fairly represent
real situations. According to the data obtained during the pilot
experiments, the rapping puff concentration has a significant
vertical gradient and the rapping puff consists of large particles.
Thus, in a multi-section precipitator the rapping puffs from
upstream sections may be recollected in the lower portion of a
downstream section and later transferred from this lower position
into the hoppers without making any significant contribution to
outlet emissions. In this case, outlet fields would contribute
the largest portion of the rapping emission, and the percentage
emission due to rapping should be relatively independent of the
size of the unit. Estimates based on this approach are tabulated
in Table 2.
The second procedure was developed by Francis. 1 ** This
approach assumes that the fraction of collected dust per section
reentrained and emitted is the same for every section, that the
between-rap sectional collection efficiency is constant from one
section to another, and that the recollection efficiency for the
rapping puffs is the same as the between-rap sectional collection
efficiency. According to Francis" procedure, the penetration of
any downstream section is equal to the penetration of the first
section (Pi) raised to the power of the number of preceding sec-
tions plus one. Thus, the penetration of the Nth section (PN) is
given by:
P= PN (3)
141
-------
Table 2. EFFECT OP RAPPING REENTRAINMENT ON EFFICIENCY FOR RAPPING LOSSES
ACCOUNTING FOR 20%, 30%, 40% AND 50% OF EMISSIONS FOR VARIOUS OVERALL EFFICIENCIES
(first approach)
K)
SCA* required
Assumed Penetration
percent of Due to
total pene- rapping
tration due reentrain-
to rapping ment.
reentrainment %
3.75
2.25
50 0.75
0.25
0.05
3.00
1.80
40 0.60
0.20
0.04
2.25
1.35
30 0.45
0.15
0.03
1.50
0.90
20 0.30
0.40
0.08
Without
rapping
reentrain-
ment.
%
3.75
2.25
0.75
0.25
0.05
4.50
2.70
0.90
0.60
0.06
5.25
3.15
1.05
0.35
0.07
6.00
4.60
1.20
0.10
0.02
Total
%
7.50
4.50
1.50
0.50
0.10
7.50
4.50
1.50
0.50
0.10
7.50
4.50
1.50
0.50
0.10
7.50
4.50
1.50
0.50
0.10
Assumed
Overall
Efficiency
n
%
92.50
95.50
98.50
99.50
99.90
92.50
95.50
98.50
99.50
99.90
92.50
95.50
98.50
99.50
99.90
92.50
95.50
98.50
99.50
99.90
Efficiency
without
rapping re-
entrainment ,
TlR, %
96.25
97.75
99.25
99.75
99.95
95.50
97.30
99.10
99.70
99.94
94.75
96.85
98.95
99.65
99.93
94.00
96.40
98.80
99.90
99.92
Increase in
TI without
rapping re-
entrainment ,
%
3.90
2.31
0.76
0.25
0.05
3.24
1.88
0.61
0.20
0.04
2.43
1.41
0.46
0.15
0.03
1.62
0.94
0.30
0.10
0.02
For n without
rapping
re-
entrainment ,
m2/
m3/sec
26.2
41.8
68.9
26.2
41.8
68.9
26.2
41.8
68.9
26.2
41.8
68.9
(ft2/
1000 ft3/
sec)
(133)
(212)
(350)
(133)
(212)
(350)
(133)
(212)
(350)
(133)
(212)
(350)
For HE
with
or for n
rapping
reentrainment ,
m2/
mVsec
16.2
21.3
35.5
53.2
81.7
18.7
32.9
50.2
78.8
30.5
47.3
75.8
29.2
68.9
73.3
(ft2/
1000 ft3/
sec)
( 82)
(108)
(180)
(270)
(415)
( 95)
(167)
(255)
(400)
(155)
(240)
(385)
(148)
(350)
(372)
Additional SCA
required
because
of reentrainment
m2/
m'/sec
9.3
11.4
12.8
6.7
8.5
9.8
4.5
5.5
6.9
3.0
3.5
4.3
(ft2/
1000 ft3/
sec)
(47)
(58)
(65)
(34)
(43)
(50)
(23)
(28)
(35)
(15)
(18)
(22)
Addi-
tional
SCA,
%
353
27.4
18.6
25.6
20.3
14.3
16.5
13.2
10.0
11.3
8.5
6,3
•Specific Collection Area (SCA) based on data given by Gooch, et al.
(Figure 26 computed performance curves at 20 nA/cm ).
-------
Table 3. EFFECT OP DAPPING REENTRAINMENT ON PERFORMANCE FOR VARIOUS REENTRAINMENT
PERCENTAGES AS A FUNCTION OF NUMBER OF SECTIONS AND A CONSTANT COLLECTION EFFICIENCY PER SECTION.
Assumed
percent of
collected
material
per sec-
tion re-
entrained
5.9
j
i
2 3
0 97
SCA* Needed
Penetration
Number
of
Sections
1
2
3
4
1
2
3
4
1
2
3
4
Due to
rapping
reentrain-
ment.
%
6.2
1.02
0.136
0.0193
2.23
0.28
0.027
0.0023
0.90
0.10
0.009
0.0007
Without
rapping
reentrain-
ment.
%
5.2
0.27
0.014
0.0007
5.20
0.27
0.014
0.0007
5.20
0.27
0.014
0.0007
Total
%
11.4
1.29
0.150
0.020
7.43
0.55
0.041
0.0030
6.1
0.37
0.023
0.0014
Percent of
penetration
due to rap-
ping reen-
trainment.
%
52
79
91
97
30
51
66
77
15
27
39
50
Efficiency
1,
%
88.6
98.7
99.85
99.98
92.57
99.45
99.959
99.997
93.9
99.63
99.97
99.9986
Assumed
efficiency
H without
rapping re-
entrainment ,
%
94.8
99.73
99.986
99.9993
94.80
99.73
99.986
99.9993
94.8
99.73
99.986
99.9993
Increase
Without rapping
in n with reentrainment to
no rapping obtain n,
reentrain-
ment.
%
6.54
1.03
0.14
0.02
2.35
0.28
0.03
0.002
0.95
0.10
0.016
0.001
mV
mVsec
28.0
62.1
98.5
40.4
83.7
108.5
46.3
91.0
108.3
(ft2/
1000 ft3/
sec)
(142)
(315)
(500)
(205)
(425)
(550)
(235)
(462)
(550)
With
rapping
reentrainment
•n2/
mVsec
51.6
105.4
108.3
51.6
105.4
108.3
51.6
105.4
108.3
(ft2/
1000 ft3/
sec)
(262)
(535)
(550)
(262)
(535)
(550)
(262)
(535)
(550)
Additional SCA
m2/
m3/sec
23.6
43.3
11.2
21.7
5.3
14.4
(ftV
1000 ft3/
sec)
(120)
(220)
( 57)
(110)
( 27)
( 73)
Addi-
tional
SCA,
%
84
69
28
25
11
16
*Specific Collection Area (SCA) based on data given by Gooch, et al.'
(Figure 26 computed performance curves at 20 nA/cm ) .
-------
Using this equation, the collection efficiencies of a pre-
cipitator having 1, 2, 3, or 4 sections with a between-rap sec-
tional collection efficiency of 94.8% were calculated for percent
penetrations of the first section due to rapping reentrainments
of 52%, 30%, and 15%. These percentages correspond to 5.9%,
2.3%, and 0.97%, respectively, of the collected material per
section being reentrained and emitted from a section. These
values were chosen to correspond approximately to those obtained
during the FluiDyne experiments. The results of the calculations
based on this procedure are tabulated in Table 3. According to
this procedure, for the conditions corresponding to the worst
rapping reentrainment case at FluiDyne, the contribution of rap-
ping reentrainment to total emissions would be greater than 90%
for a three or four section precipitator with 94.8% sectional
collection efficiency. For a case in which reentrainment accounts
for only 15% of the emissions for the first section, a situation
slightly better than the best obtained during the FluiDyne ex-
periments, rapping reentrainment, according to Francis1 procedure,
would account for 39% of emissions for a 3-section precipitator
and 50% of emissions for a 4-section precipitator, if the pre-
cipitator had a sectional efficiency of 94.8%. These calculations
indicate that, as the size of the precipitator increases, the
percentage contribution of rapping reentrainment to total emis-
sions increases. As previously discussed, this may not happen.
Investigations of rapping reentrainment in multi-section precipi-
tators do not appear to show a general increase in the percentage
contribution of rapping reentrainment to total emissions with
precipitator size except for large particles. To obtain an
indication of the effect of rapping reentrainment on the sizing
of a precipitator, estimates of the theoretical specific collec-
tion area needed to obtain a given collection efficiency for a
particular set of conditions were obtained from the model
developed by Gooch, et al.5 For the example chosen, a typical
particle size distribution for an ash from a coal-fired boiler
plant was selected and a precipitator operating current density
of 20 nA/cm2 was assumed (see the report of Gooch, e_t a!L. ,5 for
further details).
The theoretical specific collection area (SCA) required to
obtain a desired collection efficiency with rapping reentrainment
was compared with a theoretical SCA that would give the same
efficiency if there were no rapping emissions. This gave an
estimate of the increase in SCA needed to counterbalance the
rapping reentrainment emissions. The theoretical SCA needed to
obtain the desired collection efficiency with rapping reentrain-
ment was determined by first computing the collection efficiency
that would be obtained if there were no rapping reentrainment,
by assuming that either the percentage of emissions due to rap-
ping reentrainment or the fraction of collected dust reentrained
and emitted from a section due to rapping were known. Once this
collection efficiency was calculated and assuming that there
were no other nonidealities beside rapping reentrainment, the
144
-------
required theoretical SCA was obtained from data given by Gooch,
et al. Then, again using the data given by Gooch, et al., the'
SCA was determined that would give the desired efficiency if
there were no nonidealities such as rapping reentrainment. For
the particular cases considered, the increase in required SCA
due to the rapping reentrainment ranged from 6% to 84% (see
Tables 2 and 3). Thus, although the percentage decrease in
collection efficiency due to rapping reentrainment is relatively
small, the effect on the size required for a given collection
efficiency can be significant. The data obtained from the Flui-
Dyne pilot experiments indicates that even moderate changes in
rapping parameters can significantly affect the size of a pre-
cipitator for a given collection efficiency.
SUMMARY
Experimental investigations of rapping reentrainment were
conducted utilizing a large pilot electrostatic precipitator and
several electrostatic precipitators installed at coal-fired power
plants. The objectives of the experimental study included a
fundamental study of the mechanics of removal of dry fly ash by
rapping and the quantification of rapping reentrainment in terms
of the percentage and particle size distribution of the reen-
trained fly ash.
During the experiments the percentage of fly ash removed
from the plates was observed to depend on the mass per unit area
of ash collected on the plates, as predicted by theory. The
build-up of residual dust layers was also observed.
During the tests the contribution of rapping reentrainment
to the total emissions ranged from 6.5 to 53%, depending on the
rapping conditions and the precipitator design. A significant
decrease in rapping emissions was obtained by increasing the time
interval between raps. This decrease was due to the resulting
larger mass per unit area collected on the plates when they were
rapped. Particle size distribution measurements showed that the
mass median diameter of the particles emitted during the raps in-
creased with increased time between raps. As would be expected,
this produced lower overall emissions. The increase in the size
of the particles emitted during rapping was ascribed to an in-
crease in the extent of agglomeration of the particulate removed
from the precipitator plates with the corresponding increase in
the mass per unit area collected on the plates (thicker dust
layers) before the plates were rapped.
A major portion of the reentrained material was observed
to result from hopper "boil-up". A small portion of the dust
was observed to pass out a pilot precipitator in a short burst
at the velocity of the gas passing through the unit, while the
remaining portion of the material was observed to fall into the
hoppers, then to rebound, and finally to escape slowly over the
145
-------
baffles and out of the precipitator. This produced a significant
vertical gradient in the concentration of the dust emitted from
the precipitator due to rapping reentrainment.
The fractional collection efficiencies of the precipitators
both excluding and including rapping reentrainment emissions
were determined. Rapping reentrainment decreased the fractional
collection efficiencies for particles larger than 2 ym. Below
2 ym, rapping reentrainment did not appear to contribute signi-
ficantly to overall emissions.
Two approaches using simplified assumptions were used to
estimate the effect of rapping reentrainment on the size of a
precipitator. The first assumed that a fixed percentage of
emissions was emitted from a precipitator independent of the
size of the unit, and the second assumed that the same percentage
of collected material was reentrained and emitted from each
section and that the recollection efficiency for the reentrained
material for all sections was the same as the efficiency between
raps, with all sections having the same efficiency. The estimates
based on the above approaches showed that the increase in pre-
cipitator size needed to recover the rapping reentrainment emis-
sions can account for 6% to greater than 80% of the size of the
unit for the cases considered.
The parameters studied during the experiments were somewhat
limited. A data bank for detailed modelling of rapping should
include additional data on rapping emissions as a function of
plate accelerations and gas velocity.
The pilot experiments also indicated a need for a study of
the parameters affecting hopper "boil-up" and the development of
methods for reducing it. Additional studies on the effects of
residual dust layers for high resistivity dust and studies of
dust cohesive properties are also needed to aid in determining
optimum rapping conditions.
ACKNOWLEDGEMENTS
This work was conducted under various contracts with the
Tennessee Valley Authority, the Environmental Protection Agency,
and the Electric Power Research Institute.
REFERENCES
1. White, H. J. Industrial Electrostatic Precipitation.
Addison-Wesley, Reading, Mass., 1963, p.331.
2. Nichols, G. B., H. W. Spencer, and J. D. McCain. Rapping
Reentrainment Study. Report SORI-EAS-75-307 to Tennessee
Valley Authority, TVA Agreement TV36921A, November 1975.
146
-------
3. Gooch, J. P., H. W. Spencer, J. D. McCain, and G. H. Marchant.
A Study of the Performance of a Full Scale Electrostatic
Precipitator with a High Specific Collection Area. Report
to Electric Power Research Institute, EPRI Project RP 413-1,
December 1975.
4. Spencer, H. W. A Study of Rapping Reentrainment in a Nearly
Full Scale Pilot Electrostatic Precipitator. EPA-600/2-76-140,
U.S. Environmental Protection Agency, Washington, D.C., 1976.
1976.
5. Gooch, J. P., J. R. McDonald, and S. Oglesby, Jr. A
Mathematical Model of Electrostatic Precipitation.EPA-650/
2-75-016, U.S. Environmental Protection Agency, Washington,
D.C. , 1975.
6. Tassicker, 0. J. Some Aspects of Electrostatic Precipitator
Research in Australia. J. Air Pollution Control Assoc.
2j5(2) :122-128f 1975. "
7. Tassicker, O. J. Aspects of Forces on Charged Particles in
Electrostatic Precipitators. Dissertation, Wollongong
University College, University of New South Wales, Australia,
1972.
8. Sproull, W. T. Fundamentals of Electrode Rapping in Indus-
trial Electrical Precipitators. J. Air Pollution Control
Assoc. 15_(2) :50-55, 1965.
9. Penney, G. W., and E. H. Klingler. Contact Potentials and
Adhesion of Dust. Trans. Amer. Inst. Elec. Eng. Part I
^:200-204, 1962.
10. Sproull, W. T. Minimizing Rapping Loss in Precipitators at
a Megawatt Coal-Fired Power Station. Paper No. 71-125, Air
Pollution Control Assoc. Meeting, Atlantic City, N.J., 1971.
11. Plato, H. Rapping of Collecting Plates in Electrostatic
Precipitators. Staub-Reinhalt. Luft (in English) 2£(8):
22-30, 1969.
12. Sanayev, Yu. I., and I. K. Reshidov. Study of Dust Reentrain-
ment Phenomena and Their Influence on Efficiency of Industrial
Electrostatic Precipitators. Promyshlennaya i Sanitarnaya
Ochistka Gazov, (Moscow) 1974 (1):l-5.
(Moscow USSR) 1974(1):l-5.
13. Schwartz, L. B. , and M. Lieberstein. Effect of Rapping
Frequency on the Efficiency of an Electrostatic Precipitator
at a Municipal Incinerator. Proceedings of the Fourth Annual
Environmental Engineering and Science Conference, Louisville,
Kentucky, March 4-5, 1975.
147
-------
14. Gooch, J.P., and N. L. Francis. A Theoretically Based
Mathematical Model for Calculation of Electrostatic Pre-
cipitator Performance. J. Air Pollution Control Assoc.
25(2):108-113, 1975.
148
-------
PAPER 9
FABRIC FILTRATION PERFORMANCE MODEL
Douglas W. Cooper and Vladimir Hampl
GCA Corporation
GCA/Technology Division
ABSTRACT
The goals of this program were to formulate and develop
mathematical methods for predicting and analyzing fabric filter
drag (pressure drop per unit face velocity) and fabric filter
collection efficiency as a function of particle size. The liter-
ature with respect to filtration is vast, but the efforts to
model fabric filtration have been few. Most attempts have used a
linear approximation to the increase of fabric drag with added
weight per area, and those who attempted to model collection
efficiency did so with models based upon arrays of cylinders.
The woven fabric structure is much more like an array of pores
(before the cake is formed) and this representation has given
rise to a new, nonlinear model of fabric drag and a new model of
fabric collection efficiency, both presented here. The nonlinear
drag model fits the experimental data much better than does the
linear. The efficiency model appears to give results which are
both reasonable and informative. These models have been combined
to form a preliminary computer program for an entire multi-bag
filtration installation, and results of such modeling are pre-
sented for both flow and efficiency variables. Also presented
are design equations which are readily adaptable for a programm-
able calculator. Finally, we indicate several areas in which
further investigation seems warranted in analyzing cleaning and
seepage.
INTRODUCTION
This presentation indicates our progress toward development
of a predictive model for industrial fabric filtration, with
special emphasis on the use of fabric filtration to control the
particulate emissions of coal-fired power plants. Primary goals
are the prediction of collection efficiency versus particle size
and the relationship between the flow per unit area and the
Dr. Cooper is now at the Harvard School of Public Health,
665 Huntington Ave., Boston, MA.
149
-------
pressure drop for a fabric filtration installation. The follow-
ing factors can be expected to influence efficiency and the
pressure/flow relationship:
1. Dust - chemical composition, particle size distribution,
particle shape, particle phase, particle concentration.
2. Gas - chemical composition (especially moisture content),
temperature, pressure.
3- Fabric - material, weave (including fill and warp
counts), finish, history (especially accumulated dust).
4. Cleaning operations - cleaning type, intensity, fre-
quency, duration.
Ideally, we would like to be able to predict fabric filter drag
as well as efficiency for all combinations of the above, but this
is a very ambitious task. In forming a useful model, one treads
a narrow path between untractable complexity and impractical
simplicity - this seems particularly difficult in modeling fabric
filtration.
DESCRIPTION OF A FILTRATION SYSTEM
Figure 1 shows a very simple schematic of a fabric filter
installation. A dust-laden flow of gas enters the filter instal-
lation with a volumetric flow rate Q and a concentration c. The
flow is divided among n compartments, the bags within these com-
partments having average dust weight per'unit area of fabric
given by W.. The pressure drop across the itn compartment's
bags is given by the equation
Api = QiSi/Ai'
in which Q. is the volume flow rate, A. is the cloth face area,
and Si is the drag caused by the fabrii and any accumulated dust.
The ratio S./A.^ is analogous to electrical resistance (with Q.
analogous to current and Ap. analogous to voltage). The total
flow, Q, is the sum of the individual flows, Q. (as long as tem-
perature and pressure corrections are made). dsually the com-
partments are in parallel and so constructed that the pressure
drop at any given time is the same across all of them, so that
Ap. just becomes Ap.
150
-------
The relationship between the pressure drop and the volume
flow for a particular installation will depend upon the locus of
the intersections of the system fan curve and the system resis-
tance curve,1 each of which can be expressed as volume rate of
flow versus pressure drop. Usually, one of the following con-
ditions holds approximately for the installation:
The fan produces a constant volume of flow, and pressure
drop changes with system resistance.
The fan produces a constant pressure drop, and the vol-
ume flow changes with system resistance.
OBJECTIVES
The cost of the installation will depend upon the fabric
chosen and its total area, among other things, and one of the
operating costs will be the power consumption due to the pressure
drop across the system. Often the available choice of fabrics
will be limited, so that the major question becomes the area of
cloth needed to handle a volume flow rate of gas.
"o
Q
Ap
W,
S2
W,
W3
Q2
W4
Q3
w.
