vvEPA
United States Industrial Environmental Research EPA-600/9-84-025c
Environmental Protection Laboratory November 1984
Agency Research Triangle Park NC 27711
Research and Development
Fourth
Symposium on the.
Transfer and ^ ^ ^
Utilization of
Particulate Control
Technology:
Volume III. Economics,
Mechanical Collectors,
Coal Characteristics,
Inhalable Particulates,
Advanced Energy and
Novel Devices
-------
EPA-600/9-84-025C
November 1984
FOURTH SYMPOSIUM ON THE
TRANSFER AND UTILIZATION OF
PARTICULATE CONTROL TECHNOLOGY:
VOLUME III. ECONOMICS, MECHANICAL COLLECTORS,
COAL CHARACTERISTICS, INHALABLE PARTICULATES,
ADVANCED ENERGY AND NOVEL DEVICES
Compiled by:
F. P. Venditti, J. A. Armstrong, and Michael D. Durham
Denver Research Institute
P. 0. Box 10127
Denver, Colorado 80210
Grant Number: CR 809301
Project Officer
Dale L. Harmon
Office of Environmental Engineering and Technology
Industrial Environmental Research Laboratory
Research Triangle Park, North Carolina 27711
INDUSTRIAL ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U. S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
-------
DISCLAIMER
This document has been reviewed in accordance with U.S.
Environmental Protection Agency policy and approved for publication.
Mention of trade names or commercial products does not constitute
endorsement or recommendation for use.
11
-------
ABSTRACT
The papers in these three volumes of Proceedings were presented
at the Fourth Symposium on the Transfer and Utilization of Paticulate
Control Technology held in Houston, Texas during 11 October through 14
October 1982, sponsored by the Particulate Technology Branch of the
Industrial Environmental Research Laboratory of the Environmental
Protection Agency and coordinated by the Denver Research Institute of
the University of Denver.
The purpose of the symposium was to bring together researchers,
manufacturers, users, government agencies, educators and students to
discuss new technology and to provide an effective means for the
transfer of this technology out of the laboratories and into the hands
of the users.
The three major categories of control technologies -
electrostatic precipitators, scrubbers, and fabric filters - were the
major concern of the symposium. These technologies were discussed
from the perspectives of economics; new technical advancements in
science and engineering; and applications. Several papers dealt with
combinations of devices and technologies, leading to a concept of
using a systems approach to particulate control rather than device
control. Additional topic areas included novel control devices, high
temperature/high pressure applications, fugitive emissions,
measurement techniques, and economics and cost analysis.
Each volume of these proceedings contains a set of related
session topics to provide easy access to a unified technology area.
Since the spirit and style of the panel discussion are not
reproducible in print, the initial remarks presented by the panelists
have been included in the volume to which their input to the panel
pertained, in the interest of providing unified technological
organization.
111
-------
CONTENTS
VOLUME III - CONTENTS V
VOLUME I - CONTENTS viii
VOLUME II - CONTENTS xi
Keynote Address
PARTIOJLATE CONTROL TECHNOLOGY AND WHERE IT IS GOING XV
K.E. Yeager
Section A - Economic Comparisons
A COMPARISON OF A BAGHOUSE VS. ESP'S WITH AND WITHOUT GAS
CONDITIONING FOR LOW SULFUR COAL APPLICATIONS 1
W.H. Cole
APPLICATION OF THE BUBBLE CONCEPT TO FUEL BURNING SOURCES AT
A NAVAL INDUSTRIAL COMPLEX 12
C.S Thompson
Section B - Mechanical Collectors
CYCLONE PERFORMANCE: A COMPARISON OF THEORY WITH
EXPERIMENTS 26
J.A. Dirgo, D. Leith
HIGH FLOW CYCLONE DEVELOPMENT 41
W.B. Giles
CYCLONE SCALING EXPERIMENTS 53
W.B. Giles
TEST METHODS AND EVALUATION OF MIST ELIMINATOR CARRYOVER .... 66
V. Boscak, A. Demian
Section C - Coal Characterization
FILTRATION CHARACTERISTICS OF FLY ASHES FROM VARIOUS COAL
PRODUCING REGIONS 81
J.A. Dirgo, M.A. Grant, R. Dennis, L.S. Hovis
FLY ASH FROM TEXAS LIGNITE AND WESTERN SUBBITUMINOUS COAL:
A COMPARATIVE CHARACTERIZATION 97
D.R. Sears, S.A. Benson, D.P. McCollor, S.J. Miller
-------
USE OF FUEL DATABANKS FOR THE EFFECTIVE DESIGN OF STEAM
GENERATORS AND AQC EQUIPMENT 114
N.W. Frisch, T.P. Dorchak
Section D - Inhalable Particulate Matter
DEVELOPMENT OF INHALABLE PARTICULATE (IP) EMISSION FACTORS ... 131
D.L. Harmon
INHALABLE PARTICULATE MATTER RESEARCH COMPLETED BY
GCA/TECHNOLOGy DIVISION 141
S. Gronberg
RESULTS OF TESTING FOR INHALABLE PARTICULATE MATTER AT
MIDWEST RESEARCH INSTITUTE 154
K. Wilcox, F. Bergman, J. Kinsey, T. Cuscino
INHALABLE PARTICULATE EMISSION FACTORS TEST PROGRAMS 166
J.W. Davison,
CHARACTERIZATION OF PARTICULATE EMISSION FACTORS FOR
INDUSTRIAL PAVED AND UNPAVED ROADS 183
C. Cowherd, Jr., J.P. Reider, P.J. Englehart
CONDENSIBLE EMISSIONS MEASUREMENTS IN THE INHALABLE
PARTICULATE PROGRAM 198
A.D. Williamson, J.D. McCain
Section E - Advanced Energy Applications
GAS CLEANING AND ENERGY RECOVERY FOR PRESSURIZED FLUIDIZED
BED COMBUSTION 211
A. Brinkmann, P.M. Kutemeyer
DEMONSTRATION OF THE FEASIBILITY OF A MAGNETICALLY
STABILIZED BED FOR THE REMOVAL OF PARTICULATE AND ALKALI .... 226
L.P. Golan, J.L. Goodwin, E.S. Matulevicius
TEST RESULTS OF A HIGH TEMPERATURE, HIGH PRESSURE
ELECTROSTATIC PRECIPITATOR 241
D. Rugg, G. Rinard, J. Armstrong, T. Yamamoto, M. Durham
COAL-ASH DEPOSITION IN A HIGH TEMPERATURE CYCLONE 256
K.C. Tsao, A. Rehmat, D.M. Mason
DUST FILTRATION USING CERAMIC FIBER FILTER MEDIA — A STATE-
OF-THE-ART SUMMARY — 271
R. Chang, J. Sawyer, W. Kuby, M. Shackleton,
O.J. Tassicker, S. Drenker
vi
-------
HIGH TEMPERATURE AND PRESSURE PARTICULATE FILTERS FOR FLUID
BED COMBUSTION 282
D.F. Ciliberti, T.E. Lippert, O.J. Tassicker, S. Drenker
MOVING BED-CERAMIC FILTER FOR HIGH EFFICIENCY PARTICULATE
AND ALKALI VAPOR REMOVAL AT HIGH TEMPERATURE AND PRESSURE ... 300
D. Stelman, A.L. Kohl, C.A. Trilling
TESTING AND VERIFICATION OF GRANULAR BED FILTERS FOR REMOVAL
OF PARTICULATES AND ALKALIS 318
T.E. Lippert, D.F. Ciliberti, R. O'Rourke
BAGHOUSE OPERATION IN GEORGETOWN UNIVERSITY COAL-FIRED,
FLUIDIZED-BED BOILER PLANT, WASHINGTON, D.C 335
V. Buck, D. Suhre
Section F - Novel Devices
PARTICLE CAPTURE MECHANISMS ON SINGLE FIBERS IN THE PRESENCE
OF ELECTROSTATIC FIELDS 347
M.A. Ranade, F.L. Chen, D.S. Ensor, L.S. Hovis
PILOT DEMONSTRATION OF PARTICULATE REMOVAL USING A CHARGED
FILTER BED 362
P.H. Sorenson
PILOT DEMONSTRATION OF MAGNETIC FILTRATION WITH CONTINUOUS
MEDIA REGENERATION 370
C.E. Ball, D.W. Coy
Section G - Plenary Session
NOVEL PARTICULATE CONTROL TECHNOLOGY 386
S. Masuda
AUTHOR INDEX 406
VII
-------
VOLUME I
FABRIC FILTRATION
Section A - Fabric Filters: Fundamentals
THEORY OF THE TEMPORAL DEVELOPMENT OF PRESSURE DROP
ACROSS A FABRIC FILTER DURING CAKE INITIATION 1
E.A. Samuel
PULSE JET FILTRATION THEORY - A STATE-OF-THE-ART ASSESSMENT. . . 22
R. Dennis, L.S. Hovis
LABORATORY TECHNIQUES FOR DEVELOPING PULSE JET COLLECTORS. ... 37
R.R. Banks, J.T. Foster
OFF-LINE PULSE-JET CLEANING SYSTEM 48
T.C. Sunter
Section B - Fabric Filters; Measurement Techniques
FIELD EVALUATION OF THE DRAG OF INDIVIDUAL FILTER BAGS 62
W.T. Grubb, R.R. Banks
A DUAL-DETECTOR BETA-PARTICLE BACKSCATTER GAUGE FOR MEASURING
DUST CAKE THICKNESS ON OPERATING BAG FILTER AND ESP UNITS. ... 77
R.P. Gardner, R.P. Donovan, L.S. Hovis
MIT FLEX ENDURANCE TESTS AT ELEVATED TEMPERATURE 91
J.T. Foster, W.T. Grubb
THE ONE-POINT IN-SITU CALIBRATION METHOD FOR USING A BETA-
PARTICLE BACKSCATTER GAUGE FOR CONTINUOUSLY MEASURING DUST
CAKE THICKNESS ON OPERATING BAG FILTER AND ESP UNITS 107
R.P. Gardner, R.P. Donovan, L.S. Hovis
Section C - Fabric Filters: Coal Fired Boilers
PULSE-JET FABRIC FILTER EXPERIENCE USING NON-GLASS
MEDIA AT AIR TO CLOTH RATIOS OF 5 TO 1 ON A PULVERIZED
COAL FIRED BOILER • 121
G. Pearson, D.D. Capps
START-UP AND OPERATION OF A FABRIC FILTER CONTROLLING
PARTICULATE EMISSIONS FROM A 250 MW PULVERIZED COAL-FIRED
BOILER 132
C.B. Barranger, N. Spence, J. Saibini
Vlll
-------
VOLUME I CONTENTS (Cont.)
PERFORMANCE OF A 10 MW FABRIC FILTER PILOT PLANT AND
COMPARISON TO FULL-SCALE UNITS 148
W.B. Smith, K.M. Gushing, R.C. Carr
THE DESIGN, INSTALLATION, AND INITIAL OPERATION OF THE W.H.
SAMMIS PLANT, UNIT 3 FABRIC FILTER 164
D.R. Ross, J.R. Howard, R.M. Golightley
RESULTS FROM THE FABRIC FILTER EVALUATION PROGRAM AT
COYOTE UNIT #1 179
H.J. Peters, A.A. Reisinger, W.T. Grubb, M. Lewis
BAGHOUSE PERFORMANCE AND ASH CHARACTERIZATION AT THE
ARAPAHOE POWER STATION 192
R.S. Dahlin, D.R. Sears, G.P. Green
AN EVALUATION OF FULL-SCALE FABRIC FILTERS ON UTILITY
BOILERS 210
J.W. Richardson, J.D. McKenna, J.C. Mycock
STATUS OF SPS INVESTIGATION OF HARRINGTON STATION UNIT 2
FABRIC FILTER SYSTEM 226
R. Chambers, D. Harmon
UPDATE OF SPS PILOT BAGHOUSE OPERATION 239
R. Chambers, S. Kunka, D. Harmon
THE USE OF SONIC AIR HORNS AS AN ASSIST TO REVERSE AIR
CLEANING OF A FABRIC FILTER DUST COLLECTOR 255
A. Menard, R.M. Richards
Section D - Fabric Filters: Electrostatic Enhancement
ELECTROSTATIC STIMULATION OF REVERSE-AIR-CLEANED
FABRIC FILTERS 287
D.A. Furlong, G.P. Greiner, D.W. VanOsdell, L.S. Hovis
ELECTRICAL STIMULATION OF FABRIC FILTRATION: ENHANCEMENT BY
PARTICLE PRECHARGING 303
G.E.R. Lamb, R. Jones, W. Lee
ESFF AS A FIELD EFFECT 316
L.S. Hovis, G.H. Ramsey, R.P. Donovan
ELECTRICAL ENHANCEMENT OF FABRIC FILTRATION: PRECHARGING
VS. BAG ELECTRODES 327
R.P. Donovan, L.S. Hovis, G.H. Ramsey
PERMEABILITY OF DUST CAKES COLLECTED UNDER THE INFLUENCE OF
AN ELECTRIC FIELD 342
D.W. VanOsdell, R.P. Donovan, D.A. Furlong, L.S. Hovis
ix
-------
VOLUME I CONTENTS (Cont.)
Section E - Fabric Filters: Practical Considerations
HIGH VELOCITY FABRIC FILTRATION FOR INDUSTRIAL COAL-FIRED
BOILERS 357
G.P. Greiner, S. Delaney, L.S. Hovis
OPTIMIZING THE LOCATION OF ANTI-COLLAPSE RINGS IN FABRIC
BAGS 382
J. Musgrove
PULSE JET ON-LINE CLEANING FILTER FOR FLY ASH 420
W.G. Wellan
TOP INLET VERSUS BOTTOM INLET BAGHOUSE DESIGN 431
R.M. Jensen
UPGRADE OF FLY ASH COLLECTION CAPABILITY AT THE CROMBY
STATION 446
T.J. Ingram, R.J. Biese, R.O. Jacob
HIGH SULFUR FUEL/ FABRIC FILTER STARTUP EXPERIENCE 460
P. Hanson, L. Adair, R.N. Roop, R.B. Moyer
FUNDAMENTAL STRATEGIES FOR CLEANING REVERSE AIR BAGHOUSES. ... 482
M. Ketchuck, M.A. Walsh, O.F. Fortune,
M.L. Miller, M. Whittlesey,
Section F - Dry Scrubbers
DESIGN CONSIDERATIONS FOR BAGHOUSE - DRY SO, SCRUBBER
SYSTEMS 494
O.F. Fortune, R.L. Miller
RESULTS OF BAGHOUSE AND FABRIC TESTING AT RIVERSIDE 506
H.W. Spencer III, Y.J. Chen, M.T. Quach
REACTIVITY OF FLY ASHES IN A SPRAY DRYER/FABRIC FILTER FGD
PILOT PLANT 521
W.T. Davis, R.E. Pudelek, G.D. Reed
Section G - Plenary Session
FABRIC FILTRATION - AS IT WAS, HAS BEEN, IS NOW
AND SHALL BE 536
E.R. Frederick
AUTHOR INDEX 551
-------
VOLUME II
ELECTROSTATIC PRECIPITATION
Section A - Industrial Applications
MODELING OF WET BOTTOM AGITATOR SYSTEMS FOR ELECTROSTATIC
PRECIPITATORS ON RECOVERY BOILERS 1
M.A. Sandell, R.R. Crynack
DESIGN AND PERFORMANCE OF ELECTROSTATIC PRECIPITATORS
UTILIZING A NEW RIGID DISCHARGE ELECTRODE DESIGN 17
G.R. Gawreluk, R.L. Bump
DEVELOPMENT AND EVALUATION OF NEW PRECIPITATOR EMITTER
ELECTRODE 35
R.L. Adams/ P. Gelfand,
INDUSTRIAL APPLICATIONS OF TWO STAGE TUBULAR ELECTROSTATIC
PRECIPITATORS 51
H. Surati, M.R. Beltran
Section B - Advanced Technology
PILOT DEMONSTRATION TWO-STAGE ESP TEST RESULTS 65
P. Vann Bush, D.H. Pontius
EVALUATION OF PRECHARGERS FOR TWO-STAGE ELECTROSTATIC
PRECIPITATORS 84
G. Rinard, D. Rugg, M. Durham
INITIAL EXPERIMENTS WITH AN ELECTRON BEAM PRECIPITATOR TEST
SYSTEM 96
W.C. Finney, R.H. Davis, J.S. Clements, E.G. Trexler,
J.S. Halow, 0. Tokunaga
EXPERIMENTS WITH WIDE DUCTS IN ELECTROSTATTC PRECIPITATORS ... Ill
E. Weber
A RECONCILIATION: WIDE VERSUS NARROW SPACED COLLECTING
PLATES FOR PRECIPITATORS 126
D.G. Puttick
PULSE CORONA AS ION SOURCE AND ITS BEHAVIORS IN MONOPOLAR
CURRENT EMISSION 139
S. Masuda, Y. Shishikui
XI
-------
VOLUME II CONTENTS (Cont.)
Section C - Fundamentals
A NEW CORRECTION METHOD OF MIGRATION VELOCITY IN DEUT3CH
EFFICIENCY EQUATION FOR CONVERSION OF ELECTROSTATIC PRECIPITATOR
SIZING FROM A PILOT-SCALE TO FULL-SCALE 154
F. Isahaya
DISTORTION OF PULSE VOLTAGE WAVE FORM ON CORONA WIRES DUE TO
CORONA DISCHARGE 169
S. Masuda, H. Nakatani
ELECTROSTATIC PRECIPITATOR ANALYSIS AND SYNTHESIS 184
T. Chiang, T.W. Lugar
COMPUTER MODEL USE FOR PRECIPITATOR SIZING 194
G.W. Driggers, A.A. Arstikaitis, L.A. Hawkins
IMPROVEMENTS IN THE EPA/SRI ESP PERFORMANCE MODEL 204
M.G. Faulkner, R.B.Mosley, J.R. McDonald, L.E. Sparks
NUMERICAL SIMULATION OF THE EFFECTS OF VELOCITY FLUCTUATIONS
ON THE ELECTROSTATIC PRECIPITATOR PERFORMftNCE 218
E.A. Samuel
CORONA - INDUCED TURBULENCE 230
M. Mitchner, G.L. Leonard, S.A. Self
VELOCITY AND TURBULENCE FIELDS IN NEGATIVE CORONA
WIRE-PLATE PRECIPITATOR 243
H.P. Thomsen, P.S. Larsen, E.M. Christensen,
J.V. Christiansen
THE EFFECT OF TURBULENCE ON ELECTROSTATIC PRECIPITATOR
PERFORMftNCE 261
D.E. Stock
FACTORS LEADING TO ELECTRICAL BREAKDOWN OF RESISTIVE DUST
LAYERS AND SUSTAINED BACK CORONA 271
P.A. Lawless, L.E. Sparks
ELECTRICAL BREAKDOWN OF PARTICULATE LAYERS 288
G.B. Moslehi, S.A. Self
ELECTROMECHANICS OF PARTICULATE LAYERS 306
G.B. Moslehi, S.A. Self
LATERAL PROPAGATION OF BACK CORONA IN TWIN-ELECTRODE TYPE
PRECIPITATORS 322
S. Masuda, T. Itagaki
FIRST MEASUREMENTS OF AEROSOL PARTICLE CHARGING
BY FREE ELECTRONS 337
J.L. DuBard, M.G. Faulkner, L.E. Sparks
xii
-------
VOLUME II CONTENTS (Cont.)
Section D - Operation & Maintenance
GAS FLOW DISTRIBUTION MODEL TESTING 349
D.R. Cook, J.M. Ebrey, D. Novogoratz
AIR FLOW MDDEL STUDIES 369
L.H. Bradley
COLLECTING ELECTRODE RAPPING DESIGNED FOR HIGH EFFICIENCY
ELECTRIC UTILITY BOILER ELECTROSTATIC PRECIPITATORS 384
A. Russell-Jones, A.P. Baylis
ELECTROSTATIC PRECIPITATOR AND FABRIC FILTER OPERATING AND
MAINTENANCE EXPERIENCE 401
P.R. Goldbrunnerf W. Piulle
Section E - Conditioning
ECONOMICAL FLY ASH COLLECTION BY FLUE GAS CONDITIONING 416
E.L. Coef Jr.
EXPERIENCES AT DETROIT EDISON COMPANY WITH DECLINING
PERFORMANCE OF SULFUR TRIOXIDE FLUE GAS CONDITIONING
EQUIPMENT 430
L.A. Kasik, W.A. Rugenstein, J.L. Gibbs
ESP CONDITIONING WITH AMMONIA AT THE MONROE POWER PLANT OF
DETROIT EDISON COMPANY 444
E.B. Dismukes, J.P. Gooch, G.H. Marchant, Jr.
FLY ASH CHEMISTRY INDICES FOR RESISTIVITY AND EFFECTS ON
ELECTROSTATIC PRECIPITATOR DESIGN AND PERFORMANCE 459
H.J. Hall
Section F - Control Systems
A NEW ENERGIZATION METHOD FOR ELECTROSTATIC PRECIPITATORS
MITSUBISHI INTERMITTENT ENERGIZATION SYSTEM 474
T. Ando, N. Tachibana, Y. Matsumoto
SOME MEASURED CHARACTERISTICS OF AN ELECTROSTATIC
PRECIPITATOR OBTAINED USING A MICROCOMPUTER CONTROLLER 489
M.J. Duffy, T.S. Ng, Z. Herceg, K.L. McLean
ELECTROSTATIC PRECIPITATOR ENERGIZATION AND CONTROL SYSTEMS . . 499
K.M. Bradburn, K. Darby
Xlll
-------
VOLUME II CONTENTS (Cont.)
APPLYING MODULAR MICROCOMPUTER CONTROL ELEMENTS IN A
PRBCIPITATOR CONTROL SYSTEM 521
I.M. Wexler
Section G - Plenary Session
THE CURRENT STATUS, FUTURE DIRECTIONS, AND ECONOMIC
CONDITIONS IN THE APPLICATION OF ESP'S 534
S. Oglesby
AUTHOR INDEX 539
xiv
-------
PARTICULATE CONTROL TECHNOLOGY AND WHERE IT IS GOING
by: K. E. Yeager
Electric Power Research Institute
Palo Alto, California 94303
ABSTRACT
This keynote address underscores the key role of particulate control
technology in any practical strategy for reducing the emissions associated
with coal utilization. Its importance results from the long-standing and
successful cooperative efforts among user, supplier, and government to achieve
control methods which are as reliable, simple and low cost as possible.
Opportunities are discussed for capitalizing on this established and accepted
base to solve the current and emerging set of air pollution issues facing the
utility industry.
INTRODUCTION
Thank you Mr. Chairman, ladies and gentlemen. I particularly welcome this
opportunity to speak to you because I believe particulate control must be the
foundation for a rational emission control strategy. Before we examine where
particulate control technology is going, let us look at where it has been.
Nearly 100 years ago particulate control began in earnest to resolve the
obvious smoke emissions of heavy industry. A cooperative effort over the
ensuing years among user, supplier, and government has brought particulate
control to be an accepted, integral part of our industrial and combustion
processes .
This development has been successful because it never forgot that it was
the function of particulate control not just to clean the stack but to do so
in as reliable, simple and least costly manner as possible. In other words,
good industrial engineering practice guided the emission control effort as it
would all other aspects of commercial process development.
During the 1970s, the politicization of the environment, in my judgement,
led us away from building on this successful foundation. Instead, alterna-
tives sacrificing good commercial practice were imposed in return for promise
of maximum theoretical emission control efficiency. Legislative and adminis-
trative decisions were made which ignored the technical principles necessary
xv
-------
for successful control technology application. An adversial relationship
rapidly developed between user, supplier and government in which lawyers, not
scientists and engineers, were responsible for defining the state of control
technology. I submit that this results in the very antithesis of progress by
installing a psychology of contradiction not cooperation.
REQUIREMENTS
The basic issue in control requirements as they have evolved over the
last 25 or 30 years has been the dramatic increase in removal efficiency
demanded of particulate control. In the 1950s and even 1960s, control
requirements were in the 90 percent range. Today we're two orders of mag-
nitude higher in requirement. As a result, we are as close to zero particu-
late discharge as is measurable. This has forced a revolutionary change in
our control considerations . Where we were once well within the state of the
art in particulate control, we have now exceeded the state of the art in
precipitators . We are also evaluating fabric filtration because of its
potential cost and performance advantages. In summary, we are going from a
technology which is widely available and well known, but marginal in perfor-
mance, to a technology that promises a performance margin. However, its
current design base and reliability in the utility industry are not well
established.
As we go into the future, we are faced with toxic substances control
standards demanding not just a relative level of control, but absolute
control. This emerging requirement has also been connected to fine particu-
late. We are also challenged by a major secondary particulate issue, which
has changed the arena of argument for the application of flue gas
desulfurization and NOV control.
x.
From an engineering standpoint, we also have to be aware of some of the
control issues that may not be reflected in removal efficiency, but will have
a major impact on the way we design and operate equipment. First, pre-
construction review requires agreement on the level of control capability for
all pollutants prior to commencing construction. This may mean agreement not
just between the regulatory agency and the utility, but with interveners as
well.
Second, and perhaps more important, operation and maintenance standards
will require the technology to meet a predetermined level of reliability in
order for the power plant to be permitted to operate at all.
What are some of the issues facing the utility industry which must guide
our actions? These include, in addition to environmental requirements:
- uncertain petroleum availability
skyrocketing fuel prices
- declining demand growth rates
restricted capital investment capability
- loss of public confidence in the nuclear initiative
- withdrawal of government R&D support
rapidly increasing electrical rates.
xv i
-------
Several factors in the utility response to these issues seem clear:
1. Electric generation will depend increasingly on coal
2. Priority will be placed on improving the reliability and longevity of
existing generating capacity. This includes reducing sensitivity to
fuel quality.
3. The utility user will assume greater leadership for the selection,
development and commercialization of technology.
4. U.S. supplier priority is likely to focus more and more on competing
in the international market to maintain market size.
What does all this mean for the particulate control technology developer?
ELECTROSTATIC PRECIPITATION—This has long been the backbone of environ-
mental control in the utility industry. If we look at the impact of changing
control requirements from a precipitator standpoint, we would find that the
demand for increase in control capability has led to about a fivefold increase
in electrostatic precipitator size and at least a threefold increase in cost
in constant 1981 dollar terms. We've gone from a low-cost technology,
operating well, to a high-cost technology, often operating not very well.
More and more the public perception of environmental acceptability is
based on a clear stack, irrespective of regulatory requirements. This is also
becoming an industry measure and one I might add which has the positive
attention of the utility industry. It means that retrofit measures must be
advanced to reduce the sensitivity of the high efficiency precipitator to
changing fuel quality and resulting ash characteristics . Chemical condi-
tioning therefore has become one important solution to this issue and one
which has advanced from a black art to real engineering credibility. Second,
as reduced economic strength and slowed electricity demand growth reduces
capacity expansion, we must look even more to increasing the long-term
reliability of our equipment to last not just 30-40 years but 50 years and
more. Third, a revitalized effort to advance the technology of ESP is needed
to maintain its competitiveness in the face of ever more stringent emission
control requirements.
FABRIC FILTERS—In the face of very stringent standards and the desire to
maintain a clear stack under all conditions, fabric filters are rapidly
becoming a competitive, if not preferred, alternative for new plant applica-
tions. This is creating a healthy technical competition between fabric
filters and precipitators .
Although in use for many years, we have, however, found major
opportunities for improving the reliability, operability and cost of this
technology as it applies to the utility industry. These opportunities
encompass among others; improved aerodynamic design, bag materials, cleaning
frequency and method, optimized dust loading on bags, and electrostatic
enhancement.
xvii
-------
INTEGRATED EMISSION CONTROL—-If we look at the growth in environmental
control requirements on power plants since 1965 we find revolutionary tech-
nical and economic changes. In 1965 a precipitator was the primary control
device. Its cost was about 10% of a new plant. Today, typically 40% of the
cost of a new plant is for environmental control. By 1985 these requirements
provide the opportunity for the emission control technology tail to wag the
power plant, not only in dimensions, but also in investment.
This forces us to take a hard look at our conventional way of viewing
emission control. We can't afford to deal with each control requirement as an
individial black box. Rather, we have to integrate them in a systematic way
that provides simplification and improved reliability. Integrated Emission
Control (IEC) is, therefore, more than anything, a state of mind. As the cost
of meeting environmental regulations on a new coal fired power plant increases
to 40% or more of a billion dollar investment, I think we are all faced with
the realization that business as usual, i.e., adding piece meal more and more
auxiliary control devices is a bankrupt approach. Fundamentally, IEC means
elevating the engineering priority of environmental control to a level
equivalent to its economic importance.
Nowhere, in my judgement, are the opportunities for particulate control
technology greater. Issues which are accelerating the need for this approach
are closely associated with the total atmospheric loading of pollutants.
These include, for example, visibility/long-range transport and acid
precipitation.
Acid precipitation is the latest and most politically potent in a series
of issues that may require expanded retrofit emission control. It is
therefore imperative from a cost and practicality standpoint that we base any
resulting retrofit strategy on the existing particulate control capability.
Coal cleaning is important but can only reduce S02 emissions by about 2
million tons per year. Coal switching is likely to be politically, if not
economically, limited. Therefore, many companies in the industrial midwestern
states may be required to create a technology to achieve 40-60% SOo removal
utilizing dry removal in conjunction with existing particulate control
capabilities. Failure to do so may add billions of dollars per year to the
local cost of control.
As an example of the IEC approach, our tests indicate a strong correla-
tion between reduced NOX formation and reduced fine particulate formation.
This leads EPRI to urge combustion control opportunities which may cost-
effectively combine these two effects. Such an approach begins to integrate
particulate control with the combustion process itself.
NEW TECHNOLOGY
The present utility conditions present an opportunity to carry IEC one
step further as the utility industry actively develops and commercializes new
coal utilization technology for its next generation of plants. Foremost among
these technologies, in my judgement, is fluid bed combustion which provides an
evolutionary improvement in our use of coal. Specific advantages include:
xvi 11
-------
Increased fuel flexibility
- Less cost sensitivity to size
- Integral environmental control - except particulate.
A 100-200 MWe utility demonstration of this new steam generating tech-
nology is planned by EPRI and the utility industry for operation this
decade. This will be based on the 20 MWe TVA/EPRI engineering prototype now
successfully operating at Paducah, Kentucky.
In the pressurized form, fluid bed combustion has growth potential to
provide several additional advantages including:
Shop fabricated, barge transportable, standardized power modules for
rapidly and cost effectively adding generating capacity.
- Lowest busbar energy cost of any coal option
The key to its success, however, is reliable high pressure, high tempera-
ture filtration to insure gas turbine reliability and emission compliance.
This is an area demanding concentrated effort by the development community.
EPRI'S ROLE
It has been EPRI's role to help fill the void surrounding determination
of the cost, reliability and operability issues affecting required or proposed
control technology. Through the vehicle of large-scale tests, particularly at
our Arapahoe Test Facility and the new Shawnee AFBC Facility, we endeavor to
manage and limit the level of risk associated with commercial innovation. We
include efforts to:
- Determine under actual utility operating conditions the factors which
may limit practical application of controls.
- Where appropriate, help resolve these issues,
- Determine objectively the status of new technology.
- Avoid large and expensive commercial failures which may cost large
sums and kill otherwise promising technologies.
- Train utility technicians and operators .
- Provide a credible and realistic data base for establishing technical
policy on environmental control.
CONCLUSIONS
I applaud EPA for working to reconstruct the former spirit of cooperation
necessary for the advancement of environmental control technology. Having
just returned from Europe, I am even more convinced that throughout the
Western World and Japan, environmental progress must be made within the
fundamental economic and productivity strength of our societies.
xix
-------
Within this context are three principles which I believe are crucial to
successful advancement of environmental control.
1. The affected industry must be given greater freedom in selecting the
means of control. Over the past decade industry has too often found
itself reacting to legislative determinations of technology status
without constructive opportunity to participate in this determina-
tion. The regulator and regulated must work together in the tech-
nical arena without counterproductive administrative and legal
constraints.
2. Conversely, government must accept at least joint responsibility for
resolving with industry the practical issues of reliability, cost and
operability which limit productive use of new control technology.
These issues will not be effectively reduced by simply dumping them
in the lap of industry.
3. Particulate control technology remains the key to practical control
of the full set of emissions affecting coal utilization. The
challenge to all of you is tc capitalize on this established and
accepted base and aggressively provide the technical leadership and
ingenuity necessary to achieve practical solutions. EPRI and the
utilities look forward to working with you in this endeavor.
Thank you for your kind attention and my very best to you all.
xx
-------
A COMPARISON OF A BAGHOUSE VS. ESP'S WITH AND
WITHOUT GAS CONDITIONING FOR LOW SULFUR COAL APPLICATIONS
by: William H. Cole
Gibbs & Hill, Inc.
New York, N.Y. 10001
ABSTRACT
The new source emission standard of 0.03 lbs/10^ Btu for electric
utilities suggests that the selection of particulate removal equipment
will increasingly favor the baghouse as compared to conventional ESP's
for low sulfur coal applications. This paper investigates a third
alternative of a relatively small ESP used in conjunction with 303 gas
conditioning for new generating units, which typically require
efficiency levels of 99.70 percent or higher. The three alternatives
are compared for a 500 MW unit burning low sulfur western coal.
Emphasis is placed on the comparable economics of investment, and
present worth of annual costs including fixed charges, incremental
energy, bag replacement, sulfur feed stock, and maintenance. Cost
sensitivity is illustrated for assumed escalation rates from zero to 10
percent. A preliminary review indicates that ESP's in conjunction with
gas conditioning, offer an attractive alternative to a conventionally
sized ESP or baghouse, and may restore the dominance of ESP's in
equipment selection.
INTRODUCTION
During the past few years, there has been a significant decline in
the use of electrostatic precipitators (ESP's) as compared to an
increasing use of the baghouse by the electric utility industry. This
has resulted from a number of factors as follows:
(1) Typical efficiency levels in the 99.7 to 99.8 percent range
are generally required to satisfy the current maximum emission
regulation of 0.03 lbs/10 Btu for new generating units.
(2) The increased use of low sulfur coal, particularly of the
western variety, requires relatively large ESP's.
(3) A broad range of critical coal and ash characteristics is often
used in the precipitator equipment specification for
performance guarantees. This clearly tends to result in an
overly conservative design based on the "worst" coal.
1
-------
The combined effect of the above factors has made the baghouse cost
competitive with the precipitator for low sulfur coal applications, with
the added benefit of performance that is essentially independent of
normal variation in the coal characteristics. The trend to the use of
baghouse is substantiated by statistics from one vendor which indicate
that the baghouse obtained 14 and 17 percent of the electric utility
megawatt orders in 1979 and 1980, respectively. Even more indicative
are Industrial Gas Cleaning Institute statistics which show that the
baghouse market share for all flyash collectors increased from 24 to 45
percent of total dollar bookings during the same period (1).
There is now another option deserving of full consideration in the
process of equipment selection for new generating units. This is the
use of a relatively small ESP equipped with a sulfur burner type gas
conditioning system which introduces small quantities of sulfur trioxide
into the flue gas as the conditioning agent. Although the effectiveness
of sulfur conditioning has been known for many years, there was
insufficient demand to justify commercialization of the equipment design
prior to the Clean Air Act of 1970. When intially developed in the
early 1970's, the system appeared to have limited application to the
upgrading of substandard performance of existing precipitators designed
for relatively low efficiency, and subsequently being used to collect a
marginal or high resistivity fly ash. However, in recent years there
have been major improvements in both the conditioning process and
equipment design. The experience with more than 100 installations to
date indicates that the equipment meets all of the criteria of
automation, reliability, and availability required by the electric
utility application. On this basis, it is logical to evaluate the use
of a relatively small ESP designed for use with gas conditioning for a
new generating unit. Although systems have been designed with this
objective, this paper summarizes the results of a technical and economic
comparison of this alternative with a full size precipitator and bag-
house for a hypothetical 500 MW unit burning a low sulfur western coal.
EQUIPMENT DESCRIPTION
The only equipment requiring description is the gas conditioning
system, and space limitations require that this be brief. Suffice it to
say that numerous publications are available with details of the process
and equipment, whereas the objective of this paper is a comparison of
the technical and economic aspects of the system when used with an ESP
as compared to other alternatives.
The gas conditioning equipment required for a new 500 MW generating
unit comprises a liquid sulfur storage tank of nominal 100 ton capacity,
a metering pump skid including the unloading pump, and a single
burner/converter skid to serve both precipitators. As noted previously,
of salient importance is the control system and level of automation*
Assuming the sulfur is in the liquid state at startup, a push button
activates the electric ambient air heaters to raise the temperature of
-------
-the vanadium pentoxide catalyst to the required process temperature of
800 F. This requires approximately 4 hours from a cold start at which
point a second push button activates the sulfur feed to the burner/
converter skid for controlled feed of sulfur trioxide to the ESP inlet
probes* The feed rate is optimized following installation of the
system, and is automatically regulated by a boiler signal to maintain
the optimum injection rate over the total range of boiler load. If
desired, the feed will automatically shut off at a predetermined boiler
load at which point gas conditioning is no longer required because of
reduced gas flow rate to the ESP, and the system has a turn down ratio
of at least 20 to 1 in the event of an upset condition.
The equipment causes minimal problems with general arrangement,
particularly with a new generating unit. The sulfur storage tank and
pump skid can be remotely located, and only the converter skid should be
located as closely as possible to the ESP inlet ducts to minimize the
stainless steel piping to the inlet probes. The system also causes no
constraints on ductwork layout since only one second of treatment time
is required to condition the ash upstream of the precipitator inlet
flange.
SYSTEM DESIGN PARAMETERS
The system design and performance requirements are summarized as follows;
Unit Size 500 MW
Gas Flow Rate 1,800,000 ACFM
Gas Temperature 300 F
Heat Input 5000 MMBtu/hr
Coal Firing Rate 625,000 Ibs/hr
Maximum Emission 0.03 Ibs/MMBtu
Inlet Grain Loading 3.28 GR/ACF
Guarantee Efficiency 99.70 percent
Maximum Outlet Loading 0.010 GR/ACF
The above requirements were based on the coal and ash analyses in
Table 1 which represent a composite of several Wyoming coals from the
Powder River Basin.
TABLE 1. DESIGN COAL AND ASH ANALYSES
Coal
Moisture
Carbon
Ash*
Hydrogen
Nitrogen
Oxygen
Sulfur
Heating Value
Percent
28.0
46.0
9.0
3.4
0.8
12.3
0.5
8000 Btu/lb
Fly Ash
Silica
Aluminum Oxide
Iron Oxide
Calcium Oxide
Magnesium Oxide
Potassium Oxide
Sodium Oxide
Titanium Oxide
Sulfur Trioxide
Percent
35.0
19.0
5.5
22.0
4.4
0.4
1.0
1.0
11.3
*Assume 90 percent ash to flyash.
-------
EQUIPMENT DESIGN
In an effort to achieve objectivity, the sizing and design of all
equipment was obtained from experienced vendors* Since there is often
more disparity in baghouse design among suppliers/ this information was
obtained from two vendors, and it lends credence to the technical
comparison to note that both designs were virtually identical.
PRECIPITATOR DESIGNS
A rigid frame precipitator with hammer rapping was selected for the
full sized unit not equipped with gas conditioning. This design appears
appropriate for a high resistivity ash application based on adequate
rapping intensity, and reliability of the semi-rigid discharge
electrodes of the scalloped ribbon or twisted square type.
A mast type electrode design that can be used with overhead hammer
or magnetic impulse rapping was selected for the unit to be equipped
with gas conditioning. This is also appropriate due to the relatively
small precipitator size which can create problems with electrical
sectionalization with a rigid frame design.
A comparison of the basic design parameters is shown in Table 2.
TABLE 2. ESP DESIGN PARAMETERS
(One ESP of 2 Required)
Gas Flow @ 300 F (acfm)
Efficiency (%)
No. of Gas Passages (12 in.)
Plate Height (ft)
Effective Duct Length (ft)
Collecting Area (sq ft)
Gas Velocity (ft/sec)
Bus Sections (Series/Parallel)
Connected Ma/ 1000 sq ft
Connected Watts/sq ft
SCA (sq ft/ 1000 acfm)
Migration Velocity Wk (ft/ sec)
Migration Velocity W (ft/sec)
Rigid Electrode
(with Conditioning)
900,000
99.70
108
36
45
349,920
3.86
5/4
51
2.83
389
1.447
0.249
Rigid Frame
(without Conditioning)
900,000
99.70
80
45
87.6
629,856
4.17
6/4
38
2.10
700
0.804
0.138
The most significant difference between the two designs is obvious-
ly the SCA's (or design migration velocities) which indicate that the
precipitator equipped with gas conditioning has only 56 percent as much
collecting plate area as compared to the full size ESP. It may be
coincidental, but it is of interest to note that both SCA's agreed with
-------
that of the writer within less than 1.5 percent. The question may arise
as to a procedure for sizing precipitators designed specifically for use
with gas conditioning. A good guideline is to consider the application
as though it were a 2 percent sulfur coal. Although the latter can be
adjusted by the sulfur trioxide injection rate, a 2 percent sulfur is
generally about optimum for stability of electrical operation without
undue concern about re-entrainment.
BAGHOUSE DESIGN
The baghouse design provided by two equipment suppliers was
virtually identical, and is summarized in Table 3. Of the design
features shown, the only one specified for this paper was the use of
reverse air cleaning.
TABLE 3. BAGHOUSE DESIGN PARAMETERS
(One Baghouse of Two Required)
Gas Flow/Baghouse @ 300 F (acfm) 900,000
Efficiency (%) 99.80
No. of Compartments/Baghouse 12
No. Bags/Comp't. (3 Bag Reach) 393
Cloth Area/Baghouse {sq ft) 488,106
Bag Material (12 in. x 35 ft Bags) Glass
Bag Cleaning Reverse Air
Reverse Air/Cloth Ratio 1.75:1
Gross Air/Cloth Ratio (12 Comp'ts.) 1.84:1
Net Air/Cloth Ratio (10 Comp'ts.) 2.38:1
GAS CONDITIONING SYSTEM DESIGN
The sulfur burner gas conditioning system comprises a 100 ton
liquid sulfur storage tank, a metering pump skid, a single sulfur
burner/converter skid serving both precipitators, and two sets of sulfur
trioxide injection probes. The sulfur burner is rated at 300 Ibs/hr
with an expected maximum use rate of 190 Ibs/hr to provide a sulfur
trioxide injection rate of 25 ppm by volume. Maximum electrical
requirement is 200 kw.
ECONOMIC FACTORS
The assumed economic parameters are critical to the cost
comparison, and were obtained from a number of sources. The factors
include those typically provided to Gibbs & Hill by the electric
utilities for economic studies, such as plant life, fixed charges,
capacity charge, and interest rates. Other factors such as incremental
-------
energy cost, bag life, and cost of liquid sulfur were obtained
specifically for the requirements of this study. The economic factors
used as a basis for the cost comparison are summarized in Table 4.
TABLE 4. ECONOMIC FACTORS
Plant Life 30 yrs.
Capacity Charge $1,000/kw
Fixed Charge Rate 17 percent
R.O.I. For Present Worth 10 percent
Incremental Energy Cost $0.025/kwhr
Avg. ESP Pressure Loss 1.0 in wg
Avg. Baghouse Pressure Loss 4.5 in wg
Steam Cost (Sat. @ 45 psig) $0.003/lb
Liquid Superbrite Sulfur $160.00/ton
Bag Life 3 yrs.
Annual Maintenance Costs:
ESP's (% of Material Cost) 5.0 percent
Baghouse (% of Material Cost) 2.5 percent
Interest Rate 10 percent
Escalation Rate* 0, 5, & 10 percent
Operating Hours/Yr 8200/yr
Avg. Load Factor 75 percent
*Escalation is applied to cost of energy, sulfur, steam, replacement
filter bags, and maintenance.
Several of the above parameters which are critical to this study
are based on the following:
(1) Incremental energy charge is based on a G&H estimate for the
use of low sulfur western coal in the midwest region.
(2) A fabric filter bag life of 3 years was suggested by
equipment suppliers based on the economics of replacement
of an entire compartment at one time.
(3) The cost of liquid "superbrite" sulfur is based on two recent
price quotations including delivery to a Gulf Port. It
also includes a nominal $20 per ton inland freight charge to
the plant destination.
(4) Maintenance costs were the single most difficult cost to
assess. Utility records will sometimes include routine shift
inspection, or only special maintenance requirements. There
is insufficient experience to estimate routine maintenance
costs on a baghouse. However, it was assumed to be one half
that for a precipitator as a percentage of material cost. The
assumption of 5 percent of material cost for the gas condi-
tioning system was within $14,000 per year when compared to
detailed records on one specific installation.
-------
COST COMPARISON
The overall cost comparison of the various alternatives is made on
the basis of investment cost, and present worth of annual cost assuming
0, 5, and 10 percent escalation. In my opinion, the initial investment
cost and present worth of annual costs with zero escalation is
particularly pertinent. However, the extension of these costs to
include assumed rates of escalation provides a sensitivity analysis which
is of value in interpreting the effect of escalation on critical annual
costs.
INVESTMENT COSTS
The investment costs for the three alternatives were obtained from
the vendors based on a semi-turnkey installation. The costs include all
flange to flange material and auxiliaries including erection, with the
exception of ductwork and the ash handling system. However,
differentials in these items were examined and included in the investment
cost.
Costs for the ESP's were in accordance with the past experience of
Gibbs & Hill. However, there was a discrepancy of approximately 10
percent in the baghouse flange to flange erected cost provided by the two
vendors. This differential was minimized based on the overall semi-
turnkey price, and the two costs were averaged. It is also noted that
the average cost used for this comparison is still significantly less
than comparable costs obtained for a baghouse installation a year ago.
The investment costs for a turnkey
alternatives are summarized in Table 5.
installation of the three
TABLE 5. COMPARISON OF INVESTMENT COST (OOP's OF $)
Investment Costs
ESP/Baghouse (Material)
ESP/Baghouse (Erection)
Auxiliaries (Installed)*
Gas Cond. Syst. (Installed)
Installed System Cost
Capacity Charge
ESP/Baghouse
Gas Cond. System
Total Investment Cost
Investment Cost Diff.
ESP w/Cond.
$ 3,658
2,940
4,969
1,980
$13,547
$ 2,334
261
$16,142
Base
ESP w/o Cond.
$ 6,280
5,310
7,172
-
$18,762
$ 2,502
-
$21,264
$ 5,122
Baghouse
$ 5,772
3,975
4,981
-
$14,728
$ 3,683
-
$18,411
$ 2,269
*Includes ESP nozzles and duct manifolds, all support steel and
insulation, accessways, low voltage wiring, hopper heaters & level
alarms, and ash handling connections. Baghouse also includes interior
insulation (top & bottom), and bypass duct.
-------
As noted below the table, the scope of supply approaches the cost
of a turnkey installation. The initial intent in the cost comparison
was to eliminate considerations of ductwork and ash handling as being
base to both the precipitator and baghouse systems. However, it became
apparent that both of these factors are inherently favorable to the
baghouse, and cost adjustments were made as follows:
(1) A terminal point was assumed for ductwork at the entrance and
exit of the precipitators and baghouse. No cost was accessed
for any ductwork from these terminal points to the inlet and
outlet flanges of the baghouse. In the case of the ESP's, an
inlet manifold connecting four inlet nozzles, and a similar
outlet manifold was assumed without nozzles. The cost of the
manifolds, nozzles, support steel and insulation required for
the precipitators exceeded $1,000,000, and is included in the
investment cost for precipitators in Table 5.
(2) It was further recognized that ash handling is generally more
costly for a precipitator because of the increased number of
hopper connections. A recent cost quotation of $8000 per
hopper was assessed as an investment cost for all alterna-
tives. This resulted in a $128,000 cost differential adder
for the ESP equipped with gas conditioning as compared to the
baghouse. This cost is also included in the investment costs
in Table 5.
A review of the total investment costs in Table 5 indicates the
precipitators equipped with a gas conditioning system to be the least
costly. The cost differential in favor of this system is $2,269,000 and
$5,122,000 as compared to the baghouse and full sized precipitator,
respectively. This substantial differential will be reflected in the
present worth of annual cost for variable escalation rates to follow.
COMPARISON OF ANNUAL COST-NO ESCALATION
This comparison of annual costs is very significant since it
provides first year costs which are not affected by assumed escalation
rates. The latter can greatly distort the care taken in establishing
valid investment and annual operating cost estimates. This comparison
without escalation is shown in Table 6(a).
-------
TABLE 6(a). COMPARISON OF ANNUAL COST-NO ESCALATION (OOP's of $)
ESP w/Cond.
ESP w/o Cond. Baghouse
Total Investment
Annual Costs
$16,142
$21,264
$18,411
Fixed Charges
Incremental Energy
Sulfur
Steam
Bag Replacement
Maintenance
Total Annual Cost
Annual Cost Differential
P.W. of Annual Cost
P.W. Differential
$ 2,744
490
94
6
-
253
$ 3,587
Base
$33,814
Base
$ 3,
-
-
-
$ 4,
$
$41,
$ 7,
615
474
314
403
816
507
693
$ 3
$ 4
$
$37
$ 4
,130
526
-
-
212
144
,012
425
,821
,007
The summary in Table 6(a) indicates that the annual cost of the
precipitators equipped with gas conditioning is the least costly, with a
differential of $425,000 per year as compared to the baghouse. The
present worth of annual cost over 30 years is $4,007,000 in favor of the
gas conditioned ESP's.
COMPARISON OF P.W. OF ANNUAL COST ASSUMING 5% ESCALATION
The cost comparison shown in Table 6(b) will indicate the effect of
any annual costs that are sufficiently sensitive to escalation such that
they have a significant effect on the comparison shown in Table 6(a).
It will be noted that by assuming escalation, it is necessary to
capitalize (i.e., use the present worth) of each cost factor.
TABLE 6(b). COMPARISON OF P.W. OF ANNUAL COST - 5% ESCAL. (OOP's of $)
ESP w/Cond. ESP w/o Cond. Baghouse
Total Investment Cost $16,142 $21,264
P.W. Of Annual Costs (30 yrs)
Fixed Charges $25,867 $34,078
Incremental Energy 7,741 7,489
Sulfur 1,485
Steam 95
Bag Replacement
Maintenance 3,997 4,961
Total P.W. Of Annual Costs $39,185 $46,528
P.W. Differential Base 7,343
$18,411
$29,506
8,310
3,349
2,275
$43,440
$ 4,255
-------
The results summarized in Table 6(b) indicate the present worth of
annual cost of the ESP with gas conditioning to be $4,255,000 less
costly than the baghouse. This is a larger cost advantage than that
which resulted from the assumption of no escalation in Table 6(a). This
indicates that annual costs subject to escalation for the baghouse
system exceed those for the ESP equipped with a gas conditioning system.
COMPARISON OF P.W. OF ANNUAL COST ASSUMING 10% ESCALATION
The comparison shown in Table 6(c) is somewhat academic except to
show that by assuming an increasing escalation rate will result in a
more favorable cost position for the ESP with gas conditioning. It
should also be evident that assuming a rate greater than the return on
investment will cause a rapid acceleration in the cost advantage.
TABLE 6(c). COMPARISON OF P.W. OF ANNUAL COST - 10% ESCAL. (OOP's of $)
ESP w/Cond. ESP w/o Cond. Baghouse
Total Investment Cost $16,142 $21,264 $18,411
P.W. of Annual Costs (30 yrs)
Fixed Charges
Incremental Energy
Sulfur
Steam
Bag Replacement
Maintenance
Total P.W. Of Annual Costs
P.W. Differential
$25,867
14,700
2,820
180
-
7,590
$51,157
Base
$34,078
14,220
-
-
-
9,420
$57,718
6,561
$29,506
15,780
-
-
6,360
4,320
$55,966
$ 4,809
As expected, the present worth differential in favor of the pre-
cipitator with conditioning has increased to $4,809,000 as compared to
$4,007,000 and $4,255,000 for the case of zero and 5 percent escalation,
respectively.
SUMMARY AND CONCLUSIONS
The results of the study indicate the following summary and con-
clusions:
(1) The use of relatively small precipitators equipped with a
sulfur burner type gas conditioning system provides an
attractive alternative which should be considered in the
selection of equipment for new generating units burning a low
sulfur western coal.
10
-------
(2) The cost advantage without distortion by assumed escalation
rates indicates an annual cost savings of $425,000 per year,
and a savings in present worth of annual costs over 30 years
of $4,007,000 as compared to a baghouse. The major factor
which results in these differentials is a lower initial
investment of $2,269,000 for the total installed system
including auxiliaries.
(3) The cost differentials for the comparison in (2) above further
increase in favor of the gas conditioned ESP's with any
assumed escalation rate applied to operating costs.
(4) A comparison with the full size precipitator indicates this to
be the most costly alternative with a present worth cost
differential of $7,693,000 vs. $4,007,000 for the baghouse,
using the gas conditioned ESP's as base.
(5) Aside from the economic comparison, a primary objective of the
study was a technical evaluation of the alternative systems.
I believe that up to this point, a general consensus of
opinion would favor the baghouse from the standpoint of
reliability and maintenance. Although judgmental, it is my
opinion that gas conditioning used in conjunction with ESP's
greatly enhances the electrical operating stability and
reliability of the precipitator collecting high resistivity
ash. The automation which permits selection of an optimum
sulfur trioxide injection rate that remains constant in parts
per million with variable boiler load, should eliminate arcing
and minimize maintenance with discharge electrodes, trans-
former rectifier sets, and controls due to reduced transients.
(6) Discussions with several utility engineers about a year ago,
indicated no unusual maintenance with the sulfur burner
system, and availability was estimated at better than 98
percent. On this basis, it is not unreasonable to equate the
technical operating characteristics of gas conditioned ESP's
with the baghouse. This combined with the economic
advantages, may restore the precipitator to a dominant
position in equipment selection for the electric utilities for
low sulfur western coal applications.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency, and therefore the contents do not
necessarily reflect the views of the Agency, and no official endorsement
should be inferred.
REFERENCES
(1) Walker, A.B. and Gawreluck, G. Performance capability and utilization
of electrostatic precipitators past and future. International
Conference on Electrostatic Precipitation, October, 1981.
11
-------
APPLICATION OF THE BUBBLE CONCEPT TO FUEL BURNING SOURCES
AT A NAVAL INDUSTRIAL COMPLEX
by: CHARLES THOMPSON
ATLANTIC DIVISION
NAVAL FACILITIES ENGINEERING COMMAND
UTILITIES, ENERGY AND ENVIRONMENTAL DIVISION
NORFOLK, VIRGINIA 23511
ABSTRACT
The Norfolk Naval Shipyard, Portsmouth, Virginia, consists of a large
Industrial Ship Repair Complex. There are over 50 gas and oil-fired
industrial size boilers located in the Shipyard. These boilers serve
such diversified functions as generating power, space heating, hot water,
and process steam and ship system testing. Eight of these boilers exceed
Virginia's particulate emission limits by as much as 90 percent.
Engineering studies outlined methods to achieve compliance with Air
Pollution Control Equipment at a total cost of $9 million. A change in
the Virginia regulations for particulate emissions from Fuel Burning
Equipment in 1979 allowed a Bubble policy to be applied. This change
allowed a combination of Bubble concept and control equipment techniques
to be used. The cost savings in applying this technique was approx-
imately $6 million. Discussed are the problems and procedures in
formulating an acceptable Bubble concept policy and control program to
allow compliance for the boiler plants.
12
-------
INTRODUCTION
The Norfolk Naval Shipyard, Portsmouth, Virginia is a naval ship
repair activity. The Shipyard is located in the Hampton Roads intrastate
Air Quality Control Region of Virginia. This region is currently in
compliance with the total suspended particulate matter air standard.
The industrial activity that is accomplished at the Shipyard include
construction overall, repair, alterations, drydocking and outfitting of
ships and other water crafts. The industrial processes used that
generate air pollution are painting, sand and grit blasting, plating,
degreasing, metal forgering and boilers. The boilers provide steam for
process equipment, equipment testing, generating electricity and space
heating. The electric power generated is distributed to ships and
production loads at the pier facilities and to offices, barracks and
ships located throughout the Shipyard. The steam is distributed to
ship's pier facilities, offices, barracks and industrial processes. To
provide this power and steam requirements, it takes 56 boilers located at
numerous locations around the Shipyard. The boilers heat input ranges
from one million BTU per hour up to one hundred and fifty million BTU per
hour. The types of fuel burned includes natural gas, distillate and
residual oil and refuse. Table 1 is a list of the Shipyard's boilers and
their heat input.
The Shipyard is comprised of many land areas that are separated
from each other by waterways, railroads, public roadways, and private
property. Figure 1 shows the relationship of the Shipyard land areas to
each other. Each separate area has a different and specific function.
The New Gosport and Stanley Court areas are housing tracts for Navy
personnel. These areas are located up to three miles from the main
Shipyard Industrial Area. The South and St. Helena's Annexes are used to
mothball ships for extended layup. St. Julien's Creek Annex is now being
used for extended shipyard services such as electronic engineering,
warehouse storage and property disposal.
The Shipyard's industrial boilers are spread throughout these areas.
The housing areas and the South Annex contain numerous small oil fired
boilers that run separate steam heating systems from the main industrial
area. The two areas known as St. Julien's Creek and St. Helena's Annex
are permitted separately with the State Air Board and are therefore not
included under the Shipyard Bubble.
BUBBLE CONCEPT
The Environmental Protection Agency over the years has formulated and
approved the Bubble Concept. Today the concept is one of innovative
13
-------
:-«SVNAVAL!*f£KS^ 5
k.->.*.:«lHiPY&Rr»'-::>Ov iu
. SHIPYARD!
©^STANLEY COURT
HOUSIN©
Figure 1. Shipyard Vicinity Map
14
-------
emission trading for existing emission sources. The Bubble gives plant
engineers and managers the ability to develop less costly ways of meeting
air quality requirements. The Bubble Concept allows a plant with
multiple emission sources through compensating emission trades among the
sources to comply with regulatory air standards.
The State of Virginia endorsed a version of the Bubble Policy in
1976. The Air Board approved a modification to rule EX-3, "Particulant
Emission from Fuel Burning Equipment," of the Virginia's Air Pollution
Control Regulation. Under Section 4.31(e) the rule states, "The emission
contribution allocation for each of the fuel burning units of the
affective facilities shall be its designated portion of the maximum
allowable particulant emissions from the affective facility when operat-
ing at total capacity."(1)
APPLICATION OF THE BUBBLE
BACKGROUND
In the time period between 1979 and early 1980, the Shipyard was
attempting to bring two boiler plants into compliance with Virginia Air
Pollution Control Regulation for fuel burning sources.