Q
c
Figure 1. Schematic of n-compartment baghouse
151
-------
The cloth area required has as its criterion "operation at
an acceptable pressure drop across the cloth for a predetermined
cycle".2 One might add that this assumes that the collection
efficiency is adequate under these conditions. Thus one important
goal is to be able to predict the pressure drop for a particular
dust and fabric combination at a given ratio of air flow to cloth
area, the face velocity. The drag will depend upon how much dust
is on the filter surface, how it is distributed, its geometrical
nature in cake form, the geometry of the fabric, and the viscosity
of the gas. A second goal is to predict the emissions from such
a system.
OUTLINE OF MODEL
A procedure for calculating the pressure/flow relationship
and the efficiency of an installation can be developed by sub-
dividing the fabric area into smaller homogeneous sub-units
(compartments, bags, or areas on bags) and then doing the
following:
1. Calculate the drag (the pressure drop per unit face
velocity) for the sub-unit.
2. Determine the flow from the drag and the instantaneous
pressure drop.
3. Determine the penetration, the fraction of the particu-
late concentration reaching the sub-unit which then
penetrates to the clean air side.
4. Calculate the emissions rate from the sub-unit (penetra-
tion times inlet concentration times volume flow rate).
5. Calculate the new dust loading of the sub-unit.
6. Determine the new pressure drop or the new total flow
rate by combining the resistances of the sub-units
according to the law for the addition of parallel
resistances:
n
Q = Ap— = Ap/f
1=1
n
S = 2L, A./S. A (3)
To develop a time profile of the performance, this procedure must
be done iteratively, with any cleaning taken into account as
well.
152
-------
Developing this kind of model was our goal. With such a
model one should be able to predict the collection efficiency
and the relationship between flow and pressure drop for fabric
installations for reasonable combinations of variables relating
to dust, gas, fabric, and cleaning. Another part of the work
at GCA has been laboratory investigations by Dennis e_t al. , 3 the
results of which have influenced aspects of this model.
MODELS FOR DRAG - BACKGROUND
Much of the work in filtration has been done on systems
other than fabric filters. Of the fabric filtration work, most
of the predictive modeling has been directed at forecasting the
pressure drop given a fabric and dust combination.
Robinson e_t a_l. ,"* for example, produced a three-compartment
model. For the increase of drag in any one compartment with in-
creased weight per unit area on the filter, they used the linear
drag expression:
• s • S + K* w
in which the term S is the effective drag, the drag on the fabric
obtained by extrapolating the linear portion of the actual drag
versus weight curve (S versus W) back to W = 0. By solving the
equations for the total drag at constant pressure drop for the,
three-bag system, Robinson et al.1* fit the flow rate of the i
bag to:
Q. = a.txi (5)
where t. is the time since last cleaning. They used this flow to
obtain the loadings on each bag as a function of time. They
combined the individual compartment drags to get the total drag
for the system, developing a complicated approximate expression
that is only appropriate for a three-compartment system, although
the technique can clearly be extended to many-compartment sys-
tems.
Solbach5 also used the linear drag versus weight model,
Equation (4). For the case for which the pressure drop, Ap, is
fixed but the flow is changing, Solbach5 expressed the increase
of drag due to the increase of the dust loading as
dS = K, c V dt (6)
153
-------
and integrated this to give the velocity of the gas flowing
through a single compartment at any time (t = 0 right after
cleaning):
V = Ap/[2K2cApt + S ^ (7)
L _J
Solbach extended this to a multi-compartment system by averaging
V to obtain the expression:
\ /
(8)
This expression is derived assuming constant pressure drop and
infinitely many compartments. It is approximately correct
(assuming linear drag versus weight) for systems with much more
than a few compartments and even for constant flow in such a
system rather than for constant pressure drop. (The reason it is
nearly correct for a many-compartment system is that the parallel
flow arrangement tends to even out the pressure drop changes).
One obtains the required total filter area from the total volume
flow rate and the average face velocity
A = Q/Vav (9)
Dennis and Wilder6 derived an equation for the average drag
for a changing pressure drop:
K2W /2 (10)
c
Wc = CVavfcc
It can be shown that this result and Solbach's agree: in both
cases, the average drag is just the drag at half the final cloth
loading (W ).
c
This is not an exhaustive list of the investigators who
worked with linear models, but it is meant to present the linear
drag multi-compartment model. We note that the linear models
cannot describe the nonlinear part of the performance curves and
there remains the problem of obtaining both parameters K2 and S .
E
To predict K2, one approach is to use the Kozeny-Carman equa-
tion (or variants), as did Williams et al.,7 a version of which is8
K2 = (25/6)u(l-e)Sp2/Ppe3 (12)
154
-------
where K2 = cake resistivity, sec'1
y = gas viscosity, poise
e = filter cake porosity (fractional void volume)
S - surface-to-volume ratio of particulate material,
F cm""1
p = particle density, g/cm3.
K2 is quite dependent upon particle size (increasing for smaller
particles) and the porosity of the cake deposit (Unfortunately,
K2 is often measured in such English units as (inches WC/fpm)/
(lb/ft2)).
Fraser and Foley9 took a somewhat different approach, keep-
ing the K2 particle size dependence (and integrating over the
particle size distribution), while using correction factors for
particle shape, gas velocity and fabric-surface interaction,
following Billings and Wilder.10
The effective drag, S , is really just an extrapolated
value, artificial in that such a drag does not occur at that
(W 1 0) condition, so it is not surprising that there is no theory
for predicting it, only correlations. Often one must measure it
for the dust and fabric of immediate interest.
The only nonlinear model of drag versus weight of which we
are aware was presented by Davis et al.11 which we learned about
after beginning our rather different approach. Davis et al.lx
fit equations of the form
S = a Wb (13)
to data obtained for the nonlinear part of the drag versus weight
curve. They found correlation coefficients of greater than 0.90
for 104 of 106 data sets. Davis et al.J1 evaluated this at a
loading of 0.0244 g/cm2 (0.05 lb/ft2), where their drag data
generally became linear with W, to obtain S . They found that
both a and b could be correlated with K2, bat an analysis of
variance did not show that K2 was different for different fabrics
for a given dust, which is as expected, even though contradicting
the conclusion of Draemel.12 The model for drag presented by
Davis et al.Jl is useful because it includes the nonlinear, cake-
repair zone. The model we present also has this feature, with the
advantage that the parameters seem more amenable to theoretical
prediction rather than just empirical correlation.
NONLINEAR DRAG MODEL
The linear models of drag versus weight are incorrect for the
low dust loading portion of the filtration cycle, during which the
flow and the emissions through the bag are the greatest. The new
155
-------
fabric drag is S0. The conditioned, cleaned fabric drag is S
at residual loading W . Typically, the drag of the filter in-
creases with added wefght per area at one rate, K , as the con-
ditioned bag collects dust right after cleaning. This rate
(which is the slope of S versus W) changes until it reaches a
value K2, after which the rate remains constant (the S versus W
curve becomes linear) . The linear part of the performance curve
starts at loading WT- An improvement on the linear model would
be one that starts with S and K , approximates the transition to
the linear zone, and matcnes the linear part of the drag curve.
Viewing the filter as an assemblage of pores (rather than
cylinders) , we assumed that pores which were unclogged ("open")
had a different dust resistivity (Ko) than pores which were
caked (closed), assumed to have resistivity K2, as has the cake.
We would expect K0 and K2 to be different because the velocity in
the pores is much higher than through the cake but the particles
in the "open" pores are not packed together the way they are in
the cake. (Depending upon the magnitudes of such velocity grad-
ient and dust packing effects, K2 may be greater than K0 or
vice-versa. ) By assuming that pores are either caked or open and
that the decrease in the population of open pores is proportional
to the number of pores still open and to the increase in fabric
loading, dW, we obtained the expression:
dN = -(1/w*) N dW (14)
where N = number of open pores per face area of filter
W* = fabric loading that is characteristic of pore-caking.
The fraction of pores still open at loading W becomes:
where N0 is the number of pores per unit area in the fabric when
W = 0.
If the caked pores and open pores contribute to the drag in
proportion to their resistivities and their numbers, then it can
be shown:
|| = K0 f (W) -1- K2 g(W) (16)
and
where f(W) = exp{~W/W*} (the fraction of pores which are open)
g(W) = 1 - f(w) (the fraction of pores which are caked).
156
-------
By integrating Equation 16 we obtain:
S = S0 + K2W + (KO - K2) W*(l - e~W/W*) (17a)
in which S0 is measured new cloth drag; alternately
S = SR + K2W' + (KR - K2) W*(l - e~W'/W*) (I7b)
where W' = W --W , the difference between the loading and the
resiQual loading, w , at which S = S .
**, K.
We hope to be able to predict K and W* from dust and fabric
characteristics, but thus far we too have had to rely on correla-
tions. To show how well this form of equation can fit data from
an actual fabric, we present Figure 2, for a glass fabric filter-
ing mica dust and cleaned by shaking, and Figure 3, for Dacron
filtering fly ash and cleaned by a combination of shaking and
reverse air. Both are fabrics appropriate for high-temperature
operations. The optimal choice of W* produces excellent agree-
ment with the measured values of drag versus weight. The three
other such comparisons we made were similarly successful? Tables
1 and 2 give more information about the systems against which the
model was compared in our Interim Report to EPA.16
To use this model for a variety of filtration systems, we
need theoretical equations or predictive correlations for the
parameters K_, S , and W* . . . assuming we have the usual in-
formation about the fabric (permeability, from which comes S0)
and dust (resistivity, K2).
The value of KR is expected to depend upon the dust, the
fabric, and the cleaning, so we thought it might correlate with K2
for the dust and S0 for the fabric for cleaning by shaking. We
calculated (K2/K ) for data on glass,13'15 Dacron, 1 3 ' l ** ' e and
Nomex11* fabrics. (Aerosols of mica and fly ash were used with
the glass and the Dacron filters; the Nomex tests were with fly
ash). The correlation was
Kn = 515 K2 So1"035 (18)
R,
where S0 is in the units inches WC/fpm and K , K2 are in inches
WC-ft-min/lb. Similarly, correlations for tfte same five fabric-
dust combinations gave:
SD = 6.2 S, '!. °'747 (19)
R
(residual drag increases with fabric drag and dust resistivity) ,
and
157
-------
2.0 -
1.5
a
o 1.0
0.5
GLASS FIBER
EXPERIMENTAL
CALCULATED
_L
0 0.01 0.02 0.03 0.04 0.05
Wl Ib/ft2
Figure 2. Theoretical and experimental curves of glass fiber fitter
0.7
0.6
I 0.5
o
CO
0.4
0.3
0.2
0.1
0
0
T I
DACRON
xx.W*=0.45W|
= 0.41 W,
= 0.33 W.
CALCULATED
EXPERIMENTAL
1
0.01
0.02
! ib/ft2
0.03
Figure 3. Theoretical and experimental curves of Dacron-type filter
158
-------
TABLE 1. TECHNICAL DATA - FABRIC FILTERS SELECTED FOR VERIFICATION16
Ul
Type of Fabric
Glass fiber
Polypropylene
Dacron
Cotton
Polyacrylester
Fraser
Permeability,
Weight, ft3/nun
oz/yd2 Weave and Yarn Count 0.5 in. H20 Reference
9.06 3/1 crowfoot, filament 7.9
55 x 58
4.30 3x1 twill, filament 15.0
74 x 33
10.0 Plain, staple 55.0
30 x 28
10.0 Unnapped sateen 13.0
95 x 58
9.8 2x2 twill, spun 60.0
Spaite and Walsh13
Durham1 "*
Dennis and Wilder6
Dennis and Wilder6
Durham1 "*
39 x 35
-------
TABLE 2. EXPERIMENTAL DATA - FABRIC FILTERS SELECTED FOR VERIFICATION
Fabric Type
Glass fiber
Polypropylene
Dacron
Cotton
Polyacrylester
K2
KR SR WI
in. WC-ft-min/lb
16.0
6.1
12.4
15.0
4.6
400.0
135.4
93.8
72.0
26.3
in. WC/fpm
0.84
0.35
0.08
0.5
0.05
lb/ft"
0.0036
0.0135
0.0066
0.0075
0.030
Type of
Cleaning
Shaking
Shaking
Shaking and
reverse air
Shaking and
reverse air
and shaking
only
Shaking
Dust Type
Wet ground
mica
Fly ash
Fly ash
Fly ash
Fly ash
Remarks
Conditioned,
velocity 2 ft/min
Conditioned,
velocity 4 ft/min
Used fabric,
velocity 3 ft/min
Velocity 3 ft/min
Conditioned,
velocity 4 ft/min
Reference
Spaite and
Walsh13
Durham1 *
Dennis and
Wilder6
Dennis and
Wilder6
Durham1 *
-------
W* = 7.68 x 1(TVK20'913 So°*945 (20)
(pore caking requires more material for rather open fabrics and
more coarse dusts). Again, the parameters are in English units
(W is in Ib/ft2 and K is in the units of S/W).
It should be mentioned, at least in passing, that data from
the reported measurements of K2 we used were better correlated
with the inverse of particle mass median diameter than the
inverse squared as might be expected from the Carman-Kozeny
equation16) and showed a face velocity effect (as has been re-
ported before17 ).
To sum up: the form of our nonlinear expression fits quite
well that data for which we evaluated it. The parameters in
the model should be theoretically predictable, a problem on which
we are working, but until they are, one can use correlations such
as those presented here. Such a model promises to be an improve-
ment on the linear model because it can describe that part of
the filter performance for which flow and emissions are greatest,
the low-loading, nonlinear part.
EFFICIENCY MODELS - BACKGROUND
Before presenting our model for woven fabric filtration ef-
ficiency as a function of particle size, we will discuss several
recent efforts in that direction.
A theory of particle collection by clean cylindrical fibers
was adopted by Fraser and Foley9 for the evaluation of the collec-
tion filter efficiency, expressed as a function of three collec-
tion mechanisms. They used an equation derived by Davies18 for
low gas velocity (Re ~ 0.2) and single cylinders, incorporating
impaction, interception and diffusion. Then, they applied a cor-
rection factor to the single fiber efficiency y to account for
the effect of the neighboring fibers
Y = (1 + 4.56) (21)
M
to obtain the total collection efficiency. The penetration for a
fiber was expressed as:
Pn = exp
(22)
161
-------
Due to the difficulties related to the precise measurement of the
fiber diameter, they used an effective fiber diameter, derived by
Davies:l8
70 n 61*5L(1 + 5231*5) 1 h
3 _ J (23)
(The filter packing density is 8, its thickness is L, and the gas
viscosity is n •) That model of the penetration of the particles
through the clian fiber filter was extended to the dusted fabric
filter under the assumption that the main effect of the dust on
the fabric itself is to change the effective fiber diameter. An
equation for the change of the thickness of the filtering layer
during the filtration cycle was developed by assuming that the
dust fills up first the free volume of the filter (at the packing
density of the dust) before building up the dust cake. The resi-
dual dust loading was expressed as a function of a cleaning param-
eter and permeability.
The collection efficiency used in that model was based on
the calculation of particle trajectories around a single fiber,
derived by Davies18 for a clean fiber with a very small diameter.
It is doubtful that the same equation will hold for fiber filters
having fiber diameters ~102 ym. If we assume the average face
velocity of 2 cm/sec (4 ft/min), with fly ash particles having a
diameter of 10 ym, fiber diameters approximately 800 ym (which is
an average yarn size) and filter packing density around 0.3, then
the Stokes number equals 0.02, a lower value than the critical
Stokes number, below which the impaction mechanism does not con-
tribute to collection efficiency. (Also under such conditions
the diffusion parameter is negligible.) Then the penetration for
the particles would be 0.99, considering the interference effect
of neighboring fibers to be (1 + 4.5 B). Modeling the fabric
filter as an array of cylinders does not seem successful thus far.
Leith and First Model
By using tagged fly ash aerosols, Leith and First19 were able
to distinguish between those fly ash particles which, under labora-
tory conditions, penetrated a needled felt fabric filter immed-
iately and those particles which exhibited a delay in their pene-
tration. These researchers postulated three types of penetration
mechanisms: direct penetration, gradual seepage of the dust, and
the breakage and penetration of plugs of material in the vicinity
of pinholes.
162
-------
The experiments were done at face velocities from 5 cm/sec
(10 fpm) to 15 cm/sec (30 fpm) for dust cakes up to 60 ym thick.
These velocities are higher than those in normal use for the fil-
tration of fly ash Ul cm/sec) and these cake thicknesses are
rather low. The efficiency as a function of particle size was such
that the penetration was found to "remain relatively constant for
particles from 0.3 to 4.0 micrometers in diameter". The penetra-
tion increased with face velocity. From their figures for mass
flux versus time (at 10 cm/sec) we conclude:
1. Straight through mass flux seemed to decrease (roughly
exponentially) with deposit thickness, but may have gone
through a minimum near 20 ym thickness.
2. Seepage mass flux remained fairly constant with time and
deposit thickness.
3. Pinhole plug mass flux decreased with increasing deposit
thickness.
The fraction of the total penetration which was due to pin-
hole plugs and seepage was greater than the direct (straight
through) contribution (at 60 ym thickness) for 15 cm/sec but sub-
stantially less than the direct for 10 cm/sec, which suggests the
emissions due to seepage and pinhole plugs might be very much
less than the direct at 1 cm/sec (2 fpm).
Leith and First19 postulated the following functional forms
for the three types of penetration:
-aXb
1. Direct penetration proportional to e , where X is the
deposit thickness.
2. Constant seepage mass flux versus thickness of deposit.
—cX
3. Pinhole plug mass flux proportional to Xe (in which
a, b, c, are proportionality factors).
The correlations they found in using these equations ranged
from 0.86 to 0.91 correlation coefficients, for mechanism-by-
mechanism comparison, adding support for their proposed mechanisms.
Leith and First19 found only a weak particle size dependence
for efficiency, with the efficiency decreasing slightly as parti-
cle size increased. The particle size dependence, the velocity
dependence, and their general appraisal of the filtration process
led them to conclude:
163
-------
"Because media filtration theory does not describe the trends
in penetration found in a fabric filter, and was not developed
for the operational conditions found there, it should not be
used to predict or interpret the penetration characteristics
of fabric filters."
This conclusion underscored the need for a new approach to
the prediction of fabric filter efficiencies, especially for woven
fabrics.
A NEW MODEL FOR FABRIC FILTER COLLECTION EFFICIENCY
As indicated by our review of former models, most treatments
of filter collection efficiency were not for fabric filters but
for other kinds of filters. The synthesis attempted for fabric
filters by Fraser and Foley9 used as its basis formulas for col-
lection on cylinders. Because of the overlapping caused by
weaving and the closeness of the yarns, the flow in a fabric
filter is very unlike that of a cylinder in a uniform flow stream,
the starting point for models of filtration based on collection
by cylinders. Rather, the fabric produces a flow which is
directed through the interstices of the weave at and near the
intersections of the yarns. This much more resembles flow to an
array of apertures than it does flow around single or grouped
cylinders. Because the nature of the flow has a strong bearing
on the collection efficiency, one expects a pore model to be
inherently better than models which start from a cylinder or an
array of cylinders.
Our analysis fits into a baghouse simulation model outlined
in the flow chart in Figure 4. Once the starting data are put
into the simulation, the program calculates flow, pressure drop,
fabric loading, penetration, and emissions as a function of time
for the bags in the simulated baghouse. This section discusses
the equations and analysis which went into calculating the
emissions.
Collection Mechanisms
The model can calculate the fraction of incoming particles
which are captured in the cake, in the fabric, and in the pinholes
which go through cake and fabric. The collection mechanisms model-
led are impaction, interception, and diffusion.
164
-------
MAIN
1
T T
SUBROUTINE DATA
MODEL INITIALIZE
1
* *
SUBROUTINE SUBROUTINE
READIM READIT
,
f LOOP ON "\ SUBROUTINE
~*\ TIME ) PLOTIN
1
j * L T
-K ss0" ) sr™
<3RAGULATi>
1
ITT
SUBROUTINE SUBROUTINE SUBROUTI
FABDRG CAKDRG PINDRG
IT
f END BAG \
\ LOOP J
1 '
<3?owsLAT^>
?
f LOOP ON
V BAGS
, ,
/ LOOP ON
y_ PARTICLE SI2
V
J
\
-E J*
.•
X^CALCU LATEEN.
\PENETRATIOrJX
\
T T
IE SUBROUTINE SUBROUTINE
DIFUSN CAKIMP
T
t
SUBROUTINE
CAKDIF
~^>
V PARTICLE LOOP J
1
(END
BAG LOOP
*
"^
J
< PRINT & PLOT "\
PRESSURE, EMISSIONS, >
WEIGHT ^X
|
/C\__to/ END TIME
Vi/ *^ LOOP
*
f END
*
EXECUTE
PLOT
PROGRAM
~)
)
Figure 4. Baghouse simulation program flow chart
165
-------
Collection by Impaction, Interception, and Diffusion in the Cake. The particulate
material collected by the fabric filter builds up first in the
vicinity of the pores. As it builds, this deposit starts to form
a cake through which a substantial portion of the flow passes, and
particles are captured by the particulate matter of the cake. We
seek to model the capture of particles by the filter cake. The
geometry is quite complex where polydisperse aerosols are involved.
The uncertainties are so great with regard to geometry that avail-
able information warrants only an approximate model. We have
formed such a model on the basis of analysis of characteristic
time scales, characteristic lengths, etc.
An idealized view of the filter cake is shown in Figure 5.
The cake is viewed as a series of layers, with a total thickness
Lr in cm. The cake thickness is given by
L = W/p (1 - e)
in which W = particle loading, g/cm2
p = particle material density,
e =
g/cm3
void volume fraction, porosity.
L*
I
05
cS
(24)
Figure 5. Fabric filter cake, resting on fabric
By taking the number of particles and dividing it into the
volume of the cake, we can get the mean volume of cake per parti-
cle, thus the mean thickness of each layer:
1/3
L* = fw/Sd - e) 1 *'- VMD
in which VMD = particle volume mean diameter, cm.
(25)
166
-------
Assuming that the number of void volumes in the cake is equal
to the number of particles, we have the following void volume
mean diameter, d :
VMD (26)
The mean velocity in the open portions of the cake is given by:
U = vf/e (27)
in which vf = face velocity, cm/sec.
A characteristic velocity for flow_through the cake is the
face velocity divided by the porosity, U. A characteristic
dimension over which the flow changes direction is the mean diam-
eter of a void, d .
v
The inertial parameter (Stk, the Stokes number) can be de-
fined from these equations as:
Stk = UT/d (28)
' v
This parameter determines how large inertial impaction may
be. We make the simple assumption (e.£., see Fuchs20) that the
impaction efficiency of a single cake layer is approximately
equal to the Stokes parameter at each void volume (up to Stk = 1):
En = Stk (EIi < 1) (29)
The likelihood that a particle will penetrate through one
void volume is just one minus the efficiency:
Pnli = 1 - Ux/dv (30)
(T is defined by Equation 46) and the likelihood of penetrating
M such volumes U.<2., through M layers) is:
Pnz = (1 - pnn>M. (3D
(This assumes that the collection efficiency of each void volume
in the cake through which the aerosol flows is independent of the
earlier portions of the flow path.) The number of such layers is
approximately the thickness of the cake, L, divided by the mean
distance between particles forming the cake, L*, the mean layer
thickness:
M = L/L*. (32)
167
-------
Interception is the collection contribution due solely to the
size of the particles compared with the dimensions of the flow
channels. Even if the particles follow the streamlines of the flow,
those with centers on streamlines which come within a particle
radius of an obstacle will strike the obstacle and be captured (it
is assumed). We can approximate the likelihood of penetration in
passing through the i.th layer as that for passage through a cylin-
drical channel of diameter d :
pncVi= (1 - W2 (33)
and from this we derive an approximate expression for the pene-
tration of a particle through the dust cake of thickness L by:
Pnc = d - dp/dv)2 L/L*. (34)
The Brownian motion of particles causes them to diffuse in
a concentration gradient so as to transport particles from higher
concentration regions to lower. In the presence of a surface to
which the particles adhere, there will be a net diffusive flux of
particles to the surface due to this mechanism. The equation
governing this diffusion for a flowing aerosol is given by:8
|J + v • $n = D V2n (35)
d t
in which n = number concentration of particles of diameter d
-»• P
v = flow velocity
D = particle diffusivity
Vn = gradient operator
-»•
Vz = Laplacian operator
For sy.s.tems^of interest it is usually true that the convec-
tive term, v • Vn, is negligible. Within a void volume this
equation becomes approximately:
An/n = DAt/(0.5 d )2 (36)
for the capture due to diffusion. The time, At, is the charac-
teristic flow length divided by a characteristic flow:
At = dv/U (37)
As before, for M (=L/L*) layers in the cake, the penetration
becomes:
PnD = (1 - 4 D/U dv)L/L*. (38)
168
-------
As Fuchs has noted,20 there is no simple way for combining
correctly the collection efficiency terms (or, equivalently, the
penetration expressions). Commonly, the mechanisms are treated as
operating independently, in which case the penetration becomes
the product of the penetrations:
Pn = Pn Pn Pn . (39)
We have used this approach in our model.
Impaction and Interception at a Pore or Pinhole. Near the intersections of
the fabric yarns are openings which we call pores, the geometries
for which are discussed in the Interim Report by Dennis ejt al.3
and, to an extent, in Appendix A of our report. 6 It has been
observed that often a fraction of such pores, and/or pores caused
by other means, stay open during the filtration cycle. We denote
them "pinholes". Our model simplifies both pores and pinholes
into circular cylindrical channels in the fabric (and in the cake
as well, for pinholes). The task then becomes to calculate the
collection efficiency of such an array of channels.
Spurny e_t aj..21 presented calculations for a somewhat similar
situation in their analysis of membrane and nuclepore filters,
modeled as a highly porous array of parallel circular cylindrical
channels. It can be shown that their equations for impaction
do not go to the correct limits for low-porosity structures, which
is our situation for fabric filters having a few percent or less
free area.
Smith and Phillips22 found numerical solutions to the prob-
lem of the collection efficiency of an aperture due to impaction
and interception, but they included only capture at the face of
the material having the orifice and not along the walls of the
channel, which can be shown to be substantial in our context.
Thus, although we could use equations available in the liter-
ature for collection due to diffusion in a channel, we did not
have applicable equations for impaction and interception of parti-
cles by the pores and pinholes and had to derive approximate
expressions, discussed next.
169
-------
Figure 6 shows the geometry of the model. The pore (channel,
orifice, etc.) has a radius (or equivalent radius) R* and a
length L . The opening is one of several and the area trRo is
the tota? face area divided by the number of pores, distributed
approximately uniformly. The face velocity, v , far_away from
the pore, is just the flow through the pore, Q , divided by the
area per pore:
vf = Tr (40)
The flow is simplified by dividing it into three regions (see
Figure 6) :
I. In this region (radial distance, r, greater than RO )
there is a uniform face velocity toward the plane of the
opening, and the face velocity is given by:
vf = QpARo2 (41)
II. In this region (radial distances: R* < r < Rj, ) the flow
is the same as it would be in potential flow to a point
sink, being directed radially toward the center of the
pore face, with the velocity at distance r governed by:
v(r) = v*(R*2/r2) (42)
in which
v* = Q /2wR*2 (43)
P
III. In the third region (r < R*) and within the channel
there is assumed a uniform flow profile.
Of course, this set of flow profile assumptions is somewhat
different from what actually would take place. Still, far away
from the pore opening the flow would generally be uniform; there
would be an intermediate range for which the flow would be ap-
proximately that toward a point sink at the center of the orifice;
and converging flow at the orifice would be expected to produce
a rather uniform flow profile near the entrance before the typical
parabolic flow was established for laminar flow conditions. Thus
the flow model preserves many of the features of a more rigorous
solution of the flow velocity field.
170
-------
Figure 6. Schematic of flow model for analysis of impaction and interception
171
-------
A particle with an initial velocity of vf will travel a
distance L1 in a stationary medium, with L1 given by:20
S J. S X
= L (1 - e u/ l) (44)
where
L = T v (45)
o J_ \.w
T = Cp d 2/18y (46)
in which C is the Cunningham slip correction, p is the particle
density, and d is the particle diameter. The Variables L and
T are the partfcle stopping distance and particle relaxatiSn
time,20 and t is the time.
We will not give more details of the rather complex method
used by us to obtain efficiency due to impaction. Essentially it
involved comparing the stopping distances (L ', including mean
residence times in regions II and III) with the distance a parti-
cle would have to travel to strike the face of the pore or the
inner wall, keeping track of particles which were captured in one
region so as not to double-count them in capture in subsequent
regions.
The other mechanism considered was diffusion to pore and
pinhole walls. The formulas for the decrease in particle con-
centration in traversing a tube while diffusing have been pre-
sented by Fuchs.20 The concentration decreases approximately
exponentially with the parameter u as the argument:
y = D Lp/Vo R*2 (47)
in which the particle diffusivity is D. The approximate formula
for y much less than 1 is:20
Pn = l-E = l- 2.56 u2/3 + 1.2 y + 0.177 y*/3 (48)
Summary of Model Equations. The essential aspects are the collec-
tion mechanisms of impaction, interception, and diffusion as ap-
plied to the filter fabric, the cake, and pinholes in fabric
and/or cake.
172
-------
RESULTS OF MODELING OF FABRIC FILTRATION SYSTEMS
Introduction
This portion will present some of the results we have ob-
tained in modeling fabric filtration systems with an initial
emphasis on coal-fired power plants using fabric filtration,
such as the Sunbury plant that was one of the subjects of the
Interim Report by Dennis et al.3 The order is as follows:
1. Q and Ap versus time for a 15-bag system, using a linear
drag model.
2. Collection efficiency versus particle size and time for
a single bag with linear drag.
3. Collection efficiency versus particle size and time for
a single bag with nonlinear drag.
4. Collection efficiency versus particle size and time for
a three-bag system, using nonlinear drag.
Flow Modeling
In Reference 16 we presented the computer program for our
model. The drag equation used was linear:
S = S-, + K2 W (49)
.Ci
W = cVT (50)
where S = 0.5 in. WC/fpm = 2.5 cm WC/(cm/sec) = 250 (N/m2)/(cm/sec)
K2 = 11 in. WC/fpm/lb/ft2
=1.1 (N/m2)/(cm/sec) (g/m2)
c = 5 gr/ft3 = 11.5 g/m3
t = time since cleaning, minutes.
Either V was constant (2 fpm = 1 cm/sec face velocity)or it was ob-
tained from the pressure drop and the drag. Each bag was cleaned
instantaneously (to the residual weight level) every 30 minutes,
one bag cleaned per 2 minutes. The results are shown in Fig-
ures 7 and 8, where the information is plotted every minute so
that we are getting an instantaneous reading at two points in the
cycle, just after cleaning and midway between cleanings. From
Figure 7, we see that after about two complete cycles (60 min-
utes) , the pressure drop has come into dynamic equilibrium, with
a minimum near 1.4, a midpoint about 1.45, and an (inferred)
maximum about 1.5 in. WC (375 N/m2). For 15 bags, the range of
pressure drop values is much less than for one bag, as we
expected.
173
-------
1.50
1.40
o
g
c 1.30
CO
CO
£
OL
1.20
1.10
20
I
40 60
TIME, min
80
100
Figure 7. Test run No. 5: 15,-bag simulation pressure versus time graph
3.20
2.80
111
DC
§2.00
1.60
1.20
i
OBAG NO. 1
ABAC NO. 2
• BAG NO. 3
• BAG NO. 4
ABAC NO. 5
_L
20
_L
I
40
100
120
60 80
TIME, min
Figure 8. Test run No. 5: baghouse simulation individual flow rate graph
174
-------
Because the many-bag system greatly reduces the pressure
drop variations with time, we can approximate this new system as
follows: even though the total system is one with a constant
total flow, it behaves, bag-by-bag, almost as though each bag is
kept at a constant pressure drop (and has a variable flow). This
is significant, in turn, because we have available to us the
general expression governing the relationship between drag and
time (thus flow and time) for a bag with constant pressure drop
and variable flow, which we presented as Equation 8, from
Solbach.5
Figure 8 shows the flow rate for bags No. 5 to No. 1 (the
cleaning order was from No. 15 to No. 1). The total flow rate
remained constant, but the individual flow rates declined for
these bags as they accumulated more weight than the other bags in
the system, which were being cleaned before them. When it be-
came the turn of bag No. 5 to be cleaned, its flow rate increased,
while the others continued to drop. This process repeated itself
with the other bags, and it is interesting to note that their
flow rate versus time curves all look very much the same except
for being displaced in time.
Efficiency Modeling
Figure 9 shows the results of modeling the collection effi-
ciency of a single bag with the program discussed above and shown
in Figure 4.
10
-
10
-2
tr
i-
UU
2
1Q_
io-
10
5
I
I
O.
:—Q—8— ft —fi— Q
O 0.1 ym PARTICLES
A 0.3 Jim PARTICLES
• 1.0 um PARTICLES
& 3.0 um PARTICLES
• 10.0 um PARTICLES
$ ALL PARTICLES
15
45
60
30
TIME, min
Figure 9. Test run No. 13a baghouse simulation penetration
175
-------
The conditions for these results are listed as part of the
program output, shown in Table 3. Volume flow rate per unit area
was 2 fpm (1 cm/sec). The pore (fabric opening) sphere of in-
fluence radius is just the radius of a circle of area equal to
the area per pore in the fabric. The pore radius is the equiva-
lent open area radius of the pores. The pore length is the yarn
thickness. Density is the particle density pp, chosen to be one
(the density of fly ash would have been a better choice). S0
and K0 were not used, but rather Sg and K2 were the same values
as above. Gas viscosity was that appropriate for temperature of
350°K (= 77°C = 171°F); a higher temperature would probably be
more appropriate for coal-fired systems. The caked area (fraction
of filter from which the cake is not removed by cleaning) was set
to zero. The pinholes, which never close during the filtration
cycle, were given the same radius of influence as the pores, with
a radius estimated as twice the size of a typical pore, and with
a number per unit area corresponding to about 2 per square inch
(see Dennis e_t a_l.3) The cake dust was assumed to have a volume
mean diameter of 5 ym and a porosity of 0.5. In Table 4 we give
approximate size distribution for the Sunbury fly ash (Dennis
et a]..3) used to approximate the inlet particle size distribu-
tion; this is log normal with a mass median aerodynamic diameter
of 6 pm and a geometric standard deviation of 3.0. The values
0.1, 0.3, 1.0, 3.0, 10, 30 vim have been used as the midpoints of
equal logarithmic intervals (geometric means of the intervals)
for the size distribution. The penetration results in Figure 9
are plotted on semilogarithmic scale. The penetrations are
weighted by the mass fraction values in Table 4 to give the curve
marked "all particles".
TABLE 3. TEST RUN NO. 13A BAGHOUSE SIMULATION INPUT DATA
Number of bags: 1 bag
Cycle time: 60 min
Number of increments per cycle: 15 increments
Velocity, Q/A: 1 cm/sec
Concentration: 1.142 x 1Q~5 g/cm3
Initial drag resistance, S : 2.5 x 103 dyn/cm2/cm/sec
Cake resistance coefficient, K: 1.12 x 10s dyn/cm2/cm/sec/g/cm2
Pore sphere of influence radius: 4.78 x 10~2 cm
Pore radius: 2.5 x 10~3 cm
Pore length: 4.0 x 10~2 cm
Yarn radius: 2.0 x 10~2 cm
Particle density: 1.0 g/cm3
(continued)
176
-------
TABLE 3. (continued)
Gas viscosity:
Cake porosity:
Temperature:
Caked area (uncleaned fraction)
Pinhole radius of influence:
Pinhole radius:
Particle volume mean diameter:
Number of pinholes per cm2:
Particle information:
2.08 x 10 " poise
0.5
350°K
0.0
4.78 x 10~2 cm
5.0 x 10~3 cm
5.0 x 10"" cm
0.31
Diameter, Mass
cm Distribution
1.0
3.0
1.0
3.0
1.0
3.0
x
X
X
X
X
X
10"5
10~5
10"*
10~"
10~3
io-3
0.
0.
0.
0.
0.
0.
00
01
12
32
38
17
Diffusion
Slip Constant, Relaxation time
Correction cmz/sec (T) , sec
2.
1.
1.
1.
1.
1.
87
56
16
05
02
01
7
1
2
8
2
8
.02 x
.27 x
.85 x
.51 x
.49 x
.21 x
10~6
10~6
10~7
10~8
10~8
io-9
7.79 x
3.82 x
3.16 x
2.58 x
2.75 x
2.46 x
10 "8
IO"7
10~6
ID'5
io-1*
io-3
TABLE 4. APPROXIMATE SUNBURY FLY ASH SIZE DISTRIBUTION
*
(mass median aerodynamic diameter = 6 urn, a =3)
Size Interval,
Logarithmic Midpoint,
Geometric Mean,
ym
Aerosol
Mass In
Size Interval,
17
5
1
0
0
0.
.2
.7
.52
.17
052-
-52
-17
-5.
-1.
-0.
2
7
52
-0.17
30
10
3.
1.
0.
0.
0
0
3
1
17
38
32
12
1
«1
177
-------
The particle size with the greatest penetration (of those
tested) is 0.3 ym diameter, which is in accord with experimental
results from many different fibrous filtration studies. The ini-
tially linear slope on this scale (semilogarithmic) is consistent
with an exponential decay as the cake gets thicker (recall this
is constant velocity, thus the rate of cake build-up is constant).
Once the region is reached where much of the flow and virtually
all of the penetration is that through the pinholes which do not
cake, the penetration becomes nearly constant with cake thick-
ness. A substantial fraction of the 0.3 and 1.0 \im particles
penetrate the clean fabric, but then each of these decreases to-
ward an asymptote as the pinhole penetration predominates. The
3 pm particles apparently penetrate only the pinholes and this
penetration does not change appreciably during the cycle. The
10 and 30 urn particles do not even penetrate the fabric or the
pinholes at all (Pn < 10~5) so that the cake buildup does not
affect them; they are caught by inertial impaction and by inter-
ception by fabric pores and by pinholes. The mass penetration,
"all particles", starts at about 10% and rapidly declines to
about 1%. The leveling off of the curve for total mass penetra-
tion indicates that this penetration is almost wholly due to
pinholes, and it will thus be quite sensitive to the assumptions
made about the pinhole population. (The pressure drop versus
time, which follows the linear model, went from 2.5 cm WC (1 in.
WC) to about 7 cm WC (about 2.7 in. WC), reasonable values.)
Figure 10 has penetration versus time curves for the same
situation except that the nonlinear model of drag was used. We
started with an initial drag which was roughly that of the Sun-
bury fabric (S0 = 0.05 in. WC/fpm), then used the correlations
developed above to get the ratio of K2 to KR from S0, KR/K2 =
12.5, and to get W* = 0.004 lb/ft2 = 0.002 g/cm2. The efficiency
curves are very similar to those for the linear drag model,
except that the initial rapid change in pressure drop somewhat
changes the distribution of the flow between the fabric plus cake
and the pinholes, shifting the flow more rapidly to the pinholes,
thus producing a bit of increase in the penetration of the 3 urn
particles initially, before their penetration becomes effectively
independent of time. Again, major features include the maximum
penetration by the 0.3 jam diameter particles, the initial expo-
nential decrease for those particles being filtered out primarily
by the cake, and the important role of the pinholes as the cake
becomes thick.
178
-------
10-1 \
10
-2
<
DC
(-
111
10
-3
10
,-4
xB^|-«-l^^gri=ft~a.fl_i_i__
O 0.10 urn
& 0.3 urn
• 1.0 urn
m 3.0 Mm
D 10.0 jjm
A ALL
30
TIME, min
Figure 10. Test run No. 13b: baghouse simulation penetration versus time graph;
single bag, nonlinear drag, with pinholes
To demonstrate further the impact of the pinholes, we re-
peated the last test (nonlinear drag, etc.) but removed the pin-
holes, so that all the flow is through fabric pores or cake only.
Figure 11 is the penetration without pinholes. The most marked
difference is that collection efficiency for all particle sizes
now continues to decrease as the cake thickens, and this effi-
ciency becomes better than 99% by mass for the particles within
10 minutes. The 0.3 ym penetration is still greatest. The 0.1,
0.3, and 1.0 ym particles all show an exponential decrease with
cake thickness now. The 3 ym particles show more rapid changes
in collection at the beginning part of the cycle, where pressure
drop is changing most rapidly, and this is typical for other col-
lectors where inertial impaction is the predominant mechanism, as
it is for these 3 ym particles.
The final figure in this series is Figure 12. Here we have
used three bags rather than one, the same nonlinear drag param-
eters as before, and we have restored the pinholes. First, the
bags all have the same flow rate, then after one is cleaned, only
two have the same flow rate, and the third is much higher in
flow. Finally, all three have been cleaned once, and a pattern
develops which will repeat itself.