The sources were the Main Steam and Power Plant, Building 174 and
Salvage Fuel Fired Boiler Plant, Building 1460. These plants had both
failed stack emission tests. Table 2 shows measured versus regulation
allowed emissions for each plant. The Main Steam and Power Plant
exceeded air standards by approximately 60 percent. The Salvage Fuel
Fired Boiler also exceeded its emission limits by up to 90 percent.
The Main Steam and Power Plant consists of six boilers. The boilers
were constructed between 1939 and 1944. They were originally designed to
burn pulverzied coal and later converted to residual oil in the mid
60's. The particulant emissions from the Main Steam and Power Plant were
controlled by large diameter cyclones. The operating efficiency of these
cyclones was unknown.
The Salvage Fired Fuel Boiler plant consists of two water wall
boilers. The boilers were completed in 1977 and burned refuse, as
received, on recipicating grates. At the Salvage Fuel Fired Boiler, the
particulate emissions were controlled by a single field electrostatic
precipitator. The measured precipitator efficiency was approximately 90
percent.
15
-------
RELEVANT ENGINEERING STUDIES
During 1979 engineering studies were funded to review each situation
and present possible corrective solutions. For the Main Steam Plant,
this became a complicated and almost insurmountable task. The plant was
approximately 40 years old and had become incapable of fully supplying
the Shipyard's steam and power demands. It also was known that the effi-
ciency of the boilers had deteriorated over the years to an estimated 70
percent. A review of stack emission tests showed that unburned carbon
made up approximately 60 percent of the particulate being emitted from
the stack. The plant has a long horizontal breaching leading to a 200
foot stack. The breaching would readily fill up with fly ash. Theoreti-
cally it was felt that improving boiler efficiency could bring the plant
back into compliance with Virginia Air Pollution standards. However due
to the plants age and nebulous factors regarding equipment performance
variables and preliminary stack emission results of 29.5 Ib. per hour,
final compliance could not be guaranteed.
The Salvage Fuel Fired Boiler electrostatic precipitator was reviewed
and found to be deficient in many areas. The deficiencies noted were
power levels, size of the unit, spark rate control and transformers/
rectifying controls. (2) The electrostatic precipitator consultant
recommended a new electrostatic precipitator upstream of the existing
one. The electrostatic precipitator would act as a precleaner and
together with repairs to the existing precipitator would bring the
facilities into compliance.
The cost of these corrective items were estimated to be approximately
$6 million for the Main Steam Plant and $2.6 million for the Salvage Fuel
Fired Boiler Plant. Table 3 is a summary and break down of each items
cost. For the Main Steam Plant, the estimate shows that approximately
$2.6 million of the cost was just to repair and improve boiler effi-
ciency. The Shipyard has attempted since 1974 to bring the Main Steam
Plant into compliance. Past work on the boilers and the future work
scheduled, Table 3, was intended by the Shipyard to bring the Main Steam
and Power Plant into compliance. However, the Virginia Air Pollution
Control Board insisted that the Main Steam and Power Plant be brought
into compliance as rapidly as possible and that no more delays would be
tolerated. Therefore, a $3.7 million appropriation for an electrostatic
precipitator was included. The use of the precipitator would guarantee
compliance of the boilers with Virginia Air Standards.
While the above studies was being completed, two events took place
that impacted upon the Shipyard's decision. First, a preliminary plan
and location for a regional Trash Burning Plant was being finalized. A
decision has been made to locate this facility at the Norfolk Naval
Shipyard if a contract could be agreed to. The new facility would then
replace the existing Main Steam and Power Plant and Salvage Fuel Fired
Boiler Plant. The time schedule, even though not finalized, was
16
-------
estimated to be approximately five years for construction completion.
The Shipyard's ultimate decision was that spending $8.9 million on the
existing plants that would be shut down in five years was not the best
alternative. However, the Virginia Air Pollution Control Board would not
grant a long term, five years, variances to operate at higher particulate
emission levels for these two plants. The second event was notification
in January 1980, by the Virginia Air Pollution Control Board of their
Bubble Policy and suggested allocation of allowed particulate emissions.
The suggested allocation by the State had totaled the heat input for all
54 boilers located at the Shipyard and calculating total pounds per hour
of particulate emissions allowed. The total emissions were then divided
between boilers by the ratio of each boiler heat input divided by total
heat input of all boilers.
It was felt that this contingency placed an unequitable burden upon
the small (10 x 10/6 BTU/HR) residual oil fired boilers. Further, this
policy had reduced the overall allowed emissions at the Shipyard by
approximately 160 pounds per hour. Since the proposed Bubble terms by
Virginia Air Pollution Control Board would have to be modified, it was
considered feasible to also used the new Virginia Bubble Concept to
assist in bringing the Main Steam and Power Plant and Salvage Fuel Fired
Boiler Plant into compliance.
BUBBLE CONCEPT
There are 54 boilers located on the Shipyard in three different
distinct land areas, separated by public or private property. The
boilers consist of large industrial size boilers burning residual fuel
down to very small process steam type boiler burning natural gas. The
boilers also fall into two distinct categories, stationary-permanent, and
stationary-portable.
The portable boilers consist of both barge and skid mounted package
boilers. These boiler range in size from 1.5 to 17 million BTU per hour
heat input. The portable boilers are used to supply steam for testing
ship's steam systems. Therefore, the boilers are moved frequently within
the Shipyard from pier location to pier location. These boilers can also
moved into other sections of the Shipyard as well, such as the South
Annex and St. Helena's Annex.
The applicable Virginia Air Pollution Standard that the boilers had
to comply with involved a sliding scale of allowed emissions; for
example, when the heat input rises the amount of emissions allowed
decreases. As discussed previously, the Shipyard was concerned that as
the number of boilers increase, the total allowed emissions would
decrease, putting an unfair burden upon small Residual Oil Fired
Boilers. Finally many of the boilers were located not on the main
Shipyard property, but in separate distinguishable sites and these
boilers emissions also would be penalized.
17
-------
To take all these considerations into account, four Bubbles were
formed around the Shipyard. A Bubble was formed over the main Shipyard
land area for the stationary-permanent boilers; a Bubble was formed for
the portable boilers at the Shipyard; a Bubble was formed for the South
Annex boilers, and finally a Bubble was formed for the remote housing
site boilers. It was felt that by setting up four Bubbles a fair and
equitable solution was made between allowable emission and actual
operation of the Shipyard. Table 1 demonstrates each Bubble location,
boilers involved, the boiler heat input and the allowed emissions, both
before and after application of the Bubble.
ALLOWED EMISSIONS
The allowed emissions were calculated using the following procedure.
For the main Shipyard Bubble the boiler reference numbers 15 through 19,
24 through 25, and 2,001 through 2,008 were calculated using EPA air
emission factors from publication AP 42, and current fuel usage. These
allowed emissions were totaled and substracted from the total allowed
under the Bubble. The quantity of emissions left were then split between
the Main Steam and Power Plant, reference numbers 9 through 14, and the
Salvage Fuel Fired Boiler Plant, reference numbers SS-101 and SS-102.
For the other Shipyard areas the total emissions allowed for each Bubble
were divided between boilers base on the ratio of boiler heat input to
total heat input.
CONCLUSION
The accomplishment of the four Bubbles has allowed the Norfolk Naval
Shipyard to comply with all emission limits, at a reasonable cost and
time frame. The Bubble policy has allowed the Main Steam Plant to come
into compliance and at the same time not have to include an electrostatic
precipitator. The estimated control efficiency of the multi-cyclone
installed and the improvement in the boiler combustion will give a final
outlet emission within the 31 pounds per hour limit. Preliminary stack
testing of the Main Steam Plant has shown emissions to be 29.5 pounds an
hour. The saving due to the Bubble has been approximately $3.7 million.
The Salvage Fuel Fired Boiler compliance is to be accomplished by
upgrading the existing electrostatic precipitator and installing a
precleaning multi-cylcone. The use of the multi-cyclone is to remove
large particles that were degrading electrostatic precipitation opera-
tion. The precipitator should now operate at its fullest potential. The
cost of this project is approximately $400,000. This represents a saving
of approximately $2.2 million over installing a new electrostatic
precipitator and reconstructing and moving the existing precipitator.
18
-------
Current ongoing performance testing of the Main Steam and Power Plant
and Salvage Fuel Fired Boiler are showing new problems. It is now
expected that the Bubble will not be required for the Main Steam and
Power Plant compliance. Preliminary stack testing has shown emissions to
range between 15 and 19 pounds per hour. The current performance of the
new multi-cylcones and upgraded precipitators at the Salvage Fuel Fired
Boiler, however, has been very disappointing. Emission testing has not
yet taken place but visible emissions are at times exceeding 20 percent
opacity. It is now expected that the Bubble Concept will have to be used
to shift allowed emissions to the Salvage Fuel Fired Boiler for final
compliance.
The total allowed emissions from the Norfolk Naval Shipyard under its
State Registration No. 62040 has been decreased by application of the
Bubble Concept. The decrease has taken place even while two facilities
Main Steam Plant and Salvage Fuel Fired Boilers are allowed to increase
emission levels. The accomplishment of these two goals at the same time
have saved the U.S. taxpayers approximately $5.9 million in capital cost
and approximately $55,000 in annual operating costs under the Virginia
Bubble Concept.
The work described in this paper was not funded by the U.S. Environ-
mental Protection Agency and therefore, the contents do not necessarily
reflect the views of the agency and no official endorsement should be
inferred.
19
-------
REFERENCES
1. Stone, D., Editor, VirginiAir, Publication of Virginia Air Pollution
Control Board. Vol. 12, No. 2, June 1982.
2. Hall, H. J., Summary Analysis and Recommendations for Precipitator
System Improvement to meet State Regulations on Incinerator Gases at
Naval Shipyard, Portsmouth, Virginia, Technical Report HAR79-222,
June 1979.
20
-------
W
CO «
>
H Z CO
oj P PQ
•3 rJ
P< Q v-'
U
(0
O
CO
01
O
*
o
o
•
o
CM
*
o
in
•
00
vt
•
oo
CRIBUTION
£
o
U
z
0
CO PQ
JVI m
H 5
o w
CJ PQ /-\
Od W *"^
5, PQ CO
p. PQ
O rJ
gg^
H O
•
01
o
•
oo
CM
•
CO
0)
o
•
oo
CM
• •
CQ C0 *
*
0>
r-l
pcj
i— 1
r-l
4J
U
CO
4J
a
0
o
;*
i-i
60
i-l CO
PQ m
SM/ 1— 1
CO -*
i-l r-l
*
01
m
•
i
CM /-N
m r-l
1 r^
OS r-l
1— 1 •
01 60
*O 'O
0 r-l
0 &^
O M
^H
r4 1
0) oo
60-*
C r*
•H ^j
co a
m vo
r-l r-l
-*
^-\
r-l
1 r-
CO r-l
ON •
m 60
td TJ
1 r-l
O CQ
1— 1 > —
o
M
r-l 01
Ol -U
TJ cfl
O 01
a -c
M
.* a)
• O.
O 3
cj co
r4 O
0) -*
60
C3 ^
co Q
r~+
i— i
«
Q
0)
i-H
1— 1
U
C ^"^
M r-l
r^
•M-H
r-l
01 •
3 60
O "O
PQ i-l
PQ
S v~/
0 •
CO 60
fl *o
M r-l
3 PQ
21
-------
H
I
RATED HEAT CAPACITY TOTAL PARTICULATES TOTAL PARTICULATES
(10/6 BTU/HR) ALLOWED BEFORE BUBBLE ALLOWED UNDER BUBBLE
(LBS/HR) (LBS/HR)
PC.
w
M
SS
< M
h 2
JZ £5
3
JB rJ
Q
H P
55 O
3 S
32
C^ ^^
M Z
^3 •^C
o-
w
Z 04
W W
prf pq
W 53
&t p
i— i
o
•
o
vO
•
CO
i— i
l M
co cu
z ^
Z S
i 3
cr> d
rH
O rH
vO CU
rH TJ
P O
I S
§0
z d
oo
o
o
CN
CN
•
o
•
CO
0)
m
•
r^
•
CO
CU
vO
CN
<4-l |
0 0
CN
O rH
•H CU
W T3
•i-l O
> S
• l-l
p •"
09
CU P
> o
• i-l 4J
4-1 Q
0 S
O rH
M CO
4-1 >-l
0 CU
CU d
U U>
oo
rH
en
units
> -*
•r-l
M 1
TJ
CO
rH rl
\rH
O CO
O
CN •!-!
•^x r-l
-------
^-N
Q
W
C3
( CONTIN
2
O
J_(
H
PQ
I_J
N CONTR:
o
^
co
CO
W
CO _3
a PQ
3§
rJ PQ
ED
0 PS'-x
M W OS-
EH O fn
co
Pi CQ
Q J
3g~
H O
^
CO r-3
U PQ
EH pQ
J PQ
QJ fjq
rH PS
H O'-x
•^c ryi £5
CO
3§3
H » -^
S3
3
>•
i_j
ri
m
CM
•
O
cd
m
a\
O
CN
00
00
00
00
o
*
in
en
CN
CN
CN
00
•
CS
oo
•
CM
oo
oo
•
CN
L - PARTICULATE E
M
3
^^
i*rt
§
< Pi
0 se
EH 3
•<< EH
W FO
vO
Q •>>.
Cd O
3d
2
P*
w
5
0 W
<3 PQ
II
^4
S i 1
M
H Q
S5 O
w s
p^ F\
Hi |Z
|^t ^C
o-
u
M
U prj
2 U
-1
u 5
fa 2
u
PC!
•
cd
CU
en
en
CO
r-l
CU O
C 4-1
••H cd
& S-l
C l-( 0>
O eo 3 fi
co I-i H CU
•rl O 00
rH 4-1 CO
rH O td rH
< ,S O Cd
U
r-l rH CJ) -rl
cu cd NJ- IH
CO Vl rH 4J
cu cu ^
4J > T3 -rH CM
0) -H O rl
O O S T> 1
f^ vO
rs .
r-H in
r-H
1-H
cu
Q
^
• m
l-l
(U
r-H
•r-l
O
PQ
rH
cd
4J
c
cu
a
•H
4J
d
o
CO
cu
^ j
cd
4-1
M CTi
O CM
p t |
co
•
CM
/-^\
CU
rH
cd
4J
M
O
P-i
m
i
o
t~.
r-H
|
^f*
P-I
(U
00
l-l
cd
PQ
*
O
U
*
oo
MH
J5£
4J /-x
3 m
o ~*
Si^ r— 1
O 1
,_,
1
t*4
m
•
««^
3
en
1
CU CO
r-H &H
•r-l
O 0)
PQ 00
l-l
CO Cd
v* PQ
rl ^-^
0
,3 0
in
(3 rH
0 I
rl U
l-l
r-H
CO CU
0) T3
00
CM
rH
1
CO
m
o
en
•- I
W P-4
cu co
r— 1 pH
• H
O CU
PQ 00
M
CO Cd
^ PQ
l-l v-<
O
» 0
m
§r— 1
|
IH CJ
M
i— 1
CO CU
CU "O
1 1
ON
CN
rH
1
co
m
i-i
cu
r-H
•rl
O
PQ
ca
^i
O
a
0
u
H
CO
cu
^o
CM
rH
1
co
en
en
1
OS
co
><
cu
00
r-l
cd
PQ
0
m
i— I
I
o
1— 1
cu
•o
^
m
•
en
en
1
I-i PS
cu co
rH £}•<
•r-l
O CU
PQ 00
I-i
co cd
^ pp
V-l ^^
o
in
C rH
0 1
M 0
M
r-l
CO CU
0) T>
II
r^
CM
rH
1
co
in
•
en
M pj
cu co
••H
o cu
PQ 00
co ed
^ PQ
I. ^_^
O
3 O
m
fi rH
0 1
rl CJ
W
rH
CO CU
Q) T3
O
en
rH
1
CO
23
-------
W (0
o
cj
H
P
pa
a
o
M
co
CO
9
rJ PQ
CJ Pfj
M W .
H P3
CO
O PQ
H rJ
35
CO rJ
W PQ
rf «
S§
gg-
•35 W
Pu pq
co
(Hi
35
Q ---
W O
H rH
oo
CM
co
O
PQ
>^
rH
a.
a.
3
CO
*
a! i?
C •
•H rH 00
60 aj -a
(3 4J rH
W d pq
X CU ^-I
CU ri C
Q CU -i-l rH
C rH 4J O>
/-« QJ
4-1
rH 1^ r< (*•
CM 3 '-v
0 rH
CJ CU
>% >-. Tl
rH >, rH Q
CX CU CU SI
CUrH CU
3 C 3 >,
co cd co cu
4J rH
<# CO t« (3
1-3 >-" «
»
CU
^H
C
cd
4J
CO
• «
cd cd
cu a)
-* vO
• •
O oo
_ 1 ^t
r-l »^r
CM
• •
cd cd
cu cu
-* r*
O «N
rH rH
•
-------
TABLE 2 - EMISSIONS
MEASURED EMISSIONS STANDARD
Main steam and power plant boilers 45 Ib/hr 28 Ib/hr
Salvage fuel fired boiler plant boiler 1 12.4 Ib/hr 11 Ib/hr
boiler 2 19.8 Ib/hr 11 Ib/hr
TABLE 3 - COSTS
Main steam and power plant
1. Repair combustion controls 154,000
2. Replace dust collectors 441,000
3. Repair breeching 673,000
4. Rehabilitate boiler No. 10 421,000
5. Replace oxygen meters 15,000
6. Rehabilitate boiler No. 11 722,000
7. Repair smoke stack 206,000
8. Additional pollution control device 3,700,000
$6,332,000
Salvage fuel fired boiler plant
1. Repairs to electrostatic precipatator 110,000
2. Construct new multi-cyclones 300,000
3. Install opacity meters 15,000
4. Additional electrostatic precipitators 2,200,000
$2,625,000
25
-------
CYCLONE PERFORMANCE: A COMPARISON
OF THEORY WITH EXPERIMENTS
by: John A. Dirgo
David Leith
Harvard School of Public Health
Department of Environmental Health Sciences
665 Huntington Avenue
Boston, MA 02115
ABSTRACT
This paper describes the results of tests conducted on a Stairmand high-
efficiency cyclone. The cyclone was pilot plant scale with a design air flow of 0.14
m /s (300 cfm). Collection efficiency and pressure drop were measured over a range
of air flows at ambient temperature and pressure. An oil mist was used as a test aero-
sol because it consists of spherical drops of uniform density, which are unlikely to
bounce or re-entrain after striking the cyclone wall. At each air flow, a fractional
efficiency curve (collection efficiency vs. particle diameter) was determined. Each
experimental curve was compared with fractional efficiency curves generated by
several cyclone efficiency models. A comparison of this type is more valid than one
based on cyclone cut diameter (the particle size collected with 50 percent efficiency).
This work represents the initial phase of a study of optimized cyclone design.
INTRODUCTION
Cyclones have been used since the late 1800's for the removal of dust from indus-
trial gas streams. Because they rely on inertial forces to collect parlicles, cyclones
are inefficient collectors of particles smaller than about 5 /im in diameter. In spite of
this disadvantage, there has been a renewed interest in cyclones, particularly as
precollectors in atmospheric and pressurized fluidized bed combustion systems.
Many different types of cyclones have been built, but the reverse-flow cyclone
with a tangential inlet is most often used for industrial gas cleaning. Figure 1 shows a
typical reverse-flow cyclone. This collector can be characterized completely by eight
dimensions, which are often expressed in terms of their ratio to the cyclone body
diameter, D. Figure 1 also shows the dimension ratios and the actual dimensions for
the cyclone used in this study-the Stairmand high-efficiency design cyclone (l). This
design is one example of standard cyclone designs that have been developed. Many of
these designs arose through a trial and error approach as "the result of 'hunches' or
26
-------
Dimension Dimension
ratio (m)
D = 1.0
a = 0.5
b = 0.2
Dg = 0.5
5 = 0.5
h = 1.5
// = 4.0
B = 0.375
0.305
0.153
0.061
0.153
0.153
0.458
1.220
0.114
Figure 1. Reverse-flow cyclone with dimension ratios and dimensions for
a Stairmand high-efficiency design.
efforts to overcome operating difficulties (2)." According to Swift (3), "...cyclones have
been developed almost wholly by experiment, and it would be difficult to prove
mathematically that [they] are of the best design..."
There is no reason to assume that standard cyclone designs represent the
optimum possible performance. In fact, cyclone theories predict that substantial
improvements (increased collection efficiency at constant pressure drop or reduced
pressure drop at constant efficiency) can be obtained by altering cyclone dimensions.
This paper presents the results of the initial phase of a study of improved cyclone
design, in which fractional efficiency curves for a Stairmand high-efficiency cyclone
were determined over a range of gas flow rates. The results will be used to evaluate
the predictive capabilities of cyclone efficiency theories and to establish a "baseline"
performance level for this cyclone. Later changes in performance due to changes in
cyclone dimensions can be measured against this baseline.
THEORY
Cyclone collection efficiency theories differ greatly in complexity. Some are
almost entirely empirical while others are completely theoretical. There is a general
agreement that operating parameters of the system should be used to predict perfor-
mance, and most theories have some sort of impaction parameter grouping that
accounts for the influence of particle diameter and density, gas velocity and viscosity,
and cyclone diameter. There is less agreement on the effects of cyclone dimensions
and geometry. Some theories consider all eight cyclone dimensions while others
include as few as two.
All theories set up a balance between the outward centrifugal force on a particle
caused by the spinning gas stream and the inward drag force resisting the radially
27
-------
outward motion. By assuming various flow conditions within the cyclone, different
authors have dismissed different terms in this force balance as insignificant. Because
the relative importance of these terms will change with cyclone design and operating
conditions, it is unlikely that any one theory will accurately predict performance for
all applications (4). At least three general classes of cyclone efficiency theories have
been published (4). These are described below with examples given for each type.
CRITICAL DIAMETER: STATIC PARTICLE APPROACH
The static particle approach determines the particle diameter for which the out-
ward centrifugal force is exactly balanced by the drag force caused by gas flowing
radially inward to the cyclone core. Theoretically, these "static" particles should
remain suspended indefinitely at the boundary between the vortex and the cyclone
core; smaller particles flow to the core and out of the cyclone while larger particles
move to the cyclone wall for collection. Theories of this type predict an abrupt change
in collection efficiency from zero to 100 percent as particle diameter increases
beyond the critical size. In practice, this sharp cut is not realized because of varia-
tions in radial and tangential gas velocities along the cyclone axis, and the efficiency
for the critically sized particle is often taken as 50 percent.
An example of the static particle approach is the theory of Earth (5). Earth cal-
culates the diameter of the particle for which the centrifugal and drag forces are
equal. The resulting expression for diameter is used to determine the terminal set-
tling velocity, i>ts*. for the critically sized particle:
'
[Terms in Eq. (l) and subsequent equations are defined in the NOMENCLATURE section
at the end of the paper.] Here, vto * is used mainly as a measure of aerodynamic resis-
tance. Tangential gas velocity in the vortex, vt , is evaluated at radial position D9 / 2.
The terminal settling velocity of any particle size, vte, can be related to vta* by:
Barth gives a generalized plot of cyclone collection efficiency as a function of this
ratio.
CRITICAL DIAMETER: TIMED FLIGHT APPROACH
The timed flight approach assumes an initial radial position for particles entering
the cyclone. The critically sized particle is one that can cross the distance from this
initial radial position to the cyclone wall during its time in the cyclone. One of the
most commonly used theories of this type is the Lapple "cut diameter" theory (6).
Lapple assumes that the particle size that enters the cyclone at the inlet half-width
(D/ 2 — 6/2) and travels the distance to the wall during its residence time is collected
with 50 percent efficiency. Lapple' s expression for this particle size, called the cut
diameter, is:
160
= ^ /
V
The collection efficiency for any other particle size can be found from its ratio to the
cut diameter. A plot of fractional efficiency versus this ratio, reported by Lapple, has
28
-------
been fit to the equation (7)
This relationship was developed experimentally for a Lapple general purpose cyclone
design (6); similar relationships for other cyclone designs have not been reported.
FRACTIONAL EFFICIENCY APPROACH
While both critical diameter methods rely on generalized plots to determine the
collection efficiency for particles other than the critical size, the fractional efficiency
approach allows a direct calculation of the efficiency for any particle size. Examples
of this type of theory are the Leith-Licht model (8) and a more recent model by Dietz
(9).
Leith and Licht assume that the tangential gas velocity in the cyclone is related to
the distance from the cyclone axis by
vtrn = CONSTANT (5)
where n is the vortex exponent. Experimental studies of cyclone flow patterns have
reported values of 0.5 to 0.9 for n, with most values falling in the lower end of this
range. According to Alexander (10), the vortex exponent can be calculated from
n = l-[(7Y283)°-3(l-0.67£m°-14)] (6)
where T is the gas temperature in °K and Dm is the cyclone diameter in meters.
Leith and Licht also assume that radial gas velocity is zero and that the drag
force on particles travelling radially outward toward the cyclone wall is described by
Stokes" law. This model accounts for turbulence within the cyclone by assuming that
in any plane perpendicular to the cyclone axis, uncollected particles are uniformly
mixed. Cyclone dimensions are considered in the determination of an average
residence time for the gas in the cyclone. The resultant expression for collection
efficiency is
The influence of particle and gas properties are combined into i', a modified inertia
parameter
ppdpzvi (n + 1)
IBfiD ^ '
The term C is a dimensionless cyclone geometry parameter that depends only on the
eight cyclone dimension ratios and is independent of the size of the cyclone (8).
The Dietz (9) model represents a refinement of the Leith-Licht method and divides
the cyclone into three regions. These regions are the entrance region (the annular
space around the outlet duct at the top of the cyclone), the downflow region
(corresponding to the vortex below the level of the outlet duct), and the core region
(formed by the extension of the outlet duct to the bottom of the cyclone). Turbulence
within each region is assumed to produce a uniform radial concentration profile for
uncollected particles. To approximate a distribution of particle residence times in the
cyclone, the theory allows for the exchange of particles between the downflow and
core regions. The Dietz model determines cyclone collection efficiency as
29
-------
„. = i - (jr.-
(9)
where the subscripted A' terms are functions of particle and gas properties as well as
cyclone dimensions.
EXPERIMENTS
All experiments were carried out on the cyclone test system shown in Figure 2. In
this system, room air was pulled through an absolute filter to remove ambient parti-
cles. Air flow rate was measured by the pressure drop across a calibrated Stairmand
disc. A Stairmand high-efficiency cyclone, with D = 0.305 m, was used.
n
Arcoprime 200 (a mineral oil with pp = 860 kg/m ) was nebulized using a Laskin
nozzle aerosol generator (11) with a compressed air gauge pressure of 27.6 kPa (4
psig) and was injected through a cylindrical probe that introduced the aerosol at the
center of the duct. This aerosol generation system was chosen for several reasons.
First, because the liquid droplets produced are spherical, their aerodynamic behavior
is easily described. All of the theories described above assume spherical particles.
Second, collected liquid droplets should not bounce or re-entrain after striking the
cyclone wall. Again, cyclone collection efficiency theories assume this to be the case.
Finally, this generation system produced sufficient particles over a range of sizes
< dp < 5/Lim) where cyclone fractional efficiency should increase from ^0 to wl.
SLIDE
DAMPER
TO FAN AND
EXHAUST
EGGCRATE
STRAIGHTENER
DOWNSTREAM 1
LOCATION FOR
AEROSOL
GENERATOR
1
r
1
1
DOWNSTREAM VALVE A
r— ISOKINETIC $ ^-J
SAMPLING PROBE 1 * ( ) fUtli-
n
Nv H ROTAMETER
\ T
/I fA * J
y IT \ PMS STAIRMAHD
^ I DISC ^
iLfTT® PRESSURE
_. CYCLONE LASKIN UH GAUGE
I NOZZLE
/ AEROSOL
/ GENERATOR $. $ _ FROM
/ »*• ^ * All
f PRESSURE
DUST REGULATORS
HOPPER
FILTER
COMPRESSED
Figure 2. Schematic drawing of cyclone test system.
30
-------
Aerosol samples were taken isokinetically through sampling probes at the duct
centerline upstream and downstream of the cyclone. Particles were sized and
counted with a Particle Measuring Systems (PMS), Inc. (Boulder, CO 80301) aerosol
scattering-jspectrometer. The maximum number concentration counted was
< 10 /cm without dilution, well below the concentration for which coincidence is
significant for this instrument. To minimize sampling line losses, the PMS was moved
between the upstream and downstream sampling locations; in each case, the sampling
line consisted of a 23 cm long horizontal run from the duct centerline to the PMS inlet.
Accurate sampling of the aerosol downstream from the cyclone is difficult
because the gas flowing from the cyclone is swirling. Techniques for sampling from a
swirling gas flow are available, but because of transient velocity patterns, these tech-
niques are difficult to use at best, and unreliable at worst. To eliminate this problem,
an "egg crate" flow straightener was installed three duct diameters downstream from
the entrance to the cyclone outlet duct. According to Browne and Strauss (12), this
device will not affect flow patterns and collection within the cyclone if it is more than
two duct diameters from the outlet duct entrance. As reported by Ferguson, et al.
(13), we found that a straightener with each cell D/6 in height, width, and depth
(where D is the duct diameter) produced a flat axial velocity profile with negligible
tangential velocity downstream of the cyclone.
An efficiency determined from a measurement of aerosol concentration down-
stream of the flow straightener will reflect collection by both the cyclone and the
straightener. Accordingly, measurements of efficiency that include collection by the
straightener must be corrected. We accomplished this by injecting aerosol down-
stream of the cyclone but upstream of the flow straightener as shown in Figure 2. The
concentration and size distribution of aerosol injected at this point were assumed
identical to the concentration and size distribution of aerosol injected upstream of the
cyclone as measured at the upstream sampling location. Because the operating
parameters of the aerosol generator are not affected by its location, this assumption
is reasonable. The distance between the aerosol injection and sampling locations was
the same, both upstream and downstream of the cyclone.
Three sets of measurements, each set consisting of eight replicate samples, were
made for each test. For the first set of measurements, aerosol was injected and sam-
pled upstream of the cyclone. The second set sampled the aerosol downstream of the
cyclone after it had been injected upstream. For the third set, the aerosol was
injected between the cyclone and straightener and sampled downstream of the
straightener. For particles of any size, the combined efficiency of the cyclone and
straightener can be measured directly by
,v
(.10;
cs •
•"up, up
where N is the count rate for particles of that size, the first subscript refers to the
aerosol injection site relative to the cyclone (up for upstream and down for down-
stream), and the second subscript refers to the sampling location relative to the
cyclone. The efficiency of the straightener can also be measured directly from
The combined efficiency of the cyclone and straightener in series is related to the
individual efficiencies of each by
31
-------
By substituting Eqs. 10 and 11 into Eq. 12 and solving for r]c , cyclone efficiency can be
expressed as
*> - 1 Ny
*?c - 1 - T;
*" down .down
Prior to making the measurements described above, it was necessary to relate a
single sample, taken at the duct centerline, to the average particle count rate across
the duct. Samples were taken at six radial positions across the duct both upstream of
the cyclone and downstream of the straightener. By combining these six counts, it
was possible to estimate an average count rate for the duct. Downstream of the
straightener, the concentration profile was uniform and a single centerline sample was
representative of the duct average. Upstream profiles were less uniform (higher
counts near the center of the duct), but for each cyclone inlet velocity, a single aver-
age correction factor was sufficient to relate the centerline sample to the average
count for the duct.
Tests were conducted at cyclone inlet velocities of 5, 10, 15, 20, and 25 m/s. The
design inlet velocity for a Stairmand high-efficiency cyclone with D ~ 0.305 m is 15
m/s, corresponding to a flow rate of 0.14 m /s (300 cfm). Three separate tests were
conducted at each inlet velocity, for a total of fifteen experiments. For each experi-
ment, all particles > 1 /zm in diameter were counted and sized into intervals of width
0.75 /j,m by the PMS. Cyclone collection efficiency for the midpoint of each size inter-
val was calculated from Eq. 13, where Nvp 40^^ and Nd0wn >(jotun were the average counts
for the eight replicate samples. The total number of particles counted per test ranged
from7xl04to5xl05.
For each inlet velocity, pressure drop across the cyclone was measured. The
downstream pressure taps were located between the cyclone and the flow straightener
so that any additional pressure loss due to the straightener was not included in the
measurement.
RESULTS AND DISCUSSION
Cyclone pressure drop values for each of the five inlet velocities are shown in
Table 1. Pressure loss also can be expressed as a number of inlet velocity heads, A//,
which should be constant for all inlet velocities. The average value for hH was 5.7,
higher than the value of 5.3 reported by Stairmand (l) for this cyclone design.
TABLE 1. EXPERIMENTAL CYCLONE PRESSURE DROP
Inlet velocity Pressure Drop
(m/s) (Pa)
5
10
15
20
25
87
336
785
1407
2205
32
-------
Figure 3 shows the experimental fractional efficiency curves for all five cyclone
inlet velocities. Each data point represents the mean cyclone efficiency for the three
tests involving that particular particle diameter and inlet velocity combination The
effects of both of these parameters on cyclone efficiency are in general agreement
with theoretical predictions. For any particle diameter, Figure 3 shows that an
increase in inlet velocity, holding all other parameters constant, results in increased
fractional efficiency. For a given cyclone inlet velocity, efficiency increases with parti-
cle diameter.
Figures 4-8 compare experimental results with predictions of the four cyclone
efficiency theories discussed previously. The theoretical predictions are presented as
smooth curves. Experimental data points from Figure 3 have been replotted, without
the connecting lines, in Figures 4-8. To indicate the reproducibility of the experimen-
tal results, a 95 percent confidence interval is indicated by the *'s above and below
each data point. For the three determinations of efficiency represented by each data
., „
point, the 95 percent confidence interval is given by TJC ± ( - -p — — ) where s^ is the
variance of the three efficiency measurements and 4.30 is the appropriate value for
Student's t with two degrees of freedom. Where confidence intervals do not appear
symmetrical, it is because they have not been extended beyond the range of fractional
efficiency from 0 to 1.
For ideally collected liquid droplets, the experimental fractional efficiency curves
indicate a much sharper separation by the cyclone than predicted by most of the
theories. With no particle bounce or re-entrainment, large increases in collection
efficiency occur for relatively small increases in particle diameter. For example, at
cyclone inlet velocities of 10 and 15 m/s (Figures 5 and 6), increases of »2 /j,m in parti-
cle diameter cause fractional efficiency to increase from <0.2 to >0.8. The theories of
Lapple, Leith and Licht, and Dietz all predict much flatter fractional efficiency curves.
Only the Barth theory matches the experimental curves in steepness.
The Lapple theory underestimates collection efficiency for most of the data. Only
for cyclone inlet velocities ^10 m/s and fractional efficiency <0.2 do the Lapple pred-
ictions and the experimental results agree. One major drawback to the Lapple theory
is the use of the term Nt in the expression for cyclone cut diameter, Eq. 3. This
empirical term describes the effective number of turns made by the gas stream in the
cyclone and is necessary to calculate residence time. Lapple (6) recommended that
N, be determined experimentally for different cyclone designs, but the value of 5 that
he reported is often used.
Although Ne = 5 was used to calculate the Lapple curves in Figures 4-8, the experi-
mental data suggest that Ne is substantially higher. Using experimentally determined
cut diameters from the curves in Figure 3, experimental values of Na can be calcu-
lated from Eq. 3. Calculated values of JVa are not constant for the cyclone design as
suggested by Lapple (6), but increase from «12 at vt = 5 m/s to »30 at i^ = 25 m/s.
These "calibrated" values of N, can be used to calculate new fractional efficiency
curves for the Lapple theory. However, the new curves would fit the experimental data
only at the cut diameter, underestimating efficiency for larger particles and overes-
timating for smaller particles.
The predictions of the Dietz theory fall in the same range as the Lapple curves in
Figures 4-8. Experimental cyclone efficiency is generally much higher than predicted
by the Dietz theory, except for the lower ends of the fractional efficiency curves at
33
-------
5 M/S
10 M/S
15 M/S
20 M/S
25 M/S
2'. 00 3'. 00 1.00 5.00 6.00
PflRTICLE DIflMETER. MICROMETERS
7.00
8.00
Figure 3.
Experimental fractional efficiency curves for Stairmand high-
efficiency cyclone for inlet velocities from 5 m/s to 25 m/s.
o
o
o
CO
INLET VELOCITY:
5 M/S
BF1RTH (1956)
LEITH «. LICHT
(1972)
1.00
2.00 3'. 00 4'. 00
PRRTICLE DIflMETER,
5.00 6.00
MICROMETERS
7.00
8.00
Figure 4.
Experimental and theoretical cyclone efficiency for cyclone inlet
velocity = 5 m/s. (For Figures 4-8, open squares indicate ex-
perimentally determined efficiency; asterisks indicate 95% con-
fidence intervals; solid lines indicate theoretical predictions.)
34
-------
INLET VELOCITY
10 M/S
LEITH & LIGHT
(1972)
DIETZ (1981
LRPPLE (1951
°b.oo
i.oo
2.00 3.00 4.00 5.00 6.00
PflRTICLE DIflMETEB, MICROMETERS
7.00
8.00
Figure 5. Experimental and theoretical cyclone efficiency for cyclone
inlet velocity = 10 m/s.
o
o
•z.
UJ
o
a:
F°
00
INLET VELOCITY
15 M/S
BRRTH (1956)'
LEITH & LICHT
(1972)
DIETZ (1981)
LRPPLE (1951)
1.00 2.00 3.00 1.00 5.00 6.00
PflRTICLE OlflMETER. MICROMETERS
7.00
8 00
Figure 6. Experimental and theoretical cyclone efficiency for cyclone
inlet velocity = 15 m/s.
35
-------
o
o
z
UJ
z ..
00
o
cr
BFfRTH
LEITH 4 LIGHT C1972)'
D1ETZ (1981)
LflPPLE (1951)
INLET VELOCITY:
20 M/S
00 I.00 2.00 3.00 H.OO 5.00 6.00
PflRTICLE DIflMETER, MICROMETERS
7.00
8.00
Figure 7. Experimental and theoretical cyclone efficiency for cyclone
inlet velocity = 20 m/s.
o
o
UJ
.
0°
BRRTH (1956)
LEITH & LIGHT (1972)
DIETZ (1981)
LHPPLE (1951)
INLET VELOCITY:
25 M/S
t.OO 2.00 3.00 4.00 5.00 6.00
PflRTICLE DIflMETER.. MICROMETERS
7.00
a.oo
Figure 8. Experimental and theoretical cyclone efficiency for cyclone
inlet velocity = 25 m/s.
36
-------
inlet velocities of 5 and 10 m/s. The Dietz curves were calculated using a vortex
exponent of 0.56, obtained from Eq. 6, although Dietz (9) recommends a higher value
of 0.7. Higher values for n imply higher tangential velocity in the vortex and greater
centrifugal forces acting on the particles in the gas stream. While the use of n =0.7
would increase the efficiencies predicted by the Dietz theory, the Dietz curves in Fig-
ures 4-8 would be shifted only slightly upward from their present positions.
The theoretical curves based on the Leith-Licht cyclone model predict higher col-
lection efficiency than either the Lapple or Dietz models. For all of the inlet velocities
tested, the Leith-Licht model agrees with the experimental results in the middle of the
fractional efficiency range, TJC w 0.4 to 0.6. Since the Leith-Licht curves are flatter
than the experimental fractional efficiency curves, this model underestimates
efficiency for most particle diameters larger than the experimental d50. For smaller
particles, the model greatly overestimates cyclone efficiency.
Both the Leith-Licht and Dietz theories assume that turbulence within the cyclone
is sufficient to cause complete radial back-mixing of uncollected particles in any plane
perpendicular to the cyclone axis. This, in part, accounts for the relatively flat
theoretical fractional efficiency curves calculated from the two models. The steep
slope of the experimental curves suggests that turbulence is less important than
predicted. Recent work by Mothes and Loffler (14) and a previous study by Hejma (15)
indicate that there is a concentration gradient for particles in the vortex of the
cyclone. Mothes and Loffler found that larger particle (~3.5 /J,m) concentrations were
much higher near the cyclone wall and decreased by nearly two orders of magnitude
from the wall to the cyclone core. Smaller particle (~0.5 ;um) concentrations were
much more uniform as radial position changed. Hejma found similar results in the
cone of the cyclone, although in the cylinder above the cone, he found that dust con-
centration was nearly independent of radial position. These studies indicate that the
assumptions of complete radial back-mixing made by the Leith-Licht and Dietz
theories are not justified, at least for larger particles. Smaller particles, with less cen-
trifugal force, might be more strongly influenced by turbulence. The distinction
between large and small particles probably depends on cyclone geometry and operat-
ing conditions.
Of the four cyclone theories presented here, the Earth theory fits the experimen-
tal data best. At inlet velocities up to 15 m/s (Figures 4-6), the Earth curves fall
within the 95 percent confidence intervals for most experimental data points. At velo-
cities of 20 m/s and above (Figures 7-8), the theoretical curves are steeper than the
experimental data and there is agreement only at high collection efficiency.
One problem with the Earth approach is its reliance on a plot of collection
efficiency as a function of a ratio of terminal settling velocities (particle diameter over
critical particle) to determine the fractional efficiency for particles other than the
critical size. This curve was developed from the results of experiments with a variety
of cyclones. Other investigators using the static particle approach have found that
there is a pronounced dependence of curves of this type on the design of the cyclone
(16). While the predictions based on Earth's curve match some of the data for the
Stairmand high-efficiency cyclone used here, the applicability of the Earth theory to
other cyclone designs is uncertain.
SUMMARY
Collection efficiency for a Stairmand high-efficiency cyclone was measured exper-
imentally under carefully controlled conditions over a range of inlet velocities. An
37
-------
aerosol consisting of liquid droplets was used to minimize the possibility of particle
re-entrainment after collection in the cyclone and to provide the spherical particles
assumed by cyclone eSiciency theories. Four theories, representing three different
approaches for calculating collection by a cyclone, were compared with the data. Only
one of these theories, the method of Barth, gave predictions that were in substantial
agreement with the experimental results. The steep slope of the experimental frac-
tional efficiency curves indicates that although gas stream turbulence influences the
separation processes in the cyclone, the effects of turbulence are overestimated by
the Leith-Licht and Dietz theories.
Currently available cyclone theories can provide general guidelines on how
changes in operating conditions or cyclone dimensions will affect performance. We are
now varying some dimensions of the Stairmand high-efficiency cyclone to improve per-
formance above the baseline levels measured in this study. These results should be
useful in improving cyclone theories through calibration -- finding parameters that
can be adjusted so that the theories better predict experimental data. Efficient
optimization of cyclone design, however, requires a theory capable of more precise
predictions over a wide range of operating conditions and designs.
ACKNOWLEDGEMENTS
This work was supported by Grant No. CPE-8012968 from the National Science
Foundation. Arcoprime 200 mineral oil was supplied by Donald Hasselstrom of ARCO
Petroleum Products Company, Philadelphia, PA.
The work described in this paper was not funded by the U.S. Environmental Pro-
tection Agency and therefore the contents do not necessarily reflect the views of the
Agency and no official endorsement should be inferred.
NOMENCLATURE
a cyclone inlet height, m
B cyclone dust outlet diameter, m
& cyclone inlet width, m
C cyclone geometry parameter (see Eq. 7), dimensionless
D cyclone body diameter, m
De cyclone outlet diameter, m
particle diameter, m
'50 particle diameter collected with 50 percent efficiency, m
g gravitational acceleration, m s"
H overall cyclone height, m
h cyclone cylinder height, m
KQ,KI,KZ intermediate terms for computing cyclone efficiency
by Eq. 9, dimensionless
Ne number of turns made by gas stream in cyclone
(see Eq. 3), dimensionless
,down particle count rate downstream with downstream
aerosol injection, min"
particle count rate downstream with upstream
aerosol injection, min"
/V,™ up particle count rate upstream with upstream
aerosol injection, min
n cyclone vortex exponent, dimensionless
38
-------
3 -1
Q volume flow rate, m s
r radial distance from cyclone axis, m
S cyclone outlet duct length, m
T gas temperature, °K
Vj. cyclone inlet velocity, m s
vt tangential gas velocity in cyclone vortex, m s
vts particle terminal settling velocity, m s"
vts* terminal settling velocity of critical particle, m s"
cyclone pressure drop in inlet velocity heads, dimensionless
r/c cyclone fractional collection efficiency, dimensionless
r\s straightener fractional collection efficiency, dimensionless
TJC+S combined fractional collection efficiency of cyclone and
straightener, dimensionless
fj, gas viscosity, Pa s
pp particle density, kg m
fy cyclone inertia parameter, dimensionless
REFERENCES
1. Stairmand, C.J. The design and performance of cyclone separators. Trans. Instn.
Chem. Engrs. 29:356, 1951.
2. Jackson, R. Mechanical Equipment for Removing Grit and Dust from Gases. Che-
ney and Sons, Banbury, England, 1963. 281 pp.
3. Swift, P. Dust control in industry-2. Steam Heat. Engr. 38:453, 1959.
4. Leith, D. Cyclones. In: N.C. Pereira and L.K. Wang (eds.), Handbook of Environ-
mental Engineering. The Humana Press, Clifton, NJ, 1979. p. 61.
5. Earth, W. Design and layout of the cyclone separator on the basis on new investi-
gations. Brenn. Warme Kraft. 8:1, 1956.
6. Lapple, C.E. Processes use many collector types. Chem. Eng. 58:144,1951.
7. Theodore, L. and DePaola, V. Predicting cyclone efficiency. J, Air Pollut. Control
Assoc. 30:1132, 1980.
8. Leith, D. and Licht, W. The collection efficiency of cyclone type particle collec-
tors, a new theoretical approach. A /. Ch. E. Symposium Series. 68:196, 1972.
9. Dietz, P.W. Collection efficiency of cyclone separators. A.I. Ch.E. Journal. 27:888,
1981.
10 Alexander, R. McK. Fundamentals of cyclone design and operation. Proc. Austra-
las Inst. Min. Met. (New Series). 152-153:203, 1949.
11. Laskin, S. Submerged aerosol unit. AEC Project Quarterly Report, UR-38, Univer-
sity of Rochester, 1948.
39
-------
12. Browne, J.M. and Strauss, W. Pressure drop reduction in cyclones. Atmos.
Environ. 12:1213, 1978.
13. Ferguson, B.B., Mitchell, W.J., Reece, J.W., and Sterrett, J.D. Modeling and
straightening cyclonic flows. Paper No. 81-7.6 presented at the 74th Air Pollution
Control Association Annual Meeting, Philadelphia, PA. June 21-26, 1981.
14. Mothes, H. and Loftier, F. Investigation of cyclone grade efficiency using a light
scattering particle size measuring technique (abstract). /. Aerosol Sci. 13:184,
1982.
15. Hejma, J. Influence of turbulence on the separation process in a cyclone. Staub-
ReirthdLt. Luft. 3l(7):22, 1971.
16. Loftier, F. The calculation of centrifugal separators. Staub-Reinhalt. Luft.
30(12):!, 1970.
ADDENDUM
Since this paper was presented, it has come to our attention that the fractional
efficiency curves calculated according to the Barth theory (Figures 4-8) are incorrect.
An equation to determine vt in Eqs. 1 and 2 appears in a number of references in addi-
tion to Earth's original article. In Industrial Gas Cleaning (2nd Edition) by W. Strauss
(Pergamon Press, Elmsford, NY, 1975), the equation for the ratio of vt to the cyclone
outlet velocity gives values too high by a factor of two. Use of this secondary refer-
ence resulted in calculated vt's that were twice those predicted by Earth's theory.
Substitution of the correct values for vt shifts the Barth fractional efficiency curves in
Figures 4-8 substantially to the right. In addition, the slopes of the curves are
reduced, although the Barth theory still predicts a sharper cut than any of the other
theories presented.
Clearly, the curves based on the incorrect values for vt provide a much better fit
to the experimental data, suggesting that the maximum tangential velocities calcu-
lated by Earth's theory are too low. Although this result was discovered accidentially,
it provides a good example of the type of calibration—adjusting parameters so that the
theoretical predictions better match the data—that was discussed above.
John A. Dirgo
David Leith
February 1, 1983
40
-------
HIGH FLOW CYCLONE DEVELOPMENT
by
W. B. Giles
Mechanical Systems and Technology Laboratory
Corporate Research and Development
General Electric Company
Schenectady, New York 12301
ABSTRACT
Investigative studies of an atypical cyclone configuration, designed for high flow capacity,
were performed, focusing particularly on the design aspects of inlet flow and dust disengage-
ment. The results indicate that a performance equal to or superior to conventional design can
be achieved with a net savings in cyclone size and cost.
The design is characterized as a reverse flow cyclone with both a large inlet and a large
outlet, plus increased engagement length between the cyclone body and the exhaust duct.
Both of these features are seen as means to suppress large scale inlet turbulence. In addition,
reduced penetration is found by locating a vortex shield in the base of the cyclone. The net
result indicates an approximate two-to-one diameter reduction, relative to current art, for
equal flow capability and a slight pressure loss penalty.
41
-------
INTRODUCTION
Cyclones have long been a standard means of gas cleaning and owe their popularity to
their mechanical simplicity, small size, and low capital cost. On the debit side, application is
inhibited by relatively high operation cost, due to pressure loss, and a relatively modest col-
lection efficiency compared to alternate means of gas cleaning.
Recent interest in coal utilization has reawakened the desire to optimize the performance
of cyclones. This is especially germane for high temperature-high pressure applications of
coal combustion, gasification, and catalytic reactive processes. In these applications, the pres-
sure loss penalty is not as critical as with atmospheric systems and the capital cost of alternate
gas cleaning concepts can adversely impact overall system economics.
Unfortunately, the mechanical simplicity of the cyclone is not matched with a correspond-
ing understanding and control of the physical processes. Experience has indicated that tri-
boelectric charging of the dust can occur internal to the cyclone under certain test conditions;
thus, scaling to operational conditions can lead to overestimation of collection efficiency.
Analytic effort is inhibited by the complexity of the flow field and the influence of turbulence
on dust migration. Specifically, the well-known Leith-Licht [1] model incorporating turbulent
back mixing shows reasonable agreement with data for conventional designs. Their paper,
however, incorporates an oversimplification, corrected by Giles [2], that leads to overpredic-
tion of fines collection. Incorporating the back mixing mechanism into a multi-zone com-
puter model, Dietz [3] has demonstrated excellent predictive capability with conventional
designs but poor agreement with atypical designs. A promising effort is offered by Boysan, et
al. [4], but as yet is judged to omit significant cyclone mechanics. As a consequence, design
improvements must continue through empirical effort guided in large part by intuition.
BACKGROUND
The exploratory work and rationale for a high flow capacity cyclone design is described in
an earlier report [5]. In brief, conventional cyclone design derives largely from the work of
van Tongeren [6],1 ter Linden [7], and Stairmand [8]. Typically, this work leads to the use of
small inlets and small outlets for high performance cyclones with low flow capacity or large
inlets and large outlets for low performance cyclones of high flow capacity. Stairmand, for
example, offers two designs, one for each general design objective [8].
Earlier General Electric exploratory work [5], however, found that the Stairmand perfor-
mance data significantly underpredicts performance for the "high flow" design. The result is
that the "high flow" design receives serious consideration only as a rough-cut cyclone, e.g.,
for reducing the dust loading of coarse particulates. By improving design conditions at the
cyclone inlet and at the dust hopper, it is found that equal or superior performance may be
attained with a high flow capacity design for a net savings in cyclone size and cost. The
present report summarizes the final investigations with this high flow capacity design develop-
ment.
1. Designs marketed in the U.S. by General Electric Environmental Systems (formerly Buell), Lebanon, Pennsyl-
vania.
42
-------
CYCLONE DESIGN
The basic cyclone design is shown in Figure 1 as consisting of a specially-designed head
end mounted onto a conventional cyclone body. The various inserts used during experimen-
tation are also indicated.
Cyclone Head <».
0.8Dx0.4D Inlet
0.4Dx0.20D Inlet
Body Extension
1
t
D ,
1 R
1
1
1
••"->
i
-3/4D_
0.956
j
D
L_»
1
pJ
1
I
1
DM
-T- 21
T
J
)
1
3F
1
„ 3/4D ».
Exhaust Ext(
snsions
4/3D
2.5D
Spinup Spool
Vortex Shield
Figure 1. Schematic of high flow cyclone elements
43
-------
The nominal inlet consisted of a four-point scroll supplied by an inlet of 0.9 D x 0.45 D,
where D = 18 in. is the diameter of the cyclone body and the nominal outlet is 3/4 D. Thus
the total flow capacity is
Q = 0.405 D2 V,
where Vt is the average inlet velocity. The cyclone body had a cylindrical section of 1.33 D
and a conical section of length 2.5 D terminating at a pipe section of 3/8 D. The actual test
unit inlet, however, included an air shield feature. This allowed assessment of the use of
clean air injection between the dirty air at the wall and the central exhaust. Since this unit
was designed for 80% clean air, the dirty air had an inlet of 0.4 D x 0.2 D and the clean air
had an inlet of 0.8 D x 0.4 D. The interior baffle between these two streams had a diameter
of 0.956 D.
TEST PROCEDURE
The cyclone was tested as shown in Figure 2. Prefiltered air was supplied via a blower
with both streams metered for flow, using flow nozzles and inclined manometers, and one
stream contaminated with test dust from a blown fluidized bed. This bed provided dilute
concentrations of either CURL second stage flyash (pp = 2.7 g/cc) or nickel powder
(pp = 8.5 g/cc). Typical mass mean particle sizes were in the range of = 2.5 to 3.2 ^m.
Fractional efficiency was determined by optical measurement of input-output dust concen-
trations and size distributions. The overall collection efficiency was determined from two
PILLS V mass concentration monitors mounted with optic windows (shop air to purged
ducts). For the case of operation in the air shield mode, these sensors were mounted on the
small 6-inch diameter inlet line and the 13.5-inch diameter exhaust line downstream of a per-
forated plate, honeycomb deswirl/mixing element. The dust injector was positioned several
diameters upstream of the inlet sensor to assure turbulent mixing prior to sensing.
For particle size analysis, two Climet systems were used with isokinetic sampling1 through
0.051-inch and 0.075-inch diameter probes for the inlet and outlet, respectively. Typical dilu-
tion ratios of approximately ten of these sampled flows were found adequate to avoid coin-
cidence error in size counters. The resulting eight channels of size distribution information
are used to compute mass distribution which, together with the overall efficiency measure-
ment, provided means for calculating fractional efficiency.
The experiments consisted of various geometric options available within the context of
this equipment. In particular, the effectiveness of the air shielding feature was assessed by
measuring fractional efficiency with dirty air supplied to either the small or large inlet and
with clean filtered air supplied to the other inlet. The influence of clean air dilution was
accommodated in data reduction in both cases so that the data reported only reflects actual
collection.
The effect of engagement length was assessed by adding exhaust pipe sections of 1 D,
2/3 D, or both to the interior. The effect of spin-up was assessed by inserting a wooden
spool piece into the exhaust inlet. This allows variation of the spin-up or exhaust-to-cyclone
diameter from Del D = 3/4 to 1/2 and extended downward a length of D/3 from the lip of
the cylindrical exhaust. Thus, for example, a cylindrical extension of 2/3 D plus the addition
1. However, isokinetic sampling is not critical for size analysis.
44
-------
INLET
HONEYCOMB
FLUIDIZED BED
DUST GENERATOR
DUST HOPPER I
Figure 2. Cyclone test arrangement
of the wooden spool piece provides the same engagement length as a cylindrical exhaust
extension of 1 D for a direct comparison of the influence of increased swirl spin-up.
In addition, tests were also conducted with the insertion of a vortex shield mounted in the
conical base of the cyclone body with the apex of a 45° cone pointing upward, and finally, a
cylindrical length of 1 D was added to the existing cyclone body.
TEST RESULTS AND DISCUSSION
The experiments were conducted over a range of inlet velocity and the data reduced to the
form of fractional efficiency, TJ^, versus nondimensional inertial separative parameter, or
Stokes number,
18/4 D
45
-------
Also presented is the overall efficiency, 17 0, versus the mass average, i/». Here pp and dp are
density and size of particle, Vt is the average inlet velocity, /u, is the absolute gas viscosity, and
D is the cyclone diameter.
The degree of correlation obtained over this range in velocity is used as a test to verify the
dominance of inertial forces as the principal mechanism of collection.
Statistically it is anticipated that the best accuracy occurs in the vicinity of the mass mean
particle size. Thus it is found that the calculated fractional efficiencies tend to be low for very
large and very small particles in the population. Similarly, tests at low velocity or poor
cyclone configurations afford poor discrimination in the calculation of fractional efficiency and
tests at very high performance can lead to significant experimental scatter. Accordingly, the
data is presented with a line drawn to reflect the "best engineering" judgment.
The following summarizes results of the various experiments.
Influence of Air Shielding
Figure 3 shows the performance of a high flow capacity cyclone with and without a clean
air shield using flyash. Here the air shield data is presented as fractional efficiency, whereas
the non-air shield data is based on overall efficiency. The relatively poor performance in the
latter case prevented an accurate determination in fractional efficiency due to a lack of
discrimination between input-output size distributions. The obvious conclusion is that there
is a pronounced collection advantage in using the air shield feature.
Pressure loss measurements from inlet to ambient showed a loss of 13.5 inlet kinetic
heads, as indicated.
fx
(w
W
I
99.
98.
95. -
90.
80.h
|-
60.
40.
20.
"I 'T~l IN"
Original Air-
Shield Data
Non-Air Shield
Performance
_L
Ap/q =13.5
1.33 D
Cylindrical Body
Conical Body » 2.50 D
Dust Exhaust = 3/8 D
10
-2
10
-1
SEPARATIVE PARAMETER
p d2V.
P P 1
18pD
Figure 3. High flow cyclone: air shield influence
46
-------
Influence of Engagement Length
In an effort to improve performance in the non-air shield configuration, the axial engage-
ment length (e.g., the axial distance between inlet and exhaust) was increased by inserting an
18-inch extension to the exhaust duct and then an additional 12-inch extension. In these
tests, the wooden spool piece was removed so that DelD = 3/4 and the cyclone body
configuration remained constant. The results, shown in Figure 4, showed a marked improve-
ment in performance and an expected reduction in pressure loss. The full extension of
30 inches (engagement length = 2.0 D) showed a slight performance reduction relative to
the 18-inch extension, suggesting dust reentrainment due to increased proximity of the
exhaust to the dip leg. The data also showed less pressure loss, due to improved recovery in
the longer exhaust duct.
99.
Non-Air Shield Performance
Nickel Test Dust
Constant Cyclone Body
Engagement Length A /q
. Flyash
De/D = 1/2 (Fig- 3)
0.65 D
A - 1.33 D
B - 2.00 D
8.3
7.8
I I I I I
-3
10
SEPARATIVE PARAMETER i|)
10
-2
10
-1
p d2V.