179
-------
oio
p
cc
h-
-2
UJ1Q
10
o
" •*
w
O 0.10 um
A 0.3 Mm
• 1.0 um
• 3.0 um
D 10.0 um
\
\
10~5»
• •_'• •
15
a a'i m a
•*-
30 45
TIME, min
60
Figure 11. Test run No. 15 baghouse simulation penetration versus time graph;
single bag, nonlinear drag, without pinholes
g
S 10--=
a
10
-4
O 0.10 um
£ 0.3 um
• 1.0 urn
• 3.0 urn
D 10.0 um
A ALL
30
TIME, min
Figure 12. Test run No. 16 baghouse simulation penetration versus time graph;
three bags, nonlinear drag, with pinholes
180
-------
Figure 12 shows the changing pattern for particle penetra-
tion versus time with particle size as a parameter. The relative
order of penetration with respect to size is the same, and we
also still see the change of the penetration for the 3 ym particles
as pressure drop changes. Finally, the asymptote provided by the
pinhole mechanism is evident. The three-bag system with non-
linear drag produces nearly the same total mass penetration as the
one-bag system after 15 minutes, but the cleaning process then
returns one of the bags to low efficiency and high flow rate,
producing a major increase in penetration for the total system.
Calculator Model for Drag
In some cases, a quick and approximate estimation of an
average air-to-cloth ratio of a baghouse unit is necessary or
satisfactory. Therefore, we tried to derive a simple equation for
calculating an average air-to-cloth ratio. For our derivation we
used an expression for the average face velocity (air-to-cloth
ratio), V, which was derived by Solbach5 and independently also
by Dennis and Wilder.6 This expression is based on the linear
drag model. The average face velocity is given by Equation (8):
V =
K2 ct
2 ctcAp + SE'
- S
E
(51)
where K2 = dust resistivity, in. WC-min-ft/lb
c = dust concentration, lb/ft3
t = total filtration time, min
c
Ap = pressure drop across the baghouse, in. WC
S = effective drag, in. WC/fpm.
E
If we express the effective drag, S , as a function of the
clean fabric drag, S0, by16:
_ . q 0.672
= 7.4 S0
(52)
and K2 as a function of particle mass median diameter, d , by16:
K2 = 142.4 d
-1.06
p50
(53)
the Equation (51) becomes:
181
-------
- 7.4 S0 °'672 + ^54.76 So l'3" + 284.8 d ,Q -1'06 Apctr
V = - . ... P - —
-1 flfi
142'4 dp50 ctc
Then, using industrial data for Ap, d , S0, c and t , shown in
Table 5, we calculated the face velocity (air-to-clSth ratio) for
each baghouse. These values are shown in Table 5 along with the
operating industrial values of the face velocity. (Sources of
the data are listed elsewhere.16) The ratios of operating to
predicted velocities are very close. The average of all values
of that ratio was 2.29 so that we have a corrected estimate, V',
of the average face velocity:
V1 = V/2.29 (55)
The differences between corrected predicted and operating values
were within ± 15%. More data would be needed to improve the
reliability of our last equation, obviously.
FINAL COMMENTS
What is presented here seems the most detailed model yet of
a fabric filtration system, and it has sufficient flexibility to
have added to it a number of refinements as our information
warrants.
The areas which seem most in need of further study are
cleaning and seepage. Regarding cleaning, we would like to know
the amount and spatial distribution of the residual dust as a
function of cleaning type, intensity, duration, and frequency for
several dust/fabric combinations, especially those of applicability
to controlling fossil-fuel boiler emissions. Regarding seepage,
we would like to know what variables affect the rate at which it
occurs and what the relationships are. Of course, the applica-
tion of fabric filtration does not require such information, but
the optimization of fabric filtration does.
Accurate methods for prediction of fabric filter drag and
collection efficiency would greatly aid the design of optimal
fabric filter systems. By viewing the fabric as an array of
apertures rather than an array of cylinders, we have been able to
derive expressions describing the nonlinear dependence of filter
drag on fabric dust loading and the collection efficiency of the
fabric and the dust cake as functions of particle size. Much
work remains to be done, especially regarding cleaning of fabrics,
but we believe that this new approach can provide a useful frame-
work for future development as well as yielding useful insiqhts
concerning the factors which determine drag and collection effi-
ciency. As the model is improved and as more data from operating
182
-------
TABLE 5. PARAMETERS FOR FIVE BOILERS USING FABRIC FILTRATION
Plant
Pennsylvania Power
and Light Co. ,
Sunbury Station, Pa.
Clean Cloth
Fraser
Permeability,
fpm, 0.5 in. H20
54.3
Colorado UTE Electric 86.2
£"! Assoc., Nucla Power
^j Station, Colo.
Sorg Paper Co. ,
Middletown, Ohio
Crisp County Power
Commission, Cordele,
Georgia
Pennsylvania Power
and Light Co. ,
Holtwood Station, Pa
55.0
55.
45 - 65
(50)
Dust
Concentration,
gr/ft3
2.
1.93
3.2
2.6
7.7
Pressure Particle
Total Filtration Drop, Mass Median
Cycle Time, min -in. H2O diameter, um
33 2.5 8
29.4 4.2 8
30 4-5 12
60 4. 12
33 6.5 10*
Operating
Velocity,
fpm
1.92
2.79
2.07
2.2
2.26
Velocity
Predicted
by Eq. 54
V, fpm
3.90
6.50
6.10
4.69
4.66
*Assumed values.
-------
systems are incorporated, it is hoped we will have a method which
will facilitate the application of fabric filtration to the prob-
lem of obtaining power economically with minimal environmental
impact.
ACKNOWLEDGEMENTS
We appreciate the help we received from our colleagues at
GCA, especially the computer programming done by Roger Stern and
the comments and information given us by Richard Dennis. This
program has been funded through EPA Contract No. 68-02-1438.
The project officer is Dr. James H. Turner, Industrial Environ-
mental Research Laboratory - Research Triangle Park, N.C.
REFERENCES
1. Jorgensen, R. Fan Engineering, 7th ed, Buffalo, N.Y.,
Buffalo Forge Co., 1970.
2. Snyder, C.A. Ind. Eng. Chem. Process Des. Dev. £7:960, 1955.
3. Dennis, R., R.W. Cass, and J.E. Langley. Development of
Data Base and Fabric Filtration Models for Design of Parti-
culate Control Systems. Environmental Protection Agency,
Research Triangle Park, N.C. Contract No. 68-02-1438,
Interim Report for Task 5.1, Dec 1975. 119 pp.
4. Robinson, J.W., R.E. Harrington, and P.W- Spaite. Atmos.
Environ. 1:499-508, 1967.
5. Solbach, W. Staub (English) 29(l):28-33, 1969.
6. Dennis, R., and J. Wilder. Fabric Filter Cleaning Studies -
Appendix J. EPA-650/2-75-009, U.S. Environmental Protection
Agency, Washington, D.C., 1975.
7. Williams, C.E., T. Hatch, and L. Greenberg, Heating, Piping
and Air Conditioning 3^2:259-263, 1940.
8. Bird, R.B., W.E. Stewart, and E.N. Lightfoot. Transport
Phenomena, New York, Wiley, 1962.
9. Fraser, M.D., and G.J. Foley. A Predictive Performance
Model for Fabric Filter Systems: I - Intermittently Cleaned
Single-Compartment Systems. Presented at 67th Annual
Meeting of APCA, Denver. 1974.
10. Billings, C.E., and J. Wilder. Handbook of Fabric Filter
Technology. Vol. I. Fabric Filter Systems Study. NTIS
PB 200648, USDHEW, Washington, D.C., 1970.
184
-------
11. Davis, W.T., K.E. Noll, and P.J. LaRosa. A Predictive Per-
formance Model for Fabric Filtration Based on Pilot Plant
Studies. Proceedings of the User and Fabric Filtration
Equipment II, Specialty Conference APCA, Niagara Falls,
' »X« i 27 / J »
12. Draemel, D.C. Relationship Between Fabric Structure and
Filtration Performance in Dust Filtration. EPA-R2-73-288,
U.S. Environmental Protection Agency, Washington, D.C.,
-L -/ / j •
13. Spaite, P.W., and G.W. Walsh. Amer. Ind. Hyg. Assoc. J.
21:357-365, 1963.
14. Durham, J.R. Filtration Characteristics of Fabric Filter
Media. NAPCA, PHS, USDHEW, Interim Report, 1969.
15. Dennis, R. Reported Experimental Measurements, GCA Tests,
1975.
16. Cooper, D.W., V. Hampl, and R. Stern. Development of Data
Base and Fabric Filtration Models for Design of Particulate
Control Systems. Environmental Protection Agency, Research
Triangle Park, N.C. Contract No. 68-02-1438, Interim Re-
port for Task 5.3, January 1976.
17. Borgwardt, R.H., R.E. Harrington, and P.W. Spaite. J. Air
Pollution Control Assoc. 18.(6) :387-390, 1968.
18. Davies, C.N. Proc. Inst. Mech. Eng. 1B(5):185-213, 1952.
19. Leith, D., and M.W. First. Particle Collection by Pulse-Jet
Fabric Filter. Presented at 68th Annual APCA Meeting.
Boston. 1975.
20. Fuchs, N.A. Mechanics of Aerosols, New York, Pergamon, 1964,
21. Spurny, K.R., J.R. Lodge, Jr., E.R. Frank, and D.C.
Sheesley. Environ. Sci. Technol. 3_:453-469, 1969.
22. Smith, T.N., and C.R. Phillips. Environ. Sci. Technol.
9:564-568, 1975.
185
-------
PAPER 10
OPERATING EXPERIENCE AND PERFORMANCE AT
THE SUNBURY BAGHOUSE
Harry Spagnola
Pennsylvania Power & Light Company
James H. Turner
Environmental Protection Agency
Industrial Environmenta! Research Laboratory
PLANT - GENERAL DESCRIPTION
The Sunbury Steam Electric Station of Pennsylvania Power and
Light Company is located on the west bank of the Susquehanna River
in Shamokin Dam, Pennsylvania. It generates 402 MW of electricity
from six pulverized-coal-fired steam generators and four turbine-
generator sets. Turbines Nos. 1 and 2 are each rated at 87.5 MW.
Steam is supplied to these turbines through a header system from
Foster Wheeler anthracite-fired boilers. (A Foster Wheeler
bituminous-fired boiler and a 110 MW turbine-generator set are
also connected into this header system.) Each of the four boilers
is rated at 415,000 Ib/hour steam flow at 955°F and 1350 psig and
was placed in commercial service in 1949. The boilers burn a mix-
ture of anthracite (silt and No. 5 buckwheat), petroleum coke,
and bituminous coal. (Typical analysis of the plant coal is
given in Table 1.) The mixture varies considerably according to
the availability and the quality of the fuel. The normal mix is
75% anthracite (consisting of 90% silt and 10% No. 5 buckwheat),
20% coke, and 5% bituminous coal. The normal fuel consumption is
25 tons per hour per boiler. No. 2 fuel oil is used for boiler
fire-ups and to stabilize ignition.
186
-------
TABLE 1. TYPICAL PLANT COAL ANALYSIS
Anthracite Bituminous Petroleum Coke
Heating Value (wet basis)
(Btu/lb)
% Ash
% Moisture
% Volatile Matter
% Sulfur
8,680
28.0
16.4
8.4
0.8
12,159
15.3
5.3
27.1
2.6
14,054
1.7
6.4
13.3
5.3
PARTICULATE REMOVAL EQUIPMENT
Original particulate removal equipment consisted of a mechan-
ical-electrostatic precipitator set. Particulate removal effi-
ciency with this equipment was 90%. In 1973 baghouses were in-
stalled on each of the four anthracite-fired boilers. The elec-
trostatic precipitators were gutted due to a weight consideration.
The weight reduction allowed utilizing some of the existing pre-
cipitator steel to support the baghouse. The mechanical collec-
tors were modified by replacing the original vanes of the multi-
cone collectors with new, less efficient vanes. The effect of
this was a reduction of mechanical collector efficiency from 78%
to 70%, while the pressure drop decreased from 4.5 in. VWC to 1.5
in. VWC. The mechanical collectors were left in to maintain a
lower dust loading on the induced draft fans. It is not certain
to what degree, if any, they improve baghouse performance.
The baghouses are installed in series with the mechanical
collectors. The flue gas, after leaving the boiler economizer,
passes through regenerative air heaters, mechanical collectors,
the induced draft fans, and the baghouse. The then clean flue
gases discharge to atmosphere through a stack, 300 feet above the
ground. (Figure 1)
The baghouses were constructed by Western Precipitator Com-
pany. They were initially put into service on the following
dates:
No- 1A Baghouse - February 10, 1973
No. 2B Baghouse - March 5, 1973
No. 2A Baghouse - April 14, 1973
No. IB Baghouse - April 26, 1973
187
-------
BLANK OFF PLATES /
"V \
ECONOMIZER-
AIR HEATER—M~
AIR
FROM
FORCED DRAFT
FAN
BAG
HOUSE
MECHANICAL
COLLECTOR
INDUCED
DRAFT FAN
Figure 1. Location of dust removal equipment
BAGHOUSE - GENERAL DESCRIPTION
Each baghouse is approximately 80 feet long, 40 feet wide,
and 62 feet high, including the dust hoppers. Each baghouse con-
sists of 14 compartments arranged in two rows of seven compart-
ments each. Each filter bag compartment is 19 ft, 9 in. wide x
10 ft, 1-3/4 in. long x 40 ft high. The baghouse enclosures, in-
cluding the interior partitions, are constructed of 14-gauge mild
steel and are of all welded construction. The 14-gauge partitions
and welded construction were decided upon to insure gas tight con-
struction to permit the safe entering of isolated compartments
for routine inspections and minor maintenance while the baghouse
is in service.
The entire baghouse including the hoppers, collapsing flues,
and inlet/outlet flues up to the point of tie-in with the exist-
ing breeching are completely heat insulated and aluminum lagged.
The bag filter dust hoppers, located below the thimble floor,
are contained in an enclosure constructed of 0.032 in. thick alum-
188
-------
inum siding over steel girts. The floor in the hopper enclosure
is Checker plate in order to contain any ash which may spill during
maintenance periods or should malfunctions occur in the ash re-
moval system.
Access for maintaining the filter bags is accomplished in
the following manner. A 2-ft wide x 5-ft high insulated compart-
ment access door is provided at the thimble floor elevation for
each compartment. A checker plate walkway constructed of 3/16-
in. steel plate is provided through the center of the filter bag
rows (three rows of 15 filter bags each are located on each side
of the walkway) to facilitate changing filter bags as well as to
permit inspection to detect broken bags. A 2-ft wide x 5-ft
high insulated compartment access door and grating walkway
are provided in each compartment for access to the filter bags at
the bag tensioning framework elevation.
An inlet flue and a collapse air flue run the length of each
baghouse between the rows of hoppers. Side ducts from each flue
to the individual compartments are provided to the upper portion
of the compartment ash hoppers. The gas inlet damper and the
collapse air damper, both timer-operated, are located in the gas
inlet flue and the collapse flue, respectively, to permit cycling
the compartments through the in-service and cleaning cycles.
There is one main collapse fan for each baghouse and two spare
collapse fans, each serving two baghouses.
The outlet flue surmounts the baghouse and ties into the ex-
isting stack breeching. Each compartment has a manually operated
outlet flue damper which permits isolating a compartment during
maintenance periods. (See Figure 2 for the baghouse's general
arrangement.)
Vent stacks were installed on the top of each baghouse com-
partment in 1974 and 1975. Originally isolating a baghouse com-
partment was difficult. Since the gas inlet damper, a leaf type
damper, does not provide a positive shut off, too much heat and
gas leaked into the compartment. With a portable 7,000 cfm fan
mounted on the vent stack, enough ventilation is provided to per-
mit entry into the compartment for maintenance with the remainder
of the baghouse in service. The vents are 2 ft x 2 ft x approxi-
mately 4 ft high, constructed of 10-gauge mild steel and are
covered with 2-in. thick insulation and 0.032-in. thick aluminum
siding. A spring clamped cover is mounted on top of each vent
stack (Figure 3) .
Filter Bags
The design filter ratio was 1.919:1 cfm/ft2 of bag and the
active filter ratio (based on one compartment being out of ser-
vice at all times) was 2.067:1. Each baghouse compartment con-
tains 90 bags for a total of 1,260 bags per baghouse, giving a
total filter area of 115,668 ft2.
189
-------
OUTLET FLUE
OUTLET DAMPERS
ACCESS DOORS
SWEEP
VALVE
GAS INLET
DAMPERS
COLLAPSE
AIR DAMPER
COLLAPSE
AIR FANS
Figure 2. Baghouse general arrangement
Each filter bag is 11% in. in diameter by 30 ft, 4 in.
in length. The filter bags, Menardi Southern Company's style No.
601T, are made of glass fiber with Tuflex finish (Teflon base;
specifications are listed in Table 2). The Teflon coating acts
as a lubricant between the glass fiber filaments. It prevents
the glass fiber filaments from abrading one another, and thus
lengthens bag life. Each filter bag contains seven equally spaced
anti-deflation rings. With the rings preventing the bag from
collapsing completely during the cleaning cycle, the dust cake is
released freely and drops to the hopper below. A metal band is
sewn into the top and bottom of the bags for installing the bags.
The bags are suspended from a framework in the top of each com-
partment housing and approximately 50 Ib. tension is applied.
190
-------
1 - GAS INLET DAMPER - OPEN
2 - GAS INLET DAMPER - CLOSED
3 - BAG COLLAPSING DAMPER - OPEN
4 - BAG COLLAPSING DAMPER - CLOSED
5 - OUTLET DAMPER - OPEN
FROM
I.D. FAN
Figure 3. Gas flow through baghouse compartments during normal operation and cleaning
191
-------
TABLE 2. MENARDI FILTER BAG, STYLE NO. 601T
Fabric Glass Fiber
Average Permeability 38.5
Mullen Burst Strength (av psi) 595
Weight (oz/yd2) 10.0
Thread Count 54 x 30
Weave 3x1 Twill
Finish Teflon 10% by weight
Warp Yarn ISO's 1/2
Fill Yarn Bulked 1/4
Cleaning Cycle
The bags are cleaned by reversing the gas flow through a
compartment. This partially collapses the bags; some of the dust
cake is then released and falls to the hopper below. A collapse
fan (12,500 cfm @ 300°F and 10 in. VWC Buffalo Forge Company size
60, 1300 rpm) provides the force for the reverse gas flow. The
collapse air flow is drawn through the collapse duct and into the
collapse fan, and then is discharged into the baghouse inlet flue,
where any entrained fly ash is filtered by the bags (Figure 2).
Each compartment is cleaned in the following manner:
1. The gas inlet damper to the compartment closes, shutting
off the flow of "dirty" flue gas to this compartment.
2. The collapse damper opens, allowing a reverse flow of
"clean" flue gas from the outlet flue to be pulled through
the bags, which partially collapse, and thus are cleaned.
3. The collapse damper closes.
4. The gas inlet damper opens, returning the compartment
to the filtering mode. (Figure 3 illustrates a compart-
ment in the filtering mode and a compartment in the
cleaning mode.)
192
-------
This sequence is repeated on each of the 14 compartments.
Twice during the 14-compartment cycle the collapse duct is swept
clean. This is accomplished by opening a sweep valve located at
the end of the collapse duct (Figure 2). This allows ambient air
to be drawn through the collapse duct and fan, and thus purge any
fly ash accumulation.
The time sequence of the cleaning and sweep cycles is detailed
in Table 3. There are null periods between the operation of the
gas inlet and collapse air damper. This is done to prevent re-
entrainment by allowing for some settling time for suspended fly
ash; and also, to prevent a sudden reversal of air flow through
the bags, which would put undue stress on the bags. The air-
operated dampers are provided with a slow opening and closing de-
vice which also prevents any additional stress on the bags.
TABLE 3. CLEANING/SWEEP CYCLE TIME SEQUENCE
Elapsed time, sec
I. Clean compartments 1 through 7
one at a time
Gas Inlet Damper starts to close 0
Gas Inlet Damper closed 2.0
Collapse Air Damper starts to open 16.0
Collapse Air Damper open 17.0
Collapse Air Damper starts to close 68.0
Collapse Air Damper closed 69.5
Gas Inlet Damper starts to open 81.0
Gas Inlet Damper open 83.0
Next Gas Inlet Damper starts to close 122.0
II. Sweep Collapse Air Duct
Last Compartment Gas Inlet Damper open 0
Sweep Valve open 40
Sweep Valve closed 120
Next Compartment Gas Inlet Damper open 165
III. Clean compartments 8 through 14
one at a time.