P P 1
18yD
Figure 4. High flow cyclone: engagement length influence
Influence of Spin-Up
Since conventional wisdom teaches that higher vortex spin-up (generated by a smaller
exhaust diameter) improves performance, the cyclone was tested in the non-air shield
configuration with the same engagement length at De/D = 1/2 and 3/4. The result is shown
in Figure 5 and indicates a measurable performance penalty and high pressure loss associated
with the constricted outlet. Further increase of the engagement length in combination with
higher spin-up reduced collection efficiency and was indicative of dust reentrainment prob-
lems.
Influence of \ ortex Shield
The indication of dust reentrainment led to the insertion of a vortex shield (conical point
upward) above the dust exhaust, as illustrated in Figure 1. Comparative results shown in
Figure 6 indicate a significant performance improvement.
47
-------
99.
98.
- 95.
90'
B
z
u
S
fc.
< 60.
Z
O
40.
20.
10.
10
-4
Non-Air Shield Performance
Nickel Test Dust
Engagement Length = 1.33 D
Same Cyclone Body
4 /q - 16.2 for Dg/D = 1/2
_L I I I , L-
10-3
SEPARATIVE PARAMETER
10
-2
p d2V.
PP P l
J
-j
10
-1
Figure 5. High flow cyclone: spin-up influence
X
u
99.
95.
90.
80.
5 60.
|
I 40'
(j
t
20.
10.
10
-4
I I I " I I I I I
Vortex shield
in Dip Leg
W/0 Vortex
Shield
Non-Air Shield Performance
Flyash
De/D = 3/4
Engagement Length = 1.33 D
..-L .1 I I I I
-3
10
SEPARATIVE PARAMETER i|<
10
-2
10
-1
p d2V.
P P i
18pD
Figure 6. High flow cyclone: vortex shield influence
Test Dust Correlation
Figure 7 shows the degree of experimental accuracy obtained by using the two different
test dusts. This discrepancy of ± 10% on penetration may be due to differences in particle
reflectivity, which might affect the PILLS measurements, or ± 10% on particle size analysis
of the Climet measurements. In general, excellent correlation is indicated.
48
-------
99.
nf 98.
95.
90.
>. 80.
u
8 70.
u
t* 60.
U
40.
20.
10.
10
-4
•Flyash
Nickel
Non-Air Shield Performance
Engagement Length = 1.33 D
De/D - 3/4
Constant Cyclone Body
I.I I I i I
io-3 2 io-2
SEPARATIVE PARAMETERS i|i = j-p
10
-1
Figure 7. High flow cyclone: test dust correlation
Influence of Cylindrical Length
To assess the importance of cyclone body length, an 18-inch long cylindrical section was
introduced to extend the body. Performance with and without the extension was compared
(both using the vortex shield), and the results are shown in Figure 8. Accounting for the
differences in test dusts, one can conclude that the longer cyclone body is not beneficial if the
vortex shield is used to inhibit dust reentrainment.
X
U
•99.5
f 99.
98.
95.
90.
80. r-
u
J 60.
o
20. -
10.
10
-4
II II
Cylindrical Length = 1.33 D
Vortex Shield in Base
Electrode ^,
Flyash
,_J.-L..L. I. lAliJ
ID'3
SEPARATIVE PARAMETER
i -'"I ' i ' i r IT
Cylindrical Length
= 2.33 D
Vortex Shield
~De/D = 3/4
Engagement Length
- 1.33 D
Exhaust Electrode
Nickel Powder
L_LJ_L
10
-2
10
-1
p d2V.
MP P i
18yD
Figure 8. High flow cyclone: body length influence
49
-------
High Flow Versus Low Flow Cyclone Design
In a parallel activity, testing of a Stairmand low flow or "high efficiency" design with a
scroll inlet was performed using the same experimental techniques and instrumentation.
These results are compared in Figure 9 with the data of the present high flow design, using
increased engagement length and the base-mounted vortex shield. For the same inlet veloc-
ity, the high flow design is found to provide comparable performance at four times higher
flow capacity with only a 38% higher pressure loss. Alternately, the high flow design would
handle the same flow at one-half the diameter of the Stairmand design, thus shifting the »|»
values by a factor of two as shown in Figure 9. With this shift, it is found that the high flow
design provides substantially superior performance with a major size/cost savings. The impact
of cyclone size on capital cost is anticipated to be particularly important for high pressure
cleanup systems due to the required pressure vessels.
u
H
U
99.5
£ 99. -
98. -
95. _
90. -
80.
60.
§ «.
B
3
Comparable Performance
At Same Flow Capacity
High Flow Cyclone
Non-Air Shield with Vortex Shield-
De/D =0.75
Q = 0.405 D2Vi
Ap/q =8.3
20.
10. U
5.
ID'4
Stairmand High Efficiency Cyclone
D /D = 0.5
Q = 0.1 D2Vi
Ap/q = 6.
-3
10
SEPARATIVE PARAMETER
10
-2
10
-1
18pD
Figure 9. Low flow versus high flow cyclone performance
CONCLUSIONS
This investigation implies that typical high performance cyclone designs are inhibited in
two significant areas. With small inlets and small outlets, it appears that large scale turbulence
inhibits collection. The use of a large inlet contracting the flow into a relatively narrow annu-
lar region minimizes the production of large scale turbulence. Similarly, increased engage-
ment length improves the length-to-annulus width and suppresses the turbulence that is
formed. It is postulated that inlet vanes or baffling might play a similar role.
50
-------
The use of air shielding is found to be an alternate means of performance improvement
and is of special interest in systems that can use clean dilution gas, e.g., cyclone combustors.
The performance improvement found with the vortex shield implies that long cyclone
bodies are not necessary to avoid dust reentrainment if this device is employed. This sug-
gests that the cyclone body could be purely cylindrical with an inverted base and peripheral
dust exhaust. This would result in minimizing the swirl velocity and the radial pressure gra-
dients in the vicinity of the dust exhaust.
Contrary to conventional wisdom, it was found that increased vortex spin-up did not
improve collection with the high flow design. It is theorized that this result is conditioned by
the lack of the vortex shield during those tests, e.g., increased spin-up intensifies the propen-
sity of the vortex to induce swirl and recirculation in the dust exhaust, which increases dust
reentrainment.
Regarding analytical implications, it is noted that the importance of the inlet flow and dust
reentrainment found in this investigation is not treated by present analytical models. At the
present time, adequate means for analytically treating these two critical areas are not available,
and continued cyclone development must be guided by experimental methods.
One area not addressed in the present work is the suppression of pressure loss. It is antic-
ipated that significant pressure recovery could be achieved with an exhaust swirl diffuser.
SUMMARY
Empirical investigations show very substantial improvements in cyclone design by focusing
on minimizing inlet turbulence scale and dust reentrainment at the dust discharge. These
results indicate that a new high flow design can provide equal or superior collection with a
cyclone of half the diameter of current high efficiency cyclones with some increase in pressure
drop.
ACKNOWLEDGEMENT
The author wishes to acknowledge the support provided by K.E. Markel, Jr., Project
Manager, Coal Projects Management Division, U.S. Department of Energy, Morgantown,
West Virginia, in modifying and expanding the work scope of DOE Contract DE-AC21-80 ET
17091 whereby the above investigation was pursued.
The work described in this paper was not funded by the U.S. Environmental Protection
Agency and therefore the contents do not necessarily reflect the view of the Agency and no
official endorsement should be inferred.
REFERENCES
1. Leith, D. and Licht, W., "The Collection Efficiency of Cyclone Type Particle Collectors
- A New Theoretical Approach," AICHE Symposium Ser., Vol. 68, No. 126, p. 126, 1972.
2. Giles, W.B., General Electric TIS Report No. 76CRD023.
3. Dietz, P.W., "Collection Efficiencies of Cyclone Separators," General Electric TIS Report
No. 79CRD244, December 1979.
4. Boysan, F., Ayers, W.H., and Swithenbank, J., "Cyclone Design Fundamentals,"
University of Sheffield paper (unpublished).
51
-------
5. Giles, W.B., "High Flow Cyclones for PFBC Hot Gas Cleanup," General Electric TIS
Report No. 81CRD197, 1981.
6. van Tongeren, H., Mech. Eng'r., 57, p. 753, 1935.
7. ter Linden, A.J., "Investigations Into Cyclone Dust Collectors," Inst. of Mech. Eng'rs., J.,
Proc. Vol. 160, p. 233, June-December, 1949.
8. Stairmand, C.J., "The Design and Performance of Cyclone Separators," Trans. Inst.
Chem., Eng'rs., Vol. 29, p. 356, 1951.
52
-------
CYCLONE SCALING EXPERIMENTS
by
W. B. Giles
Mechanical Systems and Technology Laboratory
Corporate Research and Development
General Electric Company
Schenectady, New York 12301
ABSTRACT
A series of geometrically similar cyclones of conventional, high-efficiency design was test-
ed to assess the normally accepted perception that cyclones act as an inertial collection device
and therefore can be scaled from model to prototype size by an inertial separative parameter.
These tests were conducted for three different cyclone sizes of 4, 12, and 36 in. diameter over
a range of inlet velocity and at atmospheric pressure.
In using test dusts which had been shown to have a low propensity for triboelectric charg-
ing, good correlation was observed.
Tests were also conducted using a test dust which has been found to have a high propensi-
ty for triboelectric charging. The data does not correlate, has very high efficiency, and is
characterized by relatively constant overall efficiency versus cyclone flow. The latter behavior
is noted in several literature sources.
The critical user is, therefore, cautioned in the acceptance of data unless, as a minimum,
the fractional efficiency can be shown to correlate with the inertial separative parameter over a
range in velocity.
53
-------
INTRODUCTION
Cyclone art and theory has evolved around the perception of an inertial-fluid mechanic
mechanism of collection. The swirling gas flow induces a centrifugal force on the convected
dust particles, which drives the particles to the wall, where axial convection transports the
particles to the dust discharge. Thus it is expected that the ratio of centrifugal force to parti-
cle drag force, together with the geometric design, will describe the mechanics of collection.
However, earlier experience (1) has shown that natural electrostatic forces can play a highly
significant role in the collection efficiency of cyclones. This effect is to enhance performance,
particularly at low operating velocities. As a consequence, model test data can lead to highly
erroneous, and optimistic, expectations if natural electrostatics are present in the experiment.
Unfortunately much of the literature data is reported only at one test velocity, and from
this it is not possible to assess whether or not electrostatic forces are operative. A review of
the literature also discloses behavior which can now be interpreted as natural electrostatic
enhancement. For example, Petroll and Langhammer (2) show approximately flat cyclone
efficiency versus flow rate. Also Ludewig (3) notes the same effect in contradiction to expec-
tation. He also cites the literature of ter Linden (4), Barth and Trunz (5), and Rammler and
Breitling (6). Similar behavior was disclosed in discussion with Kraftwerk Union. Therefore,
a significant risk can exist in data extrapolation.
The purpose of the present investigation was to investigate a geometrically similar design
in three different sizes to assess the degree to which correlation of the data could be obtained.
In prior experiments, it was found that the phenomenon of natural electrostatic enhance-
ment was due to triboelectric charging of the dust particles brought about by particle-wall im-
pact. Specifically, a Faraday cage was used to measure the induced test dust charge level, as
shown schematically in Figure 1. In this arrangement, gas borne dust is admitted, enshroud-
ed by clean air. The Faraday cage then detects the image charge present in the plume, and
optic equipment (not shown) is used to measure dust concentration. In this manner, the
charge level per particle can be determined. It was noted that if dust was admitted directly
from the fluid bed dust generator, the charge density was low. However, by using a long
Q?
>^"" Coiled Metal Tubing
it.
Fluid Bed
Dust Generator
Figure 1. Faraday cage experiment
54
-------
length of coiled metal tubing to ensure particle-wall impact, it was found that the charge level
of PFB test dust from Exxon's Miniplant facility in Linden, New Jersey, was two orders of
magnitude higher than PFB test dust from the National Coal Board Coal Utilization Research
Laboratory (CURL) at Leatherhead, England. Furthermore, it was noted that the current
flux into the dust was comparable to the current flux from a collecting cyclone. Therefore, it
is expected that particle charging can occur internal to the cyclone given the right combination
of materials, and that this accounts for the extreme difficulty in eliminating the effect, if
present.
The mechanism of efficiency enhancement is theorized to be due to the mutual repulsion
forces of the resulting space charge.
It was also found that electrostatics can play a major role in sampling error. If the sam-
pling probe for particle analysis is electrically insulated from the transport dust, the probe be-
comes charged with very high gradients at the probe intake. This drastically reduces sample
counts and shifts the size distribution (7).
With these facts in mind, CURL fiyash was selected as the principal test dust and the sam-
pling probes were grounded to avoid the anomalies of electrostatics.
TEST MODEL
The design selected for study consisted of a Stairmand (8) high-efficiency cyclone modified
by use of a four-point scroll inlet. The inlet flow capacity is characterized by
Q = 0.1 D2 V,
with a pressure loss of 5.29 inlet kinetic heads. This design selection was based on its prom-
inence in the literature and its similarity to the work of ter Linden, van Tongeren, and others
so as to be representative of the state of the art. Three sizes were manufactured and tested at
4, 12, and 36 in. diameter, as shown in Figure 2.
. D/2
0.5Dx0.2D
D/6
1.5D f"
i
SIZES
Scroll-Top View
2.5D
.375D
1 Inches
12 inches
36 Inches
0.65D
Figure 2. Test model design: Stairmand high-efficiency with four-point inlet scroll
55
-------
SCALING PARAMETER
These experiments sought to assess the validity of the inertial separative parameter,
given as the ratio of centrifugal force to the Stokes drag force, or
18/* D
where pp is the particle density, dp is the particle diameter, V, is the average inlet velocity, ^ is
the absolute viscosity, and D is the diameter of the cyclone. The collection efficiency mea-
surements are, therefore, correlated as overall efficiency, TJO, or fractional efficiency, •»?/,
versus ¥. All of the parameters were varied except that of gas viscosity.
TEST PROCEDURE
These experiments consisted of blowing prefiltered room air at atmospheric conditions
through the cyclone. A fluid bed dust generator was used to contaminate the supply air at
relatively dilute dust loadings, several pipe diameters upstream of the test cyclone. Typical
dust distributions are shown in Figure 3.
The performance of the units was determined through optical measurements of inlet-
outlet dust loadings and inlet-outlet size distribution. PILLS V Mass Concentration Monitors1
were mounted with optic windows (purged ducts at right angles to the flow) and used to
determine inlet-outlet dust loadings. Particle size analysis was performed using two Climet
Particle Systems.2 The flow was sampled isokinetically and vacuum pumped through the in-
struments. To avoid coincidence errors with this equipment, dilution of the sample flow
stream was required. A specially developed General Electric sampling system was employed
to provide the functions of isokinetic sampling, flow measurement, dilution, and matching to
the Climet systems. Finally, calculations converted this primary data into the form of frac-
tional efficiency.
Measurements of cyclone pressure loss, from inlet to ambient, were also taken and nor-
malized by inlet kinetic head, 1/2 p V?, based on average inlet velocity.
TEST RESULTS AND DISCUSSION
The test results of the 12 in. diameter model for different inlet velocities are shown in Fig-
ure 4 for the CURL flyash and Figure 5 for nickel powder. The data for the 36 in. diameter
model is shown in Figure 6 for CURL flyash. The data set for the 4 in. diameter model is
shown in Figure 7 for CURL flyash and Figure 8 for nickel powder.
A composite of all of the fractional efficiency data is shown in Figure 9, exclusive of the
nickel data in the 4 in. diameter cyclone. Figure 10 shows a composite of the overall
efficiency data of all of the above tests.
In general, it is noted that there is good correlation of the fractional efficiency data within
the context of the experimental scatter. This indicates that inertial collection is clearly the
dominant mechanism.
Inspection, of the nickel data of the 4 in. cyclone, Figure 8, shows a marked lack of corre-
lation. Efficiencies of the higher velocity test conditions, > 20 ft/sec, are markedly lower
than similar data with flyash, Figure 7. It is theorized that this indicates a problem with parti-
cle bounce that is not scalable via the separation parameter.
1 Manufactured by Environmental Systems, Inc., 200 Tech Center Drive, Knoxville, Tennessee 37917.
2 Manufactured by Climet Instruments Company, 1320 W. Colton Avenue, Redlands, California 92373.
56
-------
Q
a
I
Dust Generator: Fluid Bed
.6 .8 1. 2. 3. 4. 6. 8. 10.
PARTICLE DIAMETER (microns)
20.
Figure 3. Typical test dust distributions at cyclone inlet
99.5
99.
98.
95.
90.
80.
F^TTT TTT
£ 50-
d 40- '-
20. i-
10.
5. t-
10
ni
Test Dust: CURL 2nd Stage Flyash
Flow Inlet Velocity P * '
ft/sec
O
a
O
A
V
cfm
720
544
385
272
172
122
J i- I
-Li_i_i_ij_i_L
io-3
__J. -i._L iJ_i I -1J_LL1_ .
io-2
120.0
90.7
£4.2
45.4
28.7
20.3
J i 1,1.1 I I I I
10
-1
SEPARATIVE PARAMETER
18uD
Figure 4. Fractional efficiency versus separative parameter for 12 in. diameter high-efficiency
cyclone
57
-------
99.98
40
20.
10.
SEPARATIVE PARAMETER \|i =
Figure 5. Fractional efficiency versus separative parameter for 12 in. diameter high-efficiency
cyclone
nf
*>
g
W
EFFIC]
§
FRACTI
S3. 3
99.
98.
95.
90.
80.
70.
60.
40.
20.
10.
5.
2.
- ' | • I • I • I I 1 1 1 1 I -^~T
^
^ D
A O
^ O A
D 0
7 A 0
^
O
*
- *v7 * Test Dust: CURL 2nd Stage Flyash _
^7 P - 2.7
-
O
§
A
4
*
i , 1 , I , I 1 1 1 I 1 , 1,1
Flow Inlet velocity
cfm ft/sec
7236 134.0
5454 101.0
4304 79.7
3202 59.3
2527 46.8
1755 32.5
no v» ijT
. 1 . 1 1 1 1 1 1 , 1 .
10
-4
-3
10
SEPARATIVE PARAMETER if
10
-2
IByD
Figure 6. Fractional efficiency versus separative parameter for 36 in. diameter high-efficiency
cyclone
58
-------
nf
X
a
H
o
M
111
d
0
B
99.9
99.8
99.5
99.
98.
95.
90.
80.
70.
60.
40.
20.
10.
5.
111 ' i ' i ' i ' i i i M i ' " i T T ~rr^ r~[~\ r~r
D
_-^
& .^^^
Curve of 12" £ 36" & /^
Cyclones :> ,_, ,/J ^ Q
/^ A
A O D*
A*
X^
Test Dust: CURL 2nd Stage Flyash -
/ P =2.7
/ P
/ C~> Flow Inlet Velocity
./ O cfm ft/sec
/ O 80 120
^ D 60 90
& O 43 65
A 30 45
^7 20 30
A 15 20
-^J ^ no vs if _j
_ _
til , 1 . 1 . 1 . I 1 1 1 1 1 . 1 . 1 . 1 . 1 1 1 1 1
10
~3
10~2 p d'v.
Fp p 1
SEPARATIVE PARAMETER * = 18yD
10
-1
Figure 7. Fractional efficiency versus separative parameter for 4 in. diameter high-efficiency
cyclone
nf
Of
y
55
td
O
H
In
h
Cd
^
O
H
1
b*
99.9
99.8
99.5
99.
98.
95.
90.
l
80.
70.
60.
40.
20.
10.
i ' i ' i ' i i i i i i ' i ' i ' i ' i i i i i. ' IA i V-I
4 ^ o D
r- V
^ AD
* V * 0
A V ^ *
/i A
~ ^7 A
O
5 ^7 0
A v Test Dust: Nickel Powder p • 8
- A Flow Inlet Velocity
cfm ft/sec
Q D 60 90
O 43 65
A 30 45
i- V 20 30
A 15 20
+ n« vs H
-
i.i ii i i i i . i i i I i i i i i . 1.1.1.
'lO'3 10~2 10"1
p dnvi
SEPARATIVE PARAMETER * - ? «*.
1
-
-
-
-
-
—
-
5
-
~
—
-
—
-
|
Figure 8. Fractional efficiency versus separative parameter for 4 in. diameter high-efficiency
cyclone
59
-------
99.98
99.9
99.8
99.5
99.
98.
90.
70.
60.
MM
o
D
O
D
D
D % •
%00 O
a P
PO A
•
O
A
D u A
O O Cf o A
O ° O
D
A
O
J . I . L I I I. U , |_, 1,1, I_LJ_U_L_
io"3 io"2
*.&
• •• • •
O 12 inch Diameter Cyclone, Flyash
• 12 inch Diameter Cyclone, Nickel
O 36 inch Diameter Cyclone, Flyaah
A 4 inch Diameter Cyclone, Flyash
I I I
_L_^i_
10
SEPARATIVE PARAMETER \|> »
Figure 9. Fractional efficiency versus separative parameter for high-efficiency cyclone
99.9
99.8
99. \-
i
98. —
95. -
1
; 90-
80. -
I
70. f-
60. 'r
10. -
r r~r ' i '"I T M n
o
D
a
o o
rrrrr -
LEGEND
36 inch Diameter Cyclone
12 inch Diameter Cyclone
4 inch Diameter Cyclone
Flyash
2.
1.
10"
...i , i... i
I.I I I I I I I , L_ I . I . I I I I I I
i«-2 i«-l
•._! . I . I . I I I I I
SEPARATIVE PARAMETER *
Figure 10. Composite of overall efficiency measurement of high-efficiency scaling cyclones
60
-------
As noted previously, the direct measurements consist of overall efficiency plus input-
output size analysis. From these measurements, the fractional efficiencies are computed. It
follows that the variance that is evident between fractional and overall efficiencies (see
Figures 4-8), particularly at the higher test velocities, is a direct consequence of the relatively
steep slope of the fractional efficiency characteristic and the polydispersed nature of the test
dust.
The final criterion of performance is, of course, the fractional efficiency. For example, the
data of overall efficiency versus separative parameter, Figure 10, suggests that the particle
bounce mechanism is scalable. Here the data suggests a continuous, universal curve. This
suggestion, however, is belied by the fractional efficiency data of Figure 8 with nickel. This
discrepancy shows that the data does not correlate with the separative parameter and indeed
indicates that some other nondimensional group is required to correlate this phenomenon.
Testing with Exxon flyash, Figure 11, shows an extreme lack of correlation over the whole
velocity range. The test dust has a known propensity toward charge generation, and the
efficiencies at low velocity are significantly higher than the corresponding data for CURL
flyash, Figure 7, except for the highest test velocity. It is concluded that triboelectric particle
charging is playing a major role in collection and that this data is not scalable.
Therefore, in order to validate model test data, experiments must be conducted over a
range in test velocity to demonstrate correlation. If reasonable agreement is found, inertial
scaling can be employed to provide a conservative estimate of prototype performance. Unfor-
tunately, if correlation is not found, methods must be found to eliminate, control, and/or
quantify the relevant phenomena.
SEPARATIVE PARAMETER
18yD
Figure 11. Fractional efficiency versus separative parameter with 4 in. diameter high-efficien-
cy cyclone
61
-------
Pressure loss measurements of the three units were normalized with inlet kinetic head, q
= 1/2 p V?, and are shown in Table 1 in comparison with the reported value by Stair -
mand (8). The results are found to be self-consistent and 11% higher than reported by Stair-
mand. This is not considered a significant effect. Difference in pressure tap location could
account for the discrepancy.
TABLE 1
Pressure Loss Coefficients (Inlet to Ambient)
Stairmand
Scaling Cyclones
4 inch model
12 inch model
36 inch model
Ap/q =
Ap/q =
Ap/q =
Ap/q =
5.3
5.7
6.0
5.8
CYCLONE DATA COMPARISON
Figure 12 compares the present data against other sources of evaluation on high-
performance designs. All have similar flow capacities, long cyclone bodies, and relatively
small exhaust-to-barrel diameter ratios, De/D. The Buell1 unit is basically similar to a van
Tongeren design. It is noted that the present data is significantly higher than the data report-
ed by Stairmand (8), whereas the recent data by CURL on a van Tongeren cyclone (9) is
significantly lower than its generic data base.
Conventional theory argues that smaller exhaust-to-barrel ratio would, by the conservation
of angular momentum, result in higher flow spin-up and hence provide higher particle centri-
fugal forces and superior collection. The relative agreement evident between the present GE
data and the CURL data suggests that there is little, if any, advantage in the higher spin-up of
the van Tongeren design (Z>e/D = 0.30). The van Tongeren or Buell design does have a
higher pressure loss coefficient of ~ 7.7 inlet kinetic heads, versus ~ 5.9 for the Stairmand
design. Parallel studies of an improved high flow design (10) also found that increased spin-
up did not enhance performance as would normally be expected. A possible explanation is
that the advantages of higher flow spin-up are offset by the increased vortex strength inducing
high reentrainment at the dust discharge. An alternate hypothesis, in the present case, is that
large-scale turbulence at the inlet inhibits performance.
TRIBOELECTRIC CHARGING
To further explore the influence of natural electrostatic enhancement, as noted in the 4 in.
diameter model, testing of the 12 in. diameter was conducted with Exxon flyash. These
results are shown in Figure 13. Again, the data with Exxon flyash show significantly higher
collection efficiency and lack of correlation, relative to the CURL flyash data. The latter is
shown transposed for comparison. For example, at ¥ = 10~2, the collection efficiency is in-
creased from = 97% to = 99.5%, a four-fold reduction in penetration due to electrostatic
effects. It is also noted that the high velocity performance loss is not evident with this model
as was seen with the 4 in. model. Apparently the bouncing mechanism is strongly dependent
on cyclone size.
Now GE Environmental Systems, Inc., Lebanon, Pennsylvania.
62
-------
99.9
99.8
99.5
99.
98.
95.
80-
70.
60-
40.
20.
10.
Buell High Performance
D /D = 0.32
Stairmand'81
High Efficiency
De/D - 1/2
Legend:
>Present CRD Data of
Stairmand High Efficiency
D /D .1/2
I . I
All Units Q = 0.1 TrvL
CURL Data - Van Tongeren
Cyclone, De/D -0.3
_JL I ,1,1.
10
10
SEPARATIVE PARAMETER <|i
10
-2
18yD
Figure 12. Comparison of high-performance cyclone data
99.9
99.8
99.5
99.
98.
95.
80.
70.
60.
B
si 40.
20.
10.
10
-4
Test Dust: EXXON Flyash, p - 2.7
I . I
I i . I . I i I I i
ID' lO
p d'V.
SEPARATIVE PARAMETER * - P P 1
18pD
'2
10
-1
Figure 13. Fractional efficiency versus separative parameter for 12 in. diameter model
63
-------
SUMMARY
It was found that the cyclone performance may or may not correlate with a simple inertial
separation parameter. On one hand, parametric variations of 9 to 1 in cyclone diameter, 3 to
1 in particle density, 5 to 1 in velocity, and 6 to 1 in particle size can show good correlation,
and hence data from model tests can be used to project prototype performance. On the other
hand, extreme variations can exist if other phenomena are operative. The most convenient,
and necessary, means of validation is to conduct experiments over a range of velocities.
Phenomena can exist that prohibit scaling from model test conditions to prototype opera-
tion. Triboelectfic charging tends to enhance performance and particle bouncing tends to de-
grade performance. Both can produce large effects and both are poorly defined, thus a priori
determination of the probability of these phenomena is not available. Present indications are
that the condition of particle bounce may be restricted to the use of very small cyclones and
high velocities. The effects of triboelectric charging are perhaps much more insidious. A
literature review suggests that anomalous behavior in many experiments can now be rational-
ized, and, unfortunately, much of the literature must be treated with considerable skepticism.
ACKNOWLEDGMENT
This work was supported by NYS ERDA, under Contract No. 344-ET-FUC-81, through
the sponsorship of General Electric's Energy Systems Programs Department. General
Electric's Corporate Research and Development provided the test models.
The work described in this paper was not funded by the U.S. Environmental Protection
Agency and therefore the contents do not necessarily reflect the view of the Agency and no
official endorsement should be inferred.
REFERENCES
1. Giles, W. B. Electrostatic separation in cyclones. In Symposium on the Transfer and Util-
ization of Paniculate Control Technology, EPA-600/7-79-044C, Vol. 3, February 1979, p.
291.
2. Petroll, J. and Langhammer, K. Comparative tests on cyclone precipitators. In Freiberger
Forschungsheft, Vol. A220, 1962, pp. 175-196.
3. Ludewig, H. Cyclone model experiments regarding the effect of the dip pipe depth on
separating efficiency and pressure drop. In Maschinenbantechnik, Vol. 7, No. 8, 1958,
pp. 416-421.
4. ter Linden, A. J. Investigations in cyclone separators In VDI Seminar, Vol. 3, 1954, VDI
Verlag.
5. Barth, W. and Trunz, K. Model test with water stream cyclone separator for predetermin-
ing removal efficiency. Z. F. Angew, Math and Mech., Vol. 30, No. 8/9, 1950.
6. Rammler, E. and Breitling, K. Comparative tests with centrifugal separators. In
Freiberger Forschungsheft, Vol. A56, 1957.
7. Giles, W. B. and Dietz, P. W. Electrostatic effects on sampling through ungrounded
probes. Second Symposium on the Transfer and Utilization of Particulate Control Tech-
nology, EPA-600/9-80-039D, Vol. 4, September 1980, p. 387.
64
-------
8. Stairmand, C. V. The design and performance of cyclone separators. Trans. Inst. Chem.
Engrs., Vol. 29, 1951, p. 356.
9. Advanced cleanup device performance design report (Task 4.3) - Volume A - Cyclone
theory and data correlation of PFB CFCC Development Program. U.S. DOE DE-AC21-
76ET10377, Dist. Category VC-90e, FE-2357-70, pp. 3-34.
10. Giles, W. B. High flow cyclone development. GE Report, to be published.
65
-------
TEST METHODS AND EVALUATIUN UF HIST ELIMINATOR CARRYOVER
by: Vladimir Boscak, Atef Oemiari
General Electric Environmental Services, Inc.
2UU North Seventh Street
Lebanon, Pennsylvania 17042
ABSTRACT
A test program was carried out at GEESI's R&U pilot plant to determine
mist eliminator efficiency, carryover luad and droplet size distribution from
a vertical flow mist eliminator. The modified EPA Method 5 was used to
determine efficiency and carryover load. The carryover load when the
scrubber was operated under standard operating conditions but without mist
eliminator washing was 28 to 60 rng/Nm3D (0.012 to 0.024 gr/SCFD). When the
bottom of the mist eliminator was washed, the carryover load above the washed
section was 70 to 160 mg/Nm3U (0.029 to O.U65 gr/SCFO). Mist eliminator
efficiency was greater than 99%. A droplet photography technique was used to
determine carryover aerosol size distribution. The average aerosol size
measured above the mist eliminator was about 100 to 200 microns. Mist elimi-
nator inlet size distribution averages about 140 microns. The carryover is
probably caused by re-entrainment from the mist eliminator blades.
66
-------
INTRODUCTION
Wet scruboing flue gas desul furization systems utilize mist eliminators
to separate and remove scrubbing-1iquid droplets (aerosols) contained in the
flue gas. Baffles or chevron elements remove aerosols by inertial forces
(1). Aerosol removal is needed to prevent carryover of suspended solids,
dissolved salts and liquid to the stack as well as to avoid incrustation and
corrosion of downstream system components.
There are two principal mist-eliminator conflyurations: horizontal gas
flow and vertical gas flow types. The vertical flow arrangement has been
regularly used in this country while the horizontal flow type is more common
in Japan and Germany. Vertical flow eliminators generally use the cnevron
type design which incorporates continuous zigzag baffling comprising 2 to 6
passes. This design is favored for strength, low gas pressure drop and cost
considerations. Typical design of the mist eliminator has vane spacing of
1.5 to 3.0 inches, plastic construction is most common, and wash systatis
typically operate intermittently to conserve plant water (2). EPRI's review
of commercial FGD operating experience lists plugging/seal ing, erosion,
corrosion and inefficient performance of mist eliminators as detracting from
high operability (3).
OPERATING EXPEDIENCE WITH MIST ELIMINATORS
When aerosols entrained in flue gas are not removed in the mist elimina-
tors, a number of problems may arise:
High particulate emission exiting stack
Deposition in ductwork and stack
Corrosion of ductwork and stack
"Rain" around stack
Corrosion - erosion of gas-gas heat exchanger
(when used for reheat)
One of the major reasons for mist eliminator operating problems in early
FGD installations was insufficient understanding of FGD process chemistry.
Poor performance of mist eliminators was a direct result of fouling of
baffles causing upset of design for flow conditions. The fouling was a con-
sequence of deposition of soft solids as well as formation of hard scale from
precipitation of solids from a CaS04 - supersaturated liquid.
Four-pass chevron type mist eliminators are used in General Electric
Environmental Services, Inc. (GEESI) FGD installations. The advantages of
this mist eliminator includes a high efficiency, low pressure drop and ease
of cleaning. This type of mist eliminator has been successfully used in most
of GEESI's full-scale installations.
One of the problems in cleaning of mist eliminators is limited amount of
available water. Occasional plugging, excessive aerosol carryover and depo-
sition has been reported at FGD installations. The causes of the problems
67
-------
were, however, traced down to operating changes like above design boiler
load, overwashing of the mist eliminators, Taulty valves and use of poor
quality washing liquor frequently with high sulfate content.
The performance of GLESl's mist eliminators has been evaluated both on
pilot and full-scale units. Figure 1 shows a full-scale mist eliminator and
wash spray banks for sequential washing (4).
Figure 1. Mist Eliminator and Wash Spray Bank
The evaluation of a domestic full-scale unit by an independent testing
laboratory indicated that mist eliminator efficiency was high even in
removing subrnicron dust. The carryover load was reported to be in the range
of 0.014 to 0.033 gr/SCFU (34-80 mg/Nm3D). Another overseas installation
reported higher carryover loads but results are somewhat doubtful because of
non-isokinetic sampling technique and use of Ca tracer. In second overseas
evaluation a pilot mist eliminator carryover particle size distribution was
determined using a combination of cascade impactor, MgU-impactor and
proprietary paper impactor, but results were inconclusive.
Since there are some reservations about the best method for measuring
carryover testing GEESI decided to perform tests on a pilot unit. The same
mist eliminator that is employed in a full-scale installation is installed in
the pilot unit.
OBJECTIVES
The general purpose of the mist eliminator test program was fourfold:
(1) Determine the best method for mist eliminator efficiency under
standard operating conditions.
68
-------
(2) Measure aerosol carryover (load) during and without washing.
(3) Determine carryover droplet size distribution.
(4) Evaluate applicability of modified EPA method 5 and droplet photo-
graphy technique for a full-scale mist eliminator test.
FLUE GAS DESULFURIZATIUN PILUT PLANT
The GEESI wet FGU pilot plant is located in Lebanon, PA. It consists of
a spray absorber (3 feet diameter), recycle tank, thickener, drum filter,
pumps and fan. The bottom of the spray tower (8 feet diameter) serves as a
delay tank from which scrubbing media is recycled to the spray absorber.
Figure 2 shows FGD pilot plant's spray absorber. Five hollow cone spray
nozzles are used for atomization of scrubbing media in a spray absorber. The
nozzles spray angle at pressure of 2U PSIG is 54° at 3 feet vertical distance
from nozzle orifice and the arithmetic mean droplet size of a slurry is about
20UU microns.
Figure 2. FGD Pilot Plant Spray Absorber
An open chevron type mist eliminator is located at top of absorber to
remove entrained slurry droplets from the gas stream. It consists of four
polypropylene single vane blades arranged to form an open design, four pass
unit. Mist eliminator wash system consist of six spray nozzles positioned in
the center of six equal areas on the bottom and another six identical nozzles
on the top of mist eliminator. These wash spray nozzles when operated under
standard conditions generate droplets in the range of 525 to 770 microns.
Hot flue gas from an oil fired furnace is mixed with atmospheric air and
passed through absorber where it is scrubbed with liquid media. Gas then
proceeds through the mist eliminator where entrained aerosol is removed prior
to discharge to the atmosphere. Desired levels of S02 is achieved through
addition of SO? from storage cylinder.
69
-------
The pH of the recirculating slurry is maintained at preset value either
by pH controller which controls the addition of lime (or limestone) to the
system or by feeding lime/limestone continuously at certain rate. To main-
tain the material balance, a small portion of slurry is bled from the recycle
system and disposed of after dewatering in a hydroclone/vacuum filter com-
bination. The filtrate from the dewatering system is recycled.
CARRYOVER TEST METHODS
After a thorough review of test methods suitable for carryover evaluation
it was decided to use a modified LPA method 5 for carryover load measurement
and droplet photography for aerosol particle size distribution determination.
MODIFIED EPA METHOD 5
A modified EPA Method 5 sampling train was used to extract gas from the
stack for carry-over load measurements. In this train, a stainless steel
sampling nozzle of an appropriate diameter is connected to a stainless steel
probe by Swagelok fittings. An S-type pitot tube and thermocouple are incor-
porated into the probe to measure gas velocity, pressure, temperature and
temperature of the probe. The probe is heated to prevent condensation. The
major distinction of modified from standard EPA Method 5 is that filter is
not used in the sampling train. A glass cyclone is contained in heated
sample box and the probe is connected to the cyclone inlet by leak-free glass
fitting and the cyclone outlet is connected to the impingers by L-type glass
connector. Greenburg-Smith type impingers are used in the impinger assembly.
Four impingers are connected in series with leak-free glass fittings. The
second impinger has a standard Greenburg-Smith tip, the other tips are
modified by replacing the standard tip with 1/2 inch inner diameter glass
tubing extending to within 1/2 inch of the flask bottom. The first and
second impinyers each contained 10U ml of de-ionized water, the third is
empty and the fourth contain known weight of silicon gel. These four
impingers was kept in an ice bath. A thermometer is placed at the outlet of
the fourth impinyer for monitoring purposes.
From the fourth impinger the extracted gas stream flowed through a sample
line, vacuum gauge, a vacuum pump and a dry gas meter. A calibrated orifice
completed the train and was used to measure instantaneous flow rates. The
dual manometer across the calibrated orifice is used to measure pressure
drop. During the test all the. parameters such as sampling time, temperature,
dry gas meter reading, pressure, etc., at each traverse point are recorded.
Velocity pressure is measured continuously and adjusted to maintain iso-
kinetic rate. Initial and final leak tests are performed on the sampling
train prior to sampling and upon completion of each test. At the end of the
test the nozzle, sampling probe and connecting tubes to the impinger are
washed and then wash liquor is added later to the impingers liquid. The
volume of water in each impinger is measured and recorded on the data sheet.
The combined liquid is kept for chemical analysis.
70
-------
CHEMICAL ANALYSIS PROCEDURE
Combined wash water and impingers liquid is adjusted to 5UO ml through
addition of deionized water.
Initial and final recycle slurry samples (before sampling and upon
completion of each test) are filtered and diluted with deionized water to
desired volume.
All samples are analyzed for chloride using volumetric analysis (Mohr
method).
Sodium and lithium are analyzed by flame atomic absorption (AA) tech-
nique. Sodium is analysed at 5tfy.O nm wave length, while lithium is analysed
at 670.8 nm. In case of lithium the samples are concentrated by evaporation
prior to analysis since concentration is below method's sensitivity.
In analysis of chloride, sodium and lithium it is assumed that all spe-
cies are present in the liquid in ionic form and the concentrations in
carryover are the same as in the slurry or wash liquid. When mist eliminator
wash liquid is labeled with sodium or lithium, the concentration of this com-
ponent in recycle slurry inadvertantly increase during the test.
Consequently a correction is made in concentration calculation to distinguish
between contributions from slurry and wash liquid. Another correction is
made to adjust for presence of these species in the deionized water.
DROPLET PHOTOGRAPHY METHOD
This method employs an automated 35 mm camera tied into a drive unit,
optical source and ultrahigh speed strobe unit to determine droplet particle
size distribution. The optic source and the strobe unit are located opposite
to the camera. Photographs are taken when the droplets passed through the
slit between the camera and the optic source.
The negatives are developed and the droplet size is determined using
proprietary analytical technique.
71
-------
Figure 3. Two Aerosols in a Typical Photograph for Droplet Photograph
Droplet photography method has been used before for mist eliminator eva-
luation and is considered an adequate method for determination of droplet
size distribution (5).
EXPERIMENTAL PROCEDURES
During mist eliminator carry-over study the pilot unit was operated under
standard operating conditions with gas velocity of about 10 ft/s and L/G
about 50. In the preliminary tests the unit was operated with gas recycle
and results were widely scattered. This was attributed to distorted gas flow
pattern above mist eliminator. A distorted flow pattern of flue gas is uni-
que to pilot plant where gas recycle is used and would not be encountered in
a commercial unit. Once the top of the tower was opened and unit was
operated on a once through basis the gasflow distribution improved but was
still somewhat skewed because of mist eliminator blades' angle. Measurement
of gas flow distribution below the mist eliminator indicated rather uniform
flow pattern. Two tests without mist eliminator wash were performed using
three 120° traverses dictated by the absorber sampling ports.
72
-------
The tests with mist eliminator wash were performed at a single sampling
point (above wash nozzle in operation) with close to average gas velocity.
Tne reason for this sampling procedure was the irregular coverage of mist
eliminator with wash liquid which would prevent representative sampling if a
traverse was used. Consequently, carry-over loads in these tests apply only
to the mist eliminator areas covered by washing liquid. In two tests with
wash the spray liquid was labeled with chloride while wash water was labeled
with sodium.
Final two tests were carried out with washing but both spray and wash
liquid were labeled with same cone, of chloride so that only total carry-over
could be established. The total sampling time in all above tests was about
one hour. No carry-over tests with top wash nozzles in operation was carried
out, because the sampling was not possible.
Droplet photography tests consisted of four runs. The first traverse was
carried out above mist eliminator perpendicular to the blades without wash.
The second run was under the mist eliminator. The third run covered an off
center traverse with and without the wash. The final test was at the
selected points (velocity of about 1U fps) with and without the wash.
RESULTS AND DISCUSSION
PRELIMINARY TESTS
In preliminary tests, correlations between mist eliminator operating
parameters and its pressure drop as well as percent coverage with liquid at
the top were established. Table 1 shows influence of gas velocity, spray
absorber L/G ratio and mist eliminator wash rate on pressure drop. Only one
bottom wash nozzle was operated in these tests. The wash rate was increased
from 0 to 2.5 gpm during runs at constant gas velocity and L/G ratio. The
major finding of these tests was that mist eliminator pressure drop is pri-
marily a function of gas velocity, L/G and wash rate have negligible effect.
As the gas velocity is increased from 7 to 10 fps the pressure drop doubles.
Visual observation of liquid coverage on the top of mist eliminator indi-
cate that under standard operating conditions without wash the top remains
dry. Only occasional droplet emerges from mist eliminator. If droplet is
large it falls back while if it is small it remains airborne and leaves the
absorber. When wash takes place the top of the mist eliminator becomes wet
and more droplets emerge as the wash rate is increased. About 25% of mist
eliminator is washed under standard operating conditions. The maximum wash
rate without significant liquid reentrainment appears to be about 1.5 gpm per
square foot.
CARRY OVER LOAD TESTS
Table 2 summarizes the results of mist eliminator carryover load tests.
73
-------
In the first two tests, rnist eliminator carryover load was determined
without washing the bottom side. In these tests the scrubbing liquid was
labeled with 70,OUO to 74,000 ppm of chloride. An eight sampling points tra-
verse was used in these tests. The carryover load without wash was in the
range of 28-32 mg/IMm3D (0.012-0.013 gr/SCFD).
In tests 3 and 4, the mist eliminator was washed from underneath and wash
liquid was labeled with sodium while spray liquid had chloride so that
carryover contribution of two liquid sources can be distinguished. The total
carryover from these tests was in the range of 71-110 mg/Nm3D (0.023-0.046
gr/SCFD). One should note that carryover from washing relates to the area
that was washed and is not representative of the total carryover.
The last two tests represent carryover load above washed area when both
wash and spray liquids were labeled with chloride. The carryover load was in
the range of 139-153 mg/Nm3D (O.OSb-0.062 gr/SCFD). Une can conclude that
washing increases carryover mass load to double to triple of the carryover
from the spray.
Since only 25% of the mist eliminator is washed at any one time under
standard operating conditions, the total carryover load averages about 63
mg/Nm3D (0.026 gr/SCFD).
MIST ELIMINATOR EFFICIENCY
The aerosol load in front of the mist eliminator was measured to deter-
mine its removal efficiency. An average of 5 tests using chloride tracer was
about 9000 my/Nm3D (3.73 gr/SCFD). If one averages mist eliminator carryover
load the removal efficiency is above 99.5%. Calvert developed an equation
for the prediction of primary collection efficiency in baffle type separators
based on inertial mechanisms (6):
E = 1 - exp - ( ut n W 0 )
UG b tan 9
where
E = fractional collection efficiency
b = distance between baffles normal to gas flow, cm
ut = drop terminal velocity, cm/s
UG = superficial gas velocity, cm/s
n = number of rows of baffles
W = width of the baffle, cm
y = angle of baffle from flow direction, radian
74
-------
Substituting values fur GEESl's tests:
E = 1 - exp - ( 21.3* ( 4 x 7.62 x 0.52 \ = 99.9%
304.8 7.62 x tan 0.52
* based on 140 micron droplet size and terminal velocity for water/air
systan (7).
The predicted theoretical efficiency is only sliyhtly higher then experi-
mental .
UROPLET PHOTOGRAPHY
Table 3 summarizes the data for four sets of tests. In the first set
aerosol particle size was measured on a six point traverse above the mist
eliminator perpendicular to the blades when no washing took place. The
average arithmetic mean diameter was 2b4 microns while the Sauter mean was
845 microns indicating wide particle size distribution. The probable reason
for the wide particle size distribution was uneven gas flow distribution due
to flue gas recycle. Gas flow measurenents above mist eliminator indicated
an area with downward flow. This downward flow is most likely to trap large
aerosols tnat were reentrained from the blades and large droplets formed from
the film on the lid of the absorber. If one excludes the aerosols measured
in the downward flow average aerosol arithmetic mean shifts to 83 microns.
The second set of droplet photography test was carried out on a five
point traverse underneath the mist eliminator where gas flow was rather uni-
form. Only 200 aerosols for each test representative for the whole popula-
tion of aerosols were selected for measurement. The aerosol average
arithmethic mean in all 5 tests was unexpectedly close with values between
139 and 143 microns.
In the third set of tests the measuranent took place above mist elimina-
tor above bottom wash nozzle with and without wash. Test without wash had
average size of 78 microns while with wash average size was about 89 microns.
In the fourth set the measuranent took place at the points with super-
ficial velocity of 10 ft/s with and without wash. The cumrnulative of tests
without wash shows 7 aerosols witn arithmetic mean of 98 microns while 13
aerosols with mean of 117 microns were measured with wash. Although the
number of aerosols in sets 3 and 4 is relatively small and insufficient for
valid statistical analysis one interesting observation can be made. If one
adds all the aerosols in these two sets without the wash the total number is
17 while with the wash total number is 3b. Based on aerosol count one can
conclude that aerosol carryover with wash is about double tnat without wash.
CONCLUSIONS
a. The pressure drop across the mist eliminator is primarily a function
of gas velocity. L/G and mist eliminator wash rate have negligible effect.
75
-------
b. Mist eliminator carry-over under normal operating conditions but
without wasn is in the range of about 28 to 60 mg/Nm3D (U.012 to 0.025
gr/bCFD).
c. When the mist eliminator is washed from the bottom carry-over above
the washed area is in the range from 7U to 16U mg/Nm3D (0.029 to O.U66
gr/SCFD).
d. Since only 2b% of mist eliminator is washed at any one time, tne
average carryover is about 63 mg/IMm3U (0.026 gr/SCFD).
e. Mist concentration at the inlet of mist eliminator averages about
y,UUU mg/DNm3 (3.37 gr/bCFD), indicating mist eliminator efficiency above 99%
with or without washing. Average aerosol size at the inlet is about 140
microns.
f. Carrover aerosol droplet size is about 200 microns. If one elimina-
tes data on droplets measured in downward flow, the average carryover size is
about 83 microns.
g. Modified EPA method b and droplet photography techniques are appli-
cable for mist eliminator evaluation on ttie full scale installations.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not necessarily
reflect the views of the Agency and no official endorsement should be
inferred.
REFERENCES
1. Ellison, W. Scrubber Demister Technology for Control of Solids Emissions
from S0;> Absorbers. Paper presented at tne EPA Symposium on the Transfer
and Utilization of Particulate Control Technology, Denver, Colorado.
July 24-18, 1978.
2. Laseke, B.A. and Uevitt, T.W. Status of FGU Systems in the United
States. Paper presented at the 29th Annual Conference of the Association
of Rural Electric Generating Cooperatives.
3. Balzhiser, K.E. R&D Status Report. Fossil Fuel and Advanced Systems
Division, EPRI. EPRI Journal 3: 3-45-47, April 1978.
4. Saleem, A. Spray Tower: The Workhorse of Flue-Gas Desulfurization.
Power, October, 198U.
5. Gavin, J.H., Hoffman, F.W. Droplet Removal Efficiency and Specific
Carryover for Liquid Entrainment Separators. Paper presented at the EPA
Second Symposium on the Transfer and Utilization of Particulate Control
Technology. July 23-27, 1979.
76
-------
b. Calvert, b. , et al. hntrainment Separators for Scrubbers. Journal of
the Air Pollution Control Association, Vol. 24, No. 10. October 1974.
7. Fuchs, N.A. The Mechanics of Aerosols. Peryamon Press Book. New York,
77
-------
TABLE 1
PRESSURt DROP THROUGH MIST ELIMINATOR
VS
OPERATING PARAMETERS
Test Gas Velocity
Number FPS
1 10
2 8
3 7
4 8
b 7
6 7
* Ratio of mist eliminator pressure drop to total pressure drop through
spray absoroer, dimensionless.
L/G
Gal/1000 ACFM
50
70
100
50
70
bU
Wash Rate
GPM
0-2.5
0-2.5
0-2.5
0-2.5
0-2.5
0-2.5
Relative *
Pressure Drop
0.21
0.16
0.10
0.16
0.10
O.ll
78
-------
TABLE 2
CARRYOVER RESULTS SUMMARY
Calculated Carry-Over
Test
No.
1
2
3
4
b
6
Wasn
Rate
Gpm
-
2.6
2.6
2.6
2.6
Scrubbing
Slurry
mg/NiTHU
27.7*
32.2*
41.2*
60.4*
Scrubbing
Slurry
gr/SCFD
0.012
0.013
0.017
0.02b
Mash
Liquid
mg/Nm^D
-
30.1**
49.4**
Wasn
Liquid
gr/SCFD
-
0.013
0.021
Tot a]
mg/Nmlu
27.7
32.3
71.3
109.8
153.1*
138. 9*
*Total
gr/SCFU
0.012
0.013
0.030
0.046
0.063
0.058
* Basis of entrainment calculation was measurement of chloride concentration
** Basis of entrainment calculation was measurement ot soaiurn concentration
79
-------
TABLE 3
DROPLET PHOTOGRAPHY RESULTS SUMMARY
Testing
Position
Above Mist Eliminator
Straight Traverse
Under Mist Eliminator
Above Mist Eliminator
Off Center
Above Mist Eliminator
Selected Points
Wash
No
No
Number
of
Points
Arithmetic
Mean
Microns
2b4
141
Sauter
Mean
Microns
84 b
Ibl
Number
of
Drops
133
200*
No
Yes
No
Yes
3
3
3
3
78
B9
98
117
83
104
145
163
10
17
7
19
* The number of droplets was higher but only 200 representative for size of
the total population of droplets were measured
80
-------
FILTRATION CHARACTERISTICS OF FLY ASHES FROM VARIOUS COAL
PRODUCING REGIONS
by: John A. Dirgo
Marc A. Grant
Richard Dennis
GCA/Technology Division
213 Burlington Road
Bedford, MA 01730
Louis S. Hovis
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
ABSTRACT
The filterability of fly ashes emitted by coal burning power stations is
described, including that of several ashes generated by low sulfur western
coal combustion that are best controlled by fabric filtration. Chemical and
mineralogical analyses of the coals were examined to determine possible
relationships between coal and ash properties and filtration behavior. Both
fly ash size and coal ash content correlated strongly with the fly ash
specific resistance coefficient, K£« Weaker, but discernible, correlations
were shown for electrical charge behavior and method of coal firing. Coal
sulfur content and ash fusion properties and chemical structures originally
expected to influence particle size showed no clear-cut effects on filtration
characteri sties.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
BACKGROUND
Reliable prediction of fabric filter performance depends upon accurate
estimation of two major variables: K£, the specific resistance coefficient
of the dust; and ac, the cleaning parameter (1,2). Although the filtration
process is influenced by many factors, K£> the parameter that describes the
gas permeability properties of a deposited dust layer, is especially important
in determining the pressure loss for fabric filter systems cleaned by reverse
air and/or mechanical shaking. Although theoretical relationships exist for
predicting K.2, it would be unwise to assume better than +50 percent accuracy.
This is due largely to the difficulty in accurately measuring key input
parameters such as particle size properties, discrete particle density, and
dust cake porosity.
81
-------
Our first step was to survey the chemical, mineralogical, and selected
physical properties of coals in conjunction with their associated seam and
rock structures to provide useful insights as to how the resultant fly ashes
might behave in fabric filter systems. At the inception of the study, the
development of quantitative relationships between parent coal properties and
the associated K£ values for the fly ashes was considered a viable
approach. Consequently, representative coal fly ash samples were sought from
many cooperating electrical utility, commercial, and industrial fabric filter
users. Coal and ash properties, including source data, were obtained from the
field sample supplier or from the literature. Specific resistance
coefficients and particle size properties for each fly ash were determined in
the laboratory, thus allowing comparison of K2 values with coal and ash
properties.
SELECTION OF REPRESENTATIVE COAL FLY ASH SAMPLES
Because the study was constrained to the investigation of 12 to 15 fly
ash samples, the development of a rationale for sample selection was very
important. The selection process, which has been discussed extensively in a
previous paper (3), is summarized next.
It was concluded that the number of fly ash samples to be investigated
for any coal type should reflect the best projections for current and future
use of that coal where fabric filters afford effective emission control.
Figure 1 shows the geographical distribution of major United States coal
fields, coal types, and the locations of coal-burning industries and utilities
now using fabric filtration for particulate collection (4). Table 1 indicates
the estimated 1980 tonnages of lignite, subbituminous, and bituminous coals
from the six major coal producing regions. (Anthracite coal accounts for less
than 1 percent of national production.) Observe that nearly 75 percent of the
coals are mined in the Eastern and Midwestern States. While Western coal
represented 25 percent of the 1980 total, a comparison with 1976
statistics (6) that show Western coals accounting for less than 20 percent of
production, suggests that the use of these coals is increasing much faster
than that for other regions. Additionally, the low sulfur content of many
Western coals makes fabric filtration a more attractive control option than
electrostatic precipitation. For the above reasons, Western coals were
weighted more heavily in the selection process; i.e., they were represented by
8 of the 14 fly ash samples tested. Despite a broad range in coal analyses,
Eastern coals from regions I, II, and III were found to be sufficiently
similar to justify treatment as a single group (3). Five of the samples
tested came from these regions, while the remaining sample depicted a region V
coal.
The distribution of fly ash samples was representative with respect to
the principal coal firing methods. Pulverized coal combustion, far more
common on the basis of tonnage consumed than stoker firing, accounted for 10
of the samples tested while only 4 were generated by stoker-fired boilers.
82
-------
TABLE 1. ESTIMATED 1980 COAL PRODUCTION BY COAL PRODUCING REGION (5)
Production
Region States 106 tons/yr
I Northern Appalachian PA,WV(n)a,OH,MD,MI 189(22.7)b
II Southern Appalachian WV(s),VA,KY(e),TN(n) 192 (23.1)
III Alabama AL, GA, TN(s) 32 (3.8)
IV Eastern Midwest KY(w),IN,IL 171 (2.0.6)
V Western Midwest AR,IA,OK,KS,MO,TX 39(4.7)
VI Western CO,WY,MT,SD,ND,UT, 209 (25.1)
NM,AZ,ID,WA,AK
aLetters in parentheses refer to north, south, east, and west.
^Numbers in parentheses refer to percent of total.
Because high sulfur content can accentuate fly ash hygroscopicity while
low sulfur contents are associated with electrical charge phenomena, coal
sulfur contents ranging from 0.35 to 3.5 percent were surveyed. Total ash
contents varying from 3.3 to 23 percent were investigated because it was
believed that higher ash contents, in conjunction with a fixed heating rate,
would reduce heat transfer to individual particles, such that large and
irregularly shaped mineral particles would be less likely to melt. Among the
many characterizing ratios for the mineral constituents of fly ash used to
predict ash slagging and fouling properties, the base-to-acid (B/A) ratio
appeared to have some predictive value through its impact on melting
temperatures. Thus, several B/A levels were also included in the samplings.
Table 2 provides a general classification of the 14 fly ash samples in
terms of the five selecting criteria discussed above: coal producing region,
boiler firing method, coal sulfur content, coal ash content, and base-to-acid
ratio. The 14 samples represented 12 different suppliers as well as a
combination of electric utility, industrial, and commercial fabric filter
installations. Table 3 lists the fly ash supplier and coal source region for
each of the samples. In addition to furnishing fly ash samples, suppliers
provided information on coal source(s), characterizing properties such as
proximate and ultimate analyses, and fly ash chemical constituents.
83
-------
TABLE 2. CLASSIFICATION OF FLY ASH SAMPLES BY SELECTION CRITERIA
Characteristic :
• Coal
#
producing region:3 I II
of samples 3 1
III IV V VI
1 018
• Boiler firing method: Pulverized
# of samples 10
• Sulfur content :b Low (<1%)
# of samples 9
• Ash content: Low (<5%)
# of samples 3
• Base/acid ratio: Low (<0.17)
# of samples 4
coal
Medium (1-3%)
4
Medium (5-15%)
9
Medium (0.17-0.33)
6
Stoker-fired
4
High (>3%)
1
High (>15%)
2
High (>0.33)
4
aRoman numerals refer to regions designated in Figure 1.
a range of values is used to characterize a specific coal or ash
property, the midpoint of that range is used to categorize the sample.
84
-------
TABLE 3. FLY ASH SUPPLIER AND COAL SOURCES
Fly ash supplier
Sample
I.D.
Coal source
state and coal region
Southwestern Public Service Co. SPS
Harrington Station
Amarillo, Texas
Texas Utilities Generating Co. TU
Monticello Station
Mt. Pleasant, Texas
Nebraska Public Power District NPPD
Kramer Station
Bellevue, Nebraska
Crisp County Power Commission CC
Cordele, Georgia
The Amalgamated Sugar Co.