Repeat cycle shown under I
IV. Sweep Collapse Air Duct
Repeat cycle shown under II
193
-------
Instrumentation
Two instruments, a Bailey bolometer and a differential
pressure recorder, are used to monitor baghouse performance
continuously.
The bolometer consists of a light bulb and a receiver located
in the breeching to the stack on each boiler. The receiver output
varies as the light intensity striking it varies, and thus is an
indication of smoke density. A 24-hour recorder located in the
plant monitors this signal. A reading of zero of scale indicates
no smoke and a reading of 100% of scale indicates 100% smoke.
The bolometers are not calibrated and just give a relative indi-
cation of smoke density.
The differential pressure across each baghouse is recorded
on a 24-hour recorder located in the plant control room. The
taps are located in the baghouse inlet and outlet flues. An
audible alarm is sounded if the differential pressure rises 1 in.
VWC above its normal value, giving the operator time to take cor-
rective action.
The differential pressure across each compartment is not
monitored. It was decided that a continuous monitoring system
would be too costly and of questionable value. There are taps
across each compartment but no instruments were installed to in-
dicate differential pressure.
Controls and Alarms
The four baghouse control panels are located in the plant
mill room. Each control panel houses two cleaning cycle timers
(one of which is a spare), a stepping switch, a cycle timer selector
switch, a collapse fan selector switch, a control power "off-on"
switch, and the compartment's status indicating lights.
The cleaning control timers and the stepping switch control
the cleaning cycle, operating the air-operated baghouse gas inlet
and collapse air dampers by activating the appropriate solenoid
valves. The cleaning cycle control timers were originally electro-
mechanical devices. Solid-state electronic timers were added in
1975 and 1976 to facilitate adjustment in the cycle time and in-
crease the reliability.
The compartment status lights indicate the positions of the
baghouse gas inlet and collapse air dampers. These indicate
whether a compartment is in the filtering or the cleaning mode.
Indication is accomplished through the use of limit switches on
the dampers.
194
-------
Adjacent to the baghouse control panels are alarm panels
with alarms to indicate a collapse fan failure or a cleaning
cycle timer failure.
Baghouse Vacuum Cleaning System
Individual vacuum cleaning systems are provided for each
baghouse. Vacuum connections have been provided in each bag
filter hopper enclosure as well as outside the compartment access
doors at the thimble floor elevation.
Baghouse Ash Removal System
The two fly ash removal systems, one per unit, are United
Conveyor Corp. vacuum type which pneumatically convey fly ash
from the ash hoppers. Air flow required for conveying the ash
is produced by an 8-ft Hydroveyor exhauster. The ash/air mixture
is transported by the conveyor to the Hydroveyor exhauster, where
the mixture is combined with water to form an ash slurry. This
slurry is discharged into an air separator tank where the convey-
ing air is vented to atmosphere while the ash slurry discharges
by gravity into the Units Nos. 1-2 slurry tank.
High pressure (300 psi) water at a rate of 750 gpm is pro-
vided at the Hydroveyor for each system by one of three new raw
water booster pumps. Plant ash disposal capabilities were in-
creased by the duplication of the existing fly ash disposal sys-
tem and also the installation of two new 2,000 gpm raw water
pumps. The new fly ash disposal system is composed of one ash
slurry tank, an ash slurry pump house, six slurry pumps (in series)
1600 gpm capacity and over 2 miles of 10-in. buried steel
pipe through which the slurry is pumped to the fly ash settling
basin.
OPERATION
The flue gas enters the baghouse at 350°F and leaves at 310°F.
The flue gas temperature normally stays within this range, except
during boiler startups and shutdowns, when the temperatures are
in a lower range, anywhere from ambient air temperatures to normal
operating temperature. Several times since 1973 the regenerative
air heater drives have failed. This has caused the baghouse inlet
temperature to rise as high as 510°F. The high temperatures were
sustained for about 1 to 2 hours. There were no discernible ef-
fects on the bags.
The design pressure drop across the baghouse and baghouse
duct work was not to exceed 6 in. VWC, which was dictated by the
capacity of the existing induced draft fans. Average draft loss
across the baghouses is as follows:
195
-------
Baghouse No. Av Draft Loss (in. VWC) Bags Installed
IB 3.5 4/26/73
2A 3.5 11/20/74
2B 3.0 2/23/75
1A 2.5 3/20/75
The baghouses are on the discharge side of the induced draft
fan and thus are under a positive pressure.
The gas flow through each baghouse is 222,000 acfm at a tem-
perature of 325°F. Baghouse inlet grain loading is approximately
2 gr/f t3 .
There is no visible stack discharge from any of the bag-
houses, except for vapor condensation during the coldest days of
the year (approximately 10°F or lower). When a large bag failure
occurs (e.g., a 2 to 3-ft tear in a single bag) there will some-
times be a very faint stack discharge (approximately 5% opacity
or less) depending on the location of the bag failure.
Boiler Start-Dps and Shutdowns
To prevent blinding of the bags with the oil used during
boiler fire-ups, before a new set of bags is placed in service
they are precoated with fly ash remaining in the boiler gas
passes. The boiler is brought on the line and the baghouse
cleaning cycle is not activated until one hour after coal is fired,
This allows an additional coating to form on the bags.
Each boiler is shut down and fired-up several times a year.
Other than the precoating of new bags, no special precautions are
taken. During normal fire-ups, shutdowns, and outages, the clean-
ing cycle is left on. The continued cleaning during outages does
not remove the initial fly ash coating but will prevent the forma-
tion of a thick cake, which can accumulate from moisture carry-
over while water lancing boiler slag deposits.
The number of fire-ups for each boiler is given below.
Boiler No. 1973 1974 1975
1A 632
IB 352
2A 335
2B 733
196
-------
Operating Problems
The major operational problem has been occasional baghouse
high differential pressure. An alarm sounds to warn the operator
of the high differential. The usual causes are either collapse
fan failure, cleaning cycle timer failure, or gas inlet or col-
lapse air damper failure. Once the problem is corrected, the
differential pressure returns to normal in a short time, approxi-
mately 30 minutes. Normally the problem is corrected without af-
fecting boiler operation. However, if the differential increases
to approximately 5 or 6 in. VWC, the boiler steam flow must be
reduced due to the loading on the induced draft fans.
Baghouse problems infrequently (about 5 or 10 times per year
for all baghouses) have required reductions in boilor load of
about 20,000 to 40,000 Ib/hr steam flow, but have never forced a
boiler off the line.
At present there are no alarms for collapse fan failure.
Flow switches were installed in the collapse fan duct to indicate
fan failure, but the switches proved to be inoperative. It is
planned to install other alarms. At present collapse fan failure
is detected from increased differential pressure signals.
When a main collapse fan fails the spare collapse fan is
put into service by opening blast gate (butterfly-type) dampers.
The spare fan is normally filled with fly ash due to leakage past
the blast gate dampers, and normally the fan must be cleaned out
before it is put into service. This can take as much as 2 to 3
hours. Originally the spare fan was isolated by sliding gate
dampers. These dampers provided a tight seal; however, they were
difficult to open and close- It took four men with a chain hoist
approximately 4 hours. There is a study underway to determine a
method to prevent fly ash accumulation in the spare fan and still
allow easy damper operation. As a temporary measure the spare
fans are pressurized with compressed air to prevent in-leakage.
If a cleaning cycle timer fails, an alarm is sounded and the
operator switches to the spare timer. Originally the timers failed
frequently. This has been alleviated by adding more reliable,
electronic timers.
Damper failures can sometimes be detected by observation of
the differential pressure chart. As the dampers open and close
the differential pressure swings. If a damper fails, the absence
of this pressure swing leaves a "gap" on the differential pres-
sure chart (Figure 4). If a high differential pressure alarms,
the dampers are routinely checked for proper operation. The
operator must go up to the baghouse and visually observe damper
operation through the complete cycle (a total of 32 minutes).
197
-------
Figure 4. Differential pressure chart - indicating a damper malfunction on 2A baghouse
(notice arrows), A normal differential pressure chart has no gaps.
When a baghouse hopper becomes plugged and fills with fly ash,
a high pressure spike, in sequence with the 32-min cleaning cycle,
is indicated on the differential pressure recorder. No opera-
tional problems are experienced. The operator must sledge-hammer
and/or air-lance the hopper to free the blockage.
Since the number of bag failures has been small, they have
not caused operational problems. Most failures are single-bag;
no more than two bags have failed at any one time. Isolating a
compartment increases the differential pressure by approximately
0.5 in. VWC. Two compartments have been taken out of service to-
gether with an increase of approximately 1.0 in. VWC in differ-
ential pressure and no reduction in boiler load was necessary.
Any number of compartments may be taken out of service without
affecting boiler load as long as the differential pressure remains
below 6.0 in. VWC.
There have been two major baghouse pressure excursions. On
Baghouse No. 2A in July 1975 after a boiler outage the draft loss
increased from 3.0 in. to 6.0 in. VWC over a 3-week period. This
was apparently due to moisture carry over to the bags from heavy
water-lancing of boiler slag deposits during the outage. Several
methods were tried to reduce the draft loss: two collapse fans
were placed in service on the baghouse, the collapse dampers were
opened and closed several times during each 2-minute compartment
cleaning time (this was an attempt to dislodge any thick dust cake
on the bags) , all hoppers were inspected and any blockage was
cleared, and the thimble floor was vibrated by pounding it with
sledge hammers to dislodge any buildup on the thimbles. (This
198
-------
buildup, a fly ash deposit reducing the 12-in. diameter thimble
size to 6 in. to 9 in. diameter, has been noted on all baghouses
but never has had any noticeable effect on baghouse performance.)
All of these methods proved unsuccessful. A compartment was iso-
lated and the bags were inspected. The dust cake on the bags was
very thick, from 1/2 in. to 3/4 in. It was decided to manually
shake the bags. Each bag was grasped between the palm of each
hand and vigorously shaken from side to side. The release of the
dust cake could be felt as the dust fell to the hopper below.
Alternately, all of the compartments were taken out of service
and the bags were manually shaken. When all compartments were re-
turned to service the pressure drop had decreased to 3.5 in. VWC
and has remained stable since.
In the fall of 1975 the pressure drop on Boiler No. IB in-
creased from 3.5 to 5.0 in. VWC over a 3-month period. The bags
were inspected during the boiler outage. The cake was at most
about 1/4 in., so the bags were not manually shaken. All other
methods used on No. 2A were tried with no success. With the new
electronic timers it is possible to readily change the cleaning
cycle time. On January 23, 1976, the time was reduced from the
30-minute cycle to a 15-minute cycle, doubling the number of col-
lapses. In the last part of February the draft loss began to de-
crease until it reached 3.5 in. VWC at the end of February. The
cleaning cycle was returned to the normal 30-minute cycle on March
16, 1976. The draft loss has remained stable at 3.5 in. VWC since,
Because of the time frame it is uncertain whether the 15-minute
cleaning cycle caused the draft loss reduction. Normally, when a
change is made to reduce draft loss the effect is noticed in
several hours. Additional background information on the Sunbury
installation may be found in References 1, 2, and 3.
MAINTENANCE
Most maintenance hours have been spent on bag replacement,
collapse fan repairs, and air-operated damper repairs.
Bag Replacement
The procedure for bag replacement is detailed in Figure 5.
During each annual boiler outage all baghouse compartments
are inspected. If an accumulation of dust is found on the com-
partment floor, each bag in the compartment is inspected for pos-
sible failure and all failed bags are replaced.
To replace a bag with the boiler in service, the associated
compartments must be isolated by closing the gas inlet and outlet
dampers. The lower and upper doors of the compartment are opened
to allow ambient air to circulate. The cover is removed from the
vent stack, and a portable 7,000-cfm fan is set on the vent stack
199
-------
HOISTING ROPE
METAL BAND SEWN
INTO FILTER BAG
CAP
INSTALL BAG END
MARKED "BOTTOM"
AT HOPPER FLOOR
LEVEL
DETAIL
INSTRUCTIONS FOR HANDLING AND INSTALLATION OF
GLASS CLOTH FILTER BAGS
HANDLING
GLASS CLOTH IS SOFT, FLEXIBLE, AND DURABLE. HOWEVER IT IS SUBJECT TO TYPES OF DAMAGE NOT ALWAYS
ASSOCIATED WITH OTHER FABRICS. CARE IN ITS HANDLING WILL BE REPAID IN GREATER EFFICIENCY AND
LONGER LIFE.
UNPACK BAGS CAREFULLY, ONE BY ONE, AS THEY ARE INSTALLED. AVOID CONTACT OF THE FABRIC WITH
PROJECTIONS OF METAL OR ROUGH WOOD WHICH MAY CAUSE SNAGS OR TEARS.
NEVER LAY BAGS ON THE THIMBLE-PLATE OF THE BAGHOUSE WHERE THEY MAY BE WALKED UPON.
THE FABRIC SHOULD NEVER BE PULLED BY GRASPING BETWEEN THUMB AND FINGERS, WHICH MAY CAUSE
SLIPPAGE OF THE YARNS.
FOLDED BAGS SHOULD NOT BE SUBJECTED TO HEAVY PRESSURE.
INSTALLATION
STEP ONE.-AFTER UNPACKING, PLACE FILTER BAG (3) IN BOX ON THIMBLE PLATE NEAR POSITION OF ASSEMBLY.
RAISE ONE END OF FILTER BAG AND INSERT CAP (2) AS SHOWN IN DETAIL (1). ROTATE CAP 90° AND PULL
UPWARD UNTIL METAL BAND SEWN INTO BAG CUFF FITS SNUGLY AGAINST ROLLED RIM OF CAP, AS SHOWN
IN DETAIL (2).
STEP TWO.-HOOK ONTO CAP (2) WITH HOISTING ROPE AND RAISE BULK OF BAG (3) OFF THIMBLE PLATE.
PLACE LOWER END OF BAG OVER THIMBLE, AS SHOWN IN DETAIL (2). PUSH BAG END DOWNWARD UNTIL
THIMBLE RIM IS COMPLETELY COVERED. THEN GENTLY AND EVENLY, MOVE BAG END UPWARD UNTIL METAL
BAND SEWN INTO BAG CUFF FITS SNUGLY AGAINST ROLLED RIM OF THIMBLE.
STEP THREE.-HOIST BAG TO FULL HEIGHT AND REPLACE ROPE WITH HANGER (1) AS SHOWN IN DETAIL(3).
TIGHTEN HANGER TO DEVELOP 50 LBS. TENSION ON BAG - BY COMPRESSING HANGER SPRING 1-1/4 INCHES.
Figure 5. Installing filter bag
200
-------
and started to provide forced ventilation. It normally takes 3
to 4 hours to ventilate the compartment sufficiently for men to
enter. It is normally still necessary to wear a mask to prevent
inhalation of sulfur dioxide. The compartment can then be entered
and the failed bag replaced. The entire procedure (isolation and
ventilating the compartment, finding the leak, replacing the bag,
and returning the compartment to service) takes approximately 6
to 8 hours.
Detecting Bag Failures
Bag failures are detected by daily observations of the bolo-
meter charts. When a bag has failed the bolometer senses the in-
creased particulate emissions; generally, there is no visible
stack discharge. The bolometer chart will indicate a periodic
spike, in sequence with the cleaning cycle. There will be a de-
crease in bolometer reading when the compartment with the
failed bag is removed from service, a spike when the compartment
is returned to service, and a settling out to a higher than normal
reading directly afterwards (Figure 6). Some spikes are not
easily discernible and a careful study of each chart is necessary.
Also some non-periodic spikes occur, but these do not indicate an
abnormal condition.
Figure 6. Bolometer chart - indicating a bag failure in compartment
No. 14 of No. 2B baghouse
201
-------
When it is determined what compartment has the failed bag,
the compartment is taken out of service. The bolometer readings
then return to normal. The compartment is entered, and each of
the ninety bags is inspected. An accumulation of fly ash on the
compartment floor is a tell-tale sign that there is an actual bag
failure in the compartment. However, bag failures have been found
in compartments with no fly ash accumulations. The bags are in-
spected by holding a flashlight at the bottom of the bag and
shining it up the side of the bag. Any tears are illuminated in
this manner. Also, a slight tap on the bag will, if there is a
failure, cause a stream of fly ash to flow from the bag which is
illuminated by the light, providing an additional check for
failures.
The failed bag is replaced, the compartment floor is cleaned,
and the compartment is returned to service.
Bag Failure Rate
As of May 1, 1976, a total of 37 bags have failed. A list
of the failures is detailed in Table 4.
TABLE 4. BAG FAILURES3
Baghouse
1973
1974
1975
To May
1976
Total
1A
IB
2A
2B
5
0
3
4
1
6
0
7
2
2
0
2
2
2
0
1
Total
Complete Bag Replacements
10
10
3
14
37
1A 3-20-75
IB Bags never replaced
2A 11-20-74 (except for compartment No. 8)
2B 2-23-75
a. Failures include neither those from faulty installation
or damage during maintenance (about 23 bags) nor test bags,
202
-------
As a precautionary measure, bags in three baghouses were re-
placed after approximately 2 years of service (dates are listed
in Table 4). The 4th baghouse, No. IB, still has its original
set of bags. The bags will remain in service until an increased
bag failure rate dictates their replacement.
Maintenance Problems
The collapse fans have been a continual problem. The fans
were not designed to handle dirty gas. There is baffling in the
hoppers to prevent the collapse fan from picking up dust. However,
this has not been effective. The fan casings and blades are worn
thin from the abrasiveness of the fly ash. This has caused exces-
sive fan vibration and gas leakage from the fan. Frequent mainten-
ance is required on the fans.
The air-operated dampers have caused continual problems.
They are located in an environment of high temperatures (150°F
to 200°F) and occasional dusting. This results in increased
maintenance and occasional damper failure.
ECONOMICS
The initial installation cost of all four baghouse installa-
tions, including the fly ash removal system, was $5,500,100. In-
stallation cost and yearly operating and maintenance costs are
detailed in Table 5. Increased mechanical and electrical main-
tenance cost in 1975 was the result of increased collapse fan
maintenance and the high pressure drop problem encountered on
Baghouse No. 2A.
TABLE 5. INSTALLATION, OPERATION AND MAINTENANCE COSTS
Bag Filter Installation Cost
Expenditure Description Total Cost, $
Western Precipitator Contract (4 Baghouses) 3,201,700.
Structures & Improvements 372,100,
Land & Land Rights 1,500.
Boiler Plant Equipment 1,190,000.
Ash Removal Equipment
Precipitator Modifications
Accessory Electrical Equipment 72,000.
Overhead 662,800.
Total Construction Cost 5,500,100,
(continued)
203
-------
TABLE 5. (continued)
Estimated Operating & Maintenance Cost,
@ $12/manhour Cost , $
Cost Description 1973 1974 1975
Collapse Fan Power Consumption 18,600. 18,600. 18,600.
Air Compressor Power Consumption 3,000. 3,000. 3,000.
Instrument Department Labor 950. 950. 1,800.
Mechanical Maintenance Labor 2,130. 5,840. 15,678.
Electrical Maintenance Labor 7,410. 3,800. 6,000.
Construction Department Labor 3,950. 2,350. 2,000.
Complete Bag Replacement
Material ($48,000/baghouse) 48,000. 96,000.
Labor ($ll,000/baghouse) 11,000. 22,000.
TOTAL COST 36,040. 93,540. 165,078.
PERFORMANCE
In order to assess the performance of a fabric filter system
as applied to a utility boiler, EPA asked for permission to per-
form efficiency tests after the Sunbury baghouses had been in
service for a few months. Permission was granted and one set of
tests was made in August 1973. "* The results of these tests were
impressive and further testing was scheduled to be done by GCA
Corporation. The results of the GCA testing, performed in January,
February, and March 1975, are discussed in this paper.2
Of greatest interest to EPA was the performance of the bag-
house in the fine particulate range, i.e., below about 3 ym.
Along with mass efficiency testing there were also measurements
made with impactors and diffusion apparatus. The problems as-
sociated with the latter two types of devices led to some rather
large error bands, but with over 30 tests made there is reasonable
confidence in the results.
GCA performed their tests after the baghouses had been in
service for approximately 2 years. Since bag changeover was to
coincide with a scheduled outage, this situation allowed testing
of bags at the end of their planned service life and also (for the
new set of bags) at the beginning of their service life.
Mass Results
For 19 tests made at several combinations of fuel mix and
boiler load, the average outlet concentration was 0.0017 gr/dscf
or in terms of emission rate: 0.0041 lb/106 Btu. These
204
-------
tests were for the bags which had been in service for 2 years.
Based on an average inlet loading of 2.6634 gr/dscf the aver-
age efficiency was 99.94%. For the new bags the average outlet
concentration was 0.0024 gr/dscf; emission rate was 0.00543
lb/10 Btu.; average inlet loading was 2.2853 gr/dscf and
average efficiency was 99.89%. The slightly higher outlet loading
was consistent with slightly lower pressure drop through the new
bags, and the new bags would be expected to have performance
similar to the old bags after sufficient stabilization time. The
results do indicate that baghouse efficiency does not deteriorate
with time, and tends to get better through the service life of
the bags.
For all 31 tests the range of outlet loadings was from 0.0013
to 0.0033 gr/dscf (0.0028 to 0.0101 lb/106 Btu).
Impactor Results
Apparent collection efficiency by particle size is shown in
Figure 7. For the size range covered, 10 to 0.1 urn, it is apparent
that the baghouse maintains nearly uniform, high reduction of
particles entering the system. All sizes in the impactor range
are collected with better than 99% apparent efficiency. There is
some question as to whether or not particles can change their
size as they go through the collection system, but the result at
the baghouse outlet is a clear stack.
Condensation Nuclei Counter Results
A condensation nuclei counter (CNC) was used to count num-
bers of particles in the range from about 0.003 to 0.5 ym. Al-
though the data are sketchy they indicate an apparent collection
efficiency of between 93.0 and 99.7% for particles in this size
range. It should be emphasized that this is a number efficiency
and not a mass efficiency.
GENERAL COMMENTS
It is obvious that the Sunbury baghouses (as do the Nucla
baghouses)5 do a fine job at collecting particulate matter. The
GCA testing also revealed that collection efficiency was not sig-
nificantly affected by changes in fuel composition, boiler load,
filtration velocity or cleaning regime within what might be nor-
mal limits of change. The baghouse seems to operate like a
smoothing device; regardless of inlet variations the outlet showed
very little change. For the changes that did occur, regression
equations were formulated which showed particle penetration to be
primarily a function of inlet concentration, percent moisture,
filtration velocity and pressure drop.
205
-------
aa.aa
5?
> 99.9
Z
LU
O
LL
S 99
Z
O
1—
y 95
_i
_i
g 90
APPARENT
o§
1
- .0 •-•-••*;
^»x-^ ^-X*^
_ — . . o*^"^
— —
—
FILTRATION VELOCITY = 2.0 ft/min.
(1.0 cm/sec)
OUTLET LOADING = 0.0017 gr/dscf
(0.0039 g/m3)
I
0.1
1 J
0
H
<
5 cc
LU
10 Lu
o.
50
1 1.0 10.0
PARTICLE DIAMETER,
Figure 7. Baghouse performance at Sunbury Steam Electric Station
206
-------
SUMMARY OF BAGHOUSE PERFORMANCE
As far as filtration performance is concerned, the Sunbury
baghouses do an excellent job. Mass efficiency is 99.9%, grade
efficiency is 99% or better in the impactor range, and the number
efficiency for particles in the CNC range appears to be between
93 and 99.7%. Changes in process and collector operating variables
have little effect on outlet concentration.
REFERENCES
1. Wagner, N.H., and D.C. Housenick. Sunbury S.E. Station -
Units No. 1 & 2 Design and Operation of a Baghouse Dust Col-
lector for a Pulverized Coal Fired Utility Boiler. Pennsyl-
vania Power & Light Co., Allentown, Pa. May 1973.
2. Cass, R.W., and R.M. Bradway. Fractional Efficiency of a
Utility Boiler Baghouse - Sunbury S.E. Station. GCA Cor-
poration. EPA-600/2-76-077a, U.S. Environmental
Protection Agency, Washington, D.C. 1976.
3. Operating Instructions for Therm-o-flex Filters for Pennsyl-
vania Power & Light Co., Sunbury, Pa., Joy Manufacturing
Company, Western Precipitation Division, February 1973.
4. Harris, D.B., and J.H. Turner. Particulate and S02/S03
Measurements Around an Anthracite Steam Generator Baghouse.
Particulate and Chemical Processes Branch, Control Systems
Laboratory, Office of Research & Development, Environmental
Protection Agency, National Environmental Research Center,
Research Triangle Park, N.C. November 8, 1973.
5. Bradway, R.M., and R.W. Cass. Fractional Efficiency of a
Utility Boiler Baghouse - Nucla Generating Plant. GCA Cor-
poration. EPA-600/2-75-013a. U.S. Environmental Protection
Agency, Washington, D.C.
207
-------
PAPER 11
SCRUBBER EXPERIENCE AT MOHAVE
John M. Johnson, Dale G. Jones,
Alexander Weir, Jr., and W. Carl Martin
Southern California Edison Co.
Seymour Calvert
Air Pollution Technology, Inc.
ABSTRACT
The performance of wet scrubbers in the removal of particu-
late was evaluated during a development test program conducted at
the coal-fired Mohave Generating Station during 1974 and 1975.
Data were obtained using three different scrubber configurations
that were each sized for 170 MW. The Vertical TCA, the Vertical
PPA and the Horizontal Cross-flow all received flue gas from a
location downstream of cold-side electrostatic precipitators.
A total of 154 inlet/outlet particulate samples were examined and
compared to determine the effect of inlet particulate grain load-
ing and variable scrubber operating conditions upon particulate
removal performance. These scrubber variables included flue gas
flowrate, circulating slurry flowrate, and internal packing
configurations.
INTRODUCTION
Between July, 1971 when pilot plant testing was initiated at
the Mohave Generating Station and July, 1975 when the 170 MW Test
Modules Program was completed, the participants of the Navajo and
Mohave Power Projects funded a scrubber development and evalua-
tion project which has significantly advanced the understanding
of scrubber technology. The results of this evaluation have been
published, in part, at several EPA Flue Gas Desulfurization Sym-
posium conferences. 1~I* The objectives of the 170 MW Test Modules
Program were to evaluate the SC>2 and particulate removal perfor-
mance of more than one full-scale scrubber, as well as to deter-
mine the practical feasibility of scrubbers for commercial service,
including cost, operating flexibility, maintenance requirements,
materials of construction, chemical control problems, and waste
disposal factors.
208
-------
The purpose of this paper is to describe the particulate
removal characteristics of the 170 MW Mohave Test Modules. During
the test program, a total of 154 inlet/outlet particulate samples
were obtained during tests of three different 170 MW scrubber con-
figurations. The scrubbers were located downstream from the cold-
side electrostatic precipitators, and were designed to treat
450,000 scfm of flue gas (equivalent to 170 MW at the Mohave Gen-
erating Station) .
TEST MODULE DESCRIPTION
The 170 MW Vertical Module, located on Unit 1, was tested in
the Turbulent Contacting Absorber (TCA) configuration and the Poly-
grid Packed Absorber (PPA) configuration. The TCA configuration,
shown in Figure 1, consisted of four stages of thermoplastic rub-
ber balls supported on stainless steel grids at four-foot inter-
vals. The balls were contained in compartments, with 15 compart-
ments at each of the four levels. Although various levels of ball
depths were tested, the compartments were initially filled to the
one-foot level with approximately 1,600,000 balls. The TCA con-
figuration was tested with both three and four stages of balls,
and at ball depths of 6, 10, and 12 inches in the four stage con-
figurations. The PPA configuration consisted of two and three
stages of fixed "eggcrate" packing supported at four-foot inter-
vals. The packing consisted of plastic grids 1% in. thick with
2-in. square openings, with the grids stacked to a depth of ap-
proximately 17 inches in each stage.
The 170 MW Horizontal Module, located on Unit 2, was tested
with 1, 2, 3, 4 and 5 spraying stages. The basic four stage con-
figuration, shown in Figure 2, consisted of crossflow spray dis-
charged from a row of 36 externally-mounted nozzles at each stage.
The Horizontal Module did not contain packing, but consisted of
four stages of crossflow spray. The slurry was cycled through
the scrubber in a counter current manner. That is, the fresh lime
slurry from the mix tank was first sprayed across the flue gas at
the fourth stage or discharge end of the scrubbing chamber. The
same liquid was successively collected and pumped to the third,
second, and first stages and successively depleted of alkalinity.
By the time the slurry reached the first stage collection hopper
for return to the lime mix tank, it was almost completely depleted
of any excess alkalinity. The recirculating slurry flow rate was
normally 9,000 gpm, but the slurry was mechanically pumped four
times per circuit for a total installed pump capacity of 36,000
gpm.
Both the Horizontal and Vertical Modules were designed for
identical test conditions, using the same sources of makeup water
and removing particulate matter resulting from the combustion of
low sulfur western coal from the Black Mesa Mine near Kayenta,
Arizona.
209
-------
SLURRY SPRAYS
FLU EGAS
FROM FAN
SUMP CHAMBER
SCRUBBED FLUE GAS
TO DEMISTER
THERMOPLASTIC
RUBBER SPHERES
6
IN.
QUIESCENT SPHERE
DEPTH OF 6 IN.PER
STAGE
Figure 1. 170 MW Vertical module (4-stage TCA)
210
-------
FLUE GAS
FROM FAN.
15 FT
• SLURRY SPRAYS.
\ \
:I O/^ o'k o'°v p'°v jr
\F M'9" MM/I% A^'X/
1
I
I
1
— ^-^
SCRUBBER
FLUE GAS
TO DEMISTER
Figure 2. 170 MW Horizontal module (4-stage)
OBSERVED SCRUBBER PERFORMANCE
Particulate Removal at Design Operation Conditions
The observed operating conditions of the three full-scale
scrubbers are indicated in Table 1. The outlet gas flowrate was
greater than the inlet due to water vapor evaporated inside the
scrubber. In addition, the Horizontal Module used an indirect
reheater which injected approximately 80,000 scfm of hot ambient
air into the outlet flue gas stream. The pressure drop indicated
in Table 1 is the flue gas pressure drop across the scrubbing
chamber itself, not including additional pressure drop due to
ductwork, mist eliminator, and reheater restrictions. The elec-
tric power consumption in Table 1 is the total electric power
measured for the entire scrubbing system, and includes control
room lighting and air conditioning, thickener rake drive, tank
agitators, process water pumps, and other miscellaneous power re-
quirements. However, these miscellaneous factors are estimated
to be about 5% of the total, while fan and slurry pump power re-
quirements accounted for the remaining 95% of the total.
211
-------
TABLE 1. OPERATING CONDITIONS AT DESIGN GAS FLOWRATE
Horizontal Vertical TCA Vertical PPA
Inlet gas flowrate, scfm
Outlet gas flowrate, scfm
Number of stages
Circulating slurry
flowrate, gpm
Scrubber gas velocity, ft/sec
Pressure drop, in. H2O
Nozzle pressure, psig
Power consumption,MW
Type of reheat
450
560
9
,000
,000
4
,450
22
1.0
35
2.6
Indirect
450
480
16
,000
,000
4
,200
12
14.0
15
3.4
Direct
450
480
27
,000
,000
3
,000
12
12.2
30
3.9
Direct
The observed outlet grain loadings obtained at the design op-
erating conditions for the Horizontal, Vertical TCA, and Vertical
PPA scrubbers are plotted as a function of inlet grain loading in
Figures 3, 4, and 5, respectively. At an inlet grain loading of
0.10 gr/scf, all three scrubber configurations demonstrated a
relatively high degree of particulate removal between 90% and 93%.
As will be pointed out in the discussion, the percentage particu-
late removal changed with inlet grain loading due to changes in
the average particle size and particle size distribution.
EFFECT OF TURNDOWN
During the Test Program, it was determined that the easiest
method of operating the scrubbers was to maintain a constant liq-
uid flowrate and allow the flue gas flowrate to change under sim-
ulated load-following conditions. From an engineering design
standpoint, this simplification avoids the severe erosion problems
encountered with slurry flow control valves. It was decided to
determine the effect on particulate removal of gas flowrate turn-
down with the liquid flowrates held at the design values shown
in Table 1.
The observed variation in percentage particulate removal
at a constant inlet grain loading of 0.10 gr/scf is plotted as a
function of the gas flowrate in Figure 6. Note that the particu-
late removal efficiency of the Horizontal Module increases as a
runction of turndown ratio.
imif. °*le emanation for this is the change in energy input per
unit of flue gas flowrate which occurs with increasing turndown
ratio. The primary method of shattering slurry droplets and thus
contacting particulate matter differs between the types of
scrubbers tested.
212
-------
0.50 —
0.20
w 0.10
a"
z
a
3 0.05
cc
a
i-
z 0.02
0.01
0.005
AVERAGE OPERATING CONDITIONS
INLET FLOW/RATE: 450,000 scfm
OUTLET FLOWRATE; 560,000 scfm
-CIRCULATING SLURRY FLOWRATE:
NOZZLE PRESSURE: 35 psig
NO. OF STAGES: 4
POWER REQUIRED: 2.6 MW
-AP: 1.0 in. H2O
GAS VELOCITY: 22 fps
0.001 0.002 0.005 0.01 0.02 0.05
OUTLET GRAIN LOADING, gr/scf
Figure 3. Inlet vs. outlet grain loading, 170 MW horizontal module
The Vertical TCA and PPA Modules require flue gas pressure
drop to increase the tendency to shatter slurry droplets. The
particulate removal was thus seen to decrease with decreasing
pressure drop. For example, at 450,000 scfm, the particulate re-
moval was about 90% at the design circulating slurry flowrates.
Under these conditions the pressure drop was 14 inches of water
in the TCA and 12.2 inches in the PPA. When the flue gas flowrate
was decreased to one-third the design value, the pressure drop
decreased to 3.5 inches of water in the TCA and 2.0 inches in
the PPA. When the gas flow was reduced maintaining a constant
circulating slurry flowrate, a given amount of flue gas was con-
tacted by three times as many droplets at one-third load as at
full load. In spite of this additional contacting, the particu-
late removal decreased to about 75% at one-third load.
213
-------
0.20 —
fe°
CJ
z
Q
go.
cc
(D
fr-
ill
0.02
0.01
.10 —
en
1 ?.
AVERAGE OPERATING CONDITIONS
INLET FLOWRATE: 450,000 scfm
OUTLET FLOWRATE: 480,000 scfm
CIRCULATING SLURRY FLOWRATE: 16,200 gpm
NOZZLE PRESSURE: 15psig
NO. OF STAGES: 4
POWER REQUIRED: 3.4 MW
AP: 14 in. H2O
GASVELOCIIY: 12fps
0.001 0.002 0.005 0.01
OUTLET GRAIN LOADING, gr/scf
0.02
0.05
Figure 4. Inlet vs. outlet grain loading, 170 MW vertical TCA module
0.20
0.10
CD
2
5
§
•z.
<
cc
(3
0.05
0.02
0.01
AVERAGE OPERATING CONDITIONS
INLET FLOWRATE: 450,000 scfm
OUTLET FLOWRATE: 480,000 scfm
CIRCULATING SLURRY FLOWRATE: 27,000,
NOZZLE PRESSURE: 30 psig
NO. OF STAGES: 3
POWER REQUIRED: 3.9 MW ft
P: 12.2 in. H2O
GAS VELOCITY: 12fps
I
I
gpm
I
I
0.001 0.002 0.005 0.01
OUTLET GRAIN LOADING, gr/scf
0.02 0.05
Figure 5. Inlet vs. outlet grain loading, 170 MW vertical PPA module
214
-------
100%
95%
90%
LU
DC
_J
D
p 85%
EC
<
Q.
80°/
75<
T
INLET GRAIN LOADING: 0.10 gr/scf
HORIZONTAL
MODULE
4 STAGES
9,000 GPM/STAGE
VERTICAL
TCA MODULE
4 STAGES
18,000 GPM
VERTICAL
PPA MODULE
3 STAGES
27,000 GPM
100 200 300 400
FLUE GAS FLOW/RATE, scfm x 103
Figure 6, Effect of turndown ratio on particulate removal
500
The Horizontal Module utilized nozzle pressure to shatter the
slurry droplets, and particulate removal was thus seen to increase
with decreasing flue gas flowrate. As the flue gas flowrate was
turned down from 450,000 scfm to 150,000 scfm, the pressure drop
across the scrubbing chamber decreased from 1.0 inches of water
to 0.10 inches. Since the liquid flowrate was constant, a given
amount of flue gas was contacted by three times as many droplets
at one-third load as at full load. The particulate removal
efficiency was observed to increase from 92.5% at full load to
96.5% at one-third load.
Effect of Circulating Slurry Flowrate
The effect of various circulating slurry flowrates on Verti-
cal TCA Module particulate removal is shown in Figure 7. It is
seen that the percent particulate removal increased as the flow-
rate of the circulating slurry was increased. For example, the
particulate removal achieved at 32,000 gpm is approximately 5%
greater than that achieved at 16,000 gpm for both the 0.10 and
0.03 gr/scf inlet grain loading values.
215
-------
100
95
90
O
LU
EC
LU 85
D
O
I-
cc
<
O.
80
75
70
OPERATING CONDITIONS
450,000 scfm
FOUR STAGES
BALL DEPTH: 10 TO 12 in.
DESIGN FLOW/RATE
I
il I
10 15 20 25 30 35
CIRCULATING SLURRY FLOW/RATE, gpm x 1000
40
Figure 7. Effect of liquid flow rate on vertical TCA module paniculate removal
With the Horizontal Module, particulate removal is also im-
proved as the circulating slurry flowrate is increased. As shown
in Figure 8, an increase in liquid flowrate from 5,000 gpm to
10,000 gpm would result in an increase in particulate removal
from 81% to 93% at an inlet grain loading of 0.10 gr/scf and from
70% to 86% at an inlet grain loading of 0.03 gr/scf. These tests
were conducted by shutting off individual nozzles to maintain 35
to 40 psig nozzle pressure, and although the spray coverage pat-
terns were undoubtedly altered, the slurry droplet size probably
did not change. The required liquid flowrate for a desired de-
gree of particulate removal can thus be estimated from Figure 8.
Miscellaneous Effects
Several miscellaneous effects were investigated during the
Test Program. For example, it was found that increasing the
static ball depth from 6 in. to 12 in. in the Vertical TCA Module
resulted in a minor increase in particulate removal.
216
-------
100
OPERATING CONDITIONS
450,000 scf m
FOUR STAGES
NOZZLE PRESSURE: 35 psig
DESIGN FLOW/RATE
I
2 4 6 8 10 12
CIRCULATING SLURRY FLOW/RATE, gpm x 1000
Figure 8. Effect of liquid flow rate on horizontal module paniculate removal
The effect of variations in nozzle pressure was studied with
the Horizontal Module. Particulate removal increased by 15 to 20
percent as the nozzle pressure was increased from 15 psig to 30
psig.
Finally, it was confirmed that increasing the number of con-
tacting stages improves the particulate removal efficiency. For
example, the particulate removal with the Vertical PPA Module in-
creased as the number of packed stages was increased from two to
three. Similar results were obtained with the Horizontal Module,
where the effect of four versus five stages was studied in detail,
A comparison of fractional collection efficiencies with four and
five stages of scrubbing indicated that the extra stage increased
the collection of particles less than 1,5 ym (aerodynamic). An
interesting description of particulate collection in spray tower
applications due to condensation effects can be found in Calvert,
et al.5 This may be a partial explanation for the high degree of
particulate removal achieved with the Horizontal Module.
217
-------
DISCUSSION
Data Acquisition and Reduction
It must be pointed out that it is not easy to obtain good
particulate data by sampling in large ductwork with cross-sections
of 200 square feet. Gas sampling crews of five persons working
under ideal conditions were able to obtain one inlet and one out-
let isokinetic particulate sample by probe traverse in one day.
All testing was subject to unit outages and weather conditions
ranging from 70 mph winds in the winter to 130° F temperatures in
the summer. All testing was conducted by probe traverse, and
electrical discharges from the metal probes to ground were ob-
served at the inlet duct due to electrical charge carried on the
inlet fly ash particles.
Grain loading measurements were conducted using an alundum
thimble with a backup millipore filter, and gas volumes were
metered and corrected for water vapor condensation in ice bath
condensers. Particle size distribution measurements were ob-
tained with both Andersen and University of Washington cascade
impactors. Most of the particle size distribution measurements
on the Vertical TCA Module were conducted with Andersen impactors
without backup filters while similar measurements on the Horizon-
tal Module were conducted primarily with University of Washington
impactors with backup filters. It was found that physically
identical impactors do not produce the same results from the same
sample and must be individually calibrated using particles of a
known size and instrumentation to measure the concentrations of
particles upstream and downstream from the impactor.