Union Riley Boiler Am A
Babcock & Wilcox Boiler Am B
Nampa, Idaho
Pennsylvania Power & Light Co. PPL
Ho Itwood Station
Holtwood, Pennsylvania
Westinghouse Hanford Co. WH
Hanford Eng. Development Lab
Richland, Washington
The Medical Center Company MCC
Cleveland, Ohio
Republic Steel Corp. RS
Warren, Ohio
Colorado-Ute Electric Assoc.
Nucla Station
Hopper Sample N(H)
Shake-down Sample N(S)
Nucla, Colorado
E.I. DuPont de Nemours & Co. D
Waynesboro Plant, No. 2 Silo
Waynesboro, Virginia
United States Steel Corp. USS
Western Steel Division
Geneva Works
Provo, Utah 85
Wyoming, VI
Texas, V
Wyoming, VI
Alabama, III
Wyoming, VI
Wyoming, VI
Pennsylvania and Delaware, I
Utah, VI
Ohio, I
Ohio, I
Colorado, VI
Colorado, VI
Kentucky and West Virginia, II
Utah and Colorado, VI
-------
ANALYSES OF COAL AND FLY ASH PROPERTIES
The test program involved two separate efforts, the first centering on
the collection of information on coal and associated fly ash properties, and
the second on the laboratory determination of the specific resistance
coefficient, K2, and the particle size parameters for each fly ash sample.
DETERMINATION OF COAL PROPERTIES AND CHEMICAL CONSTITUENTS OF FLY ASH
Fly ash suppliers provided most of the information on coal properties
and fly ash chemical composition. In general, data concerning the proximate
analysis and sulfur content of a coal were more complete than those relating
to the chemical composition of the resultant fly ash. When sample information
was missing, data specified by the fly ash suppliers on the source of their
coals (including state of origin, region, seam, and, where possible, mine)
were used as a supplemental source. The Keystone Coal Industry Manual (7) and
related publications (5,8) were instrumental in identifying coal properties
for coal beds and seams not described in suppliers' responses. These sources
were also used to identify the various companies mining certain seams where an
ash analysis was not available for the mine in question or when the seam
itself was not identified. Additional ash analyses were obtained from U.S.
Bureau of Mines publications (9,10). Whenever a range of values was cited for
a specific ash constituent, corresponding ranges in base/acid ratios were
computed. To supplement the above sources, fly ash suppliers were later
contacted to fill in critical data gaps.
LABORATORY MEASUREMENT OF K£ AND FLY ASH SIZE PROPERTIES
All experimental measurements were performed on the bench scale apparatus
shown in Figure 2. Fly ash samples were redispersed by an NBS dust generator
into a test loop from which the desired aerosol quantity was extracted
isokinetically for filtration tests or particle size analysis. To better
simulate field conditions, the manifold upstream from the fabric filter was
constructed with a bottom inlet that allowed coarser particles to settle out
much as they would in many commercial systems. The filter consisted of a
15 cm x 23 cm test panel of Teflon-coated woven glass of a type commonly used
for coal fly ash filtration. All tests were conducted at an air-to-cloth
ratio of 0.61 m/min (2 ft/min), a flow rate typifying many reverse-air-cleaned
systems. The specific resistance coefficient of the fly ash, K.2, was
determined by recording pressure loss across the filter and weighing the
filter and dust cake at various intervals (11). Final dust loadings on the
filter ranged from approximately 300 to 700 g/m^.
Particle size parameters were determined by Andersen Mark III cascade
impactor wherein samples were extracted via a short probe from the central
section of the inlet manifold. Collection at this location provided a good
approximation of the size characteristics of the fly ash actually reaching the
filter surface. Cumulative size distributions of the data were plotted on
log-probability paper for the two impactor sizings performed for each fly
ash. The aerodynamic mass median diameter (aMMD) and the geometric standard
deviation (ag) were estimated for each pair of curves, which showed
excellent agreement in most cases.
86
-------
RESULTS
Relevant coal and fly ash properties for each sample are listed in
Table 4 along with boiler type. Laboratory derived K2 values, particle size
properties, and a qualitative description of the electrostatic behavior of the
fly ash in the test system are also presented. In Table 5, correlation
coefficients are listed for selected K2 relationships with coal and fly ash
properties and the particle specific surface parameter, S$, discussed in
the following section.
K2 AND PARTICLE SIZE
An adaptation of the classical Kozeny-Carman relationship investigated by
Rudnick and First (12) and later modified for GCA applications (13) has been
used to predict K2 on the basis of theoretical considerations:
p c
where is the gas stream viscosity, So the specific surface parameter, R a
complex function of dust cake porosity, p the discrete particle density,
and Cc the Cunningham-Mi llikan slip correction. So characterizes the
surface to volume ratio for the polydisperse particle system constituting the
dust cake. So is readily computed from the size parameters derived from
cascade impactor measurements, provided that the cumulative size curve may be
approximated by a logarithmic-normal distribution; i.e. ,
S 6 1.151 logo- m
So MMD ' 10 8 (2)
where MMD refers to the true mass median diameter and ag is the geometric
standard deviation. As emphasized in earlier studies (13), Equation (1) is of
limited use as a predictor of K2, since small ( 10 percent) changes in
porosity can lead to gross (~50 percent) errors in K2 predictions. If all
terms in Equation (1) remain constant except for So> an arithmetic plot of
K2 versus S^ should appear as a straight line with its origin at zero.
One infers from this relationship that K2 must increase as the dust becomes
progressively finer and the dust cake less permeable to gas flow. In fact,
the actual K2 versus S^ graph (Figure 3) shows a strong positive
correlation as forecast by theory. The linear regression line developed by a
least squares data analysis is defined by:
K. = 0.93 + 0.42(S2) (3)
2 o
where K2 is expressed in N-min/g-m and S$ in ym~2. The r2 value
"expla
level.
87
where K2 is expressed in N-min/g-m and S$ in ym~2. The r2 value
for this equation, 0.54, (the fraction of the variation in K2 that is
"explained" by the equation) is statistically significant at the p = 0.003
i i
-------
co
cxi
W
H
§
Tp
§
Z
0
H
H
5
£-(
ij
H
CO
> * «
r-i
|j
CO
o
ON
CO
00
CM
CM
CO
^
CO
CM
Z
CO
•st
o
CO
1
CM
CO
CM
CM
^
!A
.*
O
1
CO
o
llverized
>tal 700
£ u
CM
«
CO
CO
*n
^
CO
CM
°^
r^
ON
00
O
09
>4
00
CM
O
0
CO
£
00
CM
1
CO
CO
0
llverized
>tal 1150
CLi u
CM
&
H
^
1-1
^f
cO
CM
CM
"*
ON
iA
^
CO
Z
CM
CM
O
CO
ON
CO
CO
IA
°
llverized
)tal 113
£ u
^
g
z
^^ro^co *oo co r^u-i
OCOiO^ONf-^iAO iA O0N
CMCM-*r-itAO-iiA CO *Ot-<
iriiAOOcMr-.oococM o CMCO
CMCMfMCMCOCOCMfO CO CMCM
CO*jr-OOQO*OCCCM 00 CO^J
.Ar-r-^eM^vD-a- CM r^,o
r^r^xOONCO^ONr^ co -J .corocoro co CMCO
CM co CM O O r^ so
iA O O CM ON -&
r4 " .,4 O U O W -H *-< U
41 41 0* 4) V 41 4) «U CA *J 41 £u
U U "O 41
<; « XX
o^^fcSacasz z as
i
•H
J3
§
w
01
S
tf
•*4
^
o
1
i1
a
*o
o
i
F-4
: derived
£
fMQ
|
g
CO
a
j*
i-l
M
•^»
9
• r4
M
•H
U
*i
o*
« c
M 41
4) C
*H 41
o «**
•° '^
° S
V* *J
^ >.
1 ^
d T>
5
co at
idtheses :
is procei
S -
2. S
u
c
: i
ty to
Z H
JQ U
*J
-o
S.
a
J
00
a
•H
t represei
f'4
to
j:
CO
X
*
88
-------
Despite the favorable statistics, the scatter of the data from which
Equation (1) was calculated restricts its use as a predictive tool. The point
scatter is attributed to a combination of experimental errors and real
differences in particle shape, charge, and density that can also cause
significant variations in cake porosity. Based on the observed point scatter,
the estimated 95 percent confidence envelope about the regression line has
been developed, as shown in Figure 3. If, for a certain fly ash, S^ were
measured as S.OxlO""8 cm"2, Equation (3) would predict a K£ of 3.0 N-min/g-m
as a "best estimate." However, were the K£ value for this fly ash to be
determined by actual measurement, there is a 95 percent chance that its true
value would fall within the range 0.85 to 5.2 N-min/g-m. Consequently,
Equation (3) cannot be used for design purposes, despite the statistically
significant ^2~so correlation.
TABLE 5. CORRELATION COEFFICIENTS FOR K2 WITH VARIOUS
COAL AND FLY ASH PROPERTIES
Variable
Correlation Coefficient
(r)
Specific surface parameter,
% Sulfur in coal
% Ash in coal
% Moisture in coal
% Volatile matter in coal
Base/acid ratio
0.733s
-0.146
-0.575a
-0.148
0.450
-0.050
alndicates that correlation coefficient is statistically
significantly different from zero (p 0.05).
EFFECT OF BOILER TYPE AND FIRING METHOD
The method of coal firing usually influences fly ash size properties,
with a stoker-fired boiler producing a coarser fly ash than that generated by
a pulverized-fired or a cyclone boiler. Accordingly, one expects to see a
higher K£ value for a pulverized coal fly ash compared to that for a stoker
fired effluent. Unfortunately, only semi-quantitative relationships could be
inferred from the present observations, first because of limited data, and
second, because additional factors not defined in this study can also affect
the size properties; e.g., air/fuel ratio, load level, system geometry, gas
residence time, and settlement losses.
89
-------
The mass median diameters presented in Table 4, which were determined for
resuspended fly ashes, do not show any striking differences when grouped
according to firing method, although all four stoker-fired ashes were slightly
more polydisperse. The latter effect is assumed to be the main reason for the
greater S^ values and hence K£ for the stoker-fired ashes (3.6 N-min/g-m
versus an average of 2.2 N-min/g-m for pulverized coal ashes). After adjustment
to a common value of 85, however, this K2 difference was shown to have no
statistical significance. It is emphasized, however, that for the above
comparisons to be valid in the field, the expected differences in size
properties between the original and resuspended states of the fly ash must be
the same for both firing methods.
ELECTRICAL CHARGE PROPERTIES
The presence of ionizable salts in fly ash may produce secondary
filtration effects due to electrical charging. Since the resultant charges
induced by thermal dissociation or contact electrification in the combustion
zone are essentially unipolar, particle repulsion effects within the close
confines of the dust layer may afford the advantage of lower K£ (and
decreased pressure loss) due to increased porosity. It can also be argued
that the extent to which charges induced on a fabric are able to leak off, may
also affect dust dislodgement characteristics with a charge accumulation
causing increased adhesion.
Electrostatic charging is often a problem in the resuspension of bulk
dusts by high velocity aspiration. Although all metal components of the test
system in Figure 2 were electrically grounded, there was still evidence of
electrostatic charging and deposition for some of the fly ashes tested.
Although no attempt was made to quantify the degree of apparent charging,
samples were categorized as to the presence or absence of visible
electrostatic deposition in the test system. According to Figure 3, it
appears that these eight samples (circled symbols) tend to lie below the six
samples (open symbols) displaying no obvious electrostatic effects. That is,
for a fixed value of 85, fly ashes exhibiting electrostatic effects also
possessed significantly (statistically) lower K£ properties. However,
except for the fact that four of the eight samples also represented
stoker-fired combustion, electrostatic behavior of the fly ashes could not be
related to any other coal or ash properties. It has not been determined
whether there is some intrinsic fly ash property that predisposes it to the
accumulation of electrostatic charge or whether this behavior is an artifact
of the experimental measurement system. Thus, although the electrostatic
properties of fly ash as observed here are useful in explaining observed
differences in K2, they provide no guidance in predicting them.
COAL SULFUR CONTENT
The manner in which the sulfur content of coal affects fly ash filtration
properties is not clearly understood, although it has been established that
sulfur in various forms can affect ash fluid properties. The presence of
suspected sulfate salts, as shown by the 803 assays for the coal ashes, may
decrease ash softening and fluid temperatures, or at least broaden the
temperature range over which an ash converts from the solid to the fluid state.
90
-------
If the viscosity of the molten ash is lowered sufficiently, it is
reasonable to expect that gas stream turbulence and shearing action might lead
to droplet breakup. On the other hand, those particles that have been
converted to the liquid phase but still remain highly viscous (and sticky) may
serve as irreversible collision sites for small particles undergoing Brownian
diffusion. Under these conditions, it appears that the presence of sulfates
could either increase or decrease particle size parameters depending upon
which mechanism prevailed.
If the coal sulfur content is due mainly to iron pyrite in the raw coal,
significant separation of FeS2 during coal upgrading will reduce the
"basic" phase of the ash and hence diminish the base-to-acid ratio. Under
these circumstances, one might expect an increase rather than a decrease in
the softening temperature. With less chance for particle adhesion and less
reduction in particle size due to droplet breakup, slightly coarser particles
and hence, lower K£ values, might be predicted. Although analyses of the 14
samples suggest an inverse effect, the low value for the correlation
coeffecient (-0.05) precludes assigning any statistical significance to it.
It is possible that examination of more coal samples, in which a wider range
of sulfur contents is exhibited, may afford better resolution of the
K2~sulfur correlation. Since sulfur content is usually a readily attainable
parameter, its use in a workable K2 versus sulfur relationship is attractive.
COAL ASH CONTENT
The effect of ash content upon ash fusion properties was examined to
determine if an increase in coal ash content might conceivably result in less
heat transfer to individual mineral particles for a fixed energy input, thus
slowing particle transition to the viscous and fluid states. Reduced melting
might be expected to result in generally coarser and more irregularly shaped
fly ash particles. Regardless of any possible impact upon particle size
parameters or softening temperatures, filtration demands (cloth and fan
capacity) will automatically relate to the volume of fly ash produced which,
in turn, should relate directly to the amount of mineral present in the parent
coal.
Data pairs representing 14 samples appear to support the proposed high
ash effect; i.e., a coarser dust. Other than specific surface properties,
coal ash content was the only other variable that correlated significantly
with measured K£ values. Although the data exhibit considerable scatter,
Figure 4, K£ values are seen to decrease with total ash content, as
predicted. The K£ versus ash content correlation is too broad to be of any
real value. When combined with the specific surface parameter, So, in a
multiple regression analysis, the resulting equation is:
K2 = 2.04 + 0.345 (s£) - 0.079 (% Ash) (4)
The r^ for Equation (4) is slightly higher than that for Equation 3, 0.65
versus 0.54, and the ratios for predicted to measured K£ values for this
equation fall within a slightly narrower range than those for Equation (3).
It should be noted, however, that the coefficient for (% ash) just misses
statistical significance at the p=0.05 level.
91
-------
BASE-TO-ACID RATIO
Although a number of characterizing ratios derived from the chemical
constituents of fly ash were investigated initially, only the base-to-acid
(B/A) ratio appeared to offer any predictive capabilities. The basic
components of the ash are Fe203, CaO, MgO, Na20, and 1^0 while the
acidic components are Si02> A.l203> and Ti02« Minimal ash melting
points and fluid temperatures accompanied by a more rapid transition from the
solid to the liquid phase are observed with a 1:1 mix of basic and acidic
components. As the B/A ratio becomes larger or smaller than 1/1, melting
points and fluid temperatures rise. This trend is roughly symmetrical on
either side of the unity ratio, depending only on the relative difference
between basic and acidic components. Thus, the effect of a B/A ratio of 1/2
is approximately the same as for a ratio of 2/1. B/A ratios were calculated
for 93 ash samples (representing all 6 coal producting regions) (9). When
these ratios were compared to their corresponding ash softening temperatures,
the correlation coefficient was very high (r = 0.81, p<0.001).
Although the base-to-acid ratio is clearly a reliable indicator of ash
melting and fusion properties, the expected link between melting and fusion
properties and effluent particle size and K2 could not be discerned, as
discussed earlier in this paper. Correlation coefficients for B/A with both
K2 and 85 were approximately -0.2, well below the level of significance
for a sample of this size.
SUMMARY AND CONCLUSIONS
The purpose of the research described in this paper was to investigate
possible relationships between coal and ash properties and fly ash filtration
characteristics. It was postulated that certain chemical and physical
properties of coals might have some predictive value in determining the
specific resistance coefficients, K.2, of their resultant fly ashes. Since
this parameter is an important index of fabric filter performance, a reliable
estimation method would prove valuable in the design and analysis of reverse-
air and/or mechanical-shake cleaned fabric filter systems. Six potential
correlations between K2 and selected coal and/or fly ash characteristics are
reviewed in this paper. Coal sulfur content and the base-to-acid ratio
appeared to be of little value in predicting K2«
Limited data suggest that those fly ashes bearing an appreciable net
electrostatic charge (based upon qualitative indications only) form a more
porous dust layer on the fabric surface. The latter effect would explain the
observed reduction in K2 when So, the specific surface parameter, does not
change.
The method of coal-firing, through its impact on the size properties of
the fuel entering the combustion zone, also produces a discernible change in
K2- The much coarser size of the coal charged to the grates of a stoker-
fired boiler, as compared to the 70 percent less than 200 mesh feedstock
typifying a pulverized coal boiler, was expected to generate a coarser fly ash
92
-------
and a lower l^. However, for the redispersed fly ash samples tested, those
from stoker-fired boilers were much more polydisperse and had higher values
for S^, resulting in higher K£ values for the stoker ashes. After
adjustment to a common value of SQJ pulverized coal and stoker ashes
showed no statistically significant K£ differences.
An empirical correlation between ash content and K2 was displayed,
although not at the statistical level where it could be used to establish
filter system design or operating parameters. Lower K2 values for high ash
coals appeared to confirm the hypothesis that high ash contents would allow
less heat for particle size reduction or alteration of surface properties.
A strong correlation was observed between fly ash size properties (i.e.,
particle specific surface) and K2> as predicted by theory. It should be
noted, however, that with concurrent variations in fuel preparation methods,
size reduction processes, and air to fuel ratios, fly ash size parameters may
show no relationship to parent coal properties. The fact that the equation
representing the least squares regression line permits only _+ 50 percent
estimates for K£ suggests that many coal fly ashes must share similar values
for dust cake porosity and average discrete particle density.
There are a number of factors that mitigate against accurate prediction
of fly ash filterability from coal properties alone. First, there may be
several unidentified coal properties that exert secondary effects on fly ash
filterability. The second order effects may be masked by strong first order
parameters, such as the particle size. Because of the data scatter, a large
number of samples might be required before the identity and magnitude of such
secondary effects could be established with any certainty.
Another factor working against precise quantification of the effects of
coal properties on K£ is the tremendous variability that exists in coals.
Coal seam overburdens, partings, and floors, the associated rock structures
surrounding and separating seams, contribute to this variability. However,
since coal is formed from diverse plant materials (such that heterogeneity in
structure is quite common), extensive compositional variations can occur
spatially and in depth for any given seam (14). This can pose serious
problems for coal users with strict specification requirements and also for
researchers. While the coal properties obtained from fly ash suppliers
represent the best available "average" values, there is no guarantee that the
particular fly ash sample received and tested was derived from coal with these
same "average" properties.
93
-------
REFERENCES
1. Dennis, R. and Dirgo. J.A. Comparison of Laboratory and Field Derived
K2 Values for Dust Collected on Fabric Filters. Filtration and
Separation. 18: 394-396,417, 1981.
2. Dennis, R. and Klemm, H.A. A Model for Coal Fly Ash Filtration. J. Air
Pollut. Control Assoc. 29: 230-234, 1979.
3. Dennis, R., Dirgo, J.A., and Hovis, L. S. Coal Properties and Fly Ash
Filterability. Third Symposium on the Transfer and Utilization of
Particulate Control Technology: Volume I. Control of Emissions from
Coal Fired Boilers, pp. 1-10. EPA-600/9-82-005a (NTIS PB 83-149583),
July 1982.
4. Gibbs and Hill, Inc. Coal Preparation for Combustion and Conversion.
Prepared for Electric Power Research Institute. EPRI AF-791, Project
466-1, Final Report, May 1978.
5. Nielson, G.F. (Editor-in-Chief). 1981 Coal Mine Directory: United
States and Canada. McGraw-Hill, Inc. New York, New York, 1981.
6. Energy Data Report. Coal-Bituminous and Lignite in 1976. DOE/EIA-0118/1
(1976). Prepared in the Office of Energy Data and Interpretation, U.S.
Department of Energy. December 18, 1978.
7. Nielson, G.F. (Editor-in-Chief). 1979 Keystone Coal Industry Manual.
McGraw-Hill, Inc., New York, New York, 1979.
8. Nielson, G.F. (Editor-in-Chief). U.S. Coal Mine Production by Seam -
1976. McGraw-Hill, Inc., New York, New York. 1977.
9. Abernethy, R.F., Gibson, F.H. , and Peterson, M.J. Major Ash Constituents
in U.S. Coals. U.S. Department of the Interior, Bureau of Mines, 1969.
10. Gibson, F.H. and Selvig, W.A. Analysis of Ash from United States Coals.
U.S. Bureau of Mines Bulletin No. 567. U.S. Department of the Interior,
Bureau of Mines, 1956.
11. Bubenick, D.V. , Hall, R.R., and Dirgo, J.A. Control of Particulate
Emissions from Atmospheric Fluidized-Bed Combustion with Fabric Filters
and Electrostatic Precipitators. EPA-600/7-81-105 (NTIS PB 82-115528),
June 1981.
12. Rudnick, S.N. and First, M.W. Specific Resistance (K£) of Filter Dust
Cakes: Comparison of Theory and Experiments. Third Symposium on Fabric
Filters for Particulate Collection, p. 251-288. EPA-600/7-78-087 (NTIS
PB 284969), June 1978.
94
-------
13. Dennis, R. and Klemra, H.A. Fabric Filter Model Format Change. Volume I,
Detailed Technical Report; Volume II, User's Guide. Industrial
Environmental Research Laboratory, U.S. Environmental Protection Agency,
Research Triangle Park, N.C. Report No. EPA-600/7-79-043a and -043b
(NTIS PB 293551 and 294042), February 1979.
14. Haggin, J. Interest in Coal Chemistry Intensifies. Chemical and
Engineering News. pp. 17-26, August 9, 1982.
95
-------
w ••
Is
IM
„ e « o ..
«' «' n m *i
Z 2
.5
t> o
0
0
« n
PSJ
1W /-s O
•H CM
e w
0 01 to
J4 -H
t8 U 4->
CJ -rl C
IW 0)
4J IM O
O u a
0)
m
8-5
M rH
U4
•So
•o
t-i
s
4J
O
r-l t-l
01 01
•ri CO
n
r-< 01
U «
i
•rl
O
01
*J
01
-o
tt)
l-l
(«
1 §
V)
•rl
m
Ol
Vi
..
u
01
(X
«*>
01
I
O u
4J U
m ai
0)
•O h
0) a)
N X
S^
> (0
rj 01
3 i-H
CU U
^•^ M
96
-------
FLY ASH FROM TEXAS LIGNITE AND WESTERN
SUBBITUMINOUS COAL: A COMPARATIVE CHARACTERIZATION
D. Richard Sears, Steven A. Benson,
Donald P. McCollor, and Stanley J. Miller
U.S. Department of Energy
Grand Forks Energy Technology Center
Grand Forks, North Dakota 58202
ABSTRACT
As examples, we use two Jackson group lignites from Atascosa and Fayette
Counties, Texas, and a Green River Region subbituminous coal from Routt
County, Colorado.
The composition of individual fly ash particles was determined using
scanning electron microscopy and electron microprobe, with support from x-ray
diffraction of bulk ash. Using particle sample populations large enough to
permit statistical treatment, we describe the relationship of composition to
particle size and the correlation between elemental concentrations, as well
as particle size and composition distributions. Correlations are displayed
as data maps which show the complete range of observed variation among these
parameters, emphasizing the importance of coal variability.
We next use this data to produce a population distribution of ash parti-
cle resistivities calculated with Bickelhaupt1s model. The relationship be-
tween calculated resistivity and particle size is also displayed, and the
results are compared with measured values.
97
-------
By acceptance of this article, the publisher and/or recipient
acknowledges the U.S. Government's right to retain a nonexclu-
sive royalty-free license in and to any copyright covering this
paper.
98
-------
INTRODUCTION
The ability to employ elemental analysis and mineral speciation of coal,
in advance of mining, to reliably select, design, and size a particulate con-
trol device would reap numerous benefits. Most accomplishments in this area,
particularly those of Bickelhaupt (l), Selle (2), and Frisch (3), relate to
electrostatic precipitator (ESP) performance, based on bulk coal or ash
composition. Frisch, ICK:. cit., describes an algorithm for including within-
seam coal variability in the inter-relationship of ash-to-BTU ratio, resis-
tivity, specific collection area (SCA) and efficiency.
By contrast, we have chosen to focus on populations of individual ash
particles. A goal of the GFETC program is to quantify the generic and spe-
cific properties and property distribution of low rank western coal ashes as
they relate to collectability, emission levels, and environmental insult. In
this paper we describe work which has led us from individual particle compo-
sition to individual particle resistivities and their frequency distribu-
tions .
MATERIALS AND TECHNIQUES
We have selected three specific coals for this report: two Texas lig-
nites notable for their high ash, low BTU, and high sulfur content and a low
sulfur Colorado subbituminous coal on the borderline between subbituminous A
and high volatile C bituminous. These coals are described in Table I.
Fly ash is generated using the GFETC Particulate Test Combustor (PTC).
This unit is an axially upward, pulverized coal fired furnace with a nominal
coal consumption of 75 Ibs/hour. Equipped with an electric air preheater,
provision for the usual primary, secondary, and tertiary air, and induced
draft exhaust through the selected particulate control device, the unit is
designed to generate ash characteristic of that produced in a utility boiler.
Axial firing maximizes fly ash/(bottom ash + slag) ratios. Although the PTC
is not equipped with boiler tubes, the flues are supplied with a system of
heat exchangers permitting delivery of flue gas at temperatures from ~200 to
~750°F. PTC instrumentation includes redundant real-time measurement of flue
gas temperature and gas concentrations.
Combustor operating parameters and the flue gas environment from which
fly ash was sampled are summarized in Table 2.
Ash was sampled isokinetically and simultaneously size fractionated
using a Southern Research Institute (SoRI) 5-stage multicyclone (4) which we
have modified for extractive, rather than in-stack, sampling. The unit is
enclosed in an oven regulated to stack temperature.
Bulk hopper ash is analyzed for major and minor elements using x-ray
fluorescence (XRF). Multicyclone ash fractions are analyzed both by XRF and
also by neutron activation analysis.
99
-------
TABLE 1. DESCRIPTION OF COALS INVESTIGATED
Name
Arapahoe
Ledbetter
San Miguel
Mine
Location
Power plant
Energy Mine
Routt Co. ,CO
Arapahoe Unit 3
None*
Fayette Co. ,TX
None*
San Miguel
Atascosa Co.
San Miguel
,TX
Station
Type
Analysis
(as burned)
C, % dry basis
H, % " "
N, % " "
SOI It II
> to
0, % " "
Ash, % "
H20, %
Public Service Co.
of Colorado
Denver, CO
Green River Region
Subbituminous
62.74
4.71
1.63
0.43
19.79
10.7
7.9
Heating value Btu/lb
as burned 11,058
X-ray coal ash analysis
Jackson Group
lignite
38.17
5.53
0.55
2.11
32.64
21.0
25.3
6,578
San Miguel Electric
Corporative
Jourdanton, TX
Jackson Group
lignite
46.42
3.79
0.72
2.49
27.38
19.2
13.0
7,719
Si02
A1203
Fe203
Ti02
P205
CaO
MgO
Na20
K20
S03
55.0
25.1
3.7
1.1
0.9
5.7
2.4
0.0
1.5
4.6
58.7
19.5
4.0
0.8
0.2
6.1
2.5
0.6
0.9
6.6
45.0
15.2
5.8
0.7
0.4
9.3
1.2
4.6
2.1
15.2
*Ledbetter lignite was a composite of core drillings collected at the loca-
tion of a future generating station and mine to be opened by the Lower Colo-
rado River Authority in Fayette Co., TX.
100
-------
TABLE 2. COMBUSTION CONDITIONS
Coal
Run No.
Arapahoe
AR-183
Ledbetter
LD-193
San Miguel
TL-153
Coal feed, Ib/hr. 45.2 83.6 69.5
Air feed, scfm 118 112. 149.
Flue gas composition
02, vol % 4.9 5.9 6.6
C02, vol % 13.2 12.2 13.0
N2, vol % 81.9 81.9 80.4
S02, ppm 353.5 2957. 3800
NO , ppm -1000 697. 783
H28, vol % 7.9 12.6 8.8
Inlet dust loading
gr/scf 2.43 7.28 13.1
Inlet temperature °F 296 325 272
For mineral speciation of coal and ash, an important tool is x-ray dif-
fraction applied to both fly-ash and to oxygen-plasma low temperature (150°C)
ashed coal (LTA). The LTA technique avoids the pyrolysis and minimizes the
dehydration of mineral species which occurs in conventional ashing.
We also employ chemical fractionation (5), of coal, a technique in which
sequential extractions of the coal are performed using 1M ammonium acetate
and 1M HC1. The first solution removes ion-exchangeable cations and soluble
salts; the second solution dissolves carbonates and acid soluble oxides.
Unaffected pyrite and silicates remain in the solid residue.
Ash resistivity is measured at temperature in simulated flue gas in a
laboratory unit (6). Although the PTC is equipped with an in-stack point-to-
plane resistivity instrument, the laboratory unit permits investigations over
the entire range from 200-825°F with control of S02 and H20 concentrations
over a wide range.
The major tool employed in our work has been the scanning electron
microscope/energy dispersive electron microprobe (SEM)1. SEM is applied to
1GFETC employs a JEOL JSM 35 scanning electron microscope. X-rays are de-
tected with a Kevex lithium-drifted silicon detector. Elemental analysis is
performed by means of a Tracer Northern NS-880 X-ray analyzer. (Reference to
specific brand names and models is done to facilitate understanding and
neither consitutes nor implies endorsement by the Department of Energy).
101
-------
sufficient individual particles to permit statistical analysis. To eliminate
human bias in selection of particles to be measured and analyzed, photographs
of SEM fields are overlaid with a grid. Random-number-generated grid coor-
dinates are used to select particles for analysis. SEM determination of
particle size and shape and SEM-electron microprobe analysis of major element
composition were done on particles from each of the five multicyclone stages.
The data obtained from all five stages were combined to form a population of
particles and the data were analyzed using the SEM's online computational
system.
RESULTS
Mineral species identified above XRD detection limits in the three coals
were: Arapahoe: quartz, kaolinite, and calcite; Ledbetter: quartz, kaolin-
ite, calcite, pyrite, and plagioclase; San Miguel: quartz, clinoptilolite (a
zeolite), CaS04-5H20, plagioclase, pyrite, kaolinite, and possibly calcite.
For our purposes, the most notable differences are the higher inorganic min-
eral content of the Texas lignites, and the absence of detectable pyrite in
the Arapahoe coal. Chemical fractionation results were complex and too
voluminous to report in detail here. Notable results were that Ledbetter
appears to have ~87% of its calcium associated with the organic structure
whereas San Miguel and Arapahoe had 44-53%, with most of the remainder as
carbonate. Almost 100% of the sodium in the Texas lignites was associated
with the organic structure, whereas ~75% of Arapahoe's sodium appears to be
in aluminosilicates.
XRD of the bulk fly ash reveals: Arapahoe: quartz and mullite (Ale~
SiaOis)? Ledbetter: quartz, mullite, anhydrite (CaS04), and cristobalite; San
Miguel: quartz, anhydrite, mullite, magnetite (Fe304), hematite (Fe20s) and
an unusually large quantity of amorphous material or quench growth.
The presence of mullite and cristobolite is consistent with the thermal
decomposition of kaolinite which proceeds through various stages including
the formation of mullite and cristobalite above 1095°F (7, 8, 9).
Particle size distributions of the three fly ashes were determined by
isokinetic extraction of material from the PTC flues through impactors and
the SoRI multicyclone. Multicyclone data appear in Figure 1. Salient obser-
vations are: all three ashes have large amounts of material in the, 2.5-4 |Jm
range; San Miguel has significantly more fines in the submicron range; based
on an assumed maximum particle size of ~70 jjm, mass median diameters (mmd)
for Arapahoe, Ledbetter, and San Miguel ashes are ~15, ~15, and ~28 |Jm,
respectively. Actual field tests at Arapahoe and San Miguel stations sugges-
ted mmd's of 20 and 16 (10, 11). Cumulative percentages at l(Jm agree reason-
ably well with field experience; Arapahoe: 0.18% in PTC, 0.08% in the field;
San Miguel: 0.65% in the PTC, 0.53% in the field.
SEM ash analyses produce quantitative results for twelve major and minor
elements for each particle examined, plus size and aspect ratio. From this
data, frequency distributions are calculated and displayed in Figure 2.
Sulfur showed a maximum at zero concentration for all three coals, and it is
omitted from the selection.
102
-------
20.0
10.0
4.0
(0
W
(0
0)
•| 0.4
|
O 0.2
O.I
1
7o /
/
V""0 ARAPAHOE
—A LEDGETTER
-"O SAN MIGUEL
—I I i » mi i
I 4 10
Aerodynamic size,
FIGURE 1. Cumulative particle size
distributions .
In order to calculate parti-
cle resistivity according to
the Bickelhaupt Model (1),
composition of individual
particles is required. This
information is also needed
in order to begin to under-
stand the origin of fly ash
particles in terms of coal
mineral composition and the
combustion process. Because
each particle is a serial
numbered entity in the SEM
statistical system, this
information is recoverable
for each particle. In
Figure 3, we present a few
examples of binary correla-
tions selected because they
illustrate some of the
diverse functional relation-
ships encountered. These
data maps represent 500,
300, and 200 particles of
Arapahoe, Ledbetter, and San
Miguel ashes, respectively.
The same display can be pro-
duced for concentration
versus size, if desired.
When this is done, it is
often observed that NaaO and
S03 are enhanced in the fine
particle region. Cf. Ref.
10., for example.
Although these data may be used in conjunction with x-ray diffraction to
describe some of the ash "mineralogy", in this paper we present it primarily
to illustrate and emphasize the extremely non-uniform composition of real fly
ash. Analyses of bulk ash fail to provide any suggestion of this diversity.
Consequently, there is a temptation to assume that concentration-dependent
physical properties such as resistivity are uniform for all particles. This
is a very bad assumption, as we see in Figure 4, in which we display ash
particle resistivities calculated by the Bickelhaupt model (1).
A point on these data maps should be interpreted as representing a bulk
resistivity one would calculate for a perfectly uniform ash in which all
particles have the same composition as the single particle corresponding to
the plotted point. One must not assume that any one point corresponds to the
resistivity one would measure in that single isolated particle. Bulk resis-
tivity is a complicated phenomena and the semi-empirical Bickelhaupt model
does not purport to apply to individual isolated particles.
103
-------
80
I160
')
5 40
20
Arapahoe
04 8 12 16 20 24 28
Midpoint
40
8,30
§20
Q.
10
80-,
0)
g*60
c
0)
a.
20
50
a)
S1
£30
u
0)
a.
10-
Ledbetter
Na20
04 8 12 16 20 24 28
Midpoint
MgO
0 2 4 6 8 10 12 14 16 18 20
Midpoint
0 3 6 9 12 15 18
Midpoint
-t-r
30
20-1
£10
50
c30
10
San Miguel
0 6 12 18 24 30
Midpoint
0 3 6 9 12 15 18
Midpoint
60
a
O)
240-
c
fl)
*
"^20-
30
10
4 8 12
Midpoint
0 10 20 30 40 50
Midpoint
0 10 20 30 40 50
Midpoint
0 10 20 30 40 50
Midpoint
80
§60
S 40-
20
50 n
!30
10
0 15 30 45 60 75 9O
Midpoint
50-|
FeaO
so
10
0 20 40 60 80
Midpoint
11?9,.? ,**,,*
\ If i i
0 20 40 60 80
Midpoint
FIGURE 2. Elemental composition distributions, reported as oxide, for the
three ashes.
104
-------
2
0
AI2C
50
30
10-
SiOj
100
60-
20
3?,
100-
7.5-
5.0-
2.5
0 •
<
Na2O
4-
.
2
o-
(
Arapahoe N«2
'2sli» .; : 2 -
^lii??- '
psfjJF-:?""":.
" _v. '.
.".*,.- . . o
6 3 6 ' ' 9 ' ' 12 '
K20
1 AI20
40
• ' '&'*ji&
' $ * ' ~- -*"^X.
-^•-X 2°'
. * " -*
n.
0 30 SO 90
Si02
SiO2
; 80-
1^
-^spsT- " *°
-'-.."
'" 0'
} 15 30 45 00
CaO
K2O
30-
20-
•' .- . 10-
.-' ..-.'••^•••arir^'"' *" "-*-
i 30 eo so
SI02
N«20
4-
.
§- . 2-
• . .-
'*-»-\
------ - ,, . 0 •
> 30 60 90
Si02
Na2O
Ledbetter
1QO-
NaaO vs KaO
75-
j.
.J- 50-
fas-
•
ifc- - - ' n .
0 10 20
K20
3 AI20
40 n
Al2O3vsSi02
. _V-v.
- --:-^t-5>i" • 20"
- -"r": --:--" V^i--
*.
• ' " '"^ n-
0 30 60 90
Si02
S,0
*;
:r- SiOavsCaO
-.^l . 80-
St"r"""-"'
^^Vfc"i"_-
-": :-"'c^.:-.-.. -°
-- , , n.
0 15 30 45
CaO
K2C
KjOvsSIOz 10-
.
5-
; :, ,n.jrifr i
0 30 60 90
Si02
N,2
Na20vsSI02 10"
. -' -
V. "• '
- :"n-t~.~ - - 5'
' •':-.•.— -. .:.-. -
0 ' 30 '60 30
SI02
San Miguel
: - --
;-":-i-:".
• .-"* • - **^jc
*X*' --*-*"**•
i—.* -*-
•i"1. -
0369
K2O
3
. - ^ • .v " *. ; %
' j* ""-*" 1 ' * * '
'• ' •'-' '"• "'- ""^"-"'vVSfc
-"- " - " :
r1** ~ ' . •
0 30 60 90
SIO2
J
^
*p** .-*--
^J-',+.
' ~~ • ~. '• .
"- '-. "•
0 30 60 90
CaO
. . ^&s£?'
-•! •'>**-• '*«.":-.' tKT
3 30 60 90
SiO2
3
- "-. i. .-» .
-"•*"-. ---1-".
". * *.- -'t.^'.
- '-,'• /" . ; : -; •">
• - - - * ' "**"."
-;•'. - -;-? '"•" • " -
i 30 60 90
S1O2
FIGURE 3. Examples of correlations between elemental concentrations,
datum represents a specific, individual particle.
Each
105
-------
•
"
* t
*j.
*"*: »* "
* •"' %?***
"*% A
CM O 00
OHB DOT
•
•
• ••
w* •
.
-
." •-..
J> •
:£<.
l»" A
vl*jpf-
« 5 N 0 00
OHd DOT
•
•
•
^
^
•
« • >
i
ID ^ CM O
OHH DOT
O
o
^
I
Ul
g
z
w
i i
u>
o
^
rr
r-
cc
Ul
(—
l_
Ul
ffi
Q
Ul
-1
(0
o
o
^—
Ul
o
X
a.
oc
CO U
If)
c
o
u
E
8!
£
a!
TO
I/) O
*~ 0
- C
re
Q.
-O
.in
•»
-
c
o
u
•86
c
0)
OJ
E
re
*" J
u
I
-o
.CO
• ">
'5 c
o
N U
*~ e
c
,_
.. Ul
, :. o
. .-: 5
• "^' 0-
• -^w . OC
i ~^j^V ' ^
r «^F»
CD T)- CM O CO Cl
OHd OOT
~ (ft
£
O
u
- E
o c
"CO ~
u
o
E
re
B
- r
re
a.
• o
_u7
v
W)
c
o
u
-O E
" c
k.
£
0)
<9
'€
.Ifl QJ
13
. r
re
a.
•o
• CO
*"
•S 8
•^ o
CM Si
*" E
O
• CO C)
E
re
.2
a
a.
• O
D
O
o
«^
^
_J
Ul
D
•'. «
'•• 5
• ' '. : . . <
• '-..:tn
• .*•"••. i'*-"
"v>v^ *
^"w* •
CM O CO CO
'OHU OO1
•
O
o
5
.
v
CC
- ( • uj
* * ^
. • .. ui
•v ..'.•-• ffi
_• • :±i . Q
•*"/•• :.i;'*T- ' j
-i. ...V'iiUl*
CD J CM O CO ID
OHd OO1
o
• o
• i
•
•
•'*
Ul
i.. ~
.. .', ^
. • % T." °"
' T\?.. DC
10 ^ N o co «
OHd 9OT
5
k_
o
. E
^c
o
" E
.2
~i? B
r
eo
a
-0
-*
«
5
o
o ^
"« c
h_
5
0)
r ,s
S
_(/) .y
"o
re
a.
"0
- (O
^
''I
.CM 0
" I
_c
"2
• oo -
£
re
cu
"* .S
<5
a.
-0
)
o
CO
oo
CM
/~N ^
o si
i-( r-l
C3> 0)
CM M
N»^ U
• H
CJ CQ
O
>
n) ^
tO T3
O crj
U i-l
m u co
0) r-l fl
S-i n) O
,fi U -H
4J 4->
CO CO
(D 0) S-4
X) -H 4-1
4-1 4-> C
•H 0)
>-l > U
O -H d
4-1 4-1 O
10 U
fll >»^
N W W
•H r-l
•H fin <4-l
4-> 0
CO CM O
co
CO CJ
> 0 (U
4-1 -* «
'£ H
•H d
to /-\
^ 0 i-l
00-cf T3
o to o
i-5 * ' S
,
•*
o
1— 1
f _
H4
106
-------
Flue gas concentrations employed in these calculation are given in Table
3. S03 was set at 0.29 and 0.0 ppm for Araphahoe and San Miguel because
these concentrations were observed in field measurements (10, 11). Ledbetter
S03 field data are not available; therefore its S03 was taken to be similar
to San Miguel's. The S02 and H20 concentrations are typical values observed
in actual test burns in the PTC pilot unit, but are not necessarily identical
to the values at the exact time of sample collection. Lithium was set at a
small value, for it is not detectable by XRF.
TABLE 3. CONDITIONS EMPLOYED IN RESISTIVITY CALCULATIONS
Coal
Arapahoe
Ledbetter
San Miguel
Concentration
H20 wt %
S02 ppm
S03 ppm
Field, Kv/cm
Lithium concn.
ash, wt %
8.7
353
0.29
1.5
0.01
17.4
2957
0.0
1.5
0.01
8.8
4400
0.0
1.5
0.01
In Figure 5 the resistivity data are displayed in another format which
may be more useful in estimating impact on precipitator performance. The
high and low temperatures span much of the range of cold side and hot side
precipitators. A notable feature of Figures 4 and 5 is the increasing breadth
of the resistivity distribution which accompanies a shift to lower valves, as
temperature is increased. Another unexpected feature is the markedly asym-
metric shape of the distribution at some temperatures (e.g. Arapahoe at 44l°C
and San Miguel at 144°C).
Average or consolidated resistivities, i.e., pseudo-bulk resistivities,
may be calculated from the voluminous individual particle values.
6, we display such averages as functions of temperature.
In Figure
COMPARISON OF CALCULATED AND MEASURED RESISTIVITIES
A few comparisons with measured values are possible. For two of the
coals, laboratory resistivities over a range of temperatures were obtained in
the GFETC resistivity apparatus (6). The value of the maximum resistivity
rho(max), which occurs at temperature t(max), is reported in Table 4, togeth-
er with field measurements at specified temperatures. Also included is the
value calculated for bulk field-collected ash using the Bickelhaupt Model
(10).
For the two Texas lignites, average particle resistivities calculated by
the methods of this paper agree well with measured values. The temperatures
of maximum resistivity, however, disagree substantially. For Arapahoe subbi-
107
-------
o
UJ
3
o
at
-------
14
12-
o 10 H
o
Q-\
6-\
10-
o
i
o
o
14-
12-
S 10 H
o
ARAPAHOE
112 144 182 227 283 352 441
Temperature,°C
SAN MIGUEL
112 144 182 227 283 352 441
Temperature,°C
LEDBETTER
112 144 182 227 283 352 441
Temperature,°C
FIGURE 6. Average resistivity versus temperature.
109
-------
tuminous coal, field in-situ resistivity and bulk ash calculated resistivity
are both almost an order of magnitute higher than calculated here from parti-
cle chemistry.
TABLE 4. COMPARISON OF CALCULATED AND EXPERIMENTAL RESISTIVITIES
Coal Arapahoe Ledbetter San Miguel
rho(max), ohm-cm
measured! N/A 9.5 X 1011 1.3 X 1010
calc.(this paper) 4 x 1011 1 X 1012 1 X 1010
t(max), °F (°C)
measured! N/A 272 (133) 289 (143)
calc.(this paper) 360 (182) 419 (215) 370 (188)
t(field),°F(°C) 266 (130) N/A 330 (166)
rho(field) ohm-cm
in-situ*
calc.(this paper)
bulk ash (calc)^
6 X 1011
7 X 1010
5 X 1011
N/A
N/A
N/A
1.6 X 1010
1 X 1010
0.5 X 1010
tUsing A.S.M.E. laboratory unit (6).
^Measured at spark-over (10, 11).
^Using standard Bickelhaupt method (1).
The average particle method described here is wholly dependent upon the
assumption that the particle data correspond to a representative selection of
particles. However, there are problems in both sample preparation and in
particle selection which will tend to favor larger particles over fines.
Second, the Bickelhaupt Model is an empirical relationship which may not
be applicable to these ashes or these flue gas conditions. Figure 7 displays
several resistivity calculations for San Miguel with an experimental curve
(PLAB) which is the average of thirteen laboratory measurements.
In Figure 7(>VS is the combined volume and surface resistivity and^VSA
is ^ VS combined with an acid contribution (1). These resistivities are
combined as follows:
T vs
and
t
r
vsa
110
-------
and are defined by relations of the form:
log 9v = Ci-C2 log (Li+Na)-C3 log (Fe)
+ C4 log (Mg + Ca)
log ^ s = logP
- K (H20)
so z vapor
and <=>
.SO
and K contain
no
where Cj through C4 are constants
concentration-dependent terms. We see that 9v is" dependent on elemental
composition, including elements which may be particle size dependent. Our
data for San Miguel (which do not extend into the submicron range) suggest
enhancement of S03 and Fe203 in the fines and are inconclusive for Na20. Any
sample preparation or particle selection problem which biases the size dis-
tribution away from fines will diminish the apparent average S03 and Fe203
concentrations, thereby increasing the calculated resistivity.
We do not have Arapahoe data with which to make a comparison similar to
Figure 7. Our data do show sodium to be enhanced in the fines in Arapahoe
ash (10). Damle et aJL have noted that the literature has conflicting results
for sodium in various coals (12).
O
c
o
o
14-
12-
10-
8-
6-
PVS& PVSAllowacidl
I high acid ]
SAN MIGUEL
112 144 182 227 283 352 441
Temperature,°C
FIGURE 7. Comparison of VS and VSA resistivities with laboratory resistivity
for San Miguel lignite. See Text for conditions and definitions.
Surface-sorbed S03 in equilibrium with the flue gas leads to the acid
resistivity rA. This is defined by an algorithm which differs for "eastern"
ashes, for which Ca + Mg < 3.5% or K > 1.0%, and "western" ashes, taken as
those for which Ca + Mg > 3.5% and K < 1.0%. Sorting one file of 103 San
Miguel particle compositions, we found that it consisted of 75 "eastern"
particles and 28 "western" particles. At low temperatures and high S03 con-
centrations, calculated PvSA's therefore fall into two narrow horizontal
bands in a data map such as Figure 4.
Ill
-------
In the absence of specific flue gas S03 concentration data, it is common
to assume (S03) ~ 0.004 (802). In our case, that would be 17.6 ppm S03,
corresponding to the P (high acid) in Figure 7. We have used P (low
acid) throughout this paper, corresponding to (863)= 0, which approximates
field measurements.
CONCLUSIONS
Reviewing Figure 7, it is apparent that the average particle resistivity
calculation most closely approximates the zero-acid or volume-surface resis-
tivity Pvs. The temperatures of resistivity maxima are poorly predicted,
however. Improved sample preparation and particle selection procedures which
scrupulously avoid biasing the particle size distribution may minimize the
problem.
The real value of the method lies in its ability to describe the breadth
of the resistivity distribution. Furthermore, resistivity vs. particle size
data maps allow one to estimate the variation of this breadth over the entire
range of particle sizes investigated. Although we have employed the Bickel-
haupt model, alternative models may be used if they express resistivity quan-
titatively as a function of those elemental concentrations measurable by SEM/
electron microprobe.
We feel that it is extremely important for persons sizing precipitators
to be fully aware of the degree and the importance of ash particle resistiv-
ity variability. Although this application of SEM to emission control plann-
ing is extremely time-consuming, it may be justified as a routine method by
those contemplating use of suspected "problem" coals.
ACKNOWLEDGMENTS
We wish to acknowledge the contributions of: Diane K. Rindt and A. L.
Severson in x-ray diffraction and fluorescence analyses; Francis J. Schanilec
and Clyde L. Ziegelman for particulate sampling and sizing; and Loretta A.
Weckerly and Annette Ahart for pilot plant operation and engineering support.
The work described in this paper was not funded by the U.S. Environ-
mental Protection Agency and therefore the contents do not necessarily re-
flect the views of the Agency and no official endorsement should be inferred.
112
-------
REFERENCES
1. Bickehaupt, R.E. A technique for predicting fly ash resistivity.
EPA-600/7-79-204, U.S. Environmental Protection Agency, Research Tri-
angle Park, NC, 1979. 105 pp.
2. Selle, S.J., Hess, L.L., and Sondreal, E.A. Western fly ash composition
as an indicator of resistivity and pilot ESP removal efficiency. Paper
No. 75-02.5. Presented at 1975 Meeting, Air Pollution Control Associa-
tion, Boston, MA. June 15-20, 1975.
3. Frisch, N.W. A technique for sizing electrostatic precipitators for
highly variable fuels. J. Air Pollution Control Association. 30:574,
1980.
4. Smith, W.B., Wilson, R.R. Jr., and Harris, D.B. A five-stage cyclone
system for in-situ sampling. Environmental Science and Technology. 13:
1389, 1979.
5. Miller, R.N. and Given, P.H. Variations in inorganic constituents of
some low rank coals. Ash Deposits and Corrosion Due to Impurities in
Combustion Gases. Hemisphere Publishing Co., Washington, B.C. 1977. pp
39-50.
6. Selle, S.J., Tuffe, P.H., and Gronhovd, G.H. A study of the electrical
resistivity of fly ashes from low-sulfur western coals using various
methods. Paper No. 72-107. Presented at the 1972 Meeting of the Air
Pollution Control Association, Miami Beach, FL. June 18-22, 1972.
7. Deer, W.A., Howie, R.A., and Zussman, J. Rock Forming Minerals. Vol. 3.
Sheet Silicates. Longman, London, 1976. pp. 202 ff.
8. Hulett, L.D. and Weinberger, A.H. Some etching studies of the micro-
structure and composition of large aluminosilicate particles in fly ash
from coal-burning power plants. Environmental Science and Technology.
14:965 ff, 1980.
9. Stinespring, C.D. and Stewart, G.W. The surface chemistry of alumino-
silicate particles—application to combustion stream chemistry. METC/
RI-79/7. U.S. Department of Energy, Morgantown, WV. August 1979.
10. Dahlin, R.S., Sears, D.R., and Green, G.P. Baghouse performance and ash
characterization at the Arapahoe Power Station. This Symposium, Session
A-5.
11. Dahlin, R.S. San Miguel station electrostatic precipitator technical
summary report, field test No. 2. DOE/GFETC/10225-2. U.S. Department
of Energy, Grand Forks, ND, 1982. 29pp.
12. Damle, A.S., Ensor, D.S., and Ranade, M.B. Coal combustion aerosol
formation mechanisms: A review. Aerosol Science and Technology. 1:119
ff, 1982.
113
-------
USE OF FUEL DATABANKS FOR THE EFFECTIVE DESIGN OF
STEAM GENERATORS AND AQC EQUIPMENT
by: N. W. Frisch* and T. P. Dorchak
AFFILIATED ENERGY & ENVIRONMENTAL
TECHNOLOGIES, INC.
North Branch, New Jersey
ABSTRACT
Information concerning coal properties and their variability is critical
to the proper design of steam generators and associated gas cleaning equip-
ment (precipitators, FGD and to a lesser degree, fabric filters). For
situations in which a fuel source is well defined, a databank of hundreds of
coal and ash analyses may be used to assess the variability of the fuel and
to develop critical sizing and design parameters.
This paper discusses a comprehensive computer approach which examines
fuel databank information and generates design parameters for a set of opera-
ting conditions. Fuel parameters related to boiler design and operation,
including fusion temperatures, T25Q, fouling and slagging indices, etc., may
be entered or generated for statistical analysis and presentation. Uncon-
trolled and corrected emission levels of particulate and S02, as well as acid
dew point temperatures are developed.
In the case of ESP's, the program can indicate collection area require-
ments for any type of emission rate (mass, concentration or opacity-basis).
Gas conditioning and pulse energization options are included. Both statis-
tical and a wide range of graphical outputs provide the user with desired
guidance; a typical plot would be a mine map with fuel or ESP parameters
overlaid.
* N. W. Frisch is with N. W. FRISCH ASSOCIATES, INC., in Kingston, New Jersey.
114
-------
INTRODUCTION
Information concerning coal properties and their variability is critical
to the proper design of large steam generators and associated gas cleaning
equipment for participate and sulfur oxide emission control. It is well rec-
ognized that the performance of electrostatic precipitators (ESP's) and to a
lesser degree fabric filters (F/F) commonly used for particulate collection,
are sensitive to the properties of the coal. The design of flue gas desulfur-
ization equipment critically depends on fuel characteristics such as sulfur
and chloride contents as well as the base content of the coal ash. The boiler
designer must also match the boiler design and ancillaries to the properties
and rank of the coal. There is no universal boiler/furnace that can operate
properly on all ranks of coal.
The nature of the coal is thus the singlemost important variable to be
considered in the specification of modern-day coal-fired steam generators.
Even with existing units scheduled to burn a different coal, detailed informa-
tion regarding the coal must also be considered of paramount importance.
In the case of coal from a new large mine, prior to mining, exploration
of the mining area may produce analyses and characteristics of hundreds of
cores. A number of these cores, as shown in one scheme presented in Figure
1(1), will be completely analyzed and characterized. For each of these cores,
which can be over one hundred in number in our experience, up to fifty data
points may be generated per core, excluding trace element analysis. It is ob-
vious that even for simple storage such a large data base is best placed into
computer files. More and more large coal companies in the United States are
today using computer storage. Typically, these files can produce limited sta-
tistics on key parameters for use by the purchaser or his architectural engin-
eer (A/E) to specify the fuel conditions for boiler and gas cleaning equipment.
These parameters may include ash, sulfur and Btu content of the coal, as well
as ash oxides.
As will be demonstrated shortly, such computer files are clearly inade-
quate for our objective of developing and presenting the critical parameters
necessary for the design of
• the boiler and furnace, as well as coal handling, storage
and preparation equipment
• the particulate emission control equipment, whether ESP or
F/F (in the case of the ESP, the actual size is developed)
• The flue gas desulfurization equipment
115
-------
[CORE'SAMPLE
DESCRIBE AND WEIGH
" i
~
;STAGE CRUSH TO 1/4" x 0 .
' RESERVE
I CRUSH TO No. 16 x 0
!RIFFLE !
PRESERVE
[EQUILIBRIUM MOISTURE ~l j
; SAMPLE
'RESERVE COMPOSITE! I
EQUILIBRIUM MOISTURE
HGI SAMPLE
Equilibrium Moisture
Hardgrove Grindability
Index at three moisture
levels
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
Total Moisture 1.
Proximate Analysis 2.
Btu 3.
Ultimate Analysis 4.
including Chlorine
Sulfur Forms 5.
Water Soluble Alkalies 6.
Mineral Analysis of Ash
Ash Fusibility 7.
Trace Element Analysis
Possibility FSI for
Higher Rank
'.ANALYSIS SAMPLE
Total Moisture
Ash and Sulfur
Btu
Ultimate Analysis
including Chlorine
Sulfur Forms
Water Soluble
Alkalies
Mineral Analysis
of Ash
Figure 1. Preparation and Complete Analysis of a Subbituminous Coal Core.
TYPICAL COAL DATABANK FILE DEFICIENCIES
The usual databank typically stores only measured parameters; often all
cores are not completely characterized for every fuel parameter. Thus, it is
desirable to be able to estimate the values of the missing parameters. Then
there are design parameters which are rarely measured directly, such as T25Q,
slagging index, etc. A design program must be capable of calculating these
values for a wide range of fuels.
Further, one is often interested in the value of a tentative parameter,
i.e., one which is not well-defined and whose form is subject to modification
as time passes and better information becomes available. Capability to devel-
op such parameters in an on-line mode is most desirable.
Additionally, one is often interested in relating various fuel parameters
one to another; for example, is Hardgrove Grindability Index (HGI) related to
other fuel parameters such as ash content or ash composition or coal heating
value? Clearly, a wide range of plotting capability is required, encompassing
at a minimum X-Y, X-Y-Z, including log coordinates, and also probability type
plots.
116
-------
And finally, if the program is to possess utility for sizing of AQC
equipment, specialized modules must be included. In the example in this paper
we use a fuel data base which contains information about core location, seam
thickness, coal ultimate, proximate and ash chemical analyses, as well as HGI
values. We then concern ourself with precipitator sizing for a specific
(opacity) target, taking into account the variability of the fuels, including
the seam thicknesses. (We, of course, develop the critical boiler design
parameters as well.)
BOILER DESIGN - CRITICAL COAL PARAMETERS
For each complete analysis of the core or coal sample, critical fuel
parameters are developed, as well as certain operating parameters, including
gas flow for a given generator size. The most problematical of these para-
meters, particularly in the case of Western low sulfur coals and lignites,
proves to be the ash slagging and fouling indices. Many different indices
have been proposed as recently summarized by Bryers (2). Slagging indices
based on measured viscosity temperature profiles are recognized as the most
reliable. Unfortunately, due to the expense of the 'determination, the vis-
cosity data is generally not available. AE2's COALMASTER program will calcu-
late a slagging index based on ash composition when such ash viscosity is
lacking.