Because of these differences, the particle penetration data
for the Horizontal and Vertical TCA Modules at design operating
conditions is difficult to compare directly. However, it can be
stated with a reasonable degree of confidence that both the Hor-
izontal and Vertical TCA Modules achieved approximately 65% col-
lection of the 1.5 ym (aerodynamic) particles at design operating
conditions. An examination of the slopes of Figures 3 and 4 would
also suggest that the Horizontal Module at design conditions
achieved a higher degree of collection efficiency for particles
larger than 1.5 ym (aerodynamic) than the Vertical Module at de-
sign operating conditions.
The particle size data presented in Figure 10 refers to
aerodynamic particle diameter as measured by a calibrated cascade
impactor. The aerodynamic particle diameter takes into account
particle density and shape factor. The term geometric standard
deviation is used in Figure 9 to identify the distribution of
particle sizes about an average size. For a log-normal particle
size distribution, a geometric standard deviation of two means
218
-------
0.0 0.1 0.2 0.3 0.4 0.5
I r\ I
CUT DIAMETER OF CONTROL DEVICE
AVERAGE PARTICLE SIZE AT INLET
Figure 9. Overall penetration as predicted by impaction collection theory
that the size below which 84 weight percent of the particles are
found is twice the average size. Thus, for a geometric standard
deviation of one, all the particles are the same size.
As another prerequisite to obtaining high quality, accurate
data, detailed grain loading and size distribution studies were
conducted to determine the effects of stratification in the duct-
work. It was found that below a grain loading of about 0.05 gr/
scf, a variation of only ± 15% in geometric mean particle size
and loading could be expected due to stratification at various
points in the duct.
Performance Predicted By Inertial Impaction
If interest is restricted to particle diameters more than
about 1.0 ym (aerodynamic), then Ranz and Wong6 have Shown that
collection by inertial impaction is by far the most important
collection mechanism. Particle penetration for many types of
219
-------
control equipment can be expressed as
P = exp (-Kd B) (1)
where K is a constant and
d is the aerodynamic particle size
P
Calvert7'8 has indicated that the exponent B equals 2 for most
types of packed towers, sieve plate columns, and venturi scrubbers.
If the particle penetration is integrated over the entire size
distribution, the average penetration can be determined by
PAY
(2)
where w is the particle weight.
Following Calvert, equation (2) can be solved for a log-
normal particle size distribution as a function of the inlet size
distribution parameters and the so-called "cut diameter." The
"cut diameter," dp50, is the particle diameter for which the
collection efficiency of the control hardware is 50%. Therefore,
more than half of the particles larger than the "cut diameter"
will be collected. The "cut diameter" characterizes the overall
particulate removal efficiency of a control device for known in-
let particulate characteristics. The solution to equation (2)
gives the overall penetration as a function of inlet particle
size distribution characteristics and the "cut diameter" of the
particulate control device.
Various "cut diameters" of the Horizontal and Vertical TCA
Modules are plotted in Figure 10 as a function of the percent of
electrical power required to operate the scrubber. Figure 10
was prepared from the data in Figures 7 and 8 as well as control
room measurements of electric power requirements over a wide range
of operating conditions. Since the "cut diameter" of a given
scrubber depends on the energy supplied for droplet shattering
and liquid/gas contacting, it is reasonable to expect that the
"cut diameter" would correlate with percent of the electrical
power required to operate the scrubber. A control device using
less electrical power to achieve the same "cut diameter" would
be a more efficient device in terms of particulate removal. The
data in Figure 10 includes penalties for inefficient power con-
sumption, such as pressure drop losses due to ductwork restric-
tions, pumping losses due to elevation differences and less than
100% mechanical efficiency of rotating equipment.
220
-------
3.0
£
a.
DC 00
HI
HI
5
< 1.5
0
h-
o
2 1.0
«x
AERODYNy
0 P
5 3
I I I
©
@K
/ ©^©
HORIZONTAL ®®\
450,000 scfm
3,500 TO 14,000 gpm
I I I
I I I
V 350,000 scfm
^
\ ~
SXAA' ^A
® S © ^S^ 450,000 scf m^
VERTICAL TCA A 4/
18,000 TO 36,000 gpm
I I I
0.8 1.2 1.6 2.0
PERCENT STATION ELECTRICAL POWER
REQUIRED TO OPERATE SCRUBBER
3.2
Figure 10. Comparison of power requirements to produce a given "cut diameter"
As shown in Figure 10, a "cut diameter" of 1.0 ym (aero-
dynamic) could be achieved at 450,000 scfm for an electric power
consumption factor of 1.7% with the Horizontal Module and 2.3%
with the Vertical TCA Module. Both scrubbers could achieve this
result, but one is 35% less efficient than the other. When turn-
down with the Vertical TCA Module is taken into account, the
electrical power comsumption resulting from an increased liquid
flowrate required to achieve the same "cut diameter" increases
significantly.
For the engineer, this method of predicting overall parti-
culate removal knowing the inlet particulate characteristics,
the type of scrubber and its electric power consumption character-
istics can be a useful design tool. An example of this method
is compared with test results.
Example
Horizontal 170 MW Module
As sumptions:
Gas flow = 450,000 scfm
Power to operate scrubber = 2.6 MW
Inlet grain loading =0.04 gr/scf
Average particle diameter =4.0 aerodynamic micro-
meters (ym), _
Geometric standard deviation (a ) - ^.o
221
-------
Calculate:
^ ^ 2 .6 MW ,
Power consumption factor ........ = 1.
"Cut diameter" from Figure 10 = 1.20 ym
"Cut Diameter" = 1.20 _ 0.30
Average Particle Diameter 4.0
Prediction:
Overall penetration from Figure 9 = 0.125
Particulate removal = 100 - 12.5 = 87.5% removal
Test Data:
From Figure 3 (Inlet vs. Outlet Grain Loading)
(0'°4 °°54)- X 10° = 86.5% removal
CONCLUSIONS
Several conclusions can be drawn from the data presented in
this paper:
1. For the 170 MW scrubbers tested with Mohave Generating
Station fly ash, more than 90% total particulate removal
can be achieved with mobile bed, fixed bed and spray
chamber scrubbers operating under design conditions at
450,000 scfm and an inlet grain loading at 0.1 gr/scf.
2. The effect of turning down the flue gas flowrate without
changing the liquid flowrate increases the particulate
removal of the Horizontal Module and significantly de-
creases the particulate removal of the Vertical TCA and
PPA Modules. This could limit the particulate removal
effectiveness of the Vertical Module configurations
during normal load following operation.
3. The effect of changes in liquid flowrate on particulate
removal is more pronounced with the Horizontal Module
than the Vertical TCA Module.
4. Of all the miscellaneous effects tested with the Vertical
Module, including stages of packing in the PPA configura-
tion and static ball depth in the TCA configuration,
only minor changes in particulate removal were observed.
222
-------
5. Large increases in grain loading at the scrubber inlet
cause relatively small increases in grain loading at
the scrubber outlet. This is due to high collection
efficiency for large particles. For example, at design
operating conditions the Horizontal and Vertical TCA
Modules both achieve more than 95% removal for particles
larger than 6.0 ym (aerodynamic).
6. At design operating conditions, the Vertical TCA Module
requires about 35% more electrical power than the Hori-
zontal Module to achieve the same degree of particulate
removal.
ACKNOWLEDGEMENTS
The Test Modules Program was a joint venture of the Navajo
and Mohave Power Project.participants who are listed below:
Salt River Project Agriculture Improvement
and Power District
Arizona Public Service Company
Department of Water and Power of the City of
Los Angeles
Nevada Power Company
Tucson Gas and Electric Company
Bureau of Reclamation of the U. S. Department of
the Interior
Southern California Edison Company
Funding for this program was provided by the participants in
accordance with their respective megawatt entitlements in the
Navajo and Mohave Power Projects. Southern California Edison
Company was the project manager of the Test Modules Program.
The conclusions presented in this paper represent the per-
sonal opinions of the authors and are not intended to represent
the opinions or position of any of the project participants.
223
-------
REFERENCES
1. Shapiro, J.L., and W.L. Kuo. The Mohave/Navajo Pilot Facil-
ity for Sulfur Dioxide Removal. 2nd EPA Flue Gas Desulfur-
ization Symposium, New Orleans, Louisiana, November 8, 1971.
2. Weir, A., and L.T. Papay. Scrubbing Experiments at the Mohave
Generating Station. 3rd EPA Flue Gas Desulfurization Sympos-
ium, New Orleans, Louisiana, May 14, 1973.
3. Weir, A., J.M. Johnson, D.G. Jones, and S.T. Carlisle. The
Horizontal Crossflow Scrubber. 4th EPA Flue Gas Desulfur-
ization Symposium, Atlanta, Georgia, November 4, 1974.
4. Weir, A., L.T. Papay, D.G. Jones, J.M. Johnson, and W-C.
Martin. Results of the 170 MW Test Modules Program. 5th
EPA Flue Gas Desulfurization Symposium, New Orleans,
Louisiana, March 8, 1976.
5. Calvert, S., J. Goldschmid, D. Leith, and N.C. Jhaveri.
Feasibility of Flux Force/Condensation Scrubbing for Fine
Particulate Collection. EPA-65Q/2-73-036, U.S. Environmental
Protection Agency, Washington, D.C., 1973.
6. Ranz, W.E., and J.B. Wong. Impaction of Dust and Smoke Par-
ticles. Ind. Eng. Chem. 44(6):1371-1381, 1952.
7. Calvert, S. Engineering Design of Fine Particle Scrubbers.
EPA/APT Fine Particle Scrubber Symposium, San Diego, Califor-
nia, May 28, 1974.
8. Calvert, S., J. Goldschmid, and D. Leith. Scrubber Perfor-
mance for Particle Collection. A.I.Ch.E. Symposium Series 70
(137):357, 1974.
9. Jones, D.G., A. Weir, J.M. Johnson, W.C. Martin, and S.
Calvert. Particulate Removal Characteristics of the 170 MW
Mohave Scrubbers. EPA/EPRI Symposium on Particulate Control
in Energy Processes, San Francisco, California, May 11, 1976.
224
-------
PAPER 12
EVALUATION OF A PARTICULATE SCRUBBER ON A
FULL SCALE UTILITY BOILER
D. S. Ensor
Meteorology Research, Inc.
L. E. Sparks
Industrial Environmental Research Laboratory
Environmental Protection Agency
S. Calvert
Air Pollution Technology, Inc.
D. V. Wallon and K. S. Campbell
Stearns-Roger, Inc.
ABSTRACT
The paper presents results of a performance test and engi-
neering analysis of a mobile-bed scrubber on a full-scale coal-
fired utility boiler. The scrubber nominally operated at the
design particulate removal efficiency of 95% but the concentra-
tion of submicron particles was greatly influenced by mist
entrainment. The entrainment resulted in a difference of aerosol
penetration through the scrubber as a function of elemental
composition and an outlet subnucron particle concentration in-
dependent of pressure drop through the scrubber. The engineering
analysis showed that the 1972 installed cost was $29/kW and the
annual operating cost is 0.5 mills/kwh (75% availability). An
initial decline in scrubber availability after start-up resulted
from now-corrected minor design problems. Steadily improving
reliability is attributed to the utility's providing maintenance
and solving operating problems.
INTRODUCTION
Although scrubbers are commonly used for particulate control
in many industries, they are not commonly used in the utility
industry. Because scrubbers are not sensitive to dust resistiv-
ity, they may provide an acceptable means of controlling high-
resistivity fly ash from combustion of low sulfur coal. In order
225
-------
to assess the potential of scrubbers for particulate control in
the electric power industry, Industrial Environmental Research
Laboratory-Research Triangle Park (IERL-RTP). sponsored an exten-
sive evaluation of the TCA* scrubber at Public Service Company of
Colorado's Cherokee Power Plant. A summary of the performance
tests and engineering analysis of the evaluation is presented in
this paper. Details of the tests and test results are reported
by Ensor et al1
DESCRIPTION OF SYSTEM
Power Plant
The Cherokee Power Plant is located in north Denver, Colo-
rado, near the South Platte River. The No. 3 unit has a name
plate rating of 150 MW. The gas cleaning equipment consists of
mechanical collectors, an electrostatic precipitator, and a wet
scrubber arranged in series.
The unit normally burns coal or natural gas. The coal is
mined in western Colorado and has a sulfur content of about 0.5%,
an ash content of 9%, and a heating value of 2.492 x 107J/kg
(10,712 Btu/lb).
Scrubber
The power plant was retrofitted in 1972 with a scrubber to
treat flue gas from an existing electrostatic precipitator.
The Model 6700 Turbulent Contact Absorber Scrubber was de-
signed by UOP, Air Correction Division. A diagram of the system
is shown in Figure 1. The flue gas from the precipitator passes
into two parallel induced draft fans. A bypass damper is used
to direct the flue gas either into the stack or into the scrubber.
The flue gas (under design conditions), 17,000 am3/min, at
137°C (610,000 acfm at 280°F), enters the booster fans to offset
the pressure drop through the scrubber. In the presaturator,
1440 Jl/min (380 gpm) of makeup water is sprayed into the gas to
reduce the temperature to approximately 50°C (125°F). From the
presaturator, the gas enters the scrubber. The scrubber con-
sists of three stages of fluidized beds packed with 3.8 cm
(1.5 in.) diameter plastic balls arranged into three separated
parallel scrubber sections. The two outer sections each handle
20% of the flow, while the center section handles the remaining
60%. All three sections can operate independently to provide
flexibility of operation.
*Mention of manufacturers or brand names does not imply endorse-
ment or recommendation for use by EPA.
226
-------
NJ
MECHANICAL
COLLECTOR
\
FLUE GAS
FROM
BOILER
^
^^L—
:-
«
-lfa±ari»x-
^
I
//,
//
vA,
A
'///!
m
m
DOWNSTREAM
MULTILEAF SAMPLE POINT
CONTROL 3 PARALLEL DUCTS
DAMPERS /
1 /
/'/// a -^
'////
m
m
M
'////A
STACK
UPSTREAM
SAMPLE POINT
4 PARALLEL DUCTS
V
r^
^^^
A--£=3
s
MIST
ELIMINATOR
WASH
.REHEAT STEAM
1 --VWV
BLOCK/BYPASS MAKEUP t ^ ^
DAMPER WATER— I w o -•-•-—•-
/ TJ Q =a
/ n^ "=
ISOLAIION 33 J """""
DAMPERS S3 «
// PRESATURATOR ., z -^
L ^^rpfn L_ o -
™aJf II i -=
L/y_Lj u^ik a
/
MIST
""ELIMINATORS
SCRUBBER
INDUCED DRAFT FANS SCRUBBER BOOSTER FANS \JSH
ELECTROSTATIC
PRECIPITATOR
1 (^1
£^f
/ — 3 RECIRCULATION
*— k.2/ PUMPS (5)
SLURRY DRAWOFF
Figure 1. Scrubber system
-------
The scrubbed gas passes through chevron-type mist elimi-
nators made of fiberglass reinforced plastic where entrained
droplets are removed. The mist eliminators are sprayed once a
work-shift from the top to prevent accumulation of solids.
The gas is then heated by steam coils to 85°C (185°F) before
entering the stack to prevent corrosion of the stack and duct
work and to provide plume buoyancy after discharge into the at-
mosphere. The steam coils are equipped with two sets of soot
blowers to remove fly ash from the heat transfer surfaces.
The scrubber was designed to the following specifications :
Gas Flow 1,036,000 am3/hr, 610,000
acfm 138°C (280°F)
Liquid-to-Gas Ratio 7.4 Jt/m3 (55 gal/100 ft3)
AP . 30.5 cm H20 (12 in. H2O)
Inlet Particulate Con- 0.92 g/m3 (0.40 gr/scf)
centration
Outlet Particulate Con- 0.046 g/m3 (0.02 gr/scf)
centration
Efficiency of Particulate 95%
Removal
MEASUREMENT TECHNIQUES
Three measurement techniques were used during the tests to
determine particle collection efficiency. These were: (1)
diffusional techniques using condensation nuclei counters and
screen diffusion batteries for determining concentration and
size distribution on a number basis for particles having diam-
eters less than approximately 0.2 ym; (2) inertial techniques
using Meteorology Research, Inc. (MRI) cascade impactors for
determining concentrations and size distributions on a mass basis
for particles having diameters between approximately 0.3 urn and
20 urn, and (3) standard EPA Method 5 mass train measurements for
determining outlet mass loadings. Details of the various instru-
ments and test procedures are reported by Ensor et al1.
Two Meteorology Research, Inc. (MRI) Plant Process Visi-
ometers (PPV) were installed-one at the scrubber inlet and the
other at the outlet. These instruments were used as real-time
monitors to detect upsets and to determine in-stack opacity.
Impactor runs were conducted to obtain samples for elemental
analysis by Ion-Excited X-Ray Analysis.
228
-------
All impactor runs were conducted using Apiezon L high vacuum
grease as a particle collection substrate. The Apiezon L was
applied to light-weight collection discs. All weighings were
conducted using a Cahn 400 analytical balance with a precision of
0.01 mg. Use of the Cahn balance and light-weight substrates was
essential to reduce errors due to weighing.
The test program was divided into three phases in order to
systematically evaluate the scrubber. The test phases were:
Phase 1 - Evaluation of the distribution of aerosol
mass concentration at the inlet. With the limited
resources in manpower and equipment, the evaluation
of the source was attacked one aspect at a time. All
four inlet ducts were sampled at the same time.
The cascade impactor trains were supplemented with
in-stack filter probes to establish the consis-
tency of the concentration.
Phase 2 - Evaluation of outlet distribution of
aerosol mass concentration. Similar parallel tests
were used on the outlet of the scrubber to indi-
cate the variation in the sections.
Phase 3 - Simultaneous inlet-outlet tests for
scrubber efficiency. The final phase of the field
program was the simultaneous inlet-outlet tests
with both primary and secondary tests.
The tests disclosed several problems with the scrubber which
are discussed by Ensor et al1.
SCRUBBER EFFICIENCY RESULTS
The Phase 3 tests were used to compute the efficiency of the
scrubber. Sufficient data were taken to allow selection of the
runs taken during steady-state plant conditions. The data used
in the efficiency determinations are summarized in Table 1. The
diffusion battery tests were restricted to section B at the out-
let because of the relatively normal behavior of that section
of the scrubber.
Total Mass Collection Efficiency
The mass collection efficiency was computed using only the
cascade impactor data. The results are summarized along with
some control room data in Table 1. The pressure drop across the
system was reported to indicate the magnitude of the effect of
the existence of the air reheaters. The outlet gas flow was the
total for the system as determined from the velocity traverses.
The lack of correlation between efficiency and pressure drop was
an unexpected result.
229
-------
TABLE 1. SCRUBBER COLLECTION EFFICIENCY
SECTION A
U)
O
Date
1974
11/20
11/21
12/105
12/11
12/12
Load,
MW
166
164
157
160
160
02, %
3.6
3.4
3.4
3.0
2.6
Outlet
Gas Flow,
am3 /hr
a
a
9.47 x 10s
10.2 x io5
8.78 x io5
AP
System,
cm H20
41
39
36
38
38
Ap
Bed,
cm H20
9.9
9.6
15.2
14.7
14.7
Ap
Mist
Elimin. ,
cm H2O
0.76
0.76
1.7
1.5
1.8
Eff.,
*
ND
ND
96.3
96.4
79.6
AP
System,
cm H20
45
43
41
42
44
SECTION B
AP
Bed,
cm H20
25
18
20.8
22.1
22.9
Ap
Mist
Elimin. ,
cm H20
2.5
1.8
2.5
3.2
2.5
Eff.,
*
84.7
89.9
92.6
93.2
93.1
AP
System,
cm H2O
46
44
41
44
46
SECTION C
Ap
Bed,
cm H2O
21
20
18.5
22.4
24.1
AP
Mist
Elimin. ,
cm H2O
8.3
5.1
3.8
2.5
3.8
Eff. ,
%
ND
ND
86.9
96.7
92.1
Full velocity traverses were not taken.
The control room data were incomplete. Interviews, data from other days and the log book were used to
supplement available information.
-------
The average efficiency of the December tests was 92%,
slightly lower than the design efficiency of 95%.
Mass Penetration as a Function of Particle Diameter
The particle mass penetration as a function of particle di-
ameter was determined for three days. These data are shown in
Figures 2 to 4.
Generally, the size distributions and penetrations were con-
sistent, though loadings fluctuated considerably. The maximum
penetration (minimum efficiency) was achieved in the region of
0.2 urn actual diameter, similar to results for another scrubber
as reported by Sparks et al3.