For Western coals, AE2 experience and research indicates that the best
expression for the slagging factor (Rvs) takes the following form adapted from
the work of Watt and Fereday (3), (4).
58.34 M°'5r
1 1
(2.4-C)0-5 (4 -
Table 1 below relates the level of the slagging index to the severity of
slagging.
TABLE 1. VALUES OF SLAGGING INDEX
Slagging Slagging Index
Classification RVS
Medium 0.5 - 0.99
High 1.0 - 1.99
Severe Greater than 2.00
The reader is referred to the referenced work for definition of the terms
above. The relationship is based on the spread between ^250 and T10,000 or
plastic range of the ash. The COALMASTER program also independently computes
the T25Q values based on the work of Sage and Duzy, as reported by Winegartner
(5). Subsequent examples show the above equation correctly predicts the high
slagging properties shown by some Western subbituminous coals.
117
-------
For Eastern coals, the slagging factor (Rs) may be calculated by the
well-established relationship:
Rs = B/A x % S
where B/A is the base-to-acid ratio of constituents in the coal ash
% S is percent sulfur in the coal, dry basis
Since the acidic components react under fire zone conditions with the basic
components, their balance at a ratio of approximately 1.0 produces typically
the lowest melting slag. (Other components such as the silica to alumina con-
tent may modify the ash fusion temperatures.) The B/A ratio, therefore, is an
indicator of the slagging potential of the coal ash. The sulfur content is
included since low melting sulfates are found in the slag from Eastern high
sulfur coals.
In addition to the slagging behavior expected, the designer must also
contend with the fouling characteristics of the coal. The selection, design
and placement of convection surfaces depends directly on the fouling index.
The number of blowers in these areas will also be affected. Here one must
apply different indices according to the type of ash. For Eastern coals, the
fouling factor Rf longest in use and calculable from the typical data base is
determined by the product of the base-to-acid ratio and the sodium oxide per-
centage in the coal ash.
Rf = B/A x % Na20
A refined index R' is based rather on the soluble sodium content of LTA ash, a
parameter which is typically unavailable. If the latter value is available,
the program can readily calculate this R'.
For Western coals, no factor has found general application. A general
consensus of workers in the field is that fouling is related to the sodium
oxide content of .the ash, the base-to-acid ratio and the ash content. AE2 has
performed regression analysis on a limited number but a wide range of fuel to
develop a function encompassing these three parameters in a form which
accounts for soluble sodium.
R _ K (% Na20)a (% Ash)b
f " 1 + D/(B/A)C
where a, b, c, D and K are constants
In the example that follows, the expression does successfully predict the low-
to-medium fouling characteristic of Western subbituminous coal.
An additional critical fuel parameter that impacts on the mill capacity
is the Hardgrove Index. To account for the impact of the variation of this
parameter in the coal on nominal mill capacity, AE2 uses the following ex-
118
-------
pression:
% of Rated Mill Capacity = A Log (HGI) + B
where A and B are constants for a given mill type
In our Figure 2, additional fuel parameters and related boiler design features
are summarized. Our program clearly presents all the critical coal parameters
for use by the boiler manufacturers for the design of a reliable low-
maintenance boiler of proper size and configuration.
COAL PROPERTIES RELATED
TO BOILER DESIGN
IMPACT ON P-C
FIRING UNITS
Rank
Moisture
Ash
Sulfur
Btu/lb and Ultimate Analysis
Ash Composition
Reactivity
Grindability
Abrasiveness
Furnace Size ?
Heat Release Rates - Btu/ft /hr
Heat Liberation Rates. -
Btu/ft3/nr
Burners & Blowers - No., Size,
Spacings
Hoppers - Openings, Angle
Pendant Heating Surface &
Placement Blowers
Convection Surfaces 8 Placement
Flue Gas Flow Rates
Coal Preparation J
Storage
Figure 2. Major Boiler Design Parameters Related to Coal Properties.
Flue Gas Desulfurization Design - Critical Coal Parameters
Although there are many types of systems in use today, both wet and dry,
using various basic reagents to control sulfur oxide emissions, their design
and operation are critically dependent on the sulfur content of the coal. How-
ever, in the ash of many Western coals in particular, sufficient basic compon-
ents such as calcium and magnesium oxides are present to naturally combine
with the sulfur oxides to reduce their emission rates. This retention of sul-
fur by fly ash has been related to the ash chemistry by Gronhovd (6) and by
Davis and Fiedler (7). The program uses a relationship of the Davis-Fiedler
type as a basis to calculate the uncontrolled emission rate. The level of con-
trol, the stoichiometry and the reagent usage are directly related to these
emission rates. Chlorine content in the fuel is also critical to wet process-
es affecting material selection and makeup composition. These key parameters
are presented by the COALMASTER program in both tabular statistical form and
graphical presentation for ready use in the selection and design of FGD equip-
ment. Presentation on the mine coordinates also allows the mine engineer to
avoid or blend off pockets of excessively higher sulfur content coal.
PRECIPITATOR SIZING - COAL PARAMETERS
It is generally recognized that precipitator performance is sensitive to
fuel characteristics. In the case of highly variable fuels, one is faced with
119
-------
the problem of developing an ESP size that is adequate to treat a very large
portion of the fuel without offering an economically unattractive SCA (ft2
collecting area/1000 acfm).
Sizing precipitators for a specific fuel has progressed from the use of a
migration velocity parameter which is solely dependent upon coal sulfur con-
tent (ca 1950). A number of approaches are in use today; they include use of
performance test data (analogy) (8), proprietary indices (9), migration
velocity-resistivity relationships (10), (11), comprehensive regression equa-
tions (relating W« to both fuel and operating parameters) (12), computer mod-
eling (13), pilot precipitator tests (14), combustor burns (15), and a number
of combination approaches (16). Unfortunately, the experimental effort ap-
pears to be diminishing and more and more reliance is placed on paper
approaches.
It is not an objective of this paper to discuss in detail sizing approach-
es. Rather, we are most concerned with the presentation of variable fuel data
in the ESP specification and how this data is properly treated by the ESP
designer.
Commonly, the A/E will examine the individual histograms of the coal ele-
ments and ash oxides and develop minimum and maximum values for each of the
components. (Subjective truncation is usually practiced at this point.) This
data and the mean value are presented in tabular form in the specification. A
specification of this type was prepared using published data (Table 2).
TABLE 2. EXAMPLE OF FUEL SPECIFICATION
Total
Ultimate Analysis
(% by weight, as received)
Mean Min Max
Carbon
Hydrogen
Nitrogen
Sulfur
Oxygen
Ash
Moisture
37.21
2.78
0.67
0.63
11.74
7.14
39.83
21.26
1.52
0.52
0.18
10.71
3.89
27.78
39.87
3.42
1.03
1.41
12.45
15.95
52.53
100.00
Higher Heating Value,
Btu/lb 6258 3068
7660
Equilibrium
Moisture.%35.51 24.80 4930
Hardgrove Grindability
Index 35.90 22.00 136.00
Ash Chemical Analysis
Si02
A1203
Ti02
Fe203
CaO
MgO
K20
Na20
Li20
S03
P205
Undetermined
(5
Mean
26.49
12.43
0.49
8.30
24.61
6.83
0.73
1.25
--
17.93
0.16
0.78
100.00
I by weight)
Min
4.66
2.22
0.04
2.65
8.80
2.70
0.11
0.12
--
3.77
.01
.01
Max
74.55
19.90
1.09
20.30
42.80
12.00
2.36
7.25
--
32.62
0.94
1.09
Ash Fusion Temperature, °F
Reducing
IT
ST
HT
FT
2183
2211
2237
2263
1900
1940
1960
1990
2820
2820
2820
2820
120
-------
The use of such a fuel specification presents several problems. First,
it is difficult to assign a probability of occurrence to any of the critical
fuel or ash component levels. Equally important is the fact that one cannot
reliably develop individual fuel and ash sample compositions from this type of
specification in the general situation (17). And it is these individual fuel
samples which dictate specific levels of ash loadings, ash resistivity and
particle size upon the precipitator system.
Let us examine a relatively simple situation in which only three fuel
parameters, A, B and C, influence ESP performance. High levels of A improve
performance while high levels of B or C depress performance or require higher
efficiency levels for the same design outlet level. (We may think of A as
B as CaO and C as ash/Btu ratio.)
When the designer selects the limiting values of A, B and C, using a
specification of the type shown in Table 2, the highest levels of B and C, and
the lowest level of A become the basis for the design. This, in theory, deter-
mines the worst case situation. This selection is shown on Figure 3. a.,
points 1 and 1'.
C max
C min
C2
B max
B min
3.a.
A min
B2
A max
Figure 3.
Locating the Limiting Values of the Design Parameters
(simple ESP situation)
But we are not even certain that this extreme combination occurs in the
fuel bank and if it does, whether it describes 0.1, 1 or 5% of the fuel. Al-
ternatively, a less conservative designer might choose another fuel basis for
design, possibly even the mean composition. Coy and Frisch (18) have dis-
cussed some of the implications of these alternate choices.
121
-------
If all the individual analyses are represented on Figure 3.a., we typi-
cally arrive at a plot similar to Figure 3.b. Here, many of the data points
cluster about the mean and the plot exhibits a decreasing density of data
points as one moves away from the mean. An elliptical boundary (contour
ellipse) can be drawn to encompass a large selected proportion of the fuel.
Knowing these boundaries, one is able to define a composition, namely point 2,
which has a known small probability of occurrence. Compositions A2, 62 and
62 occur together in a sample; A2 and 62 describe a relatively high resistivi-
ty ash and C2 fixes a high efficiency requirement. Typically, a significant
lower SCA requirement results for condition 2 compared to condition I1.
Techniques for locating point 2 are readily applied in some simple cases
(19). More complex situations require both definition of a significant number
of statistical parameters and also computer solution. In these real-life
situations, we prefer an alternate approach.
AE2's approach involves sizing of ESP's for each core analysis, taking
into account coal seam thickness, and all variables which affect fuel flow
rate, gas flow rate, ash loading, ash resistivity, ash particle size, gas
phase composition and density. Ash loading, ash resistivity and ash particle
size are especially important parameters. Loading and resistivity may exhibit
wide ranges within a single fuel bank. Ash loading is best characterized by
use of the ratio of ash to Btu; resistivity is estimated by the Bickelhaupt
approach (20) or alternatively for a given rank fuel the use of proprietary
regression equations can be applied in the computer program. The Bickelhaupt
approach combines the role of various ash species (Na, Li, Ca, Mg, Fe) and
gas phase components (S03 and H20) in influencing fly ash resistivity.
Clearly, a computer approach is required to handle the large data base
involved and to develop the statistics of the pertinent parameters and via
graphics, to present visually the relationships between various fuel and pre-
cipitator parameters. Especially useful are cumulative plots of required C.E.
area or similar plots of outlet loading or opacity for a given SCA ESP. Other
useful plots depict the mine area with important ESP or fuel parameters over-
laid; this permits one to locate especially difficult fuel areas which may
then be selectively mined, blended or eliminated from the mining plan.
A comprehensive example follows.
APPLICATION OF COALMASTER PROGRAM
A brief description of the use of the program follows.
First, the coal file is read; coal and ash parameters, as well as avail-
able boiler parameters are input. The program has been written in a most ac-
commodating manner to handle the various means of reporting data in use today.
Parameters not in the file are calculated by the program when feasible.
Thus, in the absence of ash fusion temperatures, the standard ones are estimat-
ed (IT, ST, HT and FT - both reducing and oxidizing).
122
-------
Operating parameters for the precipitator are requested. Options for
conditioning or pulse energization are allowed. The program performs combus-
tion calculations, resistivity and ash loading estimates, etc., and sizes for
each core. The operator then may provide a design collecting area, under the
guidance of the program. The pertinent (outlet) parameters for the specified
precipitator are then calculated.
Next the operator is provided with five options, including graphics and
statistics. The graphical display is most useful and can provide excellent
insight into various interactions between variables.
Another distinctive feature of AE2's program is the ability to define
from the keyboard any new parameter in a most general form. Then one may per-
form statistics or graphics in a manner analogous to the manner used with the
standard (built-in) parameters.
Selected output of the program is discussed in the next section.
BOILER DESIGN PARAMETERS
As shown previously, the capacity of a particular mill depends on the
Hardgrove Grindability of the coal. Of equal significance, the throughput in
terms of Btu per hour depends also on the heat content (Btu/lb) in the coal.
When the product of the mill capacity and heat content is at a minimum level,
the throughput necessarily falls to a minimum. Determining this minimum by
simply multiplying minimum levels of HGI and Btu found in the coal statistics
obviously results in an unrealistically low design throughput. Rather, the
minimum throughput is best determined by calculating the throughput for each
mill capacity. In the case of an existing unit, the adequacy of the existing
mill system can be evaluated beforehand to avoid costly mill problems. Figure
4 presents the results of our calculations.
We note that all but two of the more than one hundred calculated values
fall below the contour ellipse. Eliminating these extreme points which repre-
sent less than two percent of the coal, our design point is determined by-
point A, which corresponds to a heating value of 8200 Btu/lb and a 65% capa-
city. The contour ellipse encompassing the specified proportion of the fuel
population passes through point A and exhibits a minimum throughput of
4.97(10)5, which becomes the basis of design. Alternatively selecting point B
as the design point, corresponding values of 8500 Btu/lb and 39% are obtained,
resulting in nearly a 50% jump in design mill capacity. Our next considera-
tion might be the furnace size itself.
The size and even the type of furnace depend on several factors, but
prominent among these is the slagging character of the coal. To evaluate this
property, the COALMASTER program calculates the slagging index Rys for each of
the cores. These can then be inspected by plotting the indices as a percent-
age of the coal tonnage as in Figure 5. In this example, the index varies
from about 0.75 to 1.8, indicating for a Western type ash, medium to moderate-
ly high slagging characteristics. (Refer to (2) for the guidelines used here.)
123
-------
•••«•«•••• x-r KOl •••«••••.•
MIL UHCIIT, I • HIM HUIM WlIX
I.IIM'H I.WII.II !.««•« t.l'U-H f.UIIMl
I ......... I ......... I ......... I ......... I
!.««•« t.«K'«
I1MIIM IMCI
1.1 I. >.
M. M. 71.
...I I I..
....••
I | I I I I I I.
>.;tX.M I.IIM'M I.UIOU !.««•« f.im>« t.UII'H '.M«'IJ >.»«>M 1.«tl«M
... [Mil •« KttlH MM. III'LI V OU.
Figure 4. Relative Mill Throughput
and Heating Values for
Individual Cores.
t.s 1.2. s, it. it. st. ;». t». ». «. n. «.
Figure 5. Probability Plot of
Slagging Index.
The boiler designer would therefore select an intermediate-sized furnace
corresponding to an intermediate heat-release rate (Btu/ft^/hr). Beyond the
furnace zone, other areas of the boiler are also affected by the properties of
the ash.
In the back pass of the boiler, the positioning and spacing of the plat-
ens and tubing in the superheater-reheater and economizer sections are criti-
cally dependent on the fouling properties of the ash. Wider spacing is re-
quired to prevent bridging and plugging as the fouling index increases. The
type of airheater can also be affected. We at AE2, however, believe the foul-
ing index should also be evaluated along with the slagging characteristics. A
situation involving an evaluated medium fouling type ash with a severe slag-
ging characteristic represents a more difficult situation than one with a
medium slagging characteristic. The severe slagging ash raises furnace temp-
eratures, thereby raising back pass temperatures, accelerating tube fouling.
The simultaneous occurrence of high slagging and medium fouling is readily
determined by plotting the slagging index versus the fouling index as shown
in Figure 6. The points falling in the upper-right-hand quadrant represent
cores with high slagging and medium fouling properties. They represent about
18% of the cores, a significant portion of the fuel. The designer is there-
fore cautioned to give added weight to the indicated medium fouling eharacter-
124
-------
isties. Looking even further down the
line, the control of participate and
sulfur oxide emissions must be con-
sidered.
Flue Gas Desulfurization Parameters
As previously discussed, the re-
tention of sulfur in the ash, parti-
cularly a basic Western ash, can
significantly impact on the efficiency
demands on the FGD system. This is
clearly demonstrated in our next
example presented in Figure 7. The
sulfur emission levels anticipated
are shown in a cumulative plot of
the coal tonnage. Levels do not
exceed the old standard of 1.2 Ibs
per million Btu. In contrast, sulfur
levels in the coal do actually range
above this standard to about 1.35 Ibs
MI CIIIIIM IIVCI, LI/M ITU
I.I I. I. i.
Figure 7. Probability Plot of
Emissions Level.
• x-r not >
I f.«4X-t1 1.27K*M l.MK'M I .»»«*» 2.2UE*M Z.S2B«M 2.IUC*M 1.I47C*OI
LOU FOUUM
HIM KMtm
LU FWLIM
Kill* tuMMI
I
7.IJM-II-
I,
i,jm-*1 ».*4H H 1.I?tC*M 1,5I«»« I
••* tU21 ••• FftHIII 1HKX
HIM UMCIM
Kill* FOUL1W
MOIUH K.MSIM
l.iMI'H I.1IH-M i.l»t>M l.H7l
-------
IIL1TT PLOT M«««MW*
costs amount to over 30 million
dollars over the life of the plant.
PRECIPITATOR PARAMETERS
In this situation, a design
opacity of 20% was specified. Figure
8 is a probability plot of outlet
opacity for the design collecting area
specified by the designer. This area
was correctly specified and will ade-
quately treat about 97% of the coal.
Three isolated cores would require
extraordinary levels of CE area; it is
not appropriate to design for these
cores.
Examination of the relationship
between outlet opacity and various
coal and ash parameters indicates
that ash resistivity was the dominant
I.IM»«1 I.IMIrtl
M* III}] *M
l.t'll'tl 1.«lt«l 1.1IK*>1 I.IIMMI I.Hrt'fl I.IIXXI I.IMHI
INI III * «• Kllltllltl. M C«
Figure 9. Dependence of Outlet Opacity
on Ash Resistivity.
OITK! mcitT, I W1IMI
1.9 1. i' 1
.1..I...1 1
4.IMC**)
M.
I..,
ft. n.
.I....L.
•t. n. n.s
..I...!...!
...i i i..
II. U. >«.
tn * m cot
!....!...1.1.. .!...i
•. ts. n. ft. **.9
Figure 8. Probability Plot of Outlet
Opacity.
factor. Figure 9 shows the dependence
of outlet opacity on ash resistivity.
A similar plot demonstrated that ash
loading was not as critical.
. Figure 10 is a mine map overlaid
with ash resistivity indicators. (The
Z indicators, scaled from 0 to 9 in
order of increasing resistivity, refer
to each bore hole; the key is not
shown in our paper.) We have deline-
ated on the map two relatively low
resistivity areas and one critical
resistivity zone (>5(10)" ohm cm).
Use of the X-Y-Z plot permits one to
modify the mining plan if economically
feasible or to take appropriate action
to avoid unattractive coal (high
resistivity, high slagging ash, etc.).
126
-------
And finally, Figure 11 shows the
perversity of nature. Fuels which
require relatively low ESP collecting
areas are those which exhibit the
higher levels of fouling index. This
is a consequence of the opposing de-
pendence of ash resistivity and foul-
ing tendency upon soluble sodium con-
tent of the ash.
CONCLUSION
The large amount of data gener-
ated during the evaluation of large
coal mining properties must be handled
by computers. The computer programs
typically used, however, lack flexi-
bility, particularly in the generation
of new parameters and indices, as well
i »-t-i HOT <
I . Ml Htlltlllll. KM 01
1 CNtl.
4.4M
3.7]M»II-
I.17U.M I.MK.M !.««€•«• I.I1IM* J.SIJC'W l.UH'M I.IOMO
| 1 1 1 1 1 '•
7.MII-W-
1
t.»M-<>
<.M I.IW'tl I.V7KII J.UM'»I I.I1H-II J.IMl'll I."11-11 4.272IX1
•~ IH11 .~ I MOM.
Figure 10. Mine Map Showing Location
of High and Low Ash
Resistivity Zones.
as in the types of plots that can be
generated. The COALMASTER program was
developed to overcome these short-
comings and extend the data base where
necessary to produce basic design
parameters for large boiler installa-
tions and associated air quality con-
trol equipment. An additional objec-
tive is to size the electrostatic
precipitator required to meet any
given emission target whether
efficiency, mass or opacity.
The COALMASTER program, therefore,
extends the basis for the technical
.;';:;;»:.i';:j;«:H';:»«:M';:mjr»';:;;i{;u'j:MK:«'i:H«:M'i:H«^ and economic evaluation of new or even
Figure 11. Relationship of Required ESP
Collecting Area and Fouling
Index.
127
-------
currently mined coal deposits. Key parameters are presented in both tabular
statistical form and graphical plots for the designer of equipment in-
volved in the utilization of the resource. In light of today's high capital
costs and ever-increasing maintenance costs, it is imperative that the design
match the coal. The responsibility here is shared among the producer of the
coal, the boiler owner and his architectural engineer, as well as the vendor
of the equipment. AE2 welcomes inquiries from these parties and others re-
garding the use of the program.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore the contents do not necessarily reflect the
views of the Agency and no official endorsement should be inferred.
128
-------
REFERENCES
1. Duzy, A.F., et al. Western Coal Deposits, Pertinent Qualitative Evalua-
tions Prior to Mining and Utilization. Paper presented at the 1977
Ninth Annual Lignite Symposium, May 18, 1977, Grand Forks, N. Dakota.
2. Bryers, R.W. On-Line Measurements of Fouling and Slagging and Correla-
tion with Predictive Indices. Paper presented at the June 1981 Lignite
Symposium sponsored by D.O.E. at San Antonio, Texas.
3. Watt, J.D. and Fereday, F. The Flow Properties of Slags Formed from the
Ashes of British Coals: Part 1. Viscosity of Homogeneous Liquid Slags in
Relation to Slag Composition. J. Inst. of Fuel XL II, No. 338, 99-103,
(Mar.1969) Part 2. The Crystallizing Behaviour of the Slags, J. Inst. of
Fuel XLII No. 339, 131-134 (Apr. 1969).
4. Barrick, S.M. and Moore, G.F. Empirical Correlation of Coal Ash Viscosity
with Ash Chemical Composition. Paper presented at ASME Annual Meeting,
Dec. 5, 1976, New York, NY.
5. Winegartner, E.G. and Rhodes, B.T. An Empirical Study of the Relation of
Chemical Properties to Ash Fusion Temperatures. J. Engineering for Power
.97, 395-406 (1975).
6. Gronhovd, G.H., Tufte, P.M. and Selle, S.J. Some Studies on Stack Emis-
sions from Lignite-Fired Power Plants. Paper presented at the 1973
Lignite Symposium, May 9-10, 1973, Grand Forks, N. Dakota.
7. Davis, W.T. and Fiedler, M.A. The Retention of Sulfur in Fly Ash from
Coal-Fired Boilers. J. Air Pollution Control Association _32 395-397
(April 1982).
8. Frisch, N.W. Engineering Manual for Fly Ash Precipitators (1981).
9. Matts, S. Coal-Ash Composition and Its Effects on Precipitator Perfor-
mance. Flakt Engineering 1 No. 3 (Oct. 1977).
10. White, H.J. Industrial Electrostatic Precipitation, Addison-Wesley
Publishing Company, Inc., Reading, MA, 1963.
11. Sproull, W.T. Collecting High Resistivity Dust and Fumes. Laboratory
Performance of a Special Two-Stage Precipitator. Ind. Eng. Chem. 47
940-944 (No. 5, 1955).
12. Frisch, N.W. and Coy, D.W. Sizing Electrostatic Precipitators for High
Temperature. Paper presented at Symposium on the Changing Technology of
Electrostatic Precipitators, Adelaide, S. Australia, Nov. 8, 1974.
13. McDonald, J.R. A Mathematical Model of Electrostatic Precipitation,
Vol. I, Modeling and Programming; Vol. II, User Manual, EPA-600/7-78-111
a,b - June 1978.
129
-------
14. Saponja, W. A Systematic Approach to the Application of Electrostatic
Precipitators on Lower Sulfur Coals. Paper presented at Canadian Elec-
trical Association, Edmonton, Alberta, Oct. 1974.
15. Wagoner, C.L., Barrick, S.M., Vecci, S.J. and Piulle, W. Fuel and Ash
Evaluation to Predict Electrostatic Precipitator Performance-A Progress
Report. Paper presented at the IEEE-ASME Joint Power Generation Confer-
ence, Long Beach, CA, Sept. 18-21, 1977.
16. Tassicker, O.J. and Sproull, W.T. Improved Precipitator Technology by
Pilot Plant Testing and Evaluation of Coal Bore-Cores. Paper presented
at Particulate Control in Energy Processes Symposium held at San Fran-
cisco, California on May 11-13, 1976.
17. Engelbrecht, Heinz. Hot or Cold Precipitators for Fly Ash from Coal-
Fired Boilers. Paper presented at Symposium on Coal Utilization and Air
Pollution Control, Western PA section, Air Pollution Control Association,
Apr. 21-22, 1976, Pittsburgh, PA.
18. Coy, D.W. and Frisch, N.W. Specifying Precipitators for High Reliabil-
ity. Paper presented at Symposium on Control of Fine Particles,
September 30 - October 2, 1974, Pensacola Beach, Florida.
19. Frisch, N.W. A Technique for Sizing Electrostatic Precipitators for
Highly Variable Fuels. J. Air Pollution Control Association ^0 574-575
(1980).
20. Bickelhaupt, R.E. A Technique for Predicting Fly Ash Resistivity.
EPA-600/7-79-204, August 1979.
130
-------
DEVELOPMENT OF INHALABLE PARTICULATE (IP) EMISSION FACTORS
by: Dale L. Harmon
Industrial Environmental Research Laboratory
U. S. Environmental Protection Agency
Research Triangle Park, N. C. 27711
ABSTRACT
At the request of EPA's Office of Air Quality Planning and Standards
(OAQPS), ORD is conducting a study characterizing inhalable particle (IP)
emissions from various sources for the development of emission factors.
Three contracts were awarded in September 1979 to conduct source
characterizations for IP from major sources. The testing phase for these
contracts is near completion, and individual reports on the major sources which
will include the IP emission factors are being prepared. The IP emission
factors are based on existing particle size data and the IP source character-
ization tests. This paper gives an overview of the EPA program to develop IP
emission factors.
This paper has been reviewed in accordance with the U. S. Environmental
Protection Agency's peer and administrative review policies and approved
for presentation and publication.
131
-------
INTRODUCTION
Early in 1978, a task force on inhalable particulate (IP) emission
characterization was formed to develop a program for determining emission
factors based on cutoff size for IPs from both controlled and uncontrolled
sources. In 1978 EPA's Office of Research and Development (ORD) completed
a plan for a program to obtain the IP data as specified by a priority listing
developed by EPA's Office of Air Quality Planning and Standards (OAQPS). This
plan is the basis for the ongoing IP emission factor development program.
DISCUSSION
The IP Emissions Factor Task Force Executive Committee which was formed in
1978 has been meeting about once a month since February 1979 to direct the work
for development of IP emission factors. The committee has members from the
following EPA organizations:
Office of Air Quality Planning and Standards (OAQPS),
Industrial Environmental Research Laboratory-RTP (IERL-RTP),
Industrial Environmental Research Laboratory-CIN (IERL-CIN),
Environmental Sciences Research Laboratory (ESRL), and
Division of Stationary Source Enforcement (DSSE).
An early task for the committee was to establish a list of priority
sources to be tested for IP emissions. Since the range of sources emitting
IP was large and the time and funds available were limited, it was necessary
to make use of existing data to the maximum extent practical and test sources
which would provide the highest practical level of return. The priority
sources identified early in the program are listed in Table 1. Changes made
in the priority list since it was developed have been to add iron foundries
and industrial roads and to eliminate primary zinc smelters and incineration
from the list. At the time the original priority list was developed and
funding levels were established, funds were included for lower priority sources
to be added when identified; however, budget reductions, delays in testing, and
increased testing costs have eliminated any testing beyond that now planned.
When the IP emission factor program was first begun the Agency was exam-
ining the effects of several particulate size fractions including IP matter,
then defined as -15 ym aerodynamic equivalent diameter, fine particulate,
-2.5 ym aerodynamic equivalent diameter, and fractions between these size cut
points. (2) The IP fraction is based on particulate matter which can deposit
in the conducting airways and gas exchange areas of the human respiratory
system while breathing through the mouth. A fine fraction -2.5 Urn is based
on chemical composition and the bimodal size distribution of airborne particles
and the predominant penetration of particles -2.5 ym into the gas exchange
region of the respiratory tract. (3)
A large data bank ( > 300 test series) of particle size data on various
industrial sources was in existence prior to initiation of the IP emission
factor program. The most useful data was taken with impactors which have an
upper stage cut-off of 10 ym or less. For this data to be used to develop
132
-------
IP emission factors it was necessary to extrapolate from existing data to
15 ym. An extrapolation procedure was developed and data from the Fine
Particle Emissions Information System (FPEIS)(4), a data bank developed for
EPA containing much of the existing particle size data, has geen extrapolated
to 15 ym. This data is adequate to provide IP emission factors for some
industrial sources without additional testing.
During the time the IP emission factor program was being developed,
it became apparent that the measurements problem for obtaining IP source data
was much more difficult than originally believed. It was determined that a
sizeable effort would have to be devoted to this area before any field
measurements were actually undertaken. To this end a meeting was sponsored
by EPA's Industrial Environmental Research Laboratory and Environmental
Sciences Research Laboratory, both of Research Triangle Park (IERL-RTP and
ESRL), bringing together many of the nation's measurement experts to develop
a program to provide the necessary measurement and sampling techniques and
expertise. These measurement experts decided that a 2-stage cyclone set would
be the best device to use for point source sampling. Such a system was dev-
eloped for EPA by Southern Research Institute to provide 15 and 2.5 ^m
size cuts. Before fabrication of these cyclone sets was complete, however,
questions began to arise as to selection of a 15 ym upper limit for IPs. In
the spring of 1980 it was determined that the definition of IP might be
revised to some cut point less than 15 ym so that the cyclones could no longer
be used. At this time, the approved point source sampling method became an
impactor with a 15 ym cyclone precutter. Data is reported as a continuous
plot of emission factor vs particle size from 15 ym down. When the IP
cut point is finalized the appropriate emission factor can be read from the
graph. An example curve is shown in Figure 1.
It is necessary to obtain IP fugitive emission data on industrial
sources as well as point source data. Four test methods have been used for
sampling IP fugitive emissions depending on the characteristics of the site
to be sampled:
Quasi-Stack Sampling,
Roof Monitor Sampling,
Upwind/Downwind Sampling, and
Exposure Profiling.
Where cooling and dilution of ducted gas streams result in the formation
of condensed particles, it is necessary to measure these condensibles as part
of the IP emissions. One of the early tasks in the IP program was to develop
a condensibles sampler. A prototype device was designed, fabricated, and field-
and laboratory-tested by Southern Research Institute. It was intended that
each of the contractors doing IP emission testing for EPA would fabricate one
or more of the condensible samplers developed by Southern Research; however,
by the time the sampler was developed it was not cost effective to do this
with the limited funds remaining for testing, so all condensible testing for
the IP emission factor program is being done with the prototype unit.
Draft copies of protocols for these test methods have been developed for
those doing the IP testing. (5)(6)(7)(8)
133
-------
In September 1979, as the result of a competitive procurement action,
contracts were awarded to three IP characterization contractors to conduct
plant surveys and source assessments for IPs and to support OAQPS in the
development and implementation of an IP standard. Contracts were awarded to
GCA Corporation, Midwest Research Institute, and Acurex Corporation. The scope
of work for these contracts requires the contractor to conduct on-site plant
inspections or surveys for the purpose of defining and evaluating the partic-
ulate pollution problems and for determining the fugitive emission sources.
Following approval of a test plan developed by the contractor for the
sites selected by EPA, the contractor conducted tests for ducted particulate
matter, fugitive emissions, and condensible particulate matter as required.
When the IP characterization contracts were initiated, it was planned to
use EPA personnel from IERL-RTP and IERL-CIN who were familiar with the
industries to be tested to make test site selections. This did not work out
in most cases because many of the EPA personnel did not have adequate time to
devote to test site selection. It was necessary to direct the IP character-
ization contractors to develop priority lists for some industries and recommend
specific sites for testing. Early in the program, it was decided to try to
work on a cooperative basis with industrial organizations to gain access to
test sites rather than use Clean Air Act Section 114 letters to gain access.
Most of the industrial organizations contacted were willing to work on this
project on a cooperative basis, but working out such agreements was time con-
suming .
It was originally planned that much of the testing would be done early
in the contract period, but this was not possible. In addition to the delays
in test site selection and gaining access to test sites, significant delays
have resulted from the need to develop IP sampling equipment (including the
condensibles sampler and size selective inlets and elutriators) for use in the
fugitive measurements program.
It would have been desirable to conduct tests on several different sites
for each source type, but the number of sites tested was limited by the funds
available. Where possible, tests for IP emissions were combined with other
EPA field tests to stretch the limited funds. Also, existing data is being
used where possible to supplement IP testing. Table 2 lists the IP tests
which have been completed to date. The only testing remaining to be completed
with the available funds are tests of a steel mill EOF stack and an electric
arc furnace and condensibles testing of a coal-fired industrial boiler, batch
type asphalt plant, and (possibly) a coal-fired utility boiler.
All testing planned for paved urban roads has been completed. Testing of
unpaved roads and industrial paved roads is near completion.
Tests at seven different steel plants have been completed, and tests at
two additional steel plants are planned. Processes tested in steel plants
include the cast house, sinter plant, quench tower, basic oxygen furnace, hot
metal desulfurization, Q/BOP, paved and unpaved roads, and coal piles. Tests
at one lime plant were completed. Processes tested were a kiln controlled by
an ESP, a kiln controlled by a fabric filter, material transfer, and product
loading fugitives. IP emissions from kilns on a dry process cement plant and
a wet process cement plant have been completed.
134
-------
Tests on three ferroalloy electric arc furnaces have been completed.
The furnaces tested were producing silicon metal, ferromanganese, and ferro-
silicon. Tests at an iron foundry were completed for pouring and cooling
operations. Tests at a secondary lead smelter were completed. Processes
tested were blast furnace metal and slag tapping and charging, agglomeration
flue, refining kettle, and pigging machine. Slag and matte tapping operations
were tested at a primary copper smelter.
Emissions from a drum mix asphalt plant have been measured, and plans
are to measure the condensible emissions from a batch mix asphalt plant.
Tests were completed at two pulp and paper plants. Processes tested were
a nondirect contact evaporator type recovery furnace, a direct contact
evaporator type recovery furnace, a lime kiln, and a smelt dissolving tank.
There is a large body of existing data available on combustion source
emissions so that the only testing planned is for condensibles. Condensibles
tests have been completed on a coal-fired utility plant, an oil-fired utility
plant, and an oil-fired industrial boiler.
Test reports are being prepared for each field test, but it is not
intended to publish these reports. As the testing is completed for a given
source category, such as the ferroalloy industry, then one of the three IP
characterization contractors is given the task of preparing a source
category report on that category. This contractor will combine test
results from all contractors, extrapolate valid pre-existing data, develop
source category IP emission factors, and prepare a report which will be
used for the AP-42 type input. The source category reports, which will
contain summaries from the IP test reports, will be published as EPA reports.
Source category reports will be prepared for the following 10 source
categories:
1. Paved Roads
2. Industrial and Unpaved Roads
3. Iron and Steel
4. Ferroalloy
5. Cement and Lime
6. Primary and Secondary Nonferrous
7. Iron Foundries
8. Asphaltic Concrete
9. Kraft Pulp Mills
10. Combustion
Preparation of source category reports has started on all source cate-
gories except industrial roads and combustion. First drafts of the paved
road and the ferroalloy source category reports have been completed and
reviewed by EPA. Final corrections are being made on these two reports,
after which they will be submitted for peer review and publication. All
of the source category reports should be completed by early spring of 1983.
135
-------
REFERENCES
1. Compilation of Air Pollutant Emission Factors, Third Edition and
Supplements, AP-42, U. S. Environmental Protection Agency,
Research Triangle Park, N. C., August 1977 (and later).
2. Staff Paper Outline for Particulate Matter, Office of Air Quality
Planning and Standards, U. S. Environmental Protection Agency,
Research Triangle Park, N.C., July 31, 1980.
3. F. J. Miller, et al., "Size Considerations for Establishing a Standard
for Inhalable Particles," Journal of the Air Pollution Control Association,
29(6): 610-615, 1979.
4. Reider, J. P., R. F. Hegarty, "Fine Particle Emissions Information
System: Annual Report (1979)," EPA-600/7-80-092 (NTIS PB80-195753),
U. S. Environmental Protection Agency, Research Triangle Park, N.C.,
May 1980.
5. Harris, D. B., "Procedures for Cascade Impactor Calibration and Operation
in Process Streams," Revised 1979, 2nd Draft, U. S. Environmental
Protection Agency, Research Triangle Park, N. C., 1979.
6. Protocol for the Measurement of Inhalable Particulate Fugitive
Emissions from Stationary Industrial Sources, Draft, U. S. Environmental
Protection Agency, Research Triangle Park, N. C., March 1980.
7. Wilson, R. R., W. B. Smith, Procedures Manual for Inhalable Particulate
Sampler Operation, Draft, U. S. Environmental Protection Agency,
Research Triangle Park, N. C., November 1979.
8. Williamson, A. D., Procedures Manual for Operation of the Dilution
Stack Sampling System, Draft, U. S. Environmental Protection Agency,
Research Triangle Park, N. C., October 1980.
136
-------
TABLE 1. IP EMISSION FACTOR PRIORITY SOURCES
1. Paved and Unpaved Roads
2. Iron and Steel Manufacturing
3. Secondary Lead Smelters
4. Portland Cement Manufacturing
5. Lime Manufacturing
6. Asphaltic Concrete Manufacturing
7. Ferroalloy Manufacturing
8. Primary Nonferrous Smelters
a. Copper
b. Lead
c. Zinc
d. Aluminum
9. Kraft Pulp Mills
10. Combustion (Coal/Oil)
a. Utility
b. Industrial
c. Commercial/Residential
11. Incineration
137
-------
TABLE 2. COMPLETED IP EMISSION TESTS
Source Category
Paved and Unpaved Roads
Iron and Steel
Cement and Lime
Ferroalloy
Iron Foundry
Primary and Secondary Nonferrous
Asphaltic Concrete
Pulp and Paper
Combustion
Sources Tested
Urban Paved Roads
EOF
Hot Metal Desulfurization
Paved Roads
Coal Storage Pile
Cast House
Q/BOP
Sinter Plant
Quench Tower
Lime Plant
Kiln-ESP
Kiln-fabric filter
Material Transfer
Product Loading
Cement Plant
Kiln-wet process
Kiln-dry process
Electric Arc Furnace
Silicon Metal
Ferromanganese
Ferrosilicon
Pouring and Cooling
Secondary Lead-various ducted
and fugitive sources
Primary Copper-matte tap,
slag tap, and idle ladle
Drum Mix Plant
Recovery Furnace-nondirect
contact evaporator
Recovery Furnace-direct contact
evaporator
Lime Kiln
Smelt Dissolving Tank
Coal-Fired Utility Boiler-
condensibles
Oil-Fired Utility Boiler-
condensibles
Oil-Fired Industrial Boiler-
condensibles
138
-------
4.0
2.0
b
X
B
I 1.0
of 0.8
o
u
2 0.6
0.4
UJ
>
0.2
0.1
1 Ib/ton = 0.5 kg/metric ton
0.1
0.2
6 8 10
0.4 0.6 0.8 1.0 2.0 4
PARTICLE DIAMETER, jug
Figure 1. Emission factor for controlled emissions from hot metal desulfurization plant based
on one test.
20
139
-------
IlMHALABLE PARTICULATE MATTER RESEARCH
COMPLETED BY GCA/TECHNOLOGY DIVISION
by: Stephen Gronberg
Senior Environmental Scientist
GCA/Technology Division
213 Burlington Road
Bedford, MA 01730
ABSTRACT
GCA/Technology Division has completed literature surveys and stack tests
in order to develop reliable size-specific particulate emission factors. The
majority of work concerned the iron and steel, ferroalloy and iron foundry
industries. Particulate emission rates and particle size distribution were
measured at eight facilities. Typically, tests were conducted before and
after a control device and only in-stack techniques were used. The results of
these tests and information from other test programs have been reviewed,
ranked, and included in a Source Category Report for each industry. The
Source Category Reports provide an updated section of AP-42 "A Compilation of
Emission Factors" and present background information on industry trends,
engineering specifics and control devices.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
140
-------
INTRODUCTION
The EPA Office of Research and Development contracted GCA/Technology
Division in June of 1980 to develop size-specific particulate emission factors
for the iron and steel, ferroalloy production and gray iron foundry
industries. Through literature searches, telephone surveys and other methods,
GCA prepared test plans which were a summary of the best available particulate
emissions data for each industry. Potential emission rates, control devices
and test sites were described in industry-wide Test Plans. EPA subsequently
requested permission to conduct IP tests at potential test sites of interest.
Eight sources have been voluntarily sampled using IP procedures and two more
tests programs are scheduled. Reports describing each of the completed test
programs in detail have been submitted to and reviewed by EPA.
After completing several tests, EPA issued technical directives to
compile Source Category Reports for each of the three industries. The
Ferroalloy industry tests were completed first and this Source Category Report
was prepared and submitted to EPA. Comments on the report are presently being
implemented in all three Source Category Reports. The reports present
background information for revised sections of AP-42 "A Compilation of
Emission Factors." These reports are an up-to-date summary of reliable
emissions information and are useful to States when revising Implementation
Plans and to control device manufacturers, local regulatory agencies and other
industrial personnel.
Present areas of IP work at GCA include more in-depth analyses of the
data and samples obtained during field tests. Samples of respirable and
nonrespirable particulates are being analyzed for up to 28 elements of
interest, such as cadmium, lead, iron, etc. Neutron activation and x-ray
fluorescence techniques are being used and the first report of results is
presently under evaluation. Through these analyses, source signatures will be
developed where the elemental composition of certain size particles will be
determined. Only a trial set of samples from a sinter plant test have been
analyzed; however, all sources tested as part of the IP program will
eventually be analyzed by one or both methods. In addition, volatile and
nonvolatile carbon analyses will be performed.
Draft reports of field tests are being revised in light of recent efforts
by Southern Research Institute to more accurately predict particle behavior in
the cyclones and impactors used in particle size determinations. The results
presented in this paper may therefore change slightly in the near future.
SAMPLING AND ANALYSIS PROCEDURES
GENERAL
Prior to conducting field tests, several directives were issued to
compile information on available particulate data, industry trends, control
devices and to assess the availability of sampling locations. One directive
141
-------
was issued to develop emission factors based on all particulate data stored in
the EPA "Fine Particle Emission Inventory System" (FPEIS). Approximately 750
size distribution measurements of 15 different source categories were compiled
where 350 of the measurements were performed on coal-fired boiler emissions,
100 on electric arc furnaces, 98 on lime kilns and 65 at gray iron foundries.
An example of 1 of the 17 size distribution plots compiled from FPEIS data is
provided in Figure 1. This distribution is based on 235 particle size
measurements.
Three directives were then issued to develop Test Plans for the iron and
steel, ferroalloy and iron foundry industries. Sources of emissions were
ranked according to their potential (i.e., uncontrolled) yearly emission
rates. Based on the iron and steel Test Plans, GCA developed and conducted
field sampling programs for the top three ranked source categories, sources
which in total could emit approximtely 85,000 tons per year of particulate.
In the Ferroalloy production industries, tests were conducted on three
electric arc furnaces in cooperation with a Research Triangle Institute (RTI)
program to characterize organic material emissions. The open furnaces tested
were producing silicon metal, ferrosilicon and ferromanganese. All three were
controlled by positive pressure baghouses. The major sources of emissions in
the iron foundry industry were found to be the melting operations, which were
also found to be well characterized in the literature. Tests were planned and
conducted on a foundry pouring and cooling operation, a source for which no
reliable particle size data was available.
IP PROCEDURES
Size specific emission factors are based on a combination of simultaneous
total particulate and particle size distribution measurements. Total
particulate emission rates were measured using a conventional EPA Method 5
sampling train. Particle size distribution was measured using Andersen
cascade impactors equiped with cyclone precollectors. In addition, a
dual-cyclone train was operated in order to obtain bulk size-fractionated
samples for future elemental analysis.
Prior to field sampling, a presurvey of the host site was conducted in
order to gather the necessary data and to arrange for installation of sampling
facilities. Six inch (ID) sampling ports were needed for the dual cyclones.
A detailed Test Plan for the individual site was prepared and submitted to EPA
for approval. After approval, field sampling was usually completed in 1
week. Four tests, each typically consisting of one total particulate and two
impactor runs, were usually conducted simultaneously before and after a
control device. Inlet impactor runs were much shorter than the outlet runs
and sometimes up to four inlet runs were completed during one outlet impactor
run. Several tests were conducted in cooperation with other agencies. The
EPA Office of Air Quality, Planning and Standards (OAQPS) funded the dual
cyclone tests. EPA Region III and V provided some funds through existing
contracts at GCA for iron and steel tests and as mentioned earlier, Research
Triangle Institute performed the total particulate measurements at the
142
-------
OVERALL EMISSION RATES IN
Ibs. PARTICULATE
(% ASH) ton COAL .
10
99.950
99.90
99.80
99.50
99.
98.
8 95-
W 90.
a
20.
<
^ 10.
1 5.
2.
1.
0.5
0.2
0.15
O.I
n n
ESP CONTROLLED*. 080A Ibs. PARTICULATE100 MBtu/hr with 90 percent confidence
intervals.
143
-------
Ferroalloy plants. The coke quench tower tests were funded through several
sources including, DOFASCO, the host source and Hunters, the tower baffle
system manufacturer. Typically, the added funding was used to add more depth
to the programs. Additional points were sampled, more visible emissions data
was obtained, and more in-depth process evaluations were conducted as a result.
Samples were obtained for ducted sources by isokinetic sampling of four
or more sample points within the stack or duct. Collected particulates were
weighed in order to determine concentration and size distribution. Some
particulate samples will be analyzed for elemental constituents. Sampling and
analytical data was reduced by computer. Verson 4.0 of the GCA Sampling Data
Reduction System was used to calculate all total particulate results.
Particle size distribution results were initially calculated using the Cascade
Impactor Data Reduction System (CIDRS) developed by Southern Research
Institute. A newer, interactive version of CIDRS was recently developed by
Reseach Triangle Institute. The Particulate Data Reduction and Entry System
(PADRE) is on the EPA Univac computer and is accessible to all subscribers
through a telecommunications network. A program user enters data into PADRE
in response to specific prompts whereas CIDRS is a batch type program where
cards are punched and the program is run all at once. PADRE checks each entry
and allows greater quality control of results. Cyclone precutter and dual
cyclone results are calculated using equations supplied by the designers,
Southern Research Institute, however, revisions to the cyclone calculations
may be forthcoming since additional calibration work is underway. IP protocol
called for operating the cyclone precollector/cascade impactor combination at
a flow rate that would provide a 15 micron cut size in the cyclone
precollector. In many cases, this flow rate was felt to be too low to obtain
good impaction of particulate onto the impactor glass fiber substrates since
experience dictated that a higher flow was in order. The decision was made to
operate the train in favor of the impactor and to attempt to calculate the new
size cut point in the cyclone precollector. Southern Research Institute is
investigating the behavior of the cyclone precutters at these different flow
rates.
SOURCE SIGNATURE PROCEDURES
A dual cyclone train was designed by Southern Research Institute for
particle size determinations at ducted sources. The two cyclones in series
have cut points of 15 and 2.5 microns when operated at the proper flow rates.
The addition of a backup filter provides particulate samples of three size
classifications, greater than 15, between 2.5 and 15 and smaller than
2.5 microns. The dual cyclone train was operated during most of the IP test
programs with mixed results. The size fractions derived from the dual cyclone
results usually agreed within a factor of 20 percent with the impactor
results. However, the total particulate results were often a factor of two
times different hence the size specific emission factors also varied greatly.
Reasons for the variations are numerous and include single point (cyclones
always operated at the center of the stack) versus multiple point sampling.
144
-------
The elemental analysis results from the dual cyclone runs have yet to be
finalized. Neutron Activation and X-Ray Fluorescence techniques are being
employed to determine concentrations of elements such as lead, cadmium,
calcium, magnesium, zinc, etc. When these results are available, the
elemental composition of respirable and nonrespirable particles will be
deterrainable. Presently, only preliminary results for tests conducted at a
Sinter Plant by Midwest Research Institute are completed.
SUMMARY OF RESULTS
GENERAL
Field sampling programs have been completed by GCA at one coke quench
tower, two blast furnace casthouses, one Q-BOP furnace, three ferroalloy
plants, and one iron foundry pouring and cooling line. A brief summary of the
results obtained follows.
COKE QUENCH TOWER
Dominion Foundries and Steel, Limited (DOFASOO) volunteered to host the
IP Coke Quench Tower Test Program. Coke quenching is where hot coke, which
was recently pushed out of a coke oven, is sprayed with water to stop the
coking process. Coke arrives at the quench tower at 2500°F and leaves at
200°F. Some of the water used to quench the coke is recycled and 10 to 20
percent is lost to the atmosphere during quenching. IP tests were conducted
in two phases, during clean and during dirty quench water usage. Clean water
was defined as containing less than 1500 mg/1 total dissolved solids (IDS);
dirty water as containing more than 5000 mg/1 TDS. Tests were conducted
simultaneously above and below a set of Hunters baffles installed in the
tower. Nine to 12 quenches were sampled per run and 3 to 5 runs were
completed during each water quality phase. The resulting IP emission factors
are based on the amount of coal charged to the coke oven. The results,
provided in Table 1, show the expected trends where higher emission rates
resulted from dirty quench water useage. The particle size results are
questionable in that water droplets may have caused a bias toward larger
particles. The particle size results indicate a higher fraction of large
particles (droplets) in the outlet stream than was present at the inlet to the
baffles during dirty quench water usage. This study involved the efforts of
Southern Research Institute in the measurement of condensable particle size
distributions and water analyses were performed by several other recognized
laboratories.
BLAST FURNACE CATHOUSES
Two different types of blast furnace casting emission control systems
were tested using IP procedures. Dominion Foundries and Steel, Limited
(DOFASOO), Hamilton, Ontario, volunteered to host tests at the No. 3 blast
furnace. The roof monitor opening of this conventional, older casthouse was
145
-------
g
3.
0
^^ ^^
0
m
a
ftf o> a
W w m
55
rH O
m
O 0
S3 -H a
a 3.
8 CO
CO rS rH
<• -5
pcj O
O r4
Q «)
CQ g
A -4
CO
H-l r>
Pn -H
M 4J
o *d
H rH
PM 3
co B
W 0
M
CO
09
. d>
rH O J2 -O
Si O 0)
H C rH
rJ rH 0) O.
pa (838
< w cr 3
H0»
H VM
O
0)
o.
^^
4J
4J
CO
V
H
O O CM O
-* O rH O
rH rH O O
rH O VO O
r-l
rH CM O O
CT» O f~- O
rH
rH CO 00 O
O O C7\ O
•^ >* rH i-H
t-^ co in o
CO rH
rH CO
O rH O O
O O
rH rH
o e o c
m o in o
T3 4J T3 4J
V *•"• V ^^
0 0
o) in co m
CO *^ CO 'O
I B
43 .£
»< rH S-« r-l
vO VO
co co
M
0)
4J
C0 4J
tf *J 0)
0) rH
ti f"-* -U
co C 3
fl) M O
CJ
oo r>. m rH
CO CO oo CO
rH
CO O *^ co
• • • •
ON rH O O
i— 1 CM
•^" r- 1 00 CO
• • • •
rH rH -^ O
CM CM
oo CM co m
CM rH CM O
CM CO
-^ -31 oo r*^
0 rH 0> CO
CM -*
CM •*
• •
o m o rH
o o
rH rH
0 C 0 C
in o m o
T3 4J T3 4J
V ^^* V ^^
CO O
co m co m
CO TJ CO TJ
§v ca v
a
A jQ
S~S rH S< rH
ON r^
CO CM
M
0)
10 4J
S *J OJ
OJ rH
>-, i-t 4J
4J C 3
r4 M O
,-J
O
146
-------
sealed and connected via a 10 foot diameter duct to a positive pressure
baghouse. Bethlehem Steel Cororation, Sparrows Point, Maryland, allowed tests
of the newer L blast furnace casthouse runner evacuation system. L furnace is
one of the largest in the United States and produces 10,000 tons per day.
Four casting runner systems are available, each controlled by close fitting
hoods connected to a positive pressure baghouse. At both test sites, the
slotted roof monitor baghouse outlets prevented conducting meaningful outlet
tests.
DOFASCO was tested in November 1981 and was the first IP test conducted
by GCA. During the 1 week test program, 4 IP tests were completed during
casting each consisting of 1 total particulate and 3 impactor runs. Two casts
were sampled per total particulate run. The results are lower than the
results of tests conducted in the same manner at Bethlehem Steel Corporation.
The runner evacuation system emission factors are three times higher than the
building evacuation system emission factors as shown in Table 2. Extensive
visible emission observations were conducted during both test series. No
significant differences in the apparent capture of emissions by the control
systems was noted. The runner evacuation system test results also show a
larger percentage of large particles. The proximity of the evacuating draft
to the hot metal and slag runners is a logical reason for the higher emission
factors. Note that the results presented here represent the "evacuated"
emission factors and without the evacuating action of the control systems,
emissions from a typical uncontrolled casthouse are probably less than the
results of the tests at DOFASCO.
Q-BOP FURNACE
U.S. Steel Corporation volunteered to host IP tests at their Q-BOP shop
located in Fairfield, Alabama. A Q-BOP furnace is similar to a conventional
top-blown basic oxygen furnace (EOF) however oxygen is blown in from the
bottom of the vessel. Furnace C, rated at 200 ton capacity, was tested at the
outlet side of the quencher-scrubber control system. No sampling locations
were available at the inlet to the control devices. This furnace is a closed
furnace in that the hood is directly attached to the furnace shell. The
combustion of off gasses is suppressed by limiting the amount of ambient air
infiltration to only 10 percent of that required for complete combustion.
Carbon monoxide levels of 20 percent (v/v) are typical during the oxygen blow
period. A sliding doghouse enclosure is also connected to the control system
whereby any emissions escaping the close fit hood are captured.
The results of the IP tests, shown in Figure 2 show effective
particulate control. The quencher/scrubber combination is estimated to be
99.6 percent efficient based on an assumed inlet loading of 15 Ibs particulate
per ton of steel produced. The particle size results show the expected
penetration of particles smaller than 10 microns. Below this size, the
particulate does not have the inertia to penetrate droplets of water in the
control device.
The results presented were calculated using two different assumptions
regarding the validity of the backup filter weight gain. Due to the potential
for particle bounce within the impactor, the backup filter weight gain is
147
-------
CO
a
o
hH
rA
VJ
CO
M
O
z
M
H
^
0
TV!
LX-)
0
>
T3
O
m S
S ^
in
fii i
ly r— i
l~~*
U
•r-(
4J
(-1
to
fc tfl
8
to
4J
• H
C
D
a)
u
)-l
3
O
w
^O CM
O r-l r-4
• • •
in r-» o CTN in o
r—i i— i ^
0 0
r-H CM CM
• • •
in CT\ o in \o o
CM CM r_| f^
••d* ^D
r-l r-l CM
m o o o o o
CO t
oo CM in co
• • • •
r^ \o o ^o CM o
vo r~> CM co
r^
ON O
r-i co co
• • •
O -3" O O O r-l
0 r-4 O O
r-l r-l r-l in
4J 4J
CO CD
0 1 to 0 1 CO
m I u m I u
"O O 13 C5
v J mcv j ma
CJ T3O <3 T3O
win wPcom v tj
03 "Or-l T-l tO Tj -1-1
cfl V -C -Q cfl , , t-l J3
0 r^C!SVXir-.q!
XS O J3 O
S~S r-l 4J ^? r-l JJ
4J
CO
CO tO
r-t
• PQ
O
"Z 0
0)
80) A 0)
O 0) O
c/j tfl r-i to
-------
TOTAL PARTICULATE Ibs PARTlCULATE
EMISSION RATE =0-028 fon STEEL PROOUCED
99.990
99.950
99.90
99.80
99.50
99.
98.
UJ
IM
in
o
u
Z
UJ
u
cc
LU
a
Crt
«3
2
LU
>
>-
<
_J
r>
2
D
u
95.
90.
80.
70.
60.
50.
40.
30.
20.
10.
5.
2.
I.
0.5
0.2
0.15
O.I
0.0
A - AVERAGE OF 8 IMPACTOR RUNS
50% ASSUMPTION
• = AVER AGE OF 8 IMPACTOR RUNS
100% ASSUMPTION
MAXIMUM
MINIMUM
i 11
i i i i 11
0.027
0.024
0.022
0.020
0.016
0.014
0.012
0.008
0.006
Id
N
(ft
O
UJ
UJ
»-
<
or
o.
(0
UJ
0.003 >
o
UJ
u
D
O
o
Ul
(O
c
o
O
0
10'
10'
PARTICLE AERODYNAMIC DIAMETER,micrometers
RHO = i.OO
Figure 2. Average size distribution of emissions from a Q-BOP
Furnace control system, U.S. Steel, Fairfield Works.
149
-------
sometimes biased high resulting in erroneously high percentages of fine
particles. The results were first calculated assuming that 100 percent of the
weight gain was valid and then using the assumption that only half of the
weight gain was actually attributable to particles smaller than the cut size
of the previous irapaction stage. Usually the two assumptions only affect the
results in the 1 to 3 micron size ranges however the Q-BOP results were
significantly affected over the 0.6 to 10 micron range. Particle bounce may
have occurred since the impactors were operated at flow rates of 0.5 to
0.6 acfm and 0.7 acfm is considered the upper limit of good impactor
performance.
FERROALLOY INDUSTRY
Emissons from three open electric arc furnaces producing ferroalloys were
sampled using IP procedures. RTI performed the total particulate measurements
through a subcontract to Entropy Environmentalists as part of a Level 2
Environmental Assessment of organic emissions. GCA/Technology Division
personnel conducted particle size tests during the tests by Entropy.
Isokinetic sampling was conducted simultaneously at the inlets and outlets of
the baghouses controlling emissions from each furnace. The resulting
size-specific emission factors are presented in Table 3 for each test series.
The emission factors are based on the ferroalloy production rate during the
test period.
Baghouse outlet tests were conducted at poor sampling locations.