Elemental Chemical Analysis
The measurement of the concentrations of a number of ele-
ments was intended mainly to provide data to aid in the interpre-
tation of the scrubber performance. It was not planned to per-
i.o
o
oc
uj 0.1
Z
UJ
Q_
UJ
O
h-
cc
a.
0.01
••B
If
I
' DIFFUSION BATTERY RUNS NO. 8 AND NO.
> IMPACTOR RUNS NO 110 AND NO. 113
> IMPACTOR RUNS NO. 109 AND NO. 114
IIMACTORS RUNS NO. 116 AND NO. 119
MMPACTOR RUNS NO. 120 AND NO. 118
0.04 0.07 0.1
0.2 0.5 1.0
PARTICLE DIAMETER,»m
2.0
Figure 2. Combined penetrations for diffusion battery and cascade impactor (December
10, 1974)
231
-------
1.0
c
o
z
o
<
£0.1
111
z
111
HI
o
l-
oc
Q.
0.01
A DIFFUSION BATTERY RUNS
NO. 13 AND NO. 15
V IMPACTOR RUNS NO. 122 AND NO. 123
• IMPACTOR RUNS NO. 124 AND NO. 125
• IMPACTOR RUNS NO. 127 AND NO. 126
V
I
i
l
0.02 0.05 0.10 0.3 0.5 1.0
PARTICLE DIAMETER, Mm
3.0 5.0
Figure 3. Combined penetrations of diffusion battery and cascade impactor (December
11, 1974)
I.U
c
_o
u
z
z
O
<
tt 0.1
yj
Z
111
Q.
LU
O
P
oc
a.
0.01
I I ± 1 w i '
/• f" V
X "\
•** \
A *
• DIFFUSION BATTERY RUNS NO. 10 AND NO. 12
••i
AIMPACTOR RUNS NO. 129 AND NO. 133 B 1140
• IMPACTOR RUNS NO. 132 AND NO. 135 B 1300
• IMPACTOR RUNS NO. 130 AND NO. 134 A 1113
V IMPACTOR RUNS NO. 131 AND NO. 136 C 1300
II III
0.03 0.05 0.10 0.3 0.5 1.0
I
•
V *
_
» 9 —
m
,\
i
\
i
3.0 5.
PARTICLE DIAMETER,Jnm
Figure 4. Combined penetrations for diffusion battery and cascade impactor (December
12, 1974)
232
-------
form elemental balances or determine emission factors. The
samples were taken during December 10-11, 1974. The overall pen-
etrations of various elements are shown in Table 2. The
penetration as a function of particle diameter for various
elements is shown in Figure 5.
TABLE 2. PENETRATION OF ELEMENTS THROUGH THE
SCRUBBER FOR DECEMBER 10, 1974
Element
Penetration
Average Outlet
Concentration, .
micrograms/dsm3
Al
Si
S
K
Ca
Ti
V
Cr
Fe
Ni
Cu
Zn
Br
Pb
For all elements
Total mass
0.029
0.033
3.4a
0.043
0.059
0.073
0.14
i.ioa
0.18
0.95
2.9a
1.5a
0.28
0.64
0.108
0.074
326
658
1030
50
508
96
27
57
1500
33
668
501
5.7
120
aPenetrations greater than one indicate generation of particles in
flue gas by evaporation of the scrubber liquor.
bDry standard, 21.1°C, 760 mm Hg.
DISCUSSION
The scrubber performance results were analyzed in depth for
trends and to allow comparison to data in the literature.
233
-------
10'3
10
10°
PARTICLE DIAMETER,
Figure 5. Scrubber penetrations for selected elements
234
-------
Examination of the particle penetration data for each simul-
taneous pair of inlet and outlet runs shows a wide range of re-
sults, as summarized in Table 3.
TABLE 3. PARTICLE PENETRATION SUMMARY
Run
In
56
55
60
114
119
121
125
133
135
124
113
127
134
115
120
123
136
No.
Out
53
57
61
109
116
123
127
129
132
125
110
126
130
119
118
122
131
Section
B
B
B
B
B
B
B
B
B
B
A
A
A
C
C
C
C
d a
pc,
Vim
1.5
1.1
0.8
0.8
1.0
0.8
<0.5
0.8
0.8
0.6
1.0
0.8
2.0
<0.5
0.6
0.7
1.1
AP, Filter out
cm H20 Filter in
23.0
23.0
23.0
23.0
23.0
23.0
23.0
23.0
23.0
23.0
—
15.0
9.0
25.0
—
25.0
24.0
0.31
0.81
1.2
2.9
0.39
0.55
1.2
2.0
0.47
3.1
0.81
0.41
6.7
0.76
1.0
1.1
3.5
a
Scrubber cut (50% penetration) diameter.
Ratio of outlet filter to inlet filter particle concentrations
(mg/dsm3)/(mg/dsm3).
The scrubber performance cut diameter (_i.e_. / particle diam-
eter at 50% penetration), used to characterize efficiency, varied
from less than 0.6 um to 2.0 ym. No pattern of correlation
between cut diameter and other parameters such as pressure drop
was found. Thus, the variation was due to a combination of
system fluctuations and measurement errors.
Penetrations found in this study are much higher than those
reported in a previous study of the same scrubber (Calvert,
et al. **), and of another mobile bed scrubber on a coal-fired
235
-------
power plant (Statnick and Drehmel5). For example, Calvert et aJU
reported a cut diameter of 0.35 urn and a penetration at 1.0 ym of
about 0.09. Statnick and Drehmel reported that, for 25 cm of
water scrubber pressure drop, penetration did not exceed 0.05 for
any particle size and was about 0.02 for 1.0 \im particle diameter.
These and some other points are compared in Table 4 with
representative results from the present study.
TABLE 4. COMPARISON OF MOBILE BED STUDIES
Penetration at Particle
Diameter Shown
Investigators
Present
Calvert,
Statnick
et
and
al."
Drehmel5
d , urn
pc
0.8
0.35
—
0.
0
0
0
5 ym
.8
.3
.07
0.8 ym
0.5
-0.15
0.03
1.
0
0
0
0 ym
.4
.08
.02
2.
0
0
0
0 ym
.15
.02
.002
Approximate mean values for all runs, exclusive of 56/53 and
134/130.
There is evidence that the high outlet particle concentra-
tion measured is due to entrainment from the scrubber. Facts
which indicate the presence of outlet particles introduced by
entrainment are as follows:
1. Outlet filter loadings are higher than inlets, based on
the same gas sample volume, for many of the tests.
2. Penetrations in the present study are higher than those
found by others.
3. Variations in penetration are not related to gas flow
rate, pressure drop, or other known parameters, but can
be attributed to variable entrainment.
4. Penetrations for scrubber section "A" are the same as
for section "B" despite the gas pressure drop for "A"
being about half that for "B".
5. Reheater and entrainment separator operating problems
occurred during the test.
6. Elemental analyses of inlet and outlet particles showed
an apparent "generation" of particles containing soluble
elements as described in the previous section.
236
-------
The overall conclusion is that the scrubber performance data
obtained in this study are specific for the operating factors and
scrubber condition which existed during the test period. It is
not possible to establish a general mobile bed scrubber perform-
ance model from these data because of the overshadowing and un-
defined influence of liquid entrainment. It appears that both
scrubber performance and reliability could be improved by improved
entrainment separation.
ENGINEERING ANALYSIS
The objective of the engineering analysis was to assemble
the following information:
Capital costs
Operating costs
Major maintenance problems
Scrubber reliability
Estimated cost required to minimize operating problems.
Capital Costs
The total installed cost in 1972 for the scrubber was
$4,400,000. Based on the boiler nameplate rating of 150 MW, the
cost is $29/kW or, based on the rated gas volume (see above),
is $4.18/1000 m3/hr ($7.10/1000 acfm). The detailed cost itemi-
zation is given by Ensor et al.l. In 1975 dollars, the scrubber
would cost $5,800,000.
Operating Costs
The total operating costs are approximately $495,000/yr
(fourth quarter 1973 and the first three quarters of 1974) based
on 75% availability of the scrubber or 0.50 mills/kwh.
Maintenance Problems
The scrubber had a number of maintenance problems, many of
which were solved during start-up and operation. The problems
mentioned here are the more persistent ones which have defied
solution. The maintenance problems are discussed in detail by
Ensor et aJL.* . These include:
1. Breakage of Mobile Bed Contactors - The plastic mobile
bed contactors have been a chronic problem due to break-
age from wear. The desired lifetime of the spheres is
8000 hours; however, normally a lifetime of only 6000
hours was experienced.
237
-------
Moreover, damage to the pump liners and plugging of the
nozzles can result from the fragments entering the
liquor recycle piping. Screens in the scrubber hopper
have eliminated this problem with the penalty of adding
screen cleaning to the required maintenance. Public
Service Company of Colorado has tested a number of dif-
ferent packings and is currently using polyethylene
spheres in the scrubber.
2. Migration of Mobile Bed Contractors - The mobile bed con-
tactors will also migrate from one section of the
scrubber to another if an opening the width of a con-
tactor exists in the partitions. Poor distribution of
the contactors causes channeling of flue gas and a re-
duction in particle collection efficiency.
3. Guillotine Dampers - The guillotine isolation dampers
have caused problems due to breakage of the damper when
closing against a buildup of fly ash and leakage.
4. Recirculation Pumps - The recirculation pumps have been
a source of problems in the past due to mechanical fail-
ure. A new pump has been tested by Public Service
Company of Colorado and has been providing good service.
5. Reheater Section - The scrubbed gases are heated by
direct contact with three banks of steam coils. These
coils are susceptible to pluggage and corrosion in the
wet flue gas scrubber discharge. Addition of a second
set of soot blowers and drying ash during periods of the
scrubber shutdown with heavy soot blowing has had mini-
mal effect. Corrosion problems have rendered the heaters
inoperable. Both upper and lower layer reheat coils
have been removed.
6. Weather-Related Problems - The freezing of lines has
always been a problem during cold weather. All lines
must be heat-traced and drained when the scrubber is
shut down.
Another solution is to enclose the scrubber in a
weatherproof building. The major problem with this is
the possibility of leakage of flue gas into the
structure.
Scrubber Reliability
For this study, availability of the scrubber was defined as:
238
-------
hours of scrubber operation* -hours boiler was
AvaUability = burning 100% gas
hours of boiler operation -hours boiler was
burning 100% gas
The reason that the hours the boiler was burning 100% gas was
subtracted from the total hours of scrubber and boiler operation
is that the scrubber is normally either left running with only
water systems operating or completely shut down during periods of
100% gas burning operation.
It is interesting to note that the scrubber was out of ser-
vice a majority of the time during August and September 1974,
when the unit returned to mixed fuel burning. The reason for
this is that after long periods of idleness, due to either main-
tenance or 100% gas burning, the scrubber usually experienced
minor start-up problems. When start-up troubles with the scrub-
ber were encountered, every effort was made to use any available
gas on the unit. Consequently the scrubber had a poor percent-
age availability for the months involved, but the overall avail-
ability was not affected very much.
Values of availability are given for two capacity levels:
100% of capacity and 80% of capacity or greater. There is a
significant amount of time that the scrubber operated at 80% of
capacity. For example, in January of 1974, although the scrubber
operated at 100% of capacity for only 175 hours, it did operate
at 80% of capacity for 547 hours. Although the scrubber is not
treating all of the boiler flue gas, it is felt that some credit
should be given the scrubber for handling most of the flue gas.
In most cases, running at 80% of capacity will probably allow
PSCC to meet particulate emission standards.
The cumulative percent availability to date reflects the
total hours of scrubber operation since start-up. This value is
plotted by month in Figure 6. This figure shows that after an
initial drop, the availability of the scrubber has been increas-
ing steadily. This is typical of other PSCC scrubber operations.
The initial drop is attributed to the many minor design problems
which arose soon after start-up. The following increase in
availability is attributed to the solving of these problems and
the increased operating knowledge which was gained through
operating experience.
As of November 30, 1974, the scrubber had operated at 100%
capacity 59.9 percent of the time and at 80% capacity or
greater 70.9 percent of the time.
*Hours of scrubber operation is related directly to the time the
boiler was operating.
239
-------
80
70
60
LU
_J
DO
< 50
3?
LU
> 40
<
_i
D
1 30
o
O
20
10
I I I I I I I I I I 1 I I I 1 I I I I I
I I I I I I I I I J I 1 I
Z 0
1972
CQ
LU
U.
Z_lC3D-5->02
-3 D Diu O O UJ
O. <
< s
Z
D D
C5 i H >
D UJ O O
-j < W O Z
1973
1974
MONTH
Figure 6. Cherokee No. 3 scrubber accumulative availability from start-up to November
1974
-------
In order to more clearly see the increase in availability
with time, after the first year when each new month was added to
the log, only the previous 12 months availability was calculated.
This value is plotted in Figure 7. Looking at the last month
tabulated (November 1974) reveals that the scrubber operated at
100% of capacity 78.6 percent of the time and at 80% of capacity
or greater 96.7 percent of the time during the previous year.
The figure also shows that the scrubber had been operating at 80%
of capacity or greater with over 90% availability since May of
1973.
The feasibility of using the design and operating experience
to improve availability of the scrubber was investigated. The
goal was to increase the availability of the scrubber, realizing
that the maintenance required may slightly increase due to the
extra equipment. The suggested modifications are summarized
below:
Estimated Availability, %
Identical scrubber $5,800,000* 60-70
Extra 33-1/3% capacity
section 1,200,000 10-15
Indirect reheat, incre-
mental 200,000
Scrubber enclosure 90,000 5-10
Miscellaneous charges 80/000
$7,370,000 75-95
The capital cost of $7,370,000 is $49/kW for a 150 MW unit.
The important aspect of maintaining and improving the avail-
ability of the scrubber system is the attitude of the plant
operating and maintenance people in keeping the unit on line.
The steadily increasing availability of the Cherokee Scrub-
ber is an indication that Public Service of Colorado is
committed to solving the operating and maintenance problems.
*1975 dollars.
241
-------
NJ
100
90
80
70
< 60
ai
50
D
3
U
40
30
20
10
I I I I I I I 1 I I I
80% CAPACITY /
I I I I I I I I
I I I I I
8
> o
O ui
Z Q
1972
5 DC
< a.
2 <
= => 3
-5 -5 <
1973
a.
LU
CO
a 1
Q =1
CD
yj
u.
DC DC
< B.
>; 2
O
D
_ ^
UJ O O
co O Z
MONTH
1974
Figure 7. Cherokee No. 3 scrubber accumulative availability from start-up to November
1974 averaged over the previous year on a running basis
-------
CONCLUSIONS
The performance data and engineering analysis indicate that
scrubbers are capable of collecting fly ash from combustion of
low sulfur coals. The Cherokee power plant was a successful
retrofit with a scrubber to supplant the control with an existing
electrostatic precipitator. The scrubber has a small physical
space and the ability to connect with existing duct work. This
is often a problem when retrofitting other control devices.
Capital and operating costs for a scrubber are likely to be
higher than similar costs for a moderately sized electrostatic
precipitator. However, in applications where the dust resistivity
requires a large-size electrostatic precipitator, a scrubber
system is likely to be cost competitive.
ACKNOWLEDGEMENTS
The efforts of Public Service Company of Colorado personnel in
providing assistance during the source test and the engineer-
ing evaluation were vital to the success of the Project. This
work was conducted under EPA Contract 68-02-1802.
REFERENCES
1. Ensor, D.S., B.S. Jackson, S. Calvert, C. Lake, D.V. Wallon,
R.E. Nilan, K.S. Campbell, T.A. Cahill, and R.G. Flocchini.
Evaluation of a Particulate Scrubber on a Coal-Fired Utility
Boiler, EPA-600/2-75-074, NTIS PB 249562/AS, U.S. Environ-
mental Protection Agency, Washington, D.C., 1975.
2. Raben, I.A. Use of Scrubbers for Control of Emissions from
Power Boilers. U.S. Paper No. 13. Proceedings, Symposium on
Control of Fine Particulate Emissions from Industrial
Sources. San Francisco, California. January 1974.
3. Sparks, L.E., J.D. McCain, and W.B. Smith. Performance of
a Steam-Ejector Scrubber. J. Air Pollution Control Assoc.
2^:958, 1974.
4. Calvert, S., N-C. Jhaveri, and C. Yung. Fine Particle
Scrubber Performance Tests, EPA-650/2-74-093, NTIS PB
240325/AS, U.S. Environmental Protection Agency, Wash-
ington, D.C., 1975.
5. Statnick, R.M., and D.C. Drehmel. Fine Particle Control
Using Sulfur Oxide Scrubbers. 67th Meeting of the Air
Pollution Control Assoc., Denver, Colorado, Paper No. 74-231.
June 1974.
243
-------
METRIC CONVERSION FACTORS
To convert from
Ib
gr/ft3
ft3/rain (cfm)
lbs/in.2
oF
ftVlOOO cfm
in. w.g.
gallons
ft
in.
tons
in.3
ft3
gal/min
ft2
in.2
gal/1000 ft3
grams
ft/min
ounces
oz/yd2
grains
gr/ft2
Ib force
lb/ft2
in. H2O/ft/min
Btu
To
kg ,
g/m3
m3/sec
kg/m2
°C
m2/(m3/sec)
mm Hg
liters
m
m
kg
cm3
m3
I/sec
m2
cm2
1/m3
grains
cm/sec
grams
g/m2
grams
g/m2
dynes
g/cm2
cm H20/cm/sec
calories
Multiply by
0.454
2.29
0.000472
703.
(°F-32) x 5/9
0.197
1.868
3.785
0.3048
0.0254
908.
16.39
0.028
0.0631
0.0929
6.452
0.135
15.43
0.508
28.34
33.89
0.0647
0.698
4.44 x 10s
0.488
5.00
252
244
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing]
REPORT NO.
ElPA^600/7-76-016
3. RECIPIENT'S ACCESSION NO.
TITLE ANOSUBTITLE
CONFERENCE ON PARTICULATE COLLECTION
PROBLEMS IN CONVERTING TO LOW SULFUR
COALS
5. REPORT DATE
October 1976
6. PERFORMING ORGANIZATION CODE
. AUTHOR(S)
G.B. Nichols (Compiler)
8. PERFORMING ORGANIZATION REPORT NO.
. PERFORMING ORGANIZATION NAME AND ADDRESS
Southern Research Institute
2000 Ninth Avenue, South
Birmingham, Alabama 35205
10. PROGRAM ELEMENT NO.
EHE624
11. CONTRACT/GRANT NO.
68-02-2114
12. SPONSORING AGENCY NAME AND ADDRESS
13. TYPE OF REPORT AND PERIOD COVERED
EPA, Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
Proceedings: 1-9/76
14. SPONSORING AGENCY CODE
EPA-ORD
15. SUPPLEMENTARY NOTES
IERL-RTP Project Officer for this report is D. C. Drehmel, Mail
Drop 61. 919/549-8411 Ext 2925.
B. ABSTRACT fhese proceedings present papers that discuss problems encountered in
burning low-sulfur coal in electrical utility power plant boilers. Operating experience
with electrostatic precipitators (ESPs), fabric filter baghouses, and wet scrubbers
for control of fly ash stack emissions and techniques for improving performance are
described. Increasing the collection efficiency of ESPs for high-resistivity fly ash by
adding conditioning agents (e.g. , sulfur trioxide, sulfuric acid, ammonium sulfate,
sulfamic acid, and ammonia) to the flue gas is discussed, as are possible mechanisms
for their action. Experiments are described on the reentrainment of fly ash as the
result of rapping ESP collection electrodes. A mathematical model is described that
calculates the collection efficiency of an ESP as a function of particle size and oper-
ating conditions. A mathematical model is also presented for fabric filter action; it
incorporates a model of collection efficiency and a non-linear model of fabric drag.
Fractional collection efficiency data for submicron fly ash particles are given for a
baghouse. Performance data are also presented for wet scrubbers collecting fly ash
and in one experiment include the chemical composition (e.g. , content of trace ele-
ments) of the uncollected fly ash. A multiple-stream process is described for clea-
ning coal to produce fractions of low and medium sulfur contents by specific gravity
differences. ___________-™_________^^ :
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. cos AT I Field/Group
Air Pollution Fly Ash
Dust Collectors Mathematical Models
Coal Fabrics
Combustion Flue Gases
Electric Power Plants
Electrostatic Precipitators
Air Pollution Control
Stationary Sources
Particulate
Low-Sulfur Coal
Baghouses
Fabric Filters
Coal Cleaning
13B
13A
2 ID
2 IB
10B
12A
HE
IS. DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (This Report}
Unclassified
21. NO. OF PAGES
257
20. SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
245
------- |