Positive pressure baghouses with roof monitor-type outlets are used at most
ferroalloy facilities. These tests were all conducted in each of the
individual baghouse compartments, about 6 feet above the top of the bags, at a
single sample point. The flow was below detectable limits, therefore,
isokinetic sampling was impossible. Since all three facilities were tested in
the same manner, relative differences are notable; however, the overall
accuracy is questionable. Baghouse inlet tests were all conducted
isokinetically in horizontal circular ducts. Several inlet impactor runs were
completed during each outlet run. The results show a wide variation in
emission rates depending on the ferroalloy being produced and the results of
the silicon metal tests, 800 Ib/ton, were the expected results by source
personnel.
IRON FOUNDRY POURING AND COOLING
The particle size distributions of emissions from foundry melting
operations has been well documented in the literature. Sand handling, mulling
and pouring and cooling are the next largest sources of emissions at a typical
foundry. IP tests were conducted on a relatively new (1973) pouring and
cooling line at Lynchburg Foundry, Lynchburg, Virginia. Ductile iron
automotive parts were being cast during the test program.
Three stacks were tested simultaneously. Three stacks vent pouring
emissions and three vent cooling emissions directly to the atmosphere. The
total particulate results presented in Table 4 are based on one set of three
simultaneous modified Method 5 runs on each operation. A third IP test,
consisting of two runs on cooling stacks and one run on a pouring stack, was
150
-------
CO
H
3
1
PM
0
T
^™j
1
M
fa
H
^J
en
en
W
H
W
O
>2
55
^j
Q
js
CO
**
2j
|
fa
3
s
CO
co
M
S
^_j
r
l__j
o
w
PM
01
W
i ]
Uj
*
oo
s
a
/•N 1-1
g
a.
c
o m
m .
T3 CM
CO
4-1
co m
cd
•H
™
o
•1-1
S °
cd i-i
c
•u
o
cd m
i— i
co
r^i
CJ
4J
rJ
cd cd
* 8
CO
4J
'c
P
4-1
3
•d
O
PM
cf§
•H -i-l
r-l 4J
a. cd
gj cj
cd O
en i-»
00
in
CM
00
•
CM
CO
m
•
m
CO
CO
oo
CO
^
,
O^
ro
r*^
•
S
4J
1
a
£
0
4->
j
c
o
U
•r-l
r-l
• H
CO
O
^|
^4
0)
ptl
4J
0)
t --^
c
•r-l
0)
CO
3
O
,C^
&0
cd
«
CTi
a^
CO
c
r^.
i— i
CO
•
S
^
CM
CM
VO
•
m
CM
^~
•
00
CM
U
0
i-i
a
0
o
4J
Xt
j
CO
CO
co
angan
e
o
rJ
M
CO
r^
CTi
"*
o
1-1
m
^O
o
o
m
CM
00
O
i
ex
d
o
4-1
x>
r-l
cd
4J
Ci
o
o
r-t
••-1
en
CO
in
d
-------
TABLE 4. SUMMARY OF EMISSION FACTORS, LYNCHBURG FOUNDRY
Production Rate
Total Particulate Emission
Mass Concentration3
Mass Emission Rate"
Process Weight Rateb»c
Units
tons/hr
Factors
gr/dscf
Ib/hr
Ib/ton
Pouring
emissions
105
0.003
2.21
0.021
Cooling
emissions
114
0.002
2.52
0.022
Particle Size Distribution
Cumulative Mass3 % <15 Mm d50 65 63
% <2.5 urn d50 22 20
Mass Emission Rateb Ib <15 ym d50/hr 1.44 1.59
Ib <2.5 pm d50/hr 0.49 0.50
Size Specific Emission Factors0
Ib particulate <15 ym d5Q/ton 0.014 0.014
Ib particulate <2.5 ym d50/ton 0.005 0.005
aAverage of four pouring and five cooling runs.
Sum of three simultaneous runs.
cBased on the total weight of sand, cores and hot metal processed.
152
-------
completed on the last field sampling day. The mix was chosen in the field to
provide a more balanced total number of runs on both operations. When
calculating emission factors, however, the total particulate emission rate
from each of the three stacks venting each process is needed. The results
show large differences in emission rates from each stack and the potential for
variability in emission generation rates. The average of all five cooling and
four pouring impactor runs were used to determine the particle size
fractions. These results are multiplied by the average of three simultaneous
Method 5 runs to calculate emission factors.
The total particulate results appear to correlate well with the amount of
activity assocated with the area evacuated. Pouring operations were mainly
performed under the hood connected to the stack where the results are six
times greater than the results of the other tests on the stack which is over
the area usually vacant for safety purposes among other reasons. Sometimes a
second operator pours hot metal under the hoods connected to the stack for
which results are in between the results of tests. The cooling room emission
rates also appear to correlate with the location being vented. The highest
emission rates were measured where the hot flasks first enter the room and the
lowest rates were measured at the cooling room exit area. All of the results
were very low and the maximum particulate catch was 67 mg after sampling 220
dscf.
CONCLUSIONS
The emission factors developed thus far are presently being reviewed in
light of recent research on fine particle behavior in cyclones and impactors.
Calibrations are being conducted by Southern Research Institute and
modifications to the PADRE data reduction system are also underway. Most of
the IP data will be reanalyzed shortly and the emission factors presented here
may change slightly. Also in the near future, interesting conclusions
regarding the elemental nature of the respirable particulates will be
available.
153
-------
RESULTS OF TESTING FOR INHALABLE PARTICULATE MATTER
AT MIDWEST RESEARCH INSTITUTE
by: H. Kendall Wilcox
Fred J. Bergman
John Scott Kinsey
Tom Cuscino
Midwest Research Institute
Kansas City, Missouri 64110
ABSTRACT
Source test data collected by Midwest Research Institute for emissions
of inhalable particulate matter have been presented in this paper for a va-
riety of industrial categories. Test results for two cement plants (one wet
process and one dry process), one lime plant, and one asphalt paving plant
are available. Results have been presented in terms of the AP-42 format for
the relative size fractions of both controlled and uncontrolled emissions
from these processes. Such test results should be of interest to control
device manufacturers as well as those who may need to be involved in the de-
velopment of State Implementation Plans for inhalable particulate matter.
This paper has been reviewed in accordance with the U.S.
Environmental Protection Agency's peer and administra-
tive review policies and approved for presentation and
publication.
154
-------
RESULTS OF TESTING FOR INHALABLE PARTICULATE MATTER
AT MIDWEST RESEARCH INSTITUTE
INTRODUCTION
The U.S. Environmental Protection Agency (EPA) is conducting research
to characterize the emissions of the fine particles in the inhalable partic-
ulate (IP) size range for a variety of industrial sources. The purpose of
this research is to develop emission factors which are to be used if revi-
sions to the National Ambient Air Quality Standard for particulate matter
are made to address fine particles. It was originally planned to determine
size specific mass concentration on uncontrolled sources and to calculate
emissions based on established control efficiencies. However, since most
control efficiency data were based on total suspended particulate (TSP), the
scope of the program was modified to simultaneously determine uncontrolled
and controlled size specific mass concentrations.
Testing at Midwest Research Institute has included one lime plant, one
wet process cement plant, one dry process cement plant, and one drum mix as-
phalt plant. This paper presents a brief overview of the testing procedures
used to collect the data from ducted sources and the results of the measure-
ments made in terms of the emission factors for each process. More detailed
information regarding the plant processes, control devices and plant config-
urations has been reported previously. (1)(2)
GENERAL PROCEDURES
SAMPLING PROCEDURES
The basic sampling methodology used by MRI to determine the controlled
and uncontrolled emissions from each plant is that specified in the protocol
document developed for the EPA's IP program. (3) Certain modifications
were made to the standard protocol, however, when it was deemed necessary to
collect a more representative sample of the emissions.
Basically the protocol requires collecting multiple samples in each of
four quadrants of the duct as shown in Figure 1. Where suitable ductwork
meeting Method 5 requirements was available A and B patterns were used. If
suitable ductwork was not available, pattern C was used.
155
-------
Ducts with Straight Runs
B
End
Circular Duct
Square or Rectangular Duct
Ducts without Straight Runs
• - Sampling Point
Figure 1. Sampling point locations
156
-------
Two types of samples were collected at each sample point: total mass
emissions; and particle size distribution. Although EPA Reference Method 2
was used to collect preliminary velocity data, the stack was not traversed
during sampling. The total mass samples were collected isokinetically at
each point using EPA Method 5 or Method 17 techniques. The particle size
samples were collected at a constant flow rate according to manufacturer
specifications using a suitable nozzle diameter to obtain as near isokinet-
ics sampling as possible. Several types of samplers were used. For heavy
loading conditions typically found at the inlet of a control device, an
Andersen High Capacity Stack Sampler (HCSS) equipped with a Sierra 15 (Jm
preseparator was used to collect the sample for all plants except the lime
plant. For the lime plant a Brink 5 stage impactor was used with a 7 pm
cyclone as the first stage. For the Light loading conditions typically oc-
curring for controlled emissions, an Andersen Mark III impactor equipped
with the Sierra 15 \im preseparator was used. Although the sampling times
varied substantially for the inlet and outlet, an effort was made to collect
data from both the controlled and uncontrolled locations during the same
time period.
Initially, four mass runs and four size runs -were conducted from each
quadrant for a total of 16 mass and 16 particle size runs for both the con-
trolled and uncontrolled emissions. The number of samples was later reduced
for the controlled emissions to two for each quadrant rather than four.
CALCULATION OF EMISSION FACTORS
The emission factors presented in this report were calculated as fol-
lows. Total emission factors were calculated from the results of the total
mass runs (Modified Method 5 or Method 17) and the average production rate
for the process during the test period.
Emission factors for the particle size measurement were calculated by
determining the mass for each stage of the size device and calculating the
cumulative percentages of the total mass for each. These percentages were
then applied to the total mass emission factors from the modified Method 5
or Method 17 runs to obtain the emission factor for each stage.
A spline equation was used to fit the data and to extrapolate, where
required, to the desired cutpoints. (4) Emission factors were calculated
for 2.5, 10.0, and 15.0 |Jm. The particle diameter upper limit was assumed
to be 50.0 |Jm for the calculations using the spline equation.
TEST RESULTS
The test results from these studies are provided in terms of the emis-
sion factors for the total particulate emissions based on the total mass de-
terminations from the modified Method 5 and Method 17 samples. Emission
factors for < 2.5 |Jm, < 1.00 (Jm, and < 15 [im were calculated based on the
results particle size distribution obtained from the impactors and the total
mass results.
157
-------
ASPHALT PLANT
The test results from the drum-mix asphalt plant are shown in Table 1.
This plant was nearly new at the time of testing and was in excellent mechan-
ical condition. Sampling was conducted at the inlet and outlet of a bag-
house collector. Test Nos. 1 through 4 were conducted with 100% virgin ag-
gregate material. Tests A and B were conducted with approximately 34% of
the aggregate comprised of recycled asphalt material. Due to process vari-
ations and the limited number of runs, it is difficult to draw any meaning-
ful conclusions about the differences between virgin and recycled aggregate
from the data. Variations included changes in production rate and in type
of mix even within a one day period. However, even though the data was col-
lected over a period of several weeks (due to unfavorable weather conditions)
it is felt that the emission factors obtained are at least generally repre-
sentative of most drum-mix asphalt plants.
DRY PROCESS CEMENT PLANT
The emission factors for the dry process cement plant are shown in
Table 2. The kiln was operating at approximately 35 tons per hour and was
equipped with a baghouse. Material exiting the kiln passes through a 3-
stage suspension preheater system to remove the bulk of the larger material
prior to entering the baghouse. The baghouse inlet was sampled between the
preheater and the baghouse inlet. The outlet was sampled on the main exhaust
stack.
LIME PLANT
The results of testing of the lime plant are shown in Table 3. Two
kilns were operating in this plant. One kiln of 400 per day ton capacity
was equipped with three electrostatic precipitators (ESPs) arranged in par-
allel, the other of 300 ton per day capacity was equipped with one 5-cell
baghouse. Both controlled and uncontrolled emissions were measured for
each kiln. The dust control system on the conveyor belt transfer points
was also sampled.
WET PROCESS CEMENT PLANT
The emission factors for the wet process cement plant are shown in
Table 4. This plant has two rotary kilns of 35 ton/hr which are each
equipped with ESP units which exhaust through a common stack. The inlet to
one of the two ESP units was sampled as well as the outlet prior to entry
into the common stack.
During the latter part of the test the plant switched from Type II to
Type I cement. Differences in ESP performance for the two types of cement
are difficult to determine based on thelimited data available for Type I.
158
-------
TABLE 1. SUMMARY OF INHALABLE PARTICULATE EMISSION FACTORS
FOR THE ASPHALT PLANT
Test
Baghouse ns
Controlled side
Baghouse
Uncontrolled side
Q
, Pounds of particulate
Aerodynamic diameter.
No.
Ac
BC
1
2
3
4
Average
1
2
Average
matter
Emission factor
Total
25.2
16.3
37.9
37.6
30.2
27.9
30.9
0.06
0.07
0.07
per short
b
< 2.5 (Jra <
-
1.6
1.5
1.4
2.1
1.7
0.01
0.01
0.01
ton of asphalt
(lb/ton)a
b
10 |jm
.
-
7.6
7.6
6.4
7.0
7.2
0.02
0.02
0.02
paving
b
< 15 pm
.
-
8.6
8.9
7.3
7.9
8.2
0.02
0.03
0.03
produced.
T— _ J~ _
conducted with plant using recycled asphalt paving as a raw material.
Average emission factor is the arithmetic mean of the results from the
eight total mass test runs and not an average of the data shown.
TABLE 2. SUMMARY OF EMISSION FACTORS FROM A DRY PROCESS CEMENT PLANT
Emission factors
Sampling location
Baghouse
uncontrolled
side
Baghouse
controlled
side
Test Total
No. (lb/ton)a
1
2
3
4
Average
1
2
Average
220
220
210
210
220
0.62
0.89
0.76
< 2.5 |Jmb
(Ib/ton)
42
31
42
39
38
0.20
0.32
0.26
< 10.0 [jmb
(Ib/ton)
87
83
90
94
88
0.44
0.74
0.59
< 15.0 pmb
(Ib/ton)
93
90
96
99
94
0.45
0.76
0.60
(Ib/ton) = pounds per ton of product.
Aerodynamic diameter.
159
-------
TABLE 3. SUMMARY OF EMISSION FACTORS FROM A LIME PLANT
TtfiC t"
icS L
Emission factors (Ib/ton)
No. Total <
Baghouse uncontrolled side
Controlled side
Electrostatic precipitator
uncontrolled side
Controlled side
Dust collector
0
, (Ib/ton) = pounds per ton
1
2
3
4
Average
1
2
3
4
Average
1
2
3
4
Average
1
2
3
4
Average
1
2
3
4
Average
133.8
120.3
118.9
108.4
120.4
0.13
0.09
0.10
0.10
0.11
354.4
378.4
395.5
359.5
371.9
8.1
10.5
9.1
9.5
9.3
1.8
1.7
3.0
2.3
2.2
b
h
2.5 (Jm < 10 [Jin"
14.2
12.5
15.6
8.9
12.8
0.03
0.02
0.03
0.03
0.03
3.6
4.0
2.6
2.8
3.3
1.1
1.1
1.6
1.2
1.3
0.06
0.07
0.10
0.08
0.8
52.0
50.5
53.6
45.0
50.3
0.08
0.05
0.07
0.07
0.06
50.0
69.1
59.7
53.9
58.2
4.3
4.7
4.9
4.6
4.6
0.15
0.16
0.27
0.19
0.19
h
< 15 p"1
74.8
70.8
72.7
64.3
70.7
0.09
0.06
0.08
0.09
0.08
134.0
212.0
154.1
138.0
159.5
5.7
5.8
6.2
5.3
5.8
0.17
0.16
0.32
0.21
0.22
of product.
Aerodynamic diameter.
160
-------
TABLE 4. SUMMARY OF EMISSION FACTORS—COMBINED AND INDIVIDUAL
WET PROCESS CEMENT TEST RESULTS
Emission factors - cumulative %
Test Total > 2.5 [Jmb > 10.0 nmb > 15.0 |Jmb
Kiln sampling location No. (Ib/ton) (Ib/ton) (Ib/ton) (Ib/ton)
Electrostatic
Combined
precipitator
uncontrolled side
Average
Electrostatic
controlled
Average
precipitator
side
1
2
3
4
1
2
Individual
cement
1
1
1
1
1
0
0
0
cement
product
,100
,100
,400
,400
,200
.10
.17
.14
product
test results
25
22
15
18
20
0.079
0.082
0.08
test results
90
80
86
84
85
0.087
0.11
0.10
c
150
130
180
150
150
0.099
0.12
0.11
Type I cement product
Electrostatic precipitator 1,400 24 94 160
uncontrolled side
Electrostatic precipitator 0.15 0.08 0.10 0.12
controlled side
Type II cement product
Electrostatic precipitator
uncontrolled side
Electrostatic precipitator
controlled side
1,100
0.13
14
0.079
70
0.10
140
0.10
, Ib/ton = pounds per ton of product.
Aerodynamic diameter.
Aerodynamic diameter.
Data represent the average of all the sampling runs conducted during
production of the specific type of cement product.
161
-------
SUMMARY OF DATA
Two summaries of the data are shown in Tables 5 and 6. Table 5 shows a
comparison of the particle size distribution for the uncontrolled and con-
trolled emissions for each of the processes tested. It is generally expec-
ted that controlled emissions will contain higher percentages of fine par-
ticulate than the uncontrolled emissions. The data does show increased per-
centages of fine (< 2.5 pm) particulate in the controlled emissions by a fac-
tor of two or three for baghouses and from 15 to 35 for ESP's.
Table 6 presents a summary of the control device efficiencies for each
device tested. As can be seen for the lime plant the baghouse represents a
considerably better control device for this process.
162
-------
TABLE 5. SIZE DISTRIBUTION, UNCONTROLLED VS. CONTROLLED
Process
Total E.F.
Ib/ton
Size Distribution
wt. % less than stated size
< 2.5
< 10 |Jm < 15 (Jtn
Lime Plant
Baghouse
Uncontrolled
Controlled
120
0.11
11
27
42
54
59
73
ESP
Uncontrolled 372
Controlled 9.3
14
16
49
43
62
Cement, Wet Process
ESP
Uncontrolled
Controlled
1200
0.14
1.6
57
7.1
71
13
79
Cement, Dry Process
Baghouse
Uncontrolled
Controlled
220
0.76
17
34
40
78
43
79
Asphalt, Drum Mix
Baghouse
Uncontrolled
Controlled
30.9
0.07
5.5
14
23
29
27
43
163
-------
TABLE 6. CONTROL DEVICE EFFICIENCY
Lime Plant
Baghouse
Total
< 15 Mm
< 10 Mm
< 2.5 Mm
ESP
Total
< 15 Mm
< 10 Mm
< 2.5 Mm
Wet Process Cement Plant
ESP
Total
< 15 Mm
< 10 Mm
< 2.5 Mm
Dry Process Cement Plant
Baghouse
Total
< 15 Mm
< 10 Mm
< 2.5 Mm
Asphalt Drum Mix Plant
Baghouse
Total
< 15 Mm
< 10 Mm
< 2.5 Mm
Emission
Uncontrolled
120
70.7
50.3
12.8
372
159
58.2
3.3
1200
150
85
25
220
94
88
38
30.9
8.2
7.2
1.7
Factor,
Ib/ton
Controlled % Eff .
0.11
0.08
0.06
0.03
9.3
5.8
4.6
1.3
0.14
0.11
0.10
0.08
0.62
0.45
0.44
0.20
0.07
0.03
0.02
0.01
99.91
99.89
99.88
99.77
97.50
96.35
92.10
60.61
99.988 +
99.93
99.88
99.68
99.7
99.5
99.5
99.5
99.77
99.63
99.72
99.41
164
-------
REFERENCES
1. Bergman, Fred J. and Wilcox, H. Kendall. Inhalable Particulate Testing
at Cement and Lime Plants. Paper presented at the 75th Annual Meeting
of the Air Pollution Control Association, New Orleans, LA. June 20-25,
1982.
2. Kinsey, John Scott, Walker, T. and Wilcox, H. K. A Determination of
Fine Particulate Emissions from a Drum-Mix Asphalt Plant. Paper pre-
sented at the 75th Annual Meeting of the Air Pollution Control Associ-
ation, New Orleans, LA. June 20-25, 1982.
3. Wilson, R. R. and Smith, W. B. Procedures Manual for Inhalable Partic-
ulate Sampler Operation, Report No. SoRI-EAS-79-761, Southern Research
Institute, Birmingham, AL. November 30, 1979.
4. Johnson, J. W. , Clinard, G. I., Felix, L. G. and McCain, J. D. A Com-
puter-Based Cascade Impactor Data Reduction System, EPA-600/7-78-042,
U.S. Environmental Protection Agency, Washington, D.C. March 1978.
165
-------
INHALABLE PARTICULATE EMISSION FACTORS
TEST PROGRAMS
by: Jim Davison
Acurex Corporation
485 Clyde Avenue
Mountain View, California 94042
ABSTRACT
The Energy & Environmental Division of Acurex Corporation was
contracted by the Industrial Environmental Research Laboratory (IERL) of the
U.S. Environmental Protection Agency (EPA) to obtain uncontrolled/controlled
emissions data from various stationary sources of air pollution. The emission
factors derived from this data will assist in the determination of the need to
set a national ambient air quality standard for inhalable particulate matter.
An extensive series of particulate mass and particle size distribution
tests were conducted at several major sources, including Kaiser Steel (hot
metal desulfurization and BOF), and Kennecott Minerals (matte and slab
tapping).
This paper presents a review of each process, test equipment and
procedures, and test results expressed as emission factors relative to
process operations, and as a percent of the particulate emissions less than a
selected micron size.
This paper has been reviewed in accordance with
the U.S. Environmental Protection Agency's peer
and administrative review policies and approved
for presentation and publication.
166
-------
TEST EQUIPMENT
Uncontrolled and controlled participate mass emissions data was
obtained using the Acurex High Volume Stack Sampler (HVSS). Particle size
measurements were made using the SoRI two-cyclone train, the SoRI dilution
stack sampling system, and the Andersen Mark III cascade impactor.
First, a brief review of the sampling trains and test equipment used
during these programs.
All particulate mass measurements were made with the HVSS, which is an
EPA Method 5 sampler consisting of the following components:
• A 316 stainless-steel sampling nozzle (buttonhook) properly sized
for isokinetic sampling
• A 316 stainless-steel-lined sampling probe, 5-ft long, equipped
with a thermocouple to measure probe temperature, a thermocouple to
measure stack gas temperature, and an S-type pitot tube to measure
velocity pressure
• A 316 stainless-steel 3-ym cyclone for large particulate
collection (inlet test location only)
• A Teflon-coated stainless-steel filter holder containing a 142-mm
glass fiber filter
• A temperature-controlled oven to maintain the cyclone and filter
holder at 250°F
• A Teflon-lined, braided stainless-steel hose, 5-ft long to connect
the outlet of the filter holder to the inlet of the impinger train
• An impinger train containing four glass bottles to collect moisture
and condensible inorganic and organic material escaping the filter
and cyclone (bottles 1, 2 -- 250 ml distilled water, bottle 3 —
dry, bottle 4 -- silica gel)
• A 10-cfm carbon vane pump modified for very low leakage around the
shaft
• A control module, containing a dry gas meter and an orifice meter,
to monitor temperature, pressure, and flowrate throughout the
sampling train
Figure 1 illustrates the Acurex HVSS train.
The Andersen Mark III cascade impactor was equipped with a 15-ym
precutter. The sampler contains nine jet plates, each having a pattern of
precision drilled orifices. The resulting increased gas stream velocity
through the plates distribute the sample into eight fractions or particle size
167
-------
168
-------
i—I
JC
4->
•r—
S_
O
o
o.
t-
£
0)
CO
0)
169
-------
EXHAUST BLOWER
TO HEATERS, BLOWERS
TEMPERATURE SENSORS
TO ULTRAFINE
PARTICLE SIZING
SYSTEM (OPTIONAL!
.DILUTION AIR
'HEATER
CONDENSER
• DILUTION AIR
BLOWER
ICE BATH
MAIN CONTROL
FLOW, PRESSURE
MONITORS
Figure 3. Diagram of stack dilution sampling system.
170
-------
ranges. The device was mounted directly on the end of a 6-ft stainless-steel
probe connected by a stainless-steel hose to an impinger train followed by the
pump and control module (see Figure 2).
A diagram of the major components of the SoRI dilution stack sampling
system is shown in Figure 3. In normal operation, gases from the process
stream are drawn through the IP dual-cyclone sampler, in which particles with
aerodynamic diameter greater than 15 ym and those in the range 2.5 to 15 \itn
are removed in two stages. The stack gas containing the fine particle
fraction (<2.5 urn) and condensible vapors pass through the heat-traced probe
and flexible sample line and are introduced into the bottom of the cylindrical
dilution chamber. At this point the stack gas is mixed with dilution air to
form a simulated plume which flows upward through the dilution chamber,
through a standard high-volume filter which collects the fine particulate and
any new particulate formed by condensation. The diluted stream is exhausted
by a 1-hp blower or optionally by a standard high-volume blower. Stack gas
flowrate is measured by an orifice at the base of the dilution chamber.
Dilution and exhaust flow are measured by orifices in the inlet and outlet
lines, respectively.
KAISER STEEL PROGRAM
The first phase of the Kaiser Steel program included a series of
particulate mass and particle size distribution tests conducted on the
emissions from the Hot Metal Desulfurization (HMDS) Plant of Kaiser Steel in
Fontana, California. Measurements were made to quantify uncontrolled
emissions (inlet to baghouse) and controlled emissions (outlet of baghouse),
and to develop emission factors for the process.
Hot metal from the blast furnace arrives at the HMDS plant in torpedo
cars which are positioned into a partially open shed attached to the HMDS
building. Lances are inserted into as many as three torpedo cars at one time,
and a predetermined amount of calcium carbide and calcium carbonate is blown
into the hot metal using nitrogen. The hot metal is normally desulfurized by
this process to less than 0.03 percent sulfur. A stopper on the lance fits
the opening of the torpedo car to minimize emissions during the
desulfurization process. Emissions that escape are captured by a local hood
and ducted to a six-compartment, positive pressure Wheelabrator-Frye
baghouse.
Samples of the uncontrolled emissions from the HMDS process were
collected at the inlet to the baghouse as indicated in Figure 4. The sampling
ports installed in the rectangular duct included six ports (three on each
side) for particle size trains and three ports for particulate mass sampling.
Measurements were made at nine sampling points at each of these locations.
(Three points per port.)
The controlled emissions were measured at the baghouse outlet.
Sampling .ports were located on stacks 2 and 5; a 6-in. diameter port on each
stack was used for particle size determination, and two 4-in. diameter ports
were used for particulate mass sampling.
171
-------
Hot nietal
desulfurization
station hood
PIAH VIEW
6" ports Pacific exhauster fan
Reliance 200 HP
motor
a) Uncontrolled emission sampling port locations.
PLAN
VIEU
— 6" sampling ports
- Pacific exhauster fan
b) Controlled emission sampling port locations.
Figure 4. Hot metal desulfurization emissions control system.
172
-------
Measurements obtained with the EPA Method 5 particulate mass and SoRI
two-cyclone particle size trains, indicate the uncontrolled emission factor
ranged from 4.62 x 10"1 to 15.33 x 10'1 Ib (average 11.53 x 10-1) of
particulate emitted per ton of hot metal desulfurized. Thirty-two percent of
these emissions were less than 15 ym in size.
Based on measurements made with the EPA 5 and Andersen Mark III
impactor trains, the controlled emission factor ranged from 15.40 x 10-4 to
34.62 x 10-4 Ib (average 27.80 x 10~4) of particulate emitted per ton of hot
metal desulfurized. Eighty-one percent of these emissions were less than
15 ym in size.
Test results are presented in Table 1. The particulate mass removal of
the baghouse averaged 99.91 percent, and emission factors were higher for
heats requiring more desulfurization.
Phase II of the test program at Kaiser Steel involved a series of
particulate mass and particle size tests conducted on the emissions from the
basic oxygen plant (BOP). Measurements were made to quantify uncontrolled
(baghouse inlet) and controlled (baghouse outlet) secondary fugitive
emissions, and to develop emission factors for the process. The test program
included the hot metal charging and tapping portions of the BOP cycle.
Each vessel of the two-vessel plant is capable of producing 230 tons of
steel per heat. Oxygen steelmaking begins with the charging of approximately
100 tons of steel scrap. Nearly 200 tons of molten iron is charged onto the
scrap in the vessel. A door on the vessel enclosure is used to completely seal
the enclosure for capture of the fugitive emissions generated during the
charging operations. A lance is lowered into the vessel and blows oxygen into
the mixture increasing the temperature to 2,900°F. After about 40 min, molten
steel is tapped into a ladle. Fugitive emissions from tapping are also
captured in the vessel enclosure.
Emissions from the BOP steelmaking facilities are controlled by
separate primary and secondary air pollution control systems.
The primary emission control system consists of two closed-hood,
suppressed combustion systems with high-energy venturi scrubbers. Dust-laden
gases are captured in a hood above each vessel. The hood carries the gases to
a quencher and then to the scrubber. The cleaned gases, containing carbon
monoxide are burned at the top of a stack. The water used in the gas cleaning
process is then piped outside the shop to large clarifying tanks.
Particulates are removed from the water which is recycled back through the
scrubber system.
The steelmaking facility's secondary fugitive emission control system
is designed to handle all the emissions not captured by the primary system.
These include fumes generated when steel scrap and molten iron are added to
the open mouth of the steelmaking vessel and when steel is tapped into a
ladle. Fumes are drawn off from both sides of the enclosure through ductwork
which travels under the charge floor, up the outside vertical face of the shop
and across a roadway to the dirty gas fans and the baghouse. The secondary
173
-------
TABLE 1. KAISER STEEL HOT METAL DESULFURIZATION PLANT TEST RESULTS
Test location
Participate
DesulfuMzation mass
rate concentration
(tons/mln) (gr/dscf)
Partlculate
mass
emission
rate
(Ib/mln)
Emission
factor
(Ib/ton)
Percent
less
than
15 jim
Baghouse Inlet
uncontrolled
emissions3
26.89
2.3038
27.72
11.53 x 10-1 32.1
Baghouse outlet
controlled
emissions'5
65.98
13.77 x 10-3
0.17
27.80 x 10-4
'Average of tests 18 through 28
"Average of tests 15 through 17
174
-------
Furnace No. 5
Furnace
Furnace No. 6
Heat exchanger
Baghouse Inlet sampling site
(volumetric flowrate only)
Outlet sampling sites
Baghouse
Figure 5. Top view of BOP secondary fugitive particulate emission
control system ductwork and baghouse.
175
-------
fugitive emission control system, designed by PECOR, handles 630,000 acfm of
gases. This system has a separate hood and duct to capture tapping emissions.
Fugitive emissions from the hot metal transfer station and skimming station
(inside the BOP) are also captured and cleaned by this control system.
Figure 5 illustrates the location of the inlet and outlet sampling
ports in the BOP secondary fugitive particulate emission control system.
Samples of the uncontrolled secondary fugitive particulate emissions
from the hot metal charging and tapping portions of the BOP cycle were
collected in the two ducts of the secondary emission control system for each
vessel. The two ducts for each control system were designated east and west
for testing purposes and are illustrated in Figure 5. All four ducts were
manifolded together into a common duct leading to the baghouse inlet.
The particulate mass and particle size sampling ports were located on
the vertical side of each rectangular inlet duct. Each rectangular inlet duct
was 11-ft wide and 4-ft deep. The particulate mass and particle size sampling
ports for each duct consisted of four equidistant 6-in. ports while velocity
and temperature measurements were made in four 4-in. ports placed 30 in. above
the particulate mass and particle size ports.
Samples of the controlled secondary fugitive particulate emissions from
the hot metal charging and tapping portions of the BOP cycle were collected at
three of the 12 outlet stacks serving the baghouse. Budget and manpower
limitations prevented simultaneous testing of all 12 outlet stacks. Stacks 3
and 10 at the front and rear of the two rows of outlet stacks were selected as
representative sampling locations.
The particulate mass sampling ports (two ports at 90° on each of stacks
3, 4, and 10) were located 8 ft downstream from the top of the baghouse and
5 ft upstream of the stack exit. The particle size sampling port (one port,
located at 45° to the mass sampling ports) on stacks 3 and 10 were located at
the same level as the mass sampling ports. Stack 4 did not have a 6-in. port
and, hence, the 4-in. ports were used for the particle size tests as well.
All particulate mass measurements were obtained with the Acurex HVSS
and in accordance with EPA Method 5 procedures. Sampling involved strict
timing and coordination since particle size and mass determination were made
simultaneously, and were dependent on the hot metal charging and tapping times
(variable from heat to heat). Particle size determinations were made with the
SoRI two-cyclone train for all inlet tests (high grain loading), and the
Andersen Mark III impactor with a 15-ym cyclone precutter was used for the
baghouse outlet tests (very low grain loading).
Table 2 illustrates the results of testing conducted during the
changing and tapping cycles.
Based on measurements made with EPA Method 5 partfculate mass and SoRI
two-cyclone particle size trains, the uncontrolled secondary fugitive emission
factor for hot metal charging ranged from 8.08 x 10~2 to 33.34 x 10~2.lb of
particulate emitted per ton of hot metal charged (average 14.65 x 10~2) and
176
-------
TABLE 2. KAISER STEEL BOP TEST RESULTS
Hot metal
charged
Test location (tons/heat)
Baghouse Inlet 186
uncontrolled
emissions
charging3
Baghouse outlet 249
controlled
emissions
charging*3
Baghouse inlet
uncontrolled
emissions
tapping0
Baghouse outlet
controlled
emissions
tapping^
Particulate
mass
Steel emission
tapped rate
(tons/heat) (Ib/min)
27.62
•
0.1505
219 4.95
226 0.1348
Particulate
emission
factor
(Ib/ton)
14.65 x lO-2
5.97 x ID'4
14.62 x ID"2
2.58 x lO-3
Percent
less
than
15 um
56.3
63.0
49.7
40.0
'Average during 24 heats
bAverage during 16 heats
cAverage during 14 heats
''Average during 8 heats
177
-------
from 6.94 x 10-2 to 29.21 x 10-2 Ib of particulate emitted per ton of steel
tapped (average 12.48 x 10-2). The SoRI particle size showed 56.3 percent of
these emissions were less than 15 ym in size.
Based on measurements made with EPA Method 5 participate mass and
Andersen Mark III impactor trains, the controlled secondary fugitive emission
factor for hot metal charging ranged from 3.37 x 10-4 to 10.96 x 10-4 Ib of
particulates emitted per ton of hot metal charged (average 5.97 x 10-4) and
from 2.63 x 10~4 to 8.91 x 10~4 Ib of particulate emitted per ton of steel
tapped (average 4.80 x 10-4). Of these amounts, 63 percent were less than
15 urn in size.
The uncontrolled secondary fugitive emission for tapping ranged from
4.94 x 10-2 to 24.05 x 10-2 ib of particulate emitted per ton of steel tapped
(average 14.62 x 10-2). of this amount, 49.7 percent of the emissions were
less than 15 \fn in size.
The controlled secondary fugitive emission factor for tapping ranged
from 1.11 x 10-3 to 5.06 x 10-3 15 Of particulate emitted per ton of steel
tapped (average 2.58 x 10-3). Of this amount, 40 percent of the emissions
were less than 15 i/n in size.
The particulate mass removal efficiency of the baghouse averaged
99.58 percent for hot metal charging and 99.11 percent for tapping.
KENNECOTT MINERALS PROGRAM
A series of particulate mass, particle size, and sulfur dioxide tests
were conducted on the fugitive emissions generated during matte and slag
tapping operations on the reverberatory furnace of Kennecott Minerals Company
in Hayden, Arizona. These measurements were made to quantify the uncontrolled
fugitive emissions from these operations, and to develop emission factors for
the tapping process.
Both copper matte and slag are tapped by means of a tap hole in the
side of the reverb.eratory furnace. The molten material flows down a launder
chute into a ladle which is positioned one floor below the tap hole. The slag
ladle is carried by a slag hauler to the plant dump site. The approximately
18 tons of matte per ladle are used to charge the converters for concentration
of the copper to about 98 percent.
At the time of the test program, there were no particulate controls on
the hoods collecting the fugitive particulate emissions generated during the
matte and slag tap operations. The emissions captured by the hoods are ducted
through the plant roof to the atmosphere.
Figure 6 illustrates the matte sampling location. The sampling ports
consisted of two 6-in. diameter flanges and two 3-in. diameter male pipe
couplings. These ports were at the 47-ft level of the 34-in. diameter stack.
The stack is connected directly to a fan which drew air from the movable hood
over the matte tap station (the hood was lowered into place during tapping).
178
-------
Ports
83' level
Building
roof
Hole in
floor
^ \w
-v\
Slag
ladle
Matte
ladle
Figure 6. Sampling locations for matte and slag.
179
-------
Figure 6 also illustrates the slag sampling site. This site was
located very near the top of the stack on the roof of the building. Two 6-in.
flanges and two 3-in. diameter male pipe coupling ports were located at the
83-ft level of the 34-in. diameter stack.
The sampling equipment used during the particulate mass testing
compared with requirements of EPA Method 5/8. All particle size tests were
conducted with the Andersen 2000 Mark III in-stack cascade impactor fitted
with a 15-ym cyclone precutter and straight sampling nozzle.
Prior to conducting the actual tests on each stack, a series of
preliminary measurements were made to determine stack gas velocities (to size
sampling nozzles), approximate grain loadings (to avoid overloading impactors)
and the need to condition the glass fiber filter substrates in the stack gases
prior to use (to compensate for SOX filter reaction forming artifact
sulfates). The significance of the last two variables was determined by
drawing 10 ft3 of gas from each stack through the impactor which was preceded
by a flat 47-mm filter (to remove particulates but allow passage of the 503-
laden gases). The results of this preliminary measurement indicated
SOX filter reactions were not a problem at either source and, hence, in situ
conditioning of the impactor substrates was not required.
For the matte tap tests, the stack was traversed in two directions at
90° using 12 sampling points per traverse (total 24 points). Each test was
24 min in length, which was sufficient to collect at least 100 mg of
particulate in the front half of the train. Several matte tap operations were
sampled for each test and four separate tests were conducted to characterize
the particulate emissions.
For the slag tap tests, the same procedures were used (both stacks were
the same diameter) except the sampling time was extended to 48 min to collect
a sufficiently large sample. Multiple slag taps were sampled for each test
and a total of four separate tests were performed.
Particle size measurements were made at the same time as the
particulate mass tests. A total of four points (average point on each radius
of each stack) were sampled for 20 min during matte tap and 40 min during slag
tap to collect a weighable sample.
Measurements made with a combined EPA Method 5/8 train indicate that
the uncontrolled emissions from the matte tapping operations range between
0.12 and 0.17 Ib of particulate emitted for each ton of matt tapped (average
0.13 Ib/ton). Measurements made with the Andersen 2000 Mark III cascade
impactor precutter indicate that 76 percent of the particulates are <15 pm in
size.
Similar measurements made on the slag tapping emissions resulted in
uncontrolled emissions which range between 0.01 and 0.04 Ib of particulate
emitted for each ton of slag tapped (average 0.03 Ib/ton). Approximately
33 percent of these emissions were less than 15 ym in size.
Table 3 provides a summary of the test results.
180
-------
TABLE 3. KENNECOTT MINERALS REVERBERATORY FURNACE TEST RESULTS
Participate Particulate Percent
Tapping mass mass Emission less
rate concentration emission factor than
Test location (tons/min) (Gr/DSCF) (Ib/min) (Ib/ton) 15 vm
Matte hood
uncontrolled
emissions3
2.25
0.1066
0.27
0.13
76
Slag hood
uncontrolled
emissions"
1.68
0.0353
0.04
0.03
33
aAverage of tests M-l through M-4
"Average of tests S-l through S-4
181
-------
As a result of these tests it appears that fugitive emissions generated
during matte removal from the reverberatory furnace are approximately an order
of magnitude higher than slag-generated fugitive emissions and are
considerably smaller in particle size. The average fugitive particulate mass
emission factor (based on Method 5) for matte tap was 0.13 Ib/ton versus
0.03 Ib/ton for slag tap operations. Approximately 76 percent of the matte
emissions were less than 15 ym in size, whereas only 33 percent of the slag
emissions were less than 15 ym in size.
182
-------
CHARACTERIZATION OF PARTICULATE EMISSION FACTORS
FOR INDUSTRIAL PAVED AND UNPAVED ROADS
by: Chatten Cowherd, Jr.
J. Patrick Reider
Phillip J. Englehart
Midwest Research Institute
Kansas City, Missouri 64110
ABSTRACT
This paper presents the results of an expanded measurement program to
characterize uncontrolled particulate emissions generated by traffic en-
trainment of surface particulate matter from industrial paved and unpaved
roads. The emission sampling procedure used in this program provided emis-
sion factors for the following particle size ranges: < 15 |Jm, < 10 Mm> afld
< 2.5 |Jm aerodynamic diameter. Testing was performed at sites that were
representative of significant paved and unpaved road emission sources within
the following industrial categories: crushed stone and gravel processing,
primary nonferrous smelting, and asphalt and concrete batching. Measured
emissions in each particle size range were correlated with road and traffic
parameters as a preliminary step to the development of predictive emission
factor equations for industrial paved and unpaved roads. Previously col-
lected field test data for integrated iron and steel plants and surface coal
mines were also integrated into the industrial road emission factor data
bases.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
183
-------
INTRODUCTION
Over the past few years traffic-generated dust emissions from unpaved
and paved industrial roads have become recognized as a formidable source of
atmospheric particulate emissions, especially within those industries in-
volved in the mining and processing of mineral deposits. Frequently, road
dust emissions exceed emissions from other open dust sources associated with
the transfer and storage of mined materials. For example, in western
surface coal mines, dust emissions from uncontrolled unpaved roads usually
account for more than three-fourths of the total particulate emissions, in-
cluding typically controlled process sources such as crushing operations (1).
Therefore, the quantification of this source is necessary to the development
of effective strategies for the attainment and maintenance of the total sus-
pended particulate (TSP)* standards, as well as the anticipated particulate
standard based on particle size.
Although a considerable amount of field testing of industrial roads has
been performed, those studies have focused primarily on TSP emissions. Only
relatively recently has the emphasis shifted to development of size specific
emission factors in the small particle range (< 15 pm aerodynamic diameter).
The following particle size fractions have been of interest in these studies:
IP = Inhalable particulate matter consisting of particles smaller
than 15 pm in aerodynamic diameter.
PM,n = Particulate matter consisting of particles smaller than 10 (Jm
in aerodynamic diameter.
FP = Fine particulate matter consisting of particles smaller than
2.5 |Jm in aerodynamic diameter.
One major field study was directed to development of size-specific
emission factors for western surface coal mines (1). Field testing was con-
ducted at three mines, each representing a major western coal field. The
study included testing of unpaved haul roads and unpaved access roads in
the absence of dust control measures. Although the primary sampling method
for road testing was exposure profiling, the conventional upwind-downwind
method was used for a few tests. Particle size distributions were deter-
mined at two or more heights in the plume by use of dichotomous samplers
and high-volume cascade impactors with cyclone preseparators. Road dust
emission factors in the form of predictive equations were developed for the
TSP, IP, and FP fractions.
In a second study directed to evaluation of open dust source controls
in the iron and steel industry, uncontrolled emissions from paved and un-
paved roads were tested prior to application of control measures (2). The
testing was performed at two steel plants, in Ohio and Texas. Exposure pro-
filing, the primary test method, was supplemented by the use of high-volume
(*) TSP denotes the size fraction of the total airborne particulate that is
captured by a standard high volume sampler.
184
-------
cascade irapactors with cyclone precollectors for particle sizing at two
heights in the plume. Emission factors were determined for total particu-
late (TP) matter as well as IP and FP fractions.
In a related nonindustrial study, IP, PM..-, and FP emission factors
were developed for urban paved roads (3). Exposure profiling was used to
measure emissions at representative sites in the Kansas City and St. Louis
areas. Particle sizing at two heights in the plume was performed with
high-volume samplers equipped with size-specific (IP) inlets and cascade im-
pactors. A generalized emission factor equation was derived from the test
data, containing parameters that vary with particle size fraction and road
category.
This paper reports on a field study which utilized exposure profiling
to provide size-specific emission factors for uncontrolled paved and unpaved
roads within other industries having significant road dust sources. Testing
was performed at representative sites within the following industrial cate-
gories: crushed stone and gravel processing, primary nonferrous smelting,
and asphalt and concrete batching. By combining the test data from this
study with the data from the two prior industrial studies referenced above,
it was anticipated that the resulting data base would be adequate to develop
reliable emission factors for the range of road and traffic conditions which
characterize major industrial road dust sources. The sampling methodologies
and emission factors derived in this study are presented below.
SAMPLING SITE SELECTION
Plant surveys were performed within each of the specified industries to
locate suitable test sites which at the same time were representative of
road and traffic conditions within these industries.
Three major criteria were used to determine the suitability of each
candidate site for sampling of road dust emissions by the exposure profiling
technique:
1. Adequate space for sampling equipment and easy accessibility to the
area.
2. Sufficient traffic and/or road surface dust loading so that ade-
quate mass would be captured on the lightest loaded collection substrate
during a reasonable sampling time period.
3. A wide range of acceptable wind directions, taking into account:
(a) the test road orientation relative to the predominant wind directions
for the locality; and (b) possible effect of nearby structures on wind flow
across the test road.
185
-------
Table 1 gives the general geographical location of test sites within
each industry, the distribution of tests performed, and the sampling periods
for each industry. Note that this study also entailed testing of rural un-
paved roads. For purposes of comparison, Table 1 also lists appropriate
data for the three prior studies. It is apparent that the data base repre-
sented in this table represents a diversity of industrial settings and
seasonal conditions.
TABLE 1. FIELD TEST MATRIX
Industrial
category
Asphalt batching
Concrete batching
Copper smelting
Sand and gravel
processing
Stone crushing
Rural roads
Surface coal
mining*
Iron and steelt
Urban roads^
Test site
location
Missouri
Missouri
Arizona
Colorado
Kansas
Kansas
Kansas
Missouri
Colorado
Montana
North Dakota
New Mexico
Ohio
Texas
Missouri
Kansas
Illinois
Road tests
Unpaved
0
0
3
0
3
5
6
4
2
12
10
7
7
0
0
0
0
conducted
Paved
4
3
3
3
0
0
0
0
0
0
0
0
7
4
11
5
3
Sampling period
Oct. 1981
Nov. 1981
Apr. 1982
Apr. 1982
July 1982
Dec. 1981
Aug. -Sept. 1981
Mar. 1982
Apr. 1982
Aug. , Dec. 1979
Oct. 1979
July-Aug. 1980
July 1980
Oct. -Nov. 1980
July 1981
Feb. -Mar. 1980
May 1980
Feb. -Mar. 1980
May 1980
* Reference 1.
t Reference 2.
£ Reference 3.
186
-------
SAMPLING EQUIPMENT
A variety of sampling equipment was utilized in this study to measure
particulate emissions, roadway surface particulate loadings, and traffic
characteristics.
The primary tool for quantification of emissions was the MRI exposure
profiler, which was developed under EPA Contract No. 68-02-0619 (4). Nor-
mally, the exposure profiler was positioned at a distance of 5 m from the
downwind edge of the road. The profiler consisted of a portable mast (6 m
height) supporting an array of five sampling heads spaced at 1 m intervals
above the ground. Each sampling head was operated as an isokinetic TP ex-
posure sampler directing passage of the flow stream through a settling
chamber (trapping particles larger than about 50 (Jm in diameter) and then
upward through a standard 8- by 10-in.* glass fiber filter positioned hori-
zontally. Sampling intakes were pointed into the wind, and sampling
velocity of each intake was adjusted to match the local mean wind speed, as
determined prior to each test. Throughout each test, wind speed was moni-
tored by recording anemometers at two heights, and the vertical profile of
wind speed was determined by assuming a logarithmic distribution. A wind
vane at the top of the mast was used to monitor wind direction.
To obtain the particle size distribution of the particulate emissions,
high-volume parallel-slot cascade impactors with cyclone preseparators were
positioned along side of the profiler at heights of 1 and 3m. At the op-
erating flow rate of 20 scfm, the cutpoints of the impactor stages were 10,
4.2, 2.1, 1.4, and 0.73 [Jm aerodynamic diameter, and the cyclone cutpoint
was 11 (Jm aerodynamic diameter. In addition, a standard high-volume air
sampler and a high-volume sampler equipped with a size-selective IP inlet
were operated at a height of 2 m.
For measurement of background particulate concentration, a standard
high-volume sampler and a high-volume sampler with an IP inlet were operated
upwind of the test road, at a height of 2 m. Care was taken to locate the
upwind samplers away from any localized upwind emission source.
Samples of the dust found on the road surface were collected as part of
each source test. In order to collect this surface dust, it was necessary
to close each traffic lane for a period of approximately 15 min. Normally,
an area that was about 0.3 m by the width of the road was sampled. A portable
vacuum cleaner was used to collect surface dust from the paved roads. The
attached brush on the collection inlet was used to abrade surface compacted
dust and to remove dust from the crevices of the road surface. Vacuuming
was preceded by broom sweeping if large aggregate was present. For the un-
paved roads, broom sweeping was used to collect samples of loose particulate
matter from the road surface. Unpaved roads were not vacuumed.
(*) Readers more familiar with metric units may use the conversion factors
at the end of this paper.
187
-------
The characteristics of the vehicular traffic during the source testing
were determined by both automatic and manual means. The vehicular charac-
teristics included: (a) total traffic count; (b) mean traffic speed; and
(c) vehicle mix.
Total vehicle count was determined by direct observation. Vehicles
were classified into functional categories keyed to the number of axles and
wheels. The average speed of the traveling vehicles was determined by di-
rect observation with verification from plant operators. The weights of
the vehicle types were estimated by consulting plant operators for indus-
trial sites. Automobile literature was used to estimate curb weights of
vehicles traveling on rural roads.
SAMPLING AND ANALYSIS PROCEDURES
The sampling and analysis procedures employed in this study were sub-
ject to the Quality Control (QC) guidelines which met or exceeded the re-
quirements specified by EPA (5,6). As part of the QC program for this
study, routine audits of sampling and analysis procedures were performed.
The purpose of the audits was to demonstrate that measurements were made
within acceptable control conditions for particulate source sampling and
to assess the source testing data for precision and accuracy. Examples of
items audited include gravimetric analysis, flow rate calibration, data
processing, and emission factor calculation.
Particulate samples were collected on Type A slotted glass fiber im-
pactor substrates and on Type AE (8- by 10-in.) glass fiber filters. To
minimize the problem of particle bounce, the glass fiber cascade impactor
substrates were greased. The grease solution was prepared by dissolving
140 g of stopcock grease in 1 liter of reagent grade toluene. No grease
was applied to the borders and backs of the substrates. The substrates
were handled, transported, and stored in specially designed frames which
protected the greased surfaces.
Prior to the initial weighing, the greased substrates and filters were
equilibrated for at least 24 hr at constant temperature and humidity in a
special gravimetrics laboratory. During weighing, the balance was checked
at frequent intervals with standard weights to ensure accuracy. The sub-
strates and filters remained in the same controlled environment for another
24 hr, after which a second analyst reweighed them as a precision check.
Substrates or filters that could not pass audit limits were discarded.
Ten percent of the substrates and filters taken to the field were used as
blanks.
Prior to equipment deployment, a number of decisions were made as to
the potential for acceptable source testing conditions. These decisions
were based on forecast information obtained from the local U.S. Weather Ser-
vice office. A specific sampling location was identified based on the anti-
cipated wind direction. Sampling would be initiated only if the wind speed
188
-------
was forecast between 4 and 20 mph. Sampling was not planned if there was a
high probability of measurable precipitation (normally > 20%) or if the road
surface was damp.
Emission sampling usually lasted about 1 hr for unpaved roads and 4 hr
for paved roads. Occasionally, sampling was interrupted due to occurrence
of unacceptable meteorological conditions and then restarted when suitable
conditions returned. The unacceptable meteorological conditions most fre-
quently encountered consisted of light winds (below 4 mph) and insufficient
angle (< 45 degrees) between mean (15-min average) wind direction and road
direction.
To prevent particulate losses, the exposed sampling media were care-
fully transferred at the end of each run to protective containers within the
MRI instrument van. Exposed filters and substrates were placed in indivi-
dual glassine envelopes and numbered file folders and then returned to the
MRI laboratory. Particulate that collected on the interior surfaces of each
exposure probe and cyclone precollector was rinsed with distilled water into
separate glass jars.
When exposed substrates and filters (and the associated blanks) were
returned from the field, they were equilibrated under the same conditions
as the initial weighing. After reweighing, 20% were audited to check pre-
cision.
The vacuum bags and the polyethylene bags containing road sweepings were
weighed to determine total net mass collected. Then the dust was removed
from the bags and was dry sieved. The screen sizes used for the dry sieving
process were the following: 3/8-in., 4, 10, 20, 40, 100, 140, and 200 mesh.
The material passing a 200 mesh screen is referred to as silt content.
The vertical distributions of the product of plume concentration and
mean wind speed were numerically integrated to calculate emission factors.
The cyclone/cascade impactor sampler combinations provided reliable point
concentrations for IP and finer particle size fractions. Plume height was
determined by extrapolation of the vertical profile of TP concentration as
measured by the MRI exposure profiler.
TEST RESULTS
Tables 2 and 3 summarize by industry category the size-specific parti-
culate emission factors determined for unpaved roads and paved roads, re-
spectively. Geometric means and geometric standard deviations of emission
factors are given because road dust emission factor data sets are found to
follow log normal rather than normal distributions (1). For purposes of
comparison, data from References 1 to 3 are also presented.
189
-------
TABLE 2. UNPAVED ROAD EMISSION FACTORS (UNCONTROLLED)
Industrial
category*
Copper smelting
Sand and gravel
processing
Stone crushing
Rural roads
Surface coal
Un
of
tests
3
2
4
9
20
Emission factor (kg/VKT)t
TP IP
x a x a
23.4 1.1 6.64 1.1
3.76 1.5 1.42 2.4
4.28 2.5 2.98 3.2
9.93 3.5 2.54 3.9
7.62 2.4 1.57 3.1
PM10
x a x
0.42 1.1 0.06
0.97 2.7 0.24
0.50 2.4 0.08
0.97 4.3 0.21
0.06
FP
a
1.9
3.2
2.2
4.8
3.2
raining (haul
trucks)^
Surface coal
mining (light/
med. duty)^
Iron and steel
(light duty)§
Iron and steel
(heavy duty)§
10
1.77 1.8 0.503 3.8
0.03 3.6
3.30 1.2 0.67 2.1 0.57 1.9 0.21 2.1
36.9 1.0 8.69 1.1 6.80 1.2 2.35 1.1
* Parenthetical notation refers to vehicle type.
t x = geometric mean; a = geometric standard deviation; VKT = vehicular km
traveled.
# Reference 1.
§ Reference 2.
190
-------
TABLE 3. PAVED ROAD EMISSION FACTORS (UNCONTROLLED)
Emission factor (kg/VKT)*
Industrial
category
Asphalt batching
Concrete batch-
of
tests
3
3
TP
x
0.60
1.22
a
1.9 0
1.7 0
IP
x
.25
.76
a
1.5 0
1.4 0
PM
x
.11
.29
FP
10 *V
a x
1.2 0 . 05
1.5 0.10
a
1.2
1.6
ing
Copper smelting
Sand and gravel
processing
3 3.25 1.5 1.16 1.7 0.79 1.8 0.17 2.0
2 1.88 1.8 0.41 2.5 0.23 2.8 0.16 4.0
Iron and steel 10
Urban roads 10
0.72 1.6 0.22 1.5 0.17 1.5 0.06 1.6
0.001 4.3 0.001 4.2 0.001 3.6
* x = geometric mean; a = geometric standard deviation; VKT = vehicular km
traveled.
Source: Reference 3.
The average small particle fractions of the unpaved and paved road
emissions for each industry are given in Table 4. It is evident that paved
road emissions contain substantially larger portions of small particles than
unpaved road emissions.
The road and traffic parameters measured during each test included:
total loading of loose surface particulate on the traveled portion of the
road; silt content of the total loading; silt loading, which is the product
of total loading and silt content; average vehicle speed; average vehicle
weight; and average number of vehicle wheels. These source characterization
parameters are summarized for unpaved and paved roads in Tables 5 and 6,
respectively. Again, data from References 1 to 3 are presented for purposes
of comparison. Although the moisture content of the road surface particu-
late was measured, it was not included as a reliable source parameter be-
cause of the difficulty of collecting a sample without altering its moisture
content.
The parameters are being evaluated as possible correction parameters
for development of predictive emission factor equations by stepwise multiple
linear regression. Previous studies have shown that predictive emission
factor equations for unpaved and paved roads substantially reduce the un-
certainty in estimating road dust emissions on a site-specific basis (1,3,7).
191
-------
TABLE 4. EMISSION FACTOR RATIOS BY INDUSTRIAL CATEGORY
Industrial
category
IP/TP
PM1Q/TP
FP/TP
Unpaved Paved Unpaved Paved Unpaved Paved
Asphalt batching - 0.42 - 0.18 - 0.083
Concrete batching - 0.62 - 0.24 - 0.082
Copper smelting 0.28 0.36 0.018 0.24 0.0026 0.052
Sand and gravel 0.38 0.22 0.26 0.12 0.064 0.085
processing
Stone crushing 0.70 - 0.12 - 0.019
Rural roads 0.26 - 0.098 - 0.021
Surface coal mining 0.21
(haul trucks)
Surface coal mining 0.28
(light/med. duty)
Iron and steel
Iron and steel
(light duty)
Iron and steel 0.24
(heavy duty)
0.0079
0.017
0.31 - 0.24 - 0.083
0.20 - 0.17 - 0.064
0.18
0.064
192
-------
£
Q"
W
h-1
O
H
52
O
^)
*"g»
o
oo
o
05
Q
w
^
^
P i
55
P^
i
CO
|
p2
*2
OH
S3
0
rH
H
N
5
13
u
fyj
2
W
O
W
0
g
o
w
m
a
PQ
^J
H
M
i-H
01
01
JS
.
si
01
d
i x
4J
X!
oo
•H
01
3
j3
01
/— \
(0
01
§
o
•p
\_^
b
1 X
"O
01
01
o<
w
J
01
**
/^
-fl
a
s-<
b
1 X
00
d
•rl
TJ
O
r-l
4J
rH
•H
co
d
01
4->
d
o
u
4-1
rH
•rl
co
y^
CU
E
00
b
1 X
^*N
5*^
\^S
b
1 X
X
d
•rl
*rj
CO
O
i-H
l-H
cO
O
H
/~\
N
B
Of,
b
1 X
IH
O
i-H
R) >,
•H M
M O
4-1 00
in D
d 4->
"O CO
d u
M
•J'
r-l CM r-t O
rH i— 1 rH i— 1
-tf -^ 00 O
i-H
•-• O CM rH
. * • «
T-H r-H rH i-H
CM 00 CM CM
CM rH
•fr -Jc
O O CO CM
rH rH i-H rH
^O oo ^ r**
l-H rH s^-
CM CO ^ O^
• • • •
rH rH rH CM
O O O O
\Q \,Q Q\ \O
• d
4J CO *|H
I-H V4 si ta
01 00 00 W TJ
e d 3 «t
W T3 -rl IH O
d w u )H
Vj CO W
0) 01 01 i-H
a. "O u d «
PJ do o ^4
O CO M 4-1 3
U W ft CO OS
-X
CM O ^J O
rH rH rH rH
00 0 ~*
o
r^ ,<
CO CO O O
i-H rH rH i-H
rH >* CM s*
-* m co CM
co oo in
i
CM CM rH
-* m o
CM rH rH |
r-
r~ co I-H
i
rH i— 1 rH
C- ,
•rl -iH 4-1
S S 3 rH rH
/•^ ^o* ot
rH Ul rH 01 /— > 4) ^
tO JC fl) • 4-) >, 4-1 >>
OO O-d W4-> W4J
0 3 0 W 3 3
ri E *d *d *d *d
oi 4J oi "^, d d
U U 4-> CO £>> CO 4J
CO rH CO ,d C, jg
«4H3 MHOO dco d 00
(4 C8 M -rl O 01 O -H
3J3 3rH rlXl rlrH
*
d
o
•H
4-1
03
•rl
^ •
OI U
T3 -rl
•O 4H
U a)
T3 4->
d
RJ 01
4J 4J
W CO
tH
U OI
•rl d
14 01
4-1 OO
U
E 0
0 4->
01
00 TJ
01
II w
3
b
01
•x rl
d oi
(0 S
01
S w
01
O i-H
•H U
(H -rl
4J J3
01 01
B >
0
01 01
00 >
•rl
II 4->
1 X R)
u
.jc
•jc ^«
193
-------
.]<
*
s
1-4
o
2
O
cj
2
^
v-/
CO
<£
o
OS
1
Oa
|
C/3
P^
W
£— <
1
1
55
o
1-H
H
^^
w
Oi
d
d
o
4-1
D
t X
o
1 X
TJ
OJ
U
p.
co
.
jd
01
/~*,
ji
a
^_,
D
1 X
00
•iH
•o
(8
O
i— 1
4-*
i-H
•H
CO
4J
a
HI
4J
d
o
u
4J
i-H
•H
CO
r~^
N
e
w
**_/
/— ^
gsg
Sfc-/
b
1 X
o
1 X
00
d
•H
•O
<8
O
rH
rH
<8
O
H
/^N
(M
E
W
D
1 X
O U
4->
W
O U
2 4-1
rH
.3 Er
M O
4-1 00
M 01
d 4->
"O «
d u
t— <
•h-
1-1 o ro co
• • • •
t— 1 rH rH t— 1
»* O 00 rH
\^ c? in *^
rH rH
•H-
o o m o
rH i— * rH rH
-
0 CO
U 4-1 4-1 «
4J Ifl rH V4
nt ^3 oi oo oo
A Ed
Oi en -a -ri
4-)4-> CM
I-H o> t-i (it ta
cfl M OI OI
A U ft T! O
OH C »H d O
W O O Q V4
< CJ U CO &
CM CM
I-H i— 1
O rH
>43 m
**o ^
• •
I-H rH
r^ co
I-H
^J
1 •
rH
1 ^
in
oo a\
CM CO
in
m CM
r~ o
VO rH
. t
i-l CM
CO 00
r •
VO O
rH rH
m ON
CM <*
CO CM
NH/
o o
rH i-H
i-H
0>
01
U9
T3
d
"* d
d A
O -O
rH P
•
d
o
•H
4-1
•rl
> 0
01 -H
T) HH
T3 fl)
tH M
01 4->
•a
d oi
ra 4J
4J CQ
CO ^4
0)
u d
•rl 01
V4 00
4J
0) O
E 4-1
o
4) "O
oo oi
01
II d
b o)
M
*». Q)
d 3
(8
01 01
E OJ
rH
u u
•rl -rl
j-i .d
4_) 4J
0) >
s
0 Oi
0) >
00 -rl
4-1
II p<
1 X U
J(
4t 4~
194
-------
As a preliminary step in the development of emission factor equations
for this expanded data base derived from this study in combination with Ref-
erences 1 to 3, a nonparametric analysis was performed. The purpose of this
analysis was to determine whether the associations between source character-
istics and emissions intensity found to be important in the earlier studies
were reflected in the expanded data base.
Spearman's (rank-order) correlation (r) was computed based on rankings
of the geometric mean values (by industry type) of the various size-specific
emission factors and corresponding source characteristics (8). Pearson's r
is typically used to test for association; however, its use can be limited by
the restrictive assumption of a bivariate (joint) normal distribution (8).
Rank correlation methods are not limited by the form of the distribution. It
should be noted that these are only first order (simple) correlations. They
do not reflect the partial correlations that are represented in earlier MRI
equations (e.g., as given in References 1, 3, and 7).
In general, the analysis confirms earlier work which indicated that
emissions were most directly related to roadway surface loading. There are
some indications that the expression of the loading parameter will change
depending upon the particle size fraction, but no significant correlations
emerge for FP emissions. This preliminary analysis also suggests that the
relationships between source characteristics and emissions intensity are
stronger for unpaved roads than for paved roads.
For unpaved roads, IP emissions correlate significantly with both silt
content (p = 0.044) and silt loading (p = 0.068); and PM^g emissions also
correlate with silt loading (p = 0.042), where p is the probability that the
relationship is due to chance. The strongest relationship was found between
the ratio PM^Q/TP and vehicle weight, as illustrated in Figure 1. This
indicates that the more intense road surface grinding by heavy vehicles pro-
duces a greater portion of small particles in the emissions.
For paved roads, the strongest correlation is found between IP emis-
sions and silt loading (p = 0.051). This is consistent with the emission
factor equation previously developed for urban streets (3).
CONCLUSIONS
The field testing program described herein was performed to expand the
small particle emission factor data base on industrial paved and unpaved
roads. The ultimate objective of this research is to provide reliable small
particle emission factors for the range of road and traffic conditions which
characterize major industrial road dust sources.
The past approach to this problem has been to develop emission factors
in the form of predictive equations which relate particulate emissions to
road and traffic parameters. Such equations have been developed for TSP
emissions from industrial roads based on combined data crossing industry
195
-------
I 2
o
"5
Q£
I 3
o
o
5
a.
Sand & Gravel
Iron & Steel
(Heavy Vehicles)
Iron & Steel
(Light Vehicles)
Stone Crushing
Rural Roads
Copper Smelting
4 3
Vehicle Weight Rank
Figure 1. Rank-order correlation for unpaved roads:
PM10
(p = 0.017)
/TP emission factor ratio versus vehicle weight
lines. These equations have been shown to be far more accurate than
single-valued averages in estimating site-specific road dust emissions.
The nonparametric analysis performed in this study indicates that the
associations between emissions intensity and source characteristics found to
be important in earlier studies, are reflected in the expanded data base.
Thus, it appears likely that reliable emission factor equations can be de-
veloped from the data base. However, the equations for different particle
size fractions will probably contain different functional dependencies on
source parameters.
ACKNOWLEDGMENT
The work upon which this paper is based was performed pursuant to EPA
Contract No. 68-02-3158. William B. Kuykendal served as EPA project officer
for the study.
196
-------
REFERENCES
1. Axetell, K. , Jr., and Cowherd, C., Jr. Improved emission factors for
fugitive dust from western surface coal mining sources - Vol. II:
Emission factors. (Draft final.) Contract 68-03-2924, W.D. 1, U.S.
EPA, Cincinnati, OH, November 1981.
2. Cuscino, T., Jr., Muleski, G. E., and Cowherd, C., Jr. Iron and steel
plant open source fugitive emission control evaluation. (Draft final.)
Contract 68-02-3177, W.A. 4, U.S. EPA, Research Triangle Park, NC,
August 1982.
3. Bohn, R., Cowherd, C., Jr., and Englehart, P. J. Paved road particulate
emissions. (Draft final.) Contract 68-02-3158, T.D. 19, U.S. EPA,
Research Triangle Park, NC, February 1982.
4. Cowherd, C., Jr., Axetell, K. , Jr., Guenther, C. M. , and Jutze, G.
Development of emission factors for fugitive dust sources.
EPA-450/3-74-037 (NTIS PB238262). U.S. EPA, Research Triangle Park,
NC, June 1974.
5. Quality Assurance Handbook for Air Pollution Measurement Systems. Vol.
II - Ambient Air Specific Methods. EPA 600/4-77-027a. U.S. EPA,
Research Triangle Park, NC, May 1977.
6. Ambient Monitoring Guidelines for Prevention of Significant Deteriora-
tion. EPA 450/2-78-019 (NTIS PB283696). U.S. EPA, Research Triangle
Park, NC, May 1978.
7. Cowherd, C., Jr., Bohn, R., and Cuscino, T., Jr. Iron and steel plant
open source fugitive emission evaluation. EPA-600/2-79-103 (NTIS
PB299385). U.S. EPA, Research Triangle Park, NC, May 1979.
8. Bhattacharyya, G. , and Johnson, R. Statistical concepts and methods.
New York: John Wylie and Sons, pp 526-533, 1977.
CONVERSION FACTORS
Readers more familiar with metric units may use the following conversion
factors:
Non-metric Times Equals metric
in. 2.54 cm
mph 1.61 km/hr (kph)
scfm 28.32 std liters/rain
197
-------
CONDENSIBLE EMISSIONS MEASUREMENTS IN THE INHALABLE PARTICULATE PROGRAM
by: Ashley D. Williamson and Joseph D. McCain
Southern Research Institute
Birmingham, AL 35255
ABSTRACT
In order to meet the EPA's inhalable particulate program goal of
obtaining measurements of condensible matter in process streams, a Stack
Dilution Sampling System was designed at Southern Research Institute
under EPA contract. The principal component of the system is a cylindri-
cal dilution chamber in which flue gas is mixed with filtered air and the
resulting aerosol-laden mixture analyzed. As the sample is cooled by
dilution, condensible vapors form particles under conditions similar to
those which occur in actual plumes. Field measurements have been per-
formed at a continuous drum mix asphalt plant, two kraft recovery boil-
ers, a coke quenching tower, and an oil-fired package boiler. Data from
these tests show that significant fractions of the total emissions at
some sources consist of condensible vapors.
This paper has been reviewed in accordance with the U.S. Environ-
mental Protection Agency's peer and administrative review policies and
approved for presentation and publication.
198
-------
INTRODUCTION
One concern in stack sampling of particulate is the realization that the
ultimate particulate emissions from a stationary source may well be greater
than those emissions measured instack. Many process streams contain signifi-
cant amounts of condensible compounds which pass through the stack in vapor
phase, but which undergo a physical change of state as the plume is diluted
and cooled, and ultimately add to the particulate emission inventory of the
ambient environment. The added mass is expected to accumulate almost en-
tirely in the fine particle range below 10-15 ym.
In order to simulate this plume condensation process, a Stack Dilution
Sampling System (SDSS) was designed and built at Southern Research Institute
under EPA contract. This system attempts to form a replica plume from dilu-
tion of extracted stack gases with filtered ambient air so that the condensed
portion of the particulate emissions can be studied. As part of its Inha-
lable Particulate (IP) Emission Factor Measurement Program, the EPA chose to
include condensible measurements at several industry sources using the SDSS.
This paper will briefly describe this instrument and report on the condens-
ibles tests performed thus far.
DESCRIPTION OF SAMPLING SYSTEM
A diagram of the SDSS is shown in Figure 1. The design and operation of
the device have been described previously (1), so will only be summarized
here. Stack gases are extracted through the EPA IP Dual Cyclone sampler
(1,2), in which particles with aerodynamic diameters greater than 15 ym are
captured in one cyclone (Cyclone X), and particles in the 2.5-15.0 ym size
range are captured in a second (Cyclone III). Particles with aerodynamic
diameters less than 2.5 ym pass through the sampler and remain in the gas
stream as it passes through a heat-traced probe and flexible hose, through a
flow metering venturi, and into the cylindrical dilution chamber. Use of the
present cyclone train as a precutter represents a design compromise. It
would be patently unrealistic to filter all particulate matter from the
undiluted stream and thereby remove centers on which condensation occurs in
the stack plume. The cut at 2.5 ym preserves most of the condensation sites
in the diluted stream while removing the larger particles which would have
greater probability of loss in the probe and sample lines. The condensible
material is thus combined with the respirable fraction «2.5 ym) of the non-
volatile particulate matter. Dilution air is forced through an ice bath
condenser, reheated to the desired temperature, filtered, and introduced tan-
gentially into the inlet assembly at the bottom of the dilution chamber. The
dilution air flow is directed upward in an annulus bordered on the outside by
the walls of the 21.3 cm ID dilution chamber and on the inside by the 4.27 cm
ID sample inlet tube. The major purpose of the condenser/heater combination
is to allow dilution air with reproducible constant temperature and humidity
to be used at all sampling locations. A "standard" dilution air at 21.1°C
and 24 percent relative humidity is used. These values are easily achieved
and within the range of typical ambient air conditions.
199
-------
o
-------
The principal component of the system is the 1.22 m long cylindrical di-
lution chamber in which mixing of sample and dilution air occurs. The mixing
mechanism chosen for the device is injection of a jet of extracted stack gas
into the center of a confined stream of dilution air moving at a lower veloc-
ity. This choice seems to be the closest approximation to a jet exhausting
from a stack exit into a nearly stagnant ambient atmosphere. The diameters
of the sample inlet tube and the overall dilution chamber were chosen to give
cross-sectional areas for sample gas and dilution air inlets to the dilution
chamber that are proportional to the mass flows of these two gas streams at
the design dilution ratio of 25:1. Thus, if the sample gas temperature were
the same as the 21 °C dilution air, the two streams would merge isokinet-
ically. Heated sample gas streams will have extra velocity, as indeed occurs
in a buoyant plume. The sample gas flowrate is constrained by the require-
ment that the first IP cyclone cut at 15 ym. The necessary flow varies with
temperature and composition of the stack gas, but is approximately equivalent
to a mass flow of 17 normal £/min over a range of typical conditions. The
total diluted flow of 425 £/min is monitored by an orifice in the exhaust
line. In practice, it is preferable to maintain this standard exhaust flow
rather than a standard dilution ratio. Gas flowrates are adjusted using the
two blowers shown in Figure 1.
IP sampling protocol for particulate measurements calls for single point
samples at the centroids of each of the four quadrants of the duct cross
section, and calls for four samples at each point. For SDSS sampling, this
protocol was modified due to cost factors, the restricted mobility of the
SDSS, and the general lack of port access when several other parallel mea-
surements are in progress. At most sites the choice was made to make three
or four total runs at two positions in the ductwork. The SDSS was run at the
centroid of one quadrant of duct cross-section. At the centroid of a second
quadrant a second train was run simultaneously. The second train consists of
the instack inhalable-particulate precollectors used on the SDSS followed by
an instack filter. For ducts without stratification of particulate matter,
the instack precollectors should collect equivalent amounts of particulate
matter. Any mass on the SDSS filter in excess of that accounted for by the
catch of the instack filter can be attributed to condensible matter. On
alternate runs, the two trains are switched in order to minimize the effects
of stratification.
At one of the test sites it was not possible to use the first IP cyclone
(Cyclone X) due to confined duct space. For these tests, the second cyclone
was used with a buttonhook nozzle. The sample flowrate for this run was cal-
culated in the usual manner as if Cyclone X were used.
FIELD SAMPLING STUDIES
In the course of the IP Emission Factor Measurement Program, six field
sampling tests have been performed using the SDSS. Data are available from
two kraft recovery boilers, a continuous drum mix asphalt plant, a coke
quench tower, and an oil-fired package boiler. Results from the sixth test,
at an oil-fired utility boiler, were not available in time to be included in
this paper.
201
-------
The first test series was performed at two kraft recovery boilers down-
stream of the electrostatic precipitator (ESP) particulate control devices at
each source. One of the two furnaces was of the direct contact evaporator
(DCE) design; the second did not use contact evaporators. The results of the
tests are shown in Table 1. At the non-DCE boiler A, only two useful runs
were possible due to ESP failure. The desired four runs were obtained at the
DCE boiler B. The runs at those sites were somewhat unusual in that a sub-
stantial fraction of the mass collected in the SDSS was found in the sample
line rinses, as shown in Table 1. The rinses at non-DCE boiler A contained
over four times the mass collected by the SDSS filter. The probe deposits at
the DCE boiler were not as high, but were still comparable to the filter
catches. The color of the evaporated residues varied from white to yellow to
red-brown.
The large sample line catches at the recovery boilers were unexpected
in view of the results of other field tests in which less than 5 percent of
the fine particulate fraction of the SDSS sample was collected in the sample
lines and in view of laboratory tests with dye aerosol which showed line de-
positions on the order of 15 percent of the filter catch. A possible explan-
ation lies in the fact that a portion of the exit end of the Method 5 probe
used for the test was found to be unheated and therefore operated at a
temperature below that of the stack gas. Condensation of volatile species
(including water vapor) on this cooled portion of the probe was quite pos-
sible under these conditions. It is also possible that some of the probe
rinse material may contain reaction products of the stainless steel sampling
probe with the corrosive stack vapors. Thus, the condensible fractions noted
in Table 1 should be taken as upper limits. In either case, there was de-
finitely a condensible or reactive component present in the exhaust streams
at both recovery boilers, and the non-contact evaporation boiler A emitted
significantly more of this component than DCE boiler B. It should be noted in
this regard that the Method 5 tests at boiler A also had large fractions in
the probe rinse and had loadings significantly higher than were measured with
instack impactors.
A second SDSS test series was performed at the quench tower of a steel
industry coke oven. Coke quenching is a cyclic process in which individual
carloads of hot coke from the oven are doused with a stream of water, giving
a burst of emissions including steam, entrained particulate matter, organic
material, quench water residue, and chemical reaction products. Sampling was
performed only during the course of each 3 to 4 minute quench. In order to
prevent accumulation of water droplets in the sampling trains, the second IP
cyclone (ill) of both trains and the instack filter of the IP train were
heated with heating tapes. Cyclone X was left unheated and served as a large
droplet scalper. All samples were taken from a single port downstream of the
demister baffles. The first three runs were conducted with the usual mode of
quench operation in which excess quench water is collected and recycled. The
fourth run sampled emissions from single-pass "clean" quench water operation.
The results of the SDSS tests are shown in Table 2. Fairly significant
instantaneous particulate loadings were observed during the quench cycles,
with most of the collected mass in the fine particle size range (under 2.5 ym).
As in the first test, a significant fraction of the SDSS sample was found in
202
-------
H
co
p.
j3
^
i
00
in
CO
Q^
^
o
•
vO
O
**
rH
p%^
rH
VO
CM
W
W O
W O
C_3 •
o ^
P^
PM
W vO
CJ CT\
O
M C^
o ^
i in
oo
CM
PH
W
^
M
W rH
rH
CO
m
CTi
CM
o
CM
CO
8
CO
rH
OO
•
CM
00
r-J
H m
rH -*
O CM
in r-.
in r^*
H m
O ON
CM rH
m co
CO co
vO
m
P^
in
rH
CM 00
CM CO
o m
rH CO
rH CM
CO CM
rH
O -*
O O^
0 rH
CM CM
CO
CO
^**) fi[ \f\
CO H S
1 1 1
CM CM CM
00
•
O
00
in vo
r^s CM co
H in m
vO rH ^^
in vo
co
00 vo
VO CTN
CM er>
vo m
rH VO
00
-3" O
f^ r^
CM CM
vO
O 1
• 1
«* 1
CM
m m
oo m
CM O
rH CM
vo OO
O CTi
CO CO
rH
00 rH
00 CM
00 rH
H CM
CO
CO
^"t PI
CO H
in in
CM
•
vO
rH
VO O
CO vO
CTN a^
rH
CO vO
ON vO
m r^
i-H
CM -*
vO O
CM CM
in oo
CM CM
rH 0
O1 OO
(^ sj-
rH rH
00
O
O
rH
CO Sf
CTi ON
•J- CO
m m
co m
CM O
CM O
oo co
00 CT>
rH rH
co
co
^"^ ft.
CO H
203
-------
CO
Q
M
H
Si
PH
O
r-r-l
5
23
w
w
0
u
PH
0
CO
H
en
w
H
g
W
H
CO
K.I
K^
CO
O
2
M
M
CM
^
.
Qj , — |
i— 1 o m
fX t> S
6 a
cfl tO "T3
17:1 o
"O
A CO 0)
0 CD rH
PI rH P.
o-u w
pi
CU O
rH -H ^~,
a. 4J a
S Cfl -r-t
cd H S
M M V~^
pj
PC!
r-^ o iri vo
• • • •
CO *3" ^J" *3"
CO CO OJ •<)•
OOO CT»^o ^* r^* co r°*^ *>o ^^
in r^ r^ co rH r^ *^ CM
MOI inm ^ooo r-Hvo
CO 01 01 rH Ol i-H H
unoi oi v£>^£> CTiUI
r— 1 St OO (Ti^ OICO
rH rH rH r-l rH
r^ rH CM rH
c^ o^j ^,0 co
• i • i • i • i
00 | CO I -vf | l-» |
CO CT\ rH m
rH rH
vooi i — ir^ ^co ooo
coco unm r^-in coo
LOI-^ rHOO \O O r~. vO
rH rH i-H rH
01-3- inco 010 r-^oi
rHCTi 011^ ^TOO rHin
O^ i-O O^ O*J ^J* ^^ C^ C^
LOO -O ^ co co r^» r^
cfl
co co co co n)
QPM Op-i QpM OP-i
COM COM COM COM
II II II II
H H oiol coco -3-^a-
•
G
O
•H
4-1
Vi
01
p.
o
CO
CO
Pi
cu
rH
00
•H
CO
O
4-1
cu
rH
O
^*
a
OJ
e
0
}_l
m
"T3
OJ
60
G
tfl
O
rt
ai
4-1
cfl
>
j~
O
G
Q)
^J
O1
204
-------
the probe rinse. In this case, the rinses were approximately equal in mass to
the filter catch. Comparison of the two trains indicates an average of 86
mg/dnm^ condensible material in the recycle runs, corresponding to 30.9
percent of the total particulate emissions. The "clean" water run contained
51 mg/dnm^ condensible; but since the overall particulate emissions were
lower for this mode, this concentration represents 44.6 percent of the total
measured emissions. These amounts of volatile material are substantial. It
should be mentioned, however, that since the instack samplers and SDSS sample
lines were heated above the average duct temperature, much of the measured
condensible fraction may represent volatile material that is re-evaporated in
the instack portions of the sampling trains.
A third test which dramatically demonstrates the presence of conden-
sible emissions was performed at a continuous drum mix asphalt plant. The
plant in question was a modern stationary unit with 325 tons per hour capac-
ity and a highly efficient particulate control baghouse. The burner is fired
by natural gas, and the tests were run under conditions of approximately 30
percent recycle aggregate feed.
As shown in Table 3, the majority of the controlled particulate emis-
sions as measured at stack temperatures are over 15 ym aerodynamic diameter
and presumably consist of resuspended rock dust. The diluted stack gas, how-
ever, contains an equivalent loading of fine particles not seen by the in-
stack IP train. This test gave the largest unambiguous ratio of condensible
to nonvolatile fine particles of any of the SDSS tests, with an 8:1 ratio of
filter catches. When the SDSS sample line washes (which average about 25
percent of the SDSS filter catches) are added, the SDSS condensible catch
represents 90 percent of the fine particle fraction and 45 percent of the
total particulate emissions.
The high condensible fraction of the SDSS filter catch allowed a strik-
ing confirmation of residual volatility of the particulate even at room tem-
perature. As shown in Figure 2, the three SDSS filters lost up to 20 percent
of their original particulate mass over a period of 4 days after sampling.
This effect is expected to be general for condensible emissions which are not
saturated in the local air and are not stabilized by oxidation, water uptake,
or other chemical reaction.
The final SDSS test for which data are available took place at the EPA
facilities at Research Triangle Park, North Carolina. A 2.5 x 10^ BTU/hr
Scotch Marine Package boiler was sampled with the SDSS and IP trains, with an
EPA Method 5 train, an Andersen Mark III impactor, and extensive continuous
gaseous emissions monitors. The boiler emissions were measured using three
fuel oils:a No. 2 distillate oil, a No. 5 residual oil with 1 percent sul-
fur, and a No. 6 residual oil containing 2.8 percent sulfur. Table 4 con-
tains the average emissions from three runs with each oil measured by the
SDSS and IP trains and by Method 5. Comparison of the SDSS and IP trains in-
dicates that significant concentrations of condensible matter can be measured
for the two residual oils. The concentrations shown in Table 4 represent
10-15 percent of the total SDSS emissions, and 30-60 percent of the fine
particle fraction of the diluted stream. Significantly, the Method 5 front-
half loadings agree closely with the SDSS, indicating that the dew point of
205
-------
H
3
PH
H
3
PC
PM
X!
a
g
CO
p3
o
H
H
8
H
O •
^s CM
O *^
X
S §•
o
>, A
O N*-'
«
3 ro
•>
c
r^ vo CM ro
1 ....
I to in r~ m
-* co m ^3-
CTN CTiOO vOCTi CM i— 1 CT\CO
in i *^* Is** vo r** ^^ oo r^ rH
H Hm HCTi CMt^ rHrH
rH i-H CM CM
r^O
HH CNCM COCO «*-*
-------
§ 5
S
CM CO ^
z z z
3 D D
OC CC CC
oo<
§
S
CM
oq to
od
SSVW 3AI1V13U
CM
d
I
207
-------
TABLE 4. RESULTS OF STACK DILUTION SAMPLING SYSTEM TESTS AT AN
OIL-FIRED PACKAGE BOILER3
Oil Device
IP
No. 2 distillate SDSS
M5
IP
No. 5 residual SDSS
M5
IP
No. 6 high sulfur SDSS
M5
Mass
Cyclone
(>2.5 ym)
<0.1
0.1
41.7
41.2
188.4
203.6
Concentration (mg/dnm3)
Filter
(<2.5 ym)
0.7
3.3
18.1
28.5
16.4
41.3
Total
0.8
3.4
5.7
59.8
69.9
71.8
204.8
244.6
253.0
Calculated
Condensible
2.5
10.0
25.0
*Each value represents the average of three runs for the oil used.
208
-------
the condensible material is higher than the 120°C Method 5 oven temperature.
In addition, the loaded SDSS filters showed a definite tendency to gain
weight when exposed to room air, indicating a hygroscopic component in the
particulate catch. These factors strongly suggest sulfuric acid vapor to be
the condensible component in this case. Soluble sulfate analyses of the SDSS
and IP filters are in progress to check this hypothesis.
In conclusion, the Stack Dilution Sampling System has been used for
testing at five sites in four source categories where condensible emissions
were anticipated during formulation of the Inhalable Particulate sampling
program. Some indication of condensible emission was found in each test.
For at least one source, particulate matter formed by condensible vapors
accounts for almost half the source contribution to the ambient total sus-
pended particulate. The condensible contribution is even greater to particu-
late matter in the size range of respiratory interest. These results indi-
cate that from many emissions sources, the contribution of condensible mate-
rial connot be ignored.
ACKNOWLEDGEMENTS
This research was supported by the U.S. Environmental Protection Agency
under Contract No. 68-02-3118, D.B. Harris, Project Officer. Field tests at
the kraft recovery boilers and at the oil-fired package boiler were per-
formed via subcontract from Acurex Corporation under EPA primary Contract
68-02-3159. Field tests at the asphalt plant were performed via subcontract
from Midwest Research Institute under their primary Contract 68-02-3158.
Tests at the coke quench tower were performed via subcontract to GCA Tech-
nology Corporation under EPA primary Contract No. 68-02-3157. EPA program
manager for all field is D.L. Harmon.
209
-------
REFERENCES
1. Williamson, A.D., and Smith, W.B. Development of a sampling train for
stack measurement of inhalable particulate. In: Third Symposium on
the Transfer and Utilization of Particulate Control Technology, Volume
IV. Atypical Applications. EPA-600/9-82-005d (NTIS PB83-149617.
U.S. Environmental Protection Agency, Research Triangle Park, North
Carolina, 1982. p.297.
2. Smith, W.B., Gushing, K.M., Wilson, R.R., Jr., and Harris, D. B.
Cyclone samplers for measuring the concentration of inhalabla particles
in process streams. J. Aer. Sci. 13:259, 1982.
210
-------
GAS CLEANING AND ENERGY RECOVERY
for
PRESSURIZED FLUIDIZED BED COMBUSTION
By: Dr. Albert Brinkmann
Gottfried Bischoff GmbH & Co.
Essen, West Germany
Mr. Peter Kutemeyer
Bischoff Environmental Systems
Pittsburgh, PA
ABSTRACT
THE DEVELOPMENT OF FLUIDIZED BED COMBUSTION
In an effort to reduce the consumption of oil and natural gas the
search for a new technology, capable of utilizing low grade fuels, as
well as more fully extracting available energy from high grade coal, led
to the development of fluidized bed combustion (FBC).
The advantages of FBC are:
A. Lower and more uniform combustion temperatures,
resulting in lower generation of NO .
X
B. Acceptably low S0_ emissions by addition of lime-
stone, thus eliminating expensive desulfurization
equipment.
C. Smaller heat exchangers, and thus smaller boilers,
due to higher heat transfer coefficients.
D. Use of low grade fuels.
The first operational FBC systems used in Germany operated at
atmospheric pressure. These classic FBC plants required a relatively
low capital investment and presented no development problems during
installation or operation. Figure 1 shows the flow diagram of a 35 MW
thermal output plant.
211
-------
Boiler with fluid bed combustion
Limestone
Coal Htghpressure
n n AS'ea
Feedwater Flu« gas
pneum. coal
feed
Co mbustion
air
fines from filter
pneum. ash recycl ing
flue ash
Figure 1: Flow scheme of a fluidized bed combustion
plant (35 MW) operating at atmospheric pressure,
,
— Classical —f-Circulati ng
fluidized bed fluid bed
Increasing
sol ids
thro ughput
Increasing expansion
Figure 2: Bed motion as a function of SSV,
212
-------
This conventional FBC system was characterized by a well defined
fluidized combustion surface. As the gas velocity of the FBC system is
increased, this well defined fluidized combustion surface transforms to
pneumatic transport. The intermediary stage is called the circulating
fluidized bed.
The advantages of the circulating fluidized bed are:
A. Further reduction in SO- emissions by use of fine grained
limestone and combustion in various stages.
B. Reduced space requirements as compared to conventional
FBC, thus greater output per unit.
Figure 2 shows the transition from the classic FBC process to the
pneumatic transport process. Further development finally resulted in a
pressurized FBC system which utilized the above mentioned advantages of
the circulating fluidized bed to a greater degree, and, in addition,
resulted in a more completely combusted ash.
A further aim of the pressurized FBC development was to operate an
open cycle, coal fired gas turbine system. A simplified schematic of such
a system is presented in figure 3. The heart of this system is the pres-
surized FBC chamber. A heat exchanger in the FBC chamber produces the
steam to drive the steam turbine.
The combustion gases from the FBC boiler are expanded by a gas turbine
after they have been cleaned of dust. Gas temperatures are the same as
the FBC boiler temperatures about 800-950°C (1500-1750°F).
Two study groups were formed in Europe to develop the two major com-
ponents of the energy recovery separately. The steam turbine process is
being developed and optimized by the International Energy Agency of the
OECD, with primary responsibility being exercised by the British Coal
Board. The gas turbine process is being developed by a study group (AGW)
which is a joint venture of the "Bergbau Forschung GmbH" and the "Vereinigte
Kesselwerke A.G." in Germany.
The ultimate goal of these groups is the development of a system
capable of generating a large amount of power using a combined gas/steam
turbine pressurized FBC process.
GAS CLEANING AND THE PRESSURIZED FBC PROCESSES
The remainder of this paper will address the problems of cleaning the
flue gas generated in a pressurized FBC process, so that this cleaned gas
can be expanded in a gas turbine.
Cleaning of the flue gas is the key to an economically viable pres-
surized FBC process because a gas turbine requires relatively clean gas in
213
-------
r
ACW
Fluid be.d
combustion
chom
lim
L..
Figure 3: Flow scheme of a pressurized fluidized
bed combustion plant.
214
-------
order to have an acceptable operating life.
The use of electrostatic precipitators and baghouses to date has not
resulted in providing a viable gas cleaning process for pressurized FBC.
Tests conducted with granular filters have been equally unsuccessful.
Because of this, a cyclone,a less efficient gas cleaning system was select-
ed in Germany for the prototype plant, in conjunction with a gas turbine
capable of operating with dust laden gas. Whether or not this solution
will be an optimum economic and technical one is doubtful.
Even though it may not be possible, in the near future, to solve the
problem of cleaning a high temperature and high pressure gas, such as that
generated by a pressurized FBC system, it does seem feasable of using such
a FBC boiler, with all its advantages, to generate steam. In this case a
heat exchanger is placed in the pressurized FBC boiler for steam generation.
In the process, the gases leaving the boiler are cooled to about AGO C
(750°F).
For many years now Bischoff has successfully operated wet scrubbers
for the cleaning of gas from high top pressure blast furnaces (B.F.) in
conjunction with energy recovery turbines that utilize the excess temp-
erature and pressure of the B.F. gas to generate electrical energy.
Below the use of such a wet gas cleaning system will be investigated
to determine its viability for the temperatures and pressures encountered
in a pressurized FBC process. A flow diagram of such a plant is presented
in figure 4. The main component of a gas cleaning plant for high top
pressure B.F. is the Bischoff annular gap scrubber, a differential pres-
sure cleaning system. The heart of the system, as the name implies, is
the annular gap element, an adjustable conical center body in a conical
shell. A cross-section is shown in figure 5.
The Bischoff annular gap scrubber, "The Bischoff" for short, is not
only able to provide a highly efficient gas cleaning system for varying
operating conditions but is also able to maintain a pre-determined gas
pressure in the system to a high degree of accuracy.
PERFORMANCE ANALYSIS
We shall consider two different processes for comparision:
A. Expansion of a high temperature gas in a gas turbine.
B. Expansion of a cooled and saturated gas in a gas
turbine.
The results presented below will consider the required and recovered
energy on the gas side of the pressurized FBC process. The calculations
are based on the following gas analysis:
215
-------
uojtsnqtuoa paq pin|j
cfl
C
o
•H
4-J
CO
3
TJ
01
•d
•H
0)
N
•H
3
CO
01
a
n)
M
td
sr
01
216
-------
Figure 5: BIschoff Annular Gap Scrubber
217
-------
C02 = 19.02% N2 - 71.112%
CO = 0.588% SO = 0.027%
02 = 4.836% H20 = 4.317%
The calculations will compare the performance of an expansion turbine
downstream of the pressurized FBC boiler to the performance of an expansion
turbine with a wet gas cleaning system inserted between the boiler and the
turbine.
Figure 6 shows an enthalpy - moisture diagram for the case of expand-
ing a high temperature gas in a gas turbine. Inlet (E) and exit (A) cond-
itions for the isentropic expansion of the gas through a turbine are
shown. The saturation lines (/ =1.0) are indicated for the FBC discharge
pressure P. = 10 bar (145 psi) and turbine discharge pressure P~ = 1.1603
bar (16.8 psi). u is the FBC discharge temperature, corresponding to
turbine inlet temperature, and 1)2 is the turbine discharge temperature.
The moisture content X is calculated from the indicated water content
in the flue gas.
Figure 7 shows the same diagram after insertion of the wet scrubber.
Point W is the gas discharge condition from the FBC boiler; point S, the
gas condition after cooling and saturation; point E, turbine inlet condi-
tions, and finally point A, turbine discharge conditions after isentropic
expansion.
u is the saturation temperature of the gas and Uj and Ug, the turbine
inlet and discharge temperatures respectively. X , X , and X are the
corresponding moisture content values.
Values for gas temperature versus pressure at the turbine discharge
for the two cases, as well as scrubber water discharge temperature are
shown in figure 8. For these calculations an inlet water temperature of
80 C (175 F) was selected based on settling tank considerations^and optim-
ization of turbine performance. A flue gas volume of 100,000 m /h (62,100
scfm) and a cooling water use of 200 m /h (880 gpm) was selected.
Figure 9 compares the power output of the two cases as a function of
FBC pressure and temperature. Also shown is the power requirement of the
FBC compressor. Turbine and compressor efficiency were selected at 85%
and 80% respectively.
Theoretical values for the thermal powers of the turbine discharge
gas and the scrubber water as a function of pressure are given in figure
10, for the without and with wet scrubbing cases respectively.
For the case of expanding a high temperature gas in a turbine, a
turbine discharge temperature of 100 C (212 F) is assumed. For the wet
scrubbing case it is assumed that the flue gas must be reheated to 75 C
218
-------
219
-------
220
-------
^1'
240
(464)
220
(428)
(392)
1RD
(356)
160
(320)
140
(284)
120
(248)
100
(212)
80
(176)
fin
(140)
(104)
?n
(68)
Watp
scru
\
\
\
-
*
NGas temperature at turbine outlet
for expansion of high temp, gases
\
\
\
^" •-
' *••••*
**
r temperature at
Dber outlet
— ,
^"*^H"
Gas temperature at turbine
\
\
\J
\
. -N/~
•- \
i ^-
^^^•*-~>
outlet for
V
x
\
X
__^--
N^
>w
\
expansion of saturated and cooled gas
^
'Jo =400
. v£ = 400
- % = 300
^v£--30C
^i^ =4
— • i? - 3
2
(29)
4
(58)
6
(87)
8
(116)
10(145)
[bar] (psj)
Figure 8: Temperature of gas at turbine outlet
and of water at scrubber outlet.
221
-------
N[KW]
12000
10000
8000
6000
4000
2000
/
Compressor
Turbine
Expansion of
high temp, gases
£ =300°C
(572°F)
,1% = 400°C
, i£ = 300°C
Turbine
Expansion of
saturated and
cooled gas
2
(29)
4
(58)
6
(87)
8
(116)
10
(U5)
m
p Ibarl(psi)
Figure 9.: Power requirements of compressor and output of turbine
VN = 100000 m3/h (62100 scfm) HT = 85% n =
222
-------
NIKWJ
6000
thermal power of
gas ,
(outlet temp. 100°C
(212°F)
4000
2000
thermal power
of scrubbing
water
2
(29)
thermal power
for reheating
to 75°Cl167°F)
4
(58)
6
(87)
8
(116)
Figure 10: Theoretical thermal powers
tfN = 100000 m3/h (62100 scfm)
Vw = 200 m3/h (880 gpm)
223
10
(US)
pi (bar]
(psi)
-------
(168 F). The energy required for this is shown.
From figure 10 we can see that FBC pressure and remaining thermal
energy in the turbine discharge gas for the case without wet scrubbing
is inversely proportional, whereas the thermal content of the scrubber
discharge water in the wet scrubbing case and the FBC pressures is directly
proportional.
Combining the results from figures 9 and 10 we obtain the values
presented in figure 11.
From this figure it is clear that as the FBC operating pressure
increases, the use of the wet gas cleaning system with a pressurized
FBC boiler becomes a viable option.
CONCLUSION
The advantages of a pressurized FBC system are enhanced with increas-
ing operating pressure. The above comparison has shown that wet cleaning
with "The Bischoff", in conjunction with such a high pressure FBC system
and an expansion turbine, is a viable alternative to the use of electro-
static precipitators, baghouses, or cyclones. This is particularly true
when it is considered that "The Bischoff" represents a proven low main-
tanance system that is in operation on almost every high top pressure B.F.
in the world today, quite a number of which also have a gas turbine for
energy recovery.
The work described in this paper was not funded by the U,S, Environ---
mental Protection Agency and therefore the contents do not necessarily
reflect the views of the Agency and no official endorsement should be
inferred.
224
-------
NIKW]
6000
5000
4000
3000
2000
1000
c
o
cr
- TOGO
Expansion of high
temperature
gas
a£=300°C
(572° F)
/\
2
(29)
high temperature
gas
Expansion
downstream of
wet cleaning
Expansion
downstream of
wet cleaning
Figure 11: Comparison of power output.
225
-------
DEMONSTRATION OF THE FEASIBILITY OF A MAGNETICALLY STABILIZED
BED FOR THE REMOVAL OF PARTICUIATE AND ALKALI
By: L. P. Golan, J. L. Goodwin, and E. S. Matulevicius
Exxon Research and Engineering Company
Florham Park, NJ 07932
ABSTRACT
This paper describes a unique panel filter bed of magnetizable
particles subjected to a magnetic field used to remove particulate in the
flue gas generated by a pressurized fluidized bed combustor (PFBC) to a
level sufficient to prevent erosion of a gas turbine. The unique
characteristics of this magnetically stabilized panel bed (MSB) high
throughput rates, high particulate capture efficiency, trace metal removal,
and use with a wide variety of coals.
This Department of Energy sponsored program is experimental in
nature concentrating on evaluating the key factors necessary for de-
monstrating the feasibility of this concept, viz., the ability of the mag-
netic material to survive the PFBC environment, and particulate removal
efficiency operating at PFBC conditions. The magnetic bed material eval-
uation phase has been completed. The results of the materials evaluation
phase indicate that coated cobalt particles are suitable for this appli-
cation. The particles have been found to possess good oxidation and
attrition resistance while maintaining magnetic properties at the elevated
temperatures. After 1000 hours of exposure to FBC flue gas no mechanical
failure of the coating has been detected while sample magnetization was re-
duced only 10-20%..
The test phase of the program has been completed. The first
phase of testing determined filter media flow rates at various field
strengths. This was followed by a series of tests to determine the gas
side pressure gradients at combinations of filter media flow and field
strength. The final sequence of tests operated the filter in the semi-
continuous mode. While these runs are still being analyzed, initial data
appears promising.
226
-------
BACKGROUND
The purpose of the program is to demonstrate the feasibility of
using a bed of magnetizeable particles subjected to a magnetic field to
remove particulates found in the flue gas generated by a PFBC. A typical
process sequency utilizing this concept is illustrated in Figure 1. Flue
gas from the PFBC combustor (1) is passed through cyclones (2) to remove
the bulk of the flyash/dolomite particulates. The flue gas is then passed
into a magnetized cross-flow bed consisting of an admixture of ferro-
magnetic particles necessary to stabilize the bed and potentially alkali
metal scavenger particles such as bauxite or alumina. The downwardly
moving bed acts as a filter for capturing particulates by impaction, inter-
ception, and diffusion. Trace quantities of sodium and potassium could be
removed by reaction or adsorption with the scavenger bed material. As the
concentration of flyash increases on the bed, it is removed from the bottom
of the bed (4) and circulated by a gas transfer line to, first, a rough cut
cyclone (5) which removes any flyash which has been detached from the bed
material and, then, if necessary, to an elutriator (6) where the remaining
flyash is removed from the bed material. Bed material is returned to the
bed (8) while the dust laden gas from the elutriator is combined with the
transfer line gas and sent to a cyclone (7) where the dust is separated
from the gas. The remaining gas is either cleaned further in a bag filter
(11) and sent to a stack, or recirculated (10) to the combustor as makeup
air.
As trace metals build up on a non-magnetic material, they are
removed by "bleeding" a side-stream of particles and subsequently removing
the scavenger bed meterial (12). This separation should be easily accomp-
lished by using a magnetic separation.
The key feature of this concept is the use of the magnetic field
to maintain integrity of the cross-flow bed. Preliminary cold flow studies
conducted at Exxon have shown that such a bed is capable of contacting the
solid bed material with gas at velocities four times greater than possible
in the cross-flow moving bed filters without the imposed magnetic field.
The advantages for the concept developed through the cold flow studies are:
• Increased throughput. The gas velocity before bed material is
entrained or "blown out" of the bed is significantly increased be-
cause of the orientation and structuring of the bed material by the
magnetic field.
• Lower pressure drop. The structuring and the orientation of the bed
results in a higher void fraction and hence lower pressure drag per
unit thickenss of the bed when compared with a conventional moving
granular bed operating at the same conditions.
227
-------
• Increased collection efficiency. The collection efficiency is in-
creased significantly over conventional beds especially for
smaller (10 Mm) particulate.
For the magnetized panel bed concept to be possible a filter media capable
of withstanding the FBC environment was needed. The materials requirements
for the filter media are demanding. Most importantly the material must be
ferromagnetic at the operating temperature of the filter (~830°C). The
Curie temperature (that temperature where a metal loses its ferromagnetic
properties) of most ferromagntic materials is lower than PFBC operating
temperatures; the sole exception being cobalt. Pure cobalt however is a
soft metal that is easily oxidized. Since the process environment is both
corrosive and oxidative, a coating must be employed to protected the
cobalt. In addition to being oxidation and corrosion resistant, the
coating must also be attrition resistant. The filter material will be
subjected to an attrition promoting environment when it leaves the panel
bed vessel and enters the elutriator cleanup cycle. A protective coating
was develped by Exxon prior to the initiation of this program which was
shown to prevent oxidation of the cobalt.
ATTRITION RESISTANCE
As part of a preliminary evaluation, the coated cobalt particles
were subjected to an attrition resistance test. The test consisted of
vigorously fluidizing a bed of the coated cobalt for 100 continuous hours
at 830°C and a fluidizing velocity of 2.1 m/s (1.6 times minimum fluidizing
velocity). SEM photomicrographs of cross-sectioned spheres showed no signs
of wear, chipping, oxidation or separation of the coating. Magnetic pro-
perty measurements indicated that the magnetic induction force was un-
affected by the 100 hours of exposure. These tests demonstrated that a
material capable of satisfying the panel bed requirements was possible.
MAGNETIC MATERIAL EVALUATION
As part of the current DOE program, a materials evaluation task
was conducted to assess the corrosion resistance of the filter media under
high temperature FBC flue gas conditions. Small samples of filter media
were exposed at FBC conditions for periods up to 1000 hours. For the test-
ing, a small 250,000 BTU/hr AFBC unit was used (Figure 2). Flue gas from
the combustion of coal passed through three sample locations yielding three
different exposure temperatures. The first site was in the combustor tower
approximately one foot above the fluidized bed; the temperature was approx-
imately 830°C. The second sample site was at the tower outlet; the temper-
228
-------
ature was approximately 740°C. The final site was downstream of the
cyclone cleanup train; the gas temperature here was approximately 615°C.
By selectively removing and replacing samples at 100 hour intervals
multiple sample exposure times at temperature were achieved. The particles
were generally spherical in shape and approximately 800-1400 pm in
diameter.
The 1000 hours was accumulated in 14 runs. The samples cooled
between runs. This was taken to be a more severe test since the particles
were subjected to cooldown and heatup stresses which would not noramlly be
encountered in actual PFBC operation.
A summary of the operating conditions for all runs is presented
in Figure 3 including flue gas composition and materials exposure
temperature. (The average flue gas composition was 02 4.1%, SC^ 310 ppm,
NO 700 ppm, CO 580 ppm and C02 16%.)
During the materials evaluation two different filter media
samples were exposed. Both materials were pure cobalt generally spherical
in shape which were coated using the same process. The difference in the
filter media was the supplier and subsequently the method of manufacture of
the cobalt particle. The samples were designated either N or P. The
material N was used during the attrition tests conducted by Exxon. The
material from a second supplier is designated by P and is the material used
as filter media in the magnetically stabilized bed.
Material N was exposed for the entire 1000 hours period. The new
material (P) was exposed for only 774 hours.
To facilitate identification of the samples with respect to their
supplier, exposure location, and length of exposure the following
nomenclature was adopted:
first letter - identify supplier, N or P
second letter - identify exposure location
A - exposed above bed at 830°C
B - exposed before cyclones at 740°C
C - exposed after cyclones at 615°C
numbers - indicate hours of exposure
229
-------
changes had occurred. The conclusion was that neither hot gas corosion of
the coating nor metallurgical changes due to the high temperature of ex-
posure affect the life of the particle.
Did the Magnetic Properties Change?
In addition to the changes in the particle metallurgical pro-
perties, changes in magnetic properties as a function of time and temper-
ature were also determined. The magnetic properties were measured by using
a Perkin-Elmer thermal microbalance, TGS II. The principle of measurement
was the Faraday method. When a small magnetic sample is placed in a non-
uniform magnetic field, a translational force is exerted on the sample
according to the following equation:
F 1 „ dH
V = ITT M dX
o
where F is the translation force in dynes, V the volume of sample in cm , M
the magnetic induction of the sample in gauss, and dH/dX the magnetic field
gradient of the non-uniform applied field in oersted/cm. A schematic dia-
gram of the setup is shown in Figure 6. A small portion of the magnetic
particle sample, S, about 10 mg, was placed in the sample pan, P, on the
microbalance. A sample of three beads was used. The beads were placed in
a row to approximate a cylinder which helped to reduce the effect of de-
polarization. The environmental tube, E, for the sample was then placed in
position and purged with nitrogen. When a steady state flow of the inert
gas was reached, the weight of the sample was measured on the microbalance
and recorded in mg. A horseshoe permanent magnet, N, was then placed
around the sample at a distance X, in mm, from the center line as shown.
The magnetic field strength, H, of the magnet was measured and the field
gradient, dH/dX, calculated. The force exerted on the sample due to the
field gradient in the direction could be measured directly on the micro-
balance in mg. The sample magnetization in M was then calculated using the
above equation.
By placing the magnet, N, at a specific positioa, a corresponding
force, F, could be measured. Without altering the set-up, the temperaure
of the sample can be varied by controlling the heater of the furnace, C.
Therefore, the temperature dependence of the magnetization could be eval-
uated. Measurements were made over a range of temperatures (20-1000°C).
The magnetic induction properties of samples of the cobalt part-
icles which had been exposed to FBC flue gas conditions for various time
periods were measured to evaluate the effect of exposure time on magnet-
ization. For these measurements, a consistent procedure was used and the
magnet was always placed in the same position. Measurements were made at
room tempeature (25°C) and at 830°C, the exposure temperature at location
A. Due to the variability of the spheres (size, shape, and magnetic
property) at least two measurements (two different sets of beads)
230
-------
were made for each sample evaluated. In some cases, the variance was large
so additional measurements were made to assure representative values were
obtained. The magnetic induction force as measured is expressed as mg of
force/tag of sample. The object of the magnetic evaluation is to determine
if the magnetization of the sample has decayed as a result of FBC expo-
sure. A normalized plot of the magnetic induction force, F, versus expo-
sure time was used to indicate whether and to what extent the magnetization
has decayed. The normalized magnetic induction force is the ratio of F at
time T, and F at zero exposure. Figures 7 and 8 show the normalized plots
for materials N and P respectively.
In Figure 7 we see the normalized ratio plot for samples N-A-119,
327, 530, 774, 1000 and N-original. Magnetic measurements were made at
25°C and 830°C. The data scatter is not surprising and can be attributed
to the particle variability. The magnetic induction force has decreased by
less than 15% over the 100 hours of exposure with no indication of decay
after the first 100 hours of exposure of flue gas. Also, there is no
'evidence of a continuing decay with time.
Figure 8 is a similar plot for the P material. After 774 hours
of exposure, there was approximately a 20% decrease in magnetization. With
both materials, it appears that the major loss of magnetization occurs
during the first 100 hours of exposure. The magnetization appears to
stabilize after the initial drop.
The temperature dependence of the magnetiziation of the sample
can be measured by continuously recording the magnetic induction force of
the sample as it is heated from room temperature to 950°C. The temperature
dependence for two materials, N-unexposed and P-unexposed, are shown in
Figure 9. Again, a normalized ratio is used, this time the force at
temperaure, FT divided by the force at room temperature, F25°C* ^e P
material is only slightly affected by the temperature indicating a Curie
temperature higher than 950°C. The normalized ratio drops to zero at the
Curie temperature. The N sample however exhibits a decrease in magnet-
ization with increasing temperature. The magnetic properties of the two
materials were affected differently by increasing temperature, this can be
attributed to the difference in the method of manufacture and the amount of
diffusion that occurred during the coating process. The trends seen here
are consistent with the other samples of materials N and P tested.
The metallurgical examination and magnetic evaluation of the
exposed samples indicate that these particles are suitable for MSB appli-
cation under the conditions at which they were evaluated. No failure of
the particle coating was seen; any defects seen in the particles were
inherent, not a result of the exposure testing. Only a limited amount of
diffusion had taken place in the most severe case. The loss of particle
magnetization which occurred during the initial 100 hours was 20% or less
which is an acceptable level since most of the change occurred early in the
test and, a lower rate of loss of magnetization, if any, could be assumed
for long exposures.
231
-------
KEY ISSUES:
Did Metalurgical Changes Occur?
All exposed samples of both materials look unaffected to the
naked eye. Selected samples were submitted for cross-sectioning and photo-
micrographs to examine the condition of the coating. Samples of material N
which were exposed for 0, 327, 530, 774 (exposure location C at 734 hours),
and 1000 hours were submitted for examination. Samples of material P which
were exposed for 0, 101, 304, 508, and 774 hours were also submitted for
examination. Figure 4 shows cross-sections of material N particle at 0,
530, and 1000 hours of exposure at 830°C. In Figure 5, the cross-sectioned
particles material P are shown for 0, 101, 304, and 774 hours of exposure
at 830°C. The fissures (grain boundaries) and small voids seen in the
particles remain from the manufacturing process and are not a result of the
exposure testing.
Comparison of the unexposed particle with the exposed particle
sample indicates no failure of the coating had occurred. It shoud be noted
that due to the method of manufacture, material N is very spherical and has
a narrow size distribution with a majority of the spheres having a 1000
micron diameter. The P material is less spherical with more rod and
elongated pieces and cover a larger size distribution (800-1400 \im).
Energy dispersion x-ray (EDX) analysis was performed on selected
particles to determine the extent of diffusion of the coating into the Co
core. This was important because diffusion of the coating material into
the cobalt would lower the magnetic Curie temperature of the cobalt appli-
cation. Two techniques were used; one was the line scan method whereby the
concentration profile of an element of interest along a particular line
trace could be obtained to indicate locations of high concentration. The
other method was semi-quantitative analyses of all the elements at a part-
icular point of interest, such as the center of particle.
The samples submitted for EDX analysis were W-original, N-A-1000,
P-B-304, and P-A-774. The EDX analyses indicate that a minimal amount of
diffusion of the coating into the cobalt core had occurred even after 1000
hours of exposure at the most severe conditions of the materials test. The
semi-quantitative EDX results also indicated that the N material contained
some Fe contamination (up to 10% in the semiquantitative tests). The P
material had an insignificant amount of Fe contamination. A small amount
of Fe is not expected to affect the overall performance of the Co part-
icles. In fact, the use of a Fe-Co alloy as the particle core material for
this application had been studied earlier because of economic consider-
ations. The result of this evaluation showed that little metallurgical
232
-------
FILTER OPERATION
The pilot plant facility for once through operation is shown in
Figure 10. The PFBC combustor which will generate the particulate ladden
flue gas has an ID of 11.4 cm. It operates at 870°C and ten atmospheres
pressure generating 1.5-2 sm /min of flue gas. Lock hoppers located above
and below the filter vessel are used to supply and collect the filter media
as it passes through the filter. Key features of the unit are:
1. Removable cyclones to allow variation in flue gas particulate load-
ing.
2. A methane injection system to make up any flue gas heat loss.
3. An electric air preheater which will bring the filter up to temper-
ature prior to filtration.
4. Particulate sampling stations before and after the filter to measure
capture efficiencies.
5. Total particulate filters for integrated capture efficiencies.
6. Extensive system to measure pressure differential through the filter
bed.
The filter bed internals are a variation on conventional designs
used in the earlier studies of panel beds. Dirty flue gas enters down the
center of filter vessel and passes radially through the openings
(Figure 11). The louvers are designed such that the material will not flow
out the openings when a modest magnetic field is applied. A radial gas
flow pattern flowing from inside to out will be used to allow a high gas
velocity at the bed inlet but a decreased velocity through the bed. The
high inlet velocity will prevent duct laydown in the flue gas distributor;
the lower velocity in the bed will improve capture efficiency. Pressure
taps have been incorporated along the length distributor and across the
distributor for measuring system pressure drop. Mathematical flow modeling
of the distributor suggests that the flow maldistribution for this design
should be less than 20% with a 20% slot open area. In addition the flue
gas jets penetrate approximately 25% of the bed depth.
PRELIMINARY RESULTS
Filtration tests for both a stationary bed and moving bed have
been completed for a range of temperatures and loadings of particulate.
Results are being evaluated. However, several important conclusions have
emerged.
233
-------
• No plugging of the filter face has occurred. The large openings
possible because of ability to hold the particles in the bed
magnetically precludes the bridging of ash found in conventional
panel bed filters.
• Overall efficiency has averaged approximately 90%. In addition, the
efficiency did not significantly vary with particulate size but
increased with loading.
• High flue gas throughputs were shown to be feasible.
• Captured particulate were shown to be easily removed from the cobalt
spheres in a subsequent cleaning operation.
Final analysis of the data is currently underway. A final report
will be issued by December 30, 1982.
ACKNOWLEDGEMENT
Funding for this program is provided by U. S. Department of
Energy under contract No. DE-AC21-ET 15055.
The work described in this paper was not funded by the U. S.
Environmental Protection Agency and therefore the contents do not
necessarily refect the views of the agency and no official endorsement
should be inferred.
234
-------
-v
SS
«nj- 6
s
4>5
'i
g
&
c
0»
TJ
0 C
*» «>«»-
« e UJ
ilg
3 >»O>
f-£? i x^.
1
o
ai
c
cr»
«
ST
i.
o
u
c
*a
o-
V
b'l
T- 1-
^ D>
— Sf_
" "7\ / -55
. *1\ / -a KS,
5
235
-------
ro
UJ
O
a.
8
19
C
o
o
o
•
t
•2
41
i.
3
o
a.
X
UJ
u
3
re
x
c
o
V)
UJ
5
0)
u.
a>
CT
L.
41
>
<
Excess
r_
0
0
•a
1
«
a.
X
Ul
>erficial
3
<
s
<
essure
a.
c
3
O£
s
a
c
3
oe
41
c
i- o c
at «— ••-
4J > L.
«_> K-
4->
i- 0}
%t
£3
$
o -a
£1 0)
< CO
CM
O »*
•
p **
cT 0.
u o.
o a.
z a.
CM£
0 0.
10 0.
*
U »«
<
• 1-
•o aj j=
4> CB 01
U. 0£ ^
*
*»
£L
cn £
•a N-
4) 41
co -x.
*
£
°G «
•2"s
41
>
i-o
V*
t—
c
S4
31?
CH
n
0
oo
n
ir>
U3
0
o
to
0
un
CM
0
o
O!
CO
,_
«*
0
u>
o
U)
0
00
in
o
CO
c
LO
p—
00
oc
c\
^
-
en
>o
u>
«*
to
oo
r*.
tn
n
LO
r^
o
0
o
o
o
0
r—
O
CM
r-
•»
n
in
0
•f
oo
in
0
CO
rr
u>
oc
C7>
CM
r-
CVJ
r«»
»•
in
to
0
in
in
o
CM
*
o
00
o
o
01
ro
1*-
IO
o
CO
00
CO
in
0
to
CM
CM
O
1—
to
i
CM
O
*
CO
U)
*
o
ro
r*
I*-
CM
ro
in
r~
O
uo
r^
I
O
CM
O>
r—
CO
•*
r*
O
ro
oo
CO
in
O
CO
in
p»
CM
CO
in
to
01
ro
to
O
00
o
*»•
in
to
0
o
in
i
o
10
in
CM
to
ro
CO
to
o
ro
O
oo
in
O
ro
CM
O
to
1
CM
ro
to
CO
CM
IO
Ol
to
00
Ol
ro
o
to
o
ro
«r
o
ro
in
O
in
CM
«f
CM
to
ro
in
10
o
ro
on
CO
LO
o
§
O
o
CM
1
to
ro
r-
to
in
ro
CM
00
o
0
to
o
ro
<•
O
in
in
o
CO
in
CM
in
ro
in
r--
o
ro
r~
CO
in
O
CO
O
CM
O
ro
*
1
O
n
ro
OO
in
o
to
in
CM
ao
CM
in
to
o
o
in
o
ro
in
o
o
CM
to
CM
in
ro
r-»
o
to
r--
00
m
o
r^
O
O
r~
r"
1
ro
t
en
in
o
to
in
CM
oo
ro
o
to
o
to
in
o
in
o
in
ro
r-.
CM
ro
ro
ro
O
in
00
00
in
o
r*.
in
O
co
i
*•
in
o
CM
r—
IO
,
to
o
CO
ro
o
to
o
CM
in
o
in
0
o
en
ro
r~
CM
ro
ro
on
O
*
to
oo
in
o
§
in
r^
CM
o
VO
o
in
*»•
in
r~
Ol
o
to
ro
to
CM
CM
ro
,_
0
r^
oo
in
O
o
CM
0>
CM
CM
1
in
r—
to
ro
«»•
O
to
on
ro
r~
Ol
o
CO
o
in
in
o
to
*»•
o
in
r«-
o
in
r^
in
CM
in
ro
ro
0
r^
CO
in
O
O
in
f\
CM
r*
CM
r»
i
01
CM
to
*
O
"o
4)
N
C
10
a
3
** s
•*• o
vt *^
o a.
Q. S
B a
in o>
<9 c
O) f
in
4) 3
3
U
236
-------
FIGURE 1
Ht>l» MAGNEHC MAIERIALS EVALUATION
PHOTOMICROGRAPHS OF CROSS-SECTIONEI»
FILTER KED MEDIA PARTICLES
UfirXf'OSF.D
530 IIRS EXPOSURE
MATERIAL N - 100X NOMINAL
£30°C
1000 IIRS EXPOSURE
WExrostD
FIGURE 5
MSB MAGNETIC MATERIALS EVALUATION
PHOTOMICROGRAPHS OF CROSS-SECTIONED
FILTER BED MEDIA PARTICLES
101 IIRS EXPOSURE
JO* HRi EXPOSURE
77« MRS EXPOSURE
MIERIAL P - IOOX NOnlNAL
850°C
237
-------
F iGi-aE *
SCHEMATIC 3UG3A.1! OF THJ ?E'.Ki::-El«» TGS1I SET Vf SS *
5ALAflCj_FC.t rtfCfiSTlC ?<)OPE!»7I
1ectromJcroo»l»nc«
Inert Gas
Inlet
Inert 3*s
Outlet
NOffWLlZEO FORCE RATIO VERSUS EXPOSURE TIME FOR MATERIAL ft
0.5
°'4
°-3
0.2
S
O.ll
HaterUT R Samples
O Exposed at position A. Measured at »°C
A Exposed at position A. Measured at 830»C
200
400
too
800
WOO
Hours of Exposure
238
-------
1.0
S 0.6
8
§ 0.5
I 0.4
I 0.5
2
0.2
0.1
0
FIGURE 8
NORMALIZED FORCE RATIO VERSUS EXPOSURE TINE
FOR MATEftJAL P
200
MATERIAL P SAMPLE EXPOSED AT 810°C
O MEASURED AT 25°C.
A MEASURED AT 830°C
«iOO
HOURS OF EXPOSURE
600
son
FIGURE 9
NOftMAUZED FORCE MI 10 VERSUS MEASUREMENT TEMPERATURE
1.2
1.0
I.
0.4
0.2
O Material P - unexposed
A Material N - onexposed
'-A
100
200
300
400 500
Tenperjture. DC
239
600
700
800
900
1000
-------
FIGURE 10
HACNETICAUV STABILIZED BED FILTRATION SYSTEM FLOW PUU
STACK
TOTAL
PAKTICUUTE
FILTERS
BUST-LADE*
FILTER
KDIA
HOPPER
PAXTICULATE
SAMPLING
STATIC*
FIGURE 11
CUTAWAY VIEW OP MSB FllTSt VESSEL AND JED
c
E
E
toltnel*
lilCtrnill mMf In*
•on Myntlc «t*Hll
(111 tulnlm ttMl •
tlKOMl (00)
240
-------
TEST RESULTS OF A HIGH TEMPERATURE. HIGH PRESSURE ELECTROSTATIC PRECIPITATOR
D. Rugg, G. Rinard, 0. Armstrong, T. Yamamoto, M. Durham;
Denver Research Institute, University of Denver
ABSTRACT
The electrostatic precipitator (ESP) is being considered as a final
gas cleanup device for pressurized fluidized bed combustion (PFBC) combined
cycle power plants. In order to investigate the practical feasibility of
ESP's applied to high temperature, high pressure (HTHP) gas streams, a
pilot scale unit has been developed. This unit has been operated over a
spectrum of gas temperatures, pressures, and dust loadings, which can be
encountered in PFBC systems.
The electrical characteristics for a wire electrode and two electrodes
designed by Research-Cottrell are reported. Flyash from the Curtiss-Wright
PFBC was redispersed in the unit for these tests. The test results are
being used to quantify the performance of HTHP ESP's and should also pro-
vide needed information for other PFBC gas cleanup devices which employ
electrostatic augmentation.
INTRODUCTION
The electrical characteristics of three corona electrodes were meas-
ured under clean and dirty conditions'in the experimental high temperature-
high pressure electrostatic precipitator (HTHP-ESP) located at the Denver
Research Institute's Cherry Creek Field Site facility. Clean voltage-
current (VI) measurements were made on each of the three corona electrodes
at several temperatures and pressures at and near the Curtiss-Wright pres-
surized fluidized bed combustor (PFBC) operating conditions of 640 kPa
(6.4 atm) and 870°C (1600°F). Dust from the Curtiss-Wright PFBC was then
redispursed into the gas stream. VI measurements were again made on the
three corona electrodes at 640 kPa and 870°C.
These electrical characteristics measurement were made to determine if
a region between corona onset and sparkover existed where stable operation
241
-------
of the ESP could be achieved. The effects upon the electrical characteris-
tics of dust on the corona electrode and collector tube were also measured.
Determining the operating field strengths and current densities that could
be achieved with each electrode were the other objectives of the test.
DESCRIPTION OF THE HTHP-ESP FACILITY
The test facility described by Rinard, et al (1981) was designed to
simulate a wide range of PFBC operating conditions for the evaluation of
electrostatic precipitation at high temperatures and pressures. Figure 1
shows a schematic of the test facility.
The HTHP-ESP is capable of operating at temperatures up to 980°C
(1800°F) and pressures up to 1 MPa (10 atm) with a flow rate of 0.074 nvVsec
(156 ACFM) at these conditions. Collector tube electrodes up to 35 cm (14
in.) in diameter can be tested in the unit.
The pressure vessel consists of a multi-sectioned, flanged carbon
steel pressure shell having an overall length of 7.52 m (24.7 ft) and an
outside diameter of 0.91 m (3.0 ft). The hot gases enter near the bottom
and exit near the top of the vessel. Except for the top and bottom sec-
tions of the pressure shell, which are cooled by means of water jackets,
the interior of the shell is lined with a castable refractory thermal in-
sulation 22.9 cm (9 in.) thick. Sufficient insulation and cooling is pro-
vided to maintain a maximum temperature of 110°C (230°F) on all carbon
steel components and welds. The vessel is designed and rated to a pressure
of 1.2 MPa (175 psig) and a temperature of 150°C (300°F). A blanket ther-
mal insulation is used between the shell sections. As a safety precaution,
the exterior of the vessel is coated with a temperature-sensitive paint.
The power supply for the HTHP-ESP is a 150 kV and 50 mA supply with
reversible polarity and is equipped with meters to measure the applied
voltage and current. A high voltage feedthrough of high density alumina
is located on the top of the ESP. Due to considerations involving the pos-
sible electrical breakdown of a dust-coated electrical insulator, a maximum
design temperature of 260°C (500°F) was established for the top vessel sec-
tion housing the high voltage feedthrough. The top section is cooled by
means of fins on the inside of the pressure shell and a water jacket out-
side. Radiation shields on the corona electrode support rod are also pro-
vided to prevent radiant heat from reaching the high voltage feedthrough.
The pressure vessel head also contains an instrumentation feedthrough and
three mechanical feedthroughs for supporting the rapping rods from which the
ESP collector tube is suspended.
The pressure vessel bottom section is funnel shaped and forms the hop-
per of the ESP. The hopper is emptied after a test series when the entire
unit is at ambient pressure and temperature. The bottom section also con-
tains a corona electrode stabilizer bar. This alumina bar is cooled by
physical connection to the water cooled shell section.
242
-------
IOOA
CURRENT
MEASUREMENT
ESP INLET
TEMPERATURE
BURNER
T/C
FUEL-
COMPRESSED
FLUIDIZED
BED
INJECTOR
DISCHARGE
TEMPERATURE
THROTTLE
VALVE
SILENCER
'ORIFICE
METER
ELECTRODE
T0 OUTLET
SAMPUNG
-ISOLATED
COLLECTOR
TUBE
TO INLET
SAMPLING
TRAIN
STABILIZER
BAR
AIR
FIGURE 1. SCHEMATIC DIAGRAM OF HTHP ESP SYSTEM
243
-------
The collector tube used during the tests was 30.5 cm (12 in.) in dia-
meter. A section 2.1 m (6.9 ft.) long between inlet and outlet is electri-
cally isolated from ground. A 100 ohm resistor is connected between the
isolated tube section and ground. The corona current to the 2.0 m? (21.7
ft^) isolated section is determined by measuring the voltage across the 100
ohm resistor.
Flyash can be injected into the HTHP-ESP by means of a specially de-
signed redispersion system. A screw feeder is used to meter the dust into
a fluidized bed of glass beads where the dust is redispersed in a pressuri-
zed air stream. The screw feeder and fluidized bed are housed in a pressure
vessel equipped with a quick disconnect flange. The redispersed dust flows
from this pressure vessel to an ESP pressure vessel injection port located
just downstream of the burner. A baffle downstream of the injection port
enhances mixing. The dust mass loading in the test stream can be varied
from 0.23 to 11.5 g/Nm3 (0.1 to 5 grains/scf).
A monitoring and safety interlock system has been designed to maintain
safe operating conditions, assure the collection of all temperature data,
and provide for an alarm and emergency shutdown should any of the critical
parameters of the HTHP-ESP facility exceed their specified limits.
THE CORONA ELECTRODE TESTS
In the corona electrode tests, the flow rate in the precipitator was
maintained at 0.078 m^/sec (165 acfm) and the velocity through the collec-
tor tube was 1.07 m/sec (3.5 ft/sec). The collector tube had a specific
collection area of 25.7 sec/m (130 ft^/kacfm). The inlet gas temperature
as well as the discharge gas temperature were measured and the gas tempera-
ture in the ESP collector tube was determined from these values.
Sketches of three corona electrodes that were tested are shown in
Figure 2. The first corona electrode was a smooth wire 7.9/mm (5/16 in.) in
diameter. The second corona electrode was a scalloped electrode designed
by Research Cottrell (R-C) for possible use in an HTHP-ESP at the Curtiss-
Wright PFBC facility (Feldman (1977)). Scalloped fins were attached to a
2.5 cm (1 in.) tube. The 2.5 cm (1 in.) scalloped fins produced an elec-
trode 7.6 cm (3 in.) in diameter. The third or star electrode which was
also designed by R-C, is similar to the scalloped electrode except that the
2.5 cm (1 in.) fins which are attached to the 2.5 cm (1 in-) tube are not
scalloped. The latter two corona electrodes were designed to be rigid so
that if the electrode were suspended by a rigid mounting at the top, it
would not require a stabilizer bar at its lower end to prevent swinging
when high voltage is applied. However, since the stabilizer bar was avail-
able in the unit, it was used to insure that the electrodes were centered.
RESULTS AND DISCUSSIONS
The operating conditions of the HTHP-ESP during the determination of
the electrical conditions of the clean corona electrodes are shown in Table
244
-------
^r
WIRE ELECTRODE
SCALLOPED
ELECTRODE
STAR
ELECTRODE
FIGURE 2. CORONA ELECTRODES
245
-------
1. These test conditions were centered around 870°C (1600°F) and 640 kPA
which produces a relative gas density of 1.65. First the temperature was
held constant at 650°C (1200°F) while the pressure was changed from 510 to
640 kPa to determine the effects of pressure. Then the pressure was held
constant at 640 kPa and the temperature was set at three additional values
to determine the effects of temperature.
TABLE 1. TEST CONDITIONS FOR CLEAN CORONA ELECTRODES
Relative
Test No. Pressure Temperature Gas Density
1 510 kPa 650°C (1200°F) 1.65
2 640 kPa 650°C (1200°F) 2.04
3 640 kPa 845°C (1550°F) 1.69
4 640 kPa 870°C (1600°F) 1.65
5 640 kPa 900°C (1650°F) 1.61
Figure 3 shows the clean VI characteristics of the wire electrode with
negative corona at a constant pressure of 640 kPa. The scales for current
density over the isolated section of the collector tube and the average
field strengths between corona electrode and collector tube are also shown.
Field strengths above 9 kV/cm and current densities up to 0.70 uA/cm^ were
recorded. The maximum voltage was limited by the power supply and not by
sparkover. The three curves show that at a constant voltage, an increase
in temperature produces an increase in current. Figure 4 shows the clean
VI characteristics of the wire electrode with positive corona at two tem-
peratures. With positive corona the current increase due to increased tem-
perature was less than with negative corona, and sparkover occurred at lower
corona currents and lower corona voltages.
The VI characteristics of the scalloped electrode with both negative
and positive corona at 660°C (1220°F) are shown in Figure 5. The diameter
of the corona electrode was assumed to be 7.6 cm (3 in.) in determining
average field strength. The current density for the scalloped electrode
was about twice as high as with the wire electrode. With positive corona
the curves are sparkover limited and with negative corona the curves were
power supply limited. For both positive and negative corona, at a constant
voltage, an increase in pressure caused a decrease in current.
Figure 6 shows the effects of temperature on the same electrode with
negative corona. At the higher temperatures, there was a measurable corona
current at low corona voltages. Figure 7 is the same as Figure 6 except
that the corona was positive. Changing the temperature from 845°C (1550°F)
to 905°C (1660°F) did not change the positive corona VI characteristics any
significant amount.
246
-------
247
-------
a
£
o
248
-------
'A1ISN3Q iN3ManO
~T
3
UJ
o:
h-
VI
z
o
o
s
UJ
o
i
i—
o
UJ
_J
UJ
s
249
-------
Figure 8 shows VI curves for the same relative gas density at two tem-
peratures. For negative corona, the current increased as the temperature
increased. For positive corona, the VI curves at constant relative gas
density are essentially independent of temperature.
Some of the VI curves for the star electrode are shown in Figure 9.
In all cases the current from the star electrode was less than the current
from the scalloped electrode. However, the data showed the same dependence
upon temperature, pressure, and relative gas density.
A comparison of the clean VI characteristics of the three electrodes
at 870°C (1600°F) is shown in Figure 10. The curves in Figure 10 show that
the current from the scalloped electrode was considerably larger than the
current from either the wire or star electrode.
After the clean VI measurements were completed, dust was injected into
the HTHP-ESP. The dust which was used in these tests was from the second
cyclone of the Curtiss-Wright PFBC system. The elemental analysis of the
ash is presented in Table 2 and the particle size distribution is shown in
Table 3.
TABLE 2. ELEMENTAL ANALYSIS OF C-W ASH
Si02
A1203
Ti02
Fe203
CaO
MgO
Na20
K20
P205
S03
TOTAL
25.46
12.72
0.36
15.96
17.08
11.22
0.21
0.77
0.10
18.12
102.00
With the ESP dirty, the measurements at 870°C (1600°F) and 640 kPa
were repeated with negative corona. These measurements are shown in Figure
11 for both dust off and dust on conditions. The reduction in current due
to particle space charge when the dust is turned on is shown. The size of
particles in an ESP in an actual PFBC operating system would be much small-
er than in the dust used for these tests. Large particles of the size
shown in Table 3 would not create large space charge between corona wire
and collector tube. Therefore the ash from the Curtiss-Wright second cyc-
lone is being ground in a fluid mill to reduce particle size. The mass
250
-------
(jiiio/trf) A1ISN3Q
I
o
I
1
251
-------
fe
VJ
s
O
UJ
§
2«i3/»i(
1N3HHTO
i
m
UJ
5
252
-------
mean diameter is being reduced to 2 or 3 microns and the maximum size is 10
to 12 microns. The smaller size particles will be used in future tests in
order to more realistically measure the effects of particle space charge.
TABLE 3. PARTICLE SIZE DISTRIBUTION OF C-W ASH
MICRON Above Stated Micron
SCREEN ANALYSIS (By Weight)
420 0.00
210 0.00
105 0.00
45 6.38
COULTER COUNTER
40.30 6.88
32.00 7.38
25.40 7.98
20.20 13.98
16.00 25.16
12.70 39.96
10.08 55.58
8.00 70.08
6.35 82.60
5.04 91.76
4.00 95.05
3.17 98.10
2.52 98.40
2.00 98.78
1.59 99.25
1.26 99.50
1.00 100.00
The current under dirty conditions increased for all three electrodes.
This increase in current under dirty conditions also appeared in the data
of Brown and Walker (1971) and Shale and Fasching (1969). Shale and Fasch-
ing were making tests at 800°C (1470°F) and 650 kPa. Injection of dust in-
to the system lowered the corona voltage from about 42 kV to 37 kV for the
same corona current. Under clean conditions, Brown and Walker operating at
900°C (1650°F) and 800 kPa, achieved voltages of 75 kV. Under dirty con-
ditions, the voltage was only 44.5 kV for the same current level. In the
measurements by DRI on the three different corona electrodes at 870°C
(1600°F) and 640 kPa, the corona voltage was lowered by almost a factor of
2 when the system was dirty. Additional tests are planned which may ex-
plain this effect of dust on the electrodes.
Corona current density as a function of average field strength is
shown for each of the three electrodes under clean and dirty conditions in
Figure 12. The highest field strengths were achieved with the star elec-
253
-------
trode. The smooth corona wire produced the next highest fields and the
scalloped electrode the lowest field strengths.
CONCLUSIONS
The measurements of electrical characteristics showed that there was a
voltage range between corona onset and sparkover where stable ESP operation
can be achieved with each of three corona electrodes using either positive
or negative corona within the range of temperatures and pressures tested.
With positive corona, sparkover occurred at slightly lower voltages and
lower current densities than with negative corona. Also, with positive
corona, the VI characteristics appear to depend upon relative gas density
while with negative corona, current increases with temperature for a con-
stant gas density.
Changing the design of the corona electrode changes the VI character-
istics of the ESP. However, the introduction of dust into the precipitator
modified the VI characteristics to a greater extent than changing the de-
sign of the corona electrode. Additional measurements will be made in an
attempt to explain the increase in current that occurs when dust is injec-
ted.
With the precipitator dirty, average field strengths up to 8 kV/cm
and current densities larger than 1.5y A/cm? were achieved.
The next tests in the evaluation of the HTHP-ESP as a final gas clean-
up device for PFBC power plants will include determining the collection
efficiency as a function of field strength and current density. Rapping
reentrainment will also be studied.
The work described in this paper was not funded
by the U.S. Environmental Protection Agency and
therefore the contents do not necessarily re-
flect the views of the Agency and no official
.endorsement should be inferred.
REFERENCES
Brown, R.F. and A.B. Walker (1971) "Feasibility Demonstration of Electro-
static Precipitation at 1700°F), Journal of the Air Pollution Control
Association 21:615-20.
Feldman, P.L. (1977) "High Temperature, High Pressure Electrostatic Pre-
cipitator", EPA/ERDA Symposium on High Temperature/Pressure Particu-
late Control, Washington, D.C.
Rinard, G., M. Durham, J. Armstrong, and R. Gyepes (1981) "The DRI High
Temperature/High Pressure Electrostatic Precipitator Test Facility",
Proceedings: High Temperature, High Pressure Particulate and Alkali
254
-------
Control in Coal Combustion Process Streams, DOE/METC Contractor's
Meeting, Morgantown, WV
Shale, C.C. and G.E. Fashing (1969) "Operating Characteristics of a High-
Temperature Electrostatic Precipitator", Bureau of Mines Report of In-
vestigations RI 7276.
255
-------
COAL-ASH DEPOSITION IN A HIGH TEMPERATURE CYCLONE
by
K. C. Tsao
University of Wisconsin-Milwaukee
Milwaukee, Wisconsin
A. Rehmat and D. M. Mason
Institute of Gas Technology
Chicago, Illinois
ABSTRACT
Experimental evidence indicated that the increase of particulate re-
moval efficiency in a high temperature agglomerating cyclone is hampered by
the formation of cyclone wall deposits. The cyclone collection efficiency
has been observed in a laboratory hot cyclone to meet the designated per-
formance when the temperature of the dust ladden gas is increased to near
its coal-ash fusion temperature. Factors that are affecting the wall deposi-
tion process are examined and estimation of the relative importance of the
operating parameters are presented. A simple mathematical model for the
wall deposition mechanism is tentatively proposed. Experimental results on
the occurrence or absence of wall deposits will be discussed.
The work described in this paper was not funded by the U.S. Environ-
mental Protection Agency and therefore the contents do not necessarily re-
flect the views of the Agency and no official endorsement should be inferred.
256
-------
INTRODUCTION
In the course of studying the performance of a high temperature agglom-
erating cyclone, the deposition and altering of flow stream at the cyclone
jet impingement surface have caused the cyclone in-operational while the
cyclone displayed a collection efficiency of 92% and higher. Attempts were
made to analyze the basic principles of adhesion at low temperature (1 -11)
and at high temperature (12-14), yet much remains to be investigated. Char-
acterization of the adhesion mechanism of coal-ash near the softening/melt-
ing point .though initiated (12, 13), of coal-ash is far from fully under-
stood.
Development of coal conversion reactors, hot gas cleaning equipment,
and desposition on gas turbine blades and other energy conversion apparatus
usually encounters the problems of clogging, erosion and ash agglomeration/
adhesion of submicron-sized particles. These difficulties arise mostly from
the high temperature and severe environment where those equipments are ex-
posed. The products of coal gas reactors are in a form of raw, low to medium
heating value fuel for industrial plant or directly for power generation.
The gas stream contains a complex element but also includes particles of ash
char and dust of added absorbent materials. The physical and chemical prop-
erties of those small and submicron particles usually deposited along its
flow path. Softening, melting and even evaporation would accelerate the
deposition process when the device is operated near the fusion temperature
of the coals.
This study was undertaken to investigate the conditions affecting depos-
ition of coal ash in the high temperature cyclone of a fluidized bed coal
gasifier, in the process the cyclone removes char from the product gas and
returns it to the bed for completion of its gasification. Some initial but
important experimental results of the coal ash deposition are reported here.
This study also presents some initial experimental observations on ad-
hesion phenomena. A simple but effective mathematical model is tentatively
proposed to analyze the relative effect of some of the factors which may en-
hance the wall deposition. Among the parameters examined, it was found that
the momentum of the particle, the gas and the wall temperatures, and the
pseudo-molten layer thickness are the important operating variables affecting
the adhesion phenomena.
EXPERIMENTAL OBSERVATION
The experimental high temperature agglomerating cyclone (14) demon-
strated previously as an effective particulate removal device is adopted for
deposition study. The experimental set-up consisted of a controllable high
temperature gas burner, an experimental cyclone made of 314 stainless steel,
an external electrical heater, a coal ash feeder, an exhaust gas analyzer
and a particulate sampler. A photograph of the experimental facilities is
257
-------
shown in Figure 1. The inlet section of the cyclone is double jacketed,
thus allows the control of cyclone wall temperature in the range between 150
to 1600°F. The burner section was operated to obtain a reducing atmosphere
as evidenced by the presence of carbon monoxide and the absence of oxygen in
the exhaust gases. The dust sample was prepared by comminuation and sieving
of ash deposits laid down in a pilot plant run. The composition of feed
dust, Table 1, is iron rich alumina silicates which were selectively deposi-
ted from the whole ash containing on the average of 20 weight percent iron
oxide as Fe203-
Table 1 Composition of Dust Sample
Composition Wt. % of Ash
Si02 44.0
A1203 15.2
Fe203 35.2
Ti02 0.85
CaO 2.26
MgQ 0.81
Na20 0.36
K20 1.76
S03 0.47
Total 100.9
A 0-38 urn fraction of the dust was fed in most of the laboratory cyclone
tests results which are reported here; and a 10-38 ym fraction was used in
some earlier tests.
During the experimental runs, a cohesive deposit appeared only at the
inlet section of cyclone where entering gas impinges at the wall. The light
grayish deposit a.dheres strongly to the metal surface and cannot be brushed
away, shook off and has to be mechanically scraped away for weighing. A
photo of such deposit is shown in Figure 2. While the cause of formation is
not truly understood, the occurrence or the absence of these impingement de-
posits have been observed and related to the cyclone operating gas and wall
temperatures as shown in Figure 3. The data of forty-three tests covering
the gas temperatures between 1650 and 2000°F and wall temperatures from 200
to 1500°F indicates, Figure 3, that a tentative demarcation line for deposi-
tion appears plausible. It suggests that the deposition at the jet impinge-
ment surface opposing gas inlet can be prevented if proper gas-wall tempera-
ture relationship is being observed. That is, the higher the particle lad-
den gas temperature, the lower must be the cyclone wall temperature. A refer-
ence line for equal gas and wall temperature is added for comparison though
experimental data on where deposition may occur has not been established.
Note that the experimental points as indicated by the half darkened square
deposit an extremely thin film of deposits, but no accumulation, was ob-
258
-------
Figure 1 Experimental set up
Figure 2 Sample impingement deposit
259
-------
2000
1800
1600
1400
1200
ff
1000
UJ
800
600
400
200
1500
OUST ATMOS
E 38-45 REDUCING O •
E 10-38 REDUCING A A
EQUAL GAS AND WALL TEMPERATURE
PROPOSED BORDER
LINE DEPOSITION
A
1600 1700 1800 1900 2000
GAS TEMPERATURE, *F
2100
2200
Figure 3 Sample iBpingement deposit
260
-------
oc
HI
Q.
UJ
O
35
30
EXPERIMENTAL DATA RUN 18 TO 28
DUST SAMPLE, "D"
DUST SIZE <38[i
CC
3
if)
25
is
ii
20
«
uj<0
CO
O
0_
LU
Q
15
10
1400
1500
1600
1700
1800
1900
GAS TEMPERATURE, °F
Figure 4 Bar graph showing deposition
rate for 'D1 dust
261
-------
LU
O
U.
CC •
21
2°>
O.
LLlS
H <
-------
served at the termination of 30 minute test runs. Two types of dust of
fractional size below 38 pm and similar chemical composition were employed,
and the data appears consistent.
Deposits at the cyclone impingement surface were collected and weighed
through a microbalance. The rate of deposition per gram of dust fed is
shown in Figures 4 and 5. Figure 4 shows the deposition rate for dust parti-
cles of less than 38 pm sieved through No. 400 mesh screen. Figure 5 was
plotted for particle sizes larger than 10 ytn but smaller than 38 urn. It is
not clear whether the deposition rate, Figure 4, of one order of magnitude
greater is due to the effect of gas velocity. The average gas velocity at
cyclone inlet is 26.3 to 37.6% higher in Run No. 18 to 28.
ANALYTICAL CONSIDERATION
The study of particle wall adhesion near the ash fusion temperature was
attempted in a laboratory hot cyclone for enhancing the particulate removal
efficiency of submicron particles. The particle trajectory under various
controlling force in conjunction with coal ash physica7-chemical properties
will determine whether they will collide and stay together as agglomerates
or whether these strike at wall surface will adhere as deposits. Consider
the collision mechanism of a particle of radius R, with molten layer thick-
ness <5, strikes at a wall, Figure 6. The force of separation that would
cause rebounding are the initial momentum of the particle prior to collision
and the thermal gradient across the gas film between the particle and the
wall surface. Forces which would foster adhesion are the surface tension of
molten layer that the particle must dash through and the Van der Waals mole-
cular force should the interfacial distance reach 100 A during collision.
Other factors such as the electrostatic and the universal attraction are con-
sidered to be negligible.
The liquid solid bridge formed during the collision and subsequent de-
tachment from a plane surface as shown in Figure 6, is of great importance
in evaluating the surface tension and the viscous drag effect. Precise de-
termination of the liquid-solid interfacial contour is difficult and not
available in open literature. We have assumed that the force due to surface
tension, F ., following reference (12, 15) is
Fst =4™ [R-(R-6)2]1/2 (1)
where R is the radius of the semi-molten particle, 6, the molten layer thick-
ness enclosing the solid particle core, and a, the surface tension of the
molten layer per unit length.
The effect of Van der Waal's attractive force is siggificant if the
interfacial particle-wall distance falls within 4 to 104 A during collision.
According to Rumpf (16), the Van der Waal's force effect F . can be approxi-
mated by. vaw
263
-------
STAGNANTGAS
FILM « IOR
RADIUS R
SOLID CORE
MOLTEN
LAYER
THICKNESS,
VELOCITY
THERMAL
EFFECT
\
WALL
BEFORE COLLIDING
INTER SURFACE
DISTANCE
DURING
/ COLLISION
I
REBOUNDING
VELOCITY
MOLECULAR ATTRACTION -
SURFACE TENSION -
MOMENTUM FOR DETACHMENT «-
THERMAL EFFECT TO INCREASE *•
MOMENTUM FOR REBOUNDING
WALL
PARTICLE
/. DETACHMENT
7
Figure 6 Schematic diagram of particle-wall collision
264
-------
C = X 'Iyv / **
vdw g (H + 2 )7
o
where hw is the Lifshitz constant, H, the distance between the particle sur-
face and wall,0and Z is the limiting molecular separation length typically
taken to be 4 A.
Detachment of a particle after striking at the relatively cooler wall
is accelerated due to the effect of a thermal force between the particle and
wall. The thermal force increases the kinetic energy of a particle toward
the plane wall prior to impaction. Epstein (17) has estimated the thermal
force, F., as
F = _ 9 "R^ (£L) (3)
t f^f AX
V'S'
where y the viscosity of gas, p , the density of gas, T, the gas temperature,
K and Kf, the thermoconductiviiies of particle and gas respectively, and
(fiT/Ax), the thermal gradient within the thermal boundary layer thickness.
The cause for rebounding and detachment after collision is the initial
momentum carried with the particle. The inertia force, F, associated with
the momentum of the particle is the rate of change of the product of parti-
cle mass, m, and velocity, V, i.e.,
Assume that the approaching velocity of a particle to be fixed in the ther-
mal boundary region of gas film next to the plane wall, the rate change of
velocity was computed as
where Ax is the thermal boundary layer thickness next to the wall and was
judicially chosen in the subsequently computation to be five particle diam-
eters.
Summing up all the effectsand take the ratio of forces that enhance ad
hesion, F ., to the forces that cause separation and rebounding, F._, we
have: ad sp
(6)
Fad
FSP
.
265
-------
100.0
50.0
10.0
5.0
Q.
CO
1.0
0.5
0.1
0.05
\
\
\
d-, = 1.0 MICRON
6-|=0.01%
AX=5.0 MICRON
MOLECULAR DISTANCE 104 A°
— — —MOLECULAR DISTANCE 4 A°
TEMPERATURE DIFFERENCE
AT=0.0°C
= 500°C
0.0
0.2
0.4
0.6
0.8
1.0
Vv1/vg
Figure 7 Effect of velocity ratio on particle
wall adhesion for 1 micron particles
at VG of 36.0 m/sec.
266
-------
Equation (6) is used for estimating the relative importance of operating
parameters on wall impingement deposition in a high temperature cyclone.
DISCUSSION AND CONCLUSIONS
The adhesion process between a particle and a plane wall has been des-
cribed physically, Figure 6, and mathematically, Equation (6). Particle-
wall deposition would sustain if F ,/F is equal to or greater than unity.
The ratio of forces that cause adhlsion^to that of separation and detach-
ment is computed and plotted versus the velocity ratio, V , the velocity of
the particle toward the wall to that of gas velocity at cyclone inlet.
Properties of particle (18) and gas stream used in the computation are: ?
hw = 5.0 ev, T = 1000°C, V = 36 m/sec, p = 0.004 gm/cm3, p =0.65x10
poise, p = 0.75 gm/cm^, a3d a = 150 dyme/'cm. ^
Assumptions were made on those plots that the particle is moving; 1)
toward the surface, 2) at same speed before and after striking at the sur-
face and 3) at a constant speed when the particle is dashing through the
stagnant gas film, AX, next to the wall. Figures 7 to 9 demonstrate the
effects of the particle momentum, the thermal gradient, the molten layer
thickness enclosing the particle and the size of particles on wall deposi-
tion.
Figure 7 shows the effect of velocity ratio and the thermal gradient on
particle wall adhesion for 1 urn diameter particle. The net result of in-
creasing the temperature difference from 0°C to 500°C between the particle
and wall increases the initial momentum of the particle toward the cooler
wall. The larger the thermal gradient, the greater will be the force of re-
bounding. This is seen by comparing the two solid curves that for FaH/F=l>
the detachment of a particle would occur at a velocity ratio of 0.12 and p
0.27 when the temperature difference is raised from 0 to 500°C. This plot
seems to agree qualitatively with observed experimental cyclone data that
deposition may be prevented by cooling the wall surface, Figure 3. The in-
tersurface distance between the particle and wall during collision appears
apparent by comparing the curves of molecular distances of 4 and 104 /\. The
net effect is that deposition would occur at velocity ratio of 0.12 rather0
than at 0.7 as particle-wall intersurface distance is varied from 104 to 4A.
Figure 8 shows the effect of molten layer thickness enclosing a particle
on wall deposition for a 10 ym diameter particle. It is realized that while
the relative importance of thermal gradient diminishes for larger size parti-
cles, the effect of molten layer thickness seems most eminent. The critical
velocity ratio at which F ./F is unity, is increased almost threefold, from
0.08 to 0.26 as the molten layer thickness is increased from 0.01% to 1%.
The negative slope of border line deposition, Figure 3, seems to indicate the
concept.
In conclusion the adhesion and rebounding of particles upon a plane wall
is found to be function of gas and wall temperature. A simple but plausible
267
-------
50.0
Q
<
10.0
5.0
MOLECULAR DISTANCE 104 A°
MOLECULAR DISTANCE 4 A°
d-, = 10.0 MICRON
AX=50.0 MICRON
AT=(0.0-500) °C
QL
-------
mathematical model is tentatively proposed in assessing the coal-ash deposi-
tion process as observed in a laboratory high temperature cyclone. The ef-
fect of particle viscous drag in terms of hydrodynamic wetting which was not
incorporated in this study, remains to be analyzed. Proper selection of the
cyclone operating parameters namely, gas and wall temperatures, and particle-
gas velocity seems to be effective in minimizing the coal-ash particle depos-
ition.
REFERENCES
1. Browne, L.W.B., "Deposition of Particle on Rough Surfaces During Turbu-
lent Gas-Flow in a Pipe," Atmospheric Environmental, Vol. 8, 1974,
pp. 801 - 816.
2. Corn, M., "Adhesion of Particles," Chap. XI in Aerosol Sciences, edited
by Davies, Academic Press, 1966, pp. 359 - 392.
3. Kneen, T. and Strauss, W., "Deposition of Dust. From Turbulent Gas
Stream," Atmosphere Environmental, Vol. 3, 1969, pp. 55 - 67.
4. Kordecki, M. C. and Orr, C., A.M.A. Archives of Environmental Health,
Vol . 7, No. 7, 1970.
5. Gardwer, G. C., "Deposition of Particles From a Gas Flowing Parallel to
a Surface," International Journal of Multiphases Flow, Vol. 2, 1975,
pp. 213 - 218.
6. Gillespie, T., "On the Adhesion of Drops and Particles on Impact at Solid
Surfaces I and II," Journal of Colloid Science, Vol. 10, 1955,
pp. 266 - 298.
7. Hocking, L.M., "The Collision Efficiency of Small Drops," Quarterly
Journal of Research for Society, Vol. 85, No. 44, 1959.
8. Lin, S. M., "Particle Deposition Due to Thermal Force in a Tube,"
Applied Scientific Research, Vol. 32, 1976, pp. 637 - 648.
9. Rouhiainin, P.O. and Stachiewicz, J.W., "On The Deposition of Small
Particles From Turbulent Streams," Journal of Heat Transfer, ASME
Transactions, 1970, pp. 169 - 177.
10. Wang, C. S., Deabor, J. J. and Lin, S.P., "Effect of Fluid Inertia on
Particle Collection," Physical Fluids, Vol. 21, No. 12, 1978,
pp. 2365 - 2366.
11. Soo, S.L., "Particle-Gas-Surface Interaction in Collection Devices,
International Journal of Multiphase Flow, Vol. 7, 1973, pp. 89 - 101.
269
-------
References (con't)
12. Tsao, K. C. and Jen, C. 0., "Coal-Ash Agglomeration Mechanism and its
Application in High Temperature Cyclones," Separation Science and
Technology, Vol. 15, No. 3, 1980, pp. 263 - 276.
13. Tsao, K. C., Tabrizi, H., Rehmat, A. and Mason, D., "Coal-Ash Agglomera-
tion Mechanism in a High Temperature Cyclone," Paper 82-WA/HT-29,
Annual Meeting American Society of Mechanical Engineers, 1982,
Phoenix, Arizona.
14. Tsao, K. C. et. al., "Particulate Collection in a High Temperature
Cyclone," Proceedings - 2nd Int'l. Symposium on Transfer and Utiliza-
tion of Particulate Control Technology, July, 1979, Denver, CO.
Vol. 4, 1980, pp. 14 - 25.
15. Zimon, A. D., Adhesion of Dust and Power, Plenum Press, New York, 1969.
16. Rumpf, H., Particle Adhesion, Chapter 7. Agglomeration 77, pp. 97 -129
1977.
17. Epstein, P., Z. Pysik, Vol. 54, p. 537, 1929.
18. U.S. Department of Energy, "Coal Conversion Systems Technical Data
Book," HCP/T2286-01, Washington, U.S. Gov. Print. Office, 1978.
270
-------
DUST FILTRATION USING CERAMIC FIBER FILTER MEDIA
— A STATE-OF-THE-ART SUMMARY —
by: R. Chang, J. Sawyer, W. Kuby, M. Shackleton
Acurex Corporation
Energy & Environmental Division
485 Clyde avenue
Mountain View, CA 94042
0. J. Tassicker, S. Drenker
Electric Power Research Institute
3412 Hillview Avenue
Palo Alto, CA 94303
ABSTRACT
Filter media suitable for use at temperatures of >1,000°F, employing
ceramic fibers in their construction, have been under development for several
years. These filter media are intended for application in the development of
energy production processes such as pressurized fluidized bed combustion
(PFBC), but will also be suitable for many diverse industrial processes.
Ceramic media development work to date has shown significant progress toward
achievement of a commercially viable high temperature filter. Tests have
shown that at high temperature, fine particles can be collected efficiently
and pressure drop can be controlled using pulse cleaning. Accelerated
durability tests produce promise for long filter life. More work is needed
in durability testing to detect application related probelms and build the
data base needed to move this important product development to
commericalization.
271
-------
INTRODUCTION
The initial development of high temperature filters were made in
response to an identified need for hot gas cleaning devices in advanced coal
conversion processes. One approach involved the direct combustion of coal in
a pressurized, fluidized bed and generating electricity by expanding the hot
flue gas through a gas turbine. Commercially proven techniques exist to
remove the particulates from the hot gas stream before passing through the
turbine. But to work effectively, they require that the pressure or
temperature be lowered, resulting in reduced energy efficiency. Under
Department of Energy (DOE) and Electric Power Research Institute (EPRI)
sponsorships, several hot gas cleanup techniques are being investigated for
direct particulate removal at temperatures up to 1,700°F and 10 atm. These
include electrocyclones, granular bed filters, electrostatic precipitators,
and ceramic filters.
With the advancement of these high-temperature particulate removal
devices, potential markets unrelated to advanced coal conversion processes
are emerging. For these markets, high-temperature particulate removal from a
gas stream offers promise for more efficient processes, waste heat recovery
and product recovery. Some examples are given in Table 1.
TABLE 1. EXAMPLES OF HIGH TEMPERATURE FILTER APPLICATIONS
Fluidized bed combustion — Turbine blade protection
Shale oil retort vapor — Vapor phase particulate removal
Wood/peat gasifiers — Particulate Removal
Catalytic cracking — Product Recovery, "expander" protection
Silicone processing — Silica dust removal in chlorosilane gas
Iron and steel industry — Waste heat recovery
In shale retorting, for example, vapor phase particulate removal of
retorting fines would produce a clean shale oil stream requiring very little
liquid phase solid removal which is difficult and expensive. In fluidized
catalytic cracking, the catalyst is recycled using cyclones in a series. In
some cases, the hot gas is then expanded across a heavy turbine called an
"expander" for energy recovery. An efficient particulate removal device
could offer better product recovery and extended turbine life.
272
-------
HIGH TEMPERATURE CERAMIC FIBER FILTERS
Fabric filters have long been used successfully for high efficiency
particulate removal from gas streams at temperatures below 500°F. The
temperature limits of the filter can be extended by the use of materials
capable of withstanding higher temperatures such as metallic or ceramic
media. Two types of ceramic filters are currently under development; a
filter with a felted media consisting of a nonwoven ceramic mat sandwiched
between retaining screens and a filter woven from ceramic yarn. Some
examples of potential ceramic filter media and their temperature capabilities
are given in Table 2. Most of the media listed are commercially available
while a few are still in the experimental and development stage. Fiberglass
filters at present are limited to temperatures less than 550°F mainly because
of the temperature limitations of the coating. Various methods are being
explored to improve the temperture resistance of the coating and the abrasion
resistance of uncoated fiberglass. Newer, stronger materials such as
zirconia fibers are also in the development stage.
Both felted and woven filters have their advantages and shortcomings.
A comparison is given in Table 3.
TABLE 2. EXAMPLES OF CERAMIC FIBER FILTER MEDIA
Felted Woven
Saffil alumina (ICI) Nextel (3M)
95% AL203, 5% S102 62% AL203, 14% B203, 24% S±
3,000°F 2,600°F
Kaowool (B&W) Astroquartz (J. P. Stevens)
47% AL203, 53% S102 99.9% S102
2,600°F 2,000°F
Fiberfrax (carborundum) Modified fiberglass
48% AL203, 52% S102 1,200°F
2,300^ developing
AB-312 (3M) Zirconia
experimental developing
273
-------
TABLE 3. COMPARISONS OF WOVEN AND FELTED CERAMIC FILTERS
Felted Woven
Very high collection efficiencies High collection efficiencies
>99.9% >99.9%
High face velocity operations Low face velocity operations
possible. Up to 20 ft/min. 1 to 6 ft/min
Requires relatively high cleaning Relatively low energy cleaning
energy required
Generally pulse cleaned Can be pulse cleaned, mechanically
shaken, or reverse air cleaned
In general, the felted filters have higher collection efficiencies and
can operate at higher face velocities than woven filters so that the size of
the overall filter unit can be reduced. However, they are more difficult to
clean and usually require higher cleaning energy compared to woven filters.
In pulse cleaned units, the filter lengths are also limited to less than
15 feet. At Acurex there are various programs to explore and develop both
types of filters for DOE, EPRI, and industrial customers.
SUMMARY OF SOME RECENT RESULTS
ACUREX BENCH-SCALE TEST RESULTS WITH FELTED FILTERS
Under DOE sponsorship, a bench scale test program was undertaken to
evaluate the felted ceramic filter concept. A schematic of the,test
facilities is given in Figure 1. The system is capable of operating at
temperatures up to 1,500°F and pressures up to 10 atm with one filter. The
types of dusts used were a variety of redispersed PFBC flyash injected via a
turntable dust feeder. The mass median diameter of the various flyash used
ranged from 5 to 20 ym. Cleaning was initiated on either a fixed time
interval basis or a set pressure drop across the filter using a
short-duration, high-pressure pulse jet of air. The duration was generally
kept at 0.05 to O.ls. Overall mass collection efficiency was determined by
measuring the amount of dust penetrating the filters versus the amount of
dust fed. To determine the amount of dust penetration, a total filter made
of a high-efficiency fiberglass mat was used downstream of the filter unit to
collect the penetrating dust. A total of about 1,000 hours of test time have
been accumulated under various conditions and a summary of the test results
are given in Table 4.
274
-------
CO LU
CO CO
ULJ (0
CC UJ
OL >
u
•H
4J
rt
e
-------
TABLE 4. BENCH SCALE TEST SUMMARY
Filter dimensions: 3 to 6 in. diameter, 3 to 8 ft. long
System pressure: 1 to 10 atm
System temperature: Ambient to 700°F
Dust loading: 0.1 to 0.5 g/ACF
Air/ cloth ratio: 5 to 18 ft/min
Maximum pressure drop across 5 to 20 in.
dirty filter before cleaning:
Baseline pressure drop: 3 in.
Collection efficiency: 99.89 to 99.999 percent
Having established high collection efficiency and good cleanability, the
durability of the filter was tested by subjecting the filter to a series of
rapid pulses in a dusty environment. A summary of the results is provided
in Table 5.
At the end of the 100,000 cleaning cycles (equivalent to 4 years of
operation cleaning every 20 minutes), the overall mass collection efficiency
was greater than 99.9 percent.
SCALEUP TESTING AT WESTINGHOUSE
Further evaluation of the filters were conducted on a larger scale
520 ACFM at Westinghouse under simulated PFBC conditions (150 psia, 1,500°F)
using redispersed flyash. The unit contained five filters, 8 inches diameter
by 5 feet and was operated under dust loadings of 1 to 2 gr/ACF and face
velocities 8 to 15 ft/min.
Cleanability of the filters was generally good over a total of 77 hours
of test time. An overall baseline pressure drop of about 8 inches ^0 was
achieved from a pressure drop set point for cleaning 15 inches of ^0.
Overall collection efficiencies were initially high (greater than
99.9 percent) but dropped after about 20 hours of testing. Still, the
collection efficiency stayed above 95 percent throughout most of the test.
At the conclusion of the test it was found that in many places along the
filters the saffil filter media had been blown away. It was determined that
a combination of overpulsing and large outer screen openings were the major
276
-------
TABLE 5. SUMMARY OF RAPID PULSE DURABILITY TESTING
Filter dimensions:
System temperature:
System pressure:
Air/ cloth ratio:
Cleaning pressure:
Cleaning duration:
Number of cleaning cycles:
Overall collection efficiency:
3 in. by
700°F
35 psia
5 ft/min
190 psia
50 msec
100,000
5 ft
>99.9 percent
cause of saffil blowout. The pulse duration was set at 250 msec, the
shortest possible setting on the timer but still significantly higher than
normal settings of about 50 msec. This caused excessive high pressure
backflow during pulsing. The large outer screen openings also did not help
to retain the fibers of the mat. The inner screen held up quite well and was
the primary reason why a relatively high collection efficiency could still be
maintained.
SUBPILOT TESTING AT CURTISS-WRIGHT
A subpilot filter unit consisting of 15 felted filters 6-inch diameter
by 8 feet long (Figure 2) was tested at the Curtiss-Wright (Wood Ridge, N.J.)
PFBC facilities.
The PFBC was started with preheat air followed by kerosene addition and
then coal. A final temperature of about 1,460°F and pressure of about
71 psig was achieved at a gas flow of about 940 ACFM to the filter vessel.
The filters operated quite well during the first 70 hours of overall
operation, including 20 hours on coal and then seemed to fail abruptly. In
terms of particulate collection (Figure 3), the filters were >99.6 percent
efficient on the average before failure. The pressure drop characteristics
of the filter (Figure 4) also show that they have good cleanability. At a
set point for cleaning 25 inches of t^O, the baseline AP seem to be
maintained at rather steady levels. During preheat, baseline was kept at 2
to 3 inches 1^0 while during coal feed operations, the baseline was around
7 inches H20.
277
-------
Figure 2. Schematic of pressure vessel for Acurex Ceramic Filter unit
installed at Curtiss Wright.
100
« 80
70
60
99.971;:
Start of
coal feed
Start system
• heatup
/
ft
Start
kerosene
flow
i i
99.9%
99. U
39.6' (20 min sample)
98.2.
90.6
89.56?.
73.-
61.6%
Two chambers
closed .
81.5% (3 and 4 closed)
82.37% (3 open; 4closed)
f ,- 78.7% (4 open, 5 closed)
• /- 73.79% (1 open. 2closed)
67.5% (2 open, 3 closed)
•
.57.52% (chamber 1 closed)
One chamber closed
-Two chambers closed
0 10 20 30 40 50 60 70 30 90 100 110 120 130 140 150
Hours of ooeration
Figure 3. Collection efficiency of filters vs. hours of operation.
278
-------
50
45
40
35
°30
20
15
10
5
0
Maximum AP
Baseline A.P
l
l
l
2 chambers closed
I I i i
10 20 30 40 50 60 70 80 90
Hours of operation
100 110 120 130 140
150
Figure 4. Pressure drop across filter with time of operation.
A detailed examination of the filters and the data indicated that the
filter inner screen failed abruptly about 70 hours into operation. The
failure occurred simultaneously with a disturbance to the PFBC which caused a
surge in airflow and perhaps fuelflow. Failure also occured when the
temperature encountered by the filter was the highest (1,460°F) since the
start of operation. Failure was therefore probably related to the
disturbance or the high temperature or both. One possibility is that the
ceramic filter is clamped at both ends of the metal cage, consequently, the
metal elongates with temperature, the filter is streched and finally tears.
Another possibility is that hot spots on the filter caused by fuel carryover
caused ash fusion and inner screen deterioration. Detailed analysis is being
conducted by electric microscopy and chemical analysis to confirm some of the
postulates. Preliminary results using scanning electron microscopy analysis
located where the inner screen failed indicated that ash fusion did occur
which embrittled the fabric. In any event, modification in the cage design
is needed to keep a constant uniform tension on the filter. This would
prevent overstretching of the filter media as well as keep dust from
accumulating under the outer screen, a phenomena which was observed with most
of the filters.
279
-------
Figure 5. Durability test rig.
280
-------
DURABILITY TESTING OF WOVEN CERAMIC FILTERS
A program sponsored by EPRI is underway to test ceramic filters woven
from 3M Nextel® yarns. The test will be conducted for 6,000 hours at 800°F
and atmospheric pressure using re-entrained PFBC dust. Thirteen filters
6 inches diameter and 5 feet long will be used. A summary of expected test
conditions is given in Table 6. While the basic objective of the test is to
evaluate filter durability under steady-state conditions, several advanced
concepts and devices such as a filter cake mass detector and a real time
particulate analyzers will also be tested. Figure 5 shows a photograph of
the durability test rig. This major test is scheduled to begin during
October 1982.
TABLE 6. CERAMIC FABRIC FILTER LIFE CYCLE TEST FACILITY
Temperature: 700K (800°F)
Pressure: 1 atm
Dust loading: 13.9 g/acm (5.9 grains/acf)
Flowrate: 24 acm/min (840 acf/min)
Air/Cloth ratio: 2.4 m/min (8 ft/min)
Number of filters: 13
Filter dimensions: 15.2 cm (6 in.) diameter
1.5 m (5 ft) long
Cleaning method: Online reverse pulse
ACKNOWLEDGEMENT
This work is partially supported by a Contract DE-AC01-80ETIT092 from
the Department of Energy, Morgantown, West Virgina and Contract RP-1336-4
from the Electric Power Research Institute, Palo Alto, California.
The work described in this paper was not funded by the US Environmental
Protection Agency and therefore the contents do not necessarily reflect the
view of the Agency and no official endorsement should be inferred.
281
-------
HIGH TEMPERATURE AND PRESSURE PARTICULATE FILTERS FOR
FLUID-BED COMBUSTION
by: D. F. Ciliberti, T. E. Lippert
Westinghouse Research and Development Center
0. J. Tassicker, S. Drenker
Electric Power Research Institute
ABSTRACT
The only technological barrier to the commercialization of
Pressurized Fluid-Bed Combustion (PFBC) is the efficient removal of
particulates at high temperature and pressure. The Electric Power
Research Institute has sponsored work at the Westinghouse Research and
Development Center to investigate several filtration devices for this
application. This effort has included high pressure and temperature
pilot-scale testing of multielement ceramic bag filters of both the
woven and felted type. The current program also includes screening
testing of high alloy sintered metal and tubular porous ceramic filter
candles at temperatures in the range of 800-900°C and at pressures of 11
atm. Subsequent to these tests long duration life testing of a single
woven ceramic bag will be carried out to optimize bag life with respect
to cleaning regimen.
The work described in this paper was not funded by the U. S.
Environmental Protection Agency and, therefore, the contents do not
necessarily reflect the views of the Agency and no official endorsement
should be inferred.
INTRODUCTION AND BACKGROUND
The successful removal of particulates from high temperature and
pressure gas streams is a goal that is important to many advanced coal
conversion technologies both from an operational process point of view
and for environmental considerations. The economics of processes such
as electric power production from low-Btu coal gasification could be
greatly enhanced by a viable hot gas cleaning system, while the
commercial development of power production from pressurized fluid-bed
combustion critically depends on an effective hot gas cleaning system
that will result in adequate turbine life. As such the hot gas cleaning
problem has been identified as the single technological barrier to the
commercial development of PFBC. To address this technology gap both the
Electric Power Research Institute (EPRI) and the Department of Energy
(DOE) recently have sponsored a broad range of research and development
programs in the area of hot gas particulate removal. These programs
have covered many variations of high temperature filters (granular,
porous ceramic, sintered metal, and ceramic fiber bag filters), as well
282
-------
as advanced cyclone concepts, electrostatic precipitators and solid
particle scrubbers.
TEST FACILITIES
Since Westinghouse is a major supplier of gas turbines we have
actively participated in such programs directed at the solution of the
hot gas cleaning problem and subsequent commercialization of combined-
cycle plant concepts. In support of this participation we have (with
the aid of DOE, EPRI, and EPA) developed two high pressure and
temperature particulate control test facilities. One is a flexible
bench-scale unit for the development of novel concepts and the other a
pilot-scale test facility in which more advanced concepts can be tested
at signifiant scale.
The bench-scale test facilities was constructed by Westinghouse
in conjunction with the execution of the DDE-sponsored program to
develop a ceramic cross-flow filter. The original operating equipment,
instrumentation, and controls were purchased and installed under the DOE
cross-flow filter contract. The test facility as configured for the
this work is shown schematically in Figure 1. The range of operating
conditions is presented in Table 1.
,-N -A
' cv».c,mi t"" ™ V ~
Figure 1. Bench-scale HTHP test loop.
283
-------
TABLE 1 - OPERATING PARAMETERS FOR THE WESTINGHOUSE BENCH SCALE
HIGH TEMPERATURE AND PRESSURE TEST FACILITY
Temperature 950°C
Pressure 18 atm
Air Flow 225 Nm3/hr
Dust Concentrations (typical) 1,000-10,000 ppm
Under EPRI sponsorship, and with DOE approval, the facility is
currently being upgraded with the intention of making it capable of
unattended operation and thereby economical for long-term testing. This
has involved the installation of a natural gas compressor for fuel
supply (in place of the bottled propane system previously used) and the
installation of other safety-related instrumentation and control.
Additionally, a minicomputer-based data aquisition and control system
has been installed to facilitate operation of the system for long
periods of time and to manage the consequent large amounts of test data.
There also is a need to test, at high temperature and high
pressure, particle removal equipment large enough to permit
extrapolation to the design of utility-scale units. Westinghouse, with
the support of both EPA and DOE, designed and constructed a test
facility at their Synthetic Fuel Division for evaluating particle
removal equipment us.ing simulated flue gases at temperatures up to 871°C
and pressures up to 15 atm. Hot gas flows up to 5.44 kg/s can be
provided. Equipment up to 1.37 m in diameter and 2.44 m in length can
be mounted within an insulated pressure shell for testing.
A functional schematic of the test passage is presented in
Figure 2. The high-pressure air for the system is supplied by either or
both of the available compressors. A 1500 kW centrifugal compressor can
supply 3.4 kg/s of air at 21 atm. A second, three-stage, 900 kW
reciprocating compressor can be run in parallel, providing flows up to
5.86 kg/s. The high pressure air flows from the compressor building to
the laboratory, where it can be heated to temperatures up to 650°C by
either of two natural-gas-fired air preheaters.
The pressurized, preheated air then flows through a combustor
where No. 2 fuel is burned to raise the gases to the desired
temperature. The combustor fuel is pumped from the fuel-blending
building where several tanks are available for blending either corrosion
inhibitors or promoters (combustible alkali organometallic compounds).
From the combustor the hot pressurized gases enter the test passage
piping. The passage piping and valving are arranged to allow a great
deal of flexibility in the manner in which the gases are introduced to
and exit from the pressure vessel, allowing virtually any device that
will fit in the pressure vessel to be tested.
284
-------
Dug. I/I8B26
Operating Conditions
Pressure - Up to 150 psig ( capability to 220 psi)
Temperature - 200 - 1600° F
Flow Rates - Up to 12 Ib/s
Vessel - 56" Dia x 110" Length
Piping - 10" Sh. 80 with 6" Inconel Liners
Air 1
Air Compressors —
Alkalis L No- 2 Fu
— G3
i — i .r_i
Fuels
Blending
Tanks Atomizir
n Air
Preheater r;
-J Process
__f Air
uel
i Combustor "
ig Air — '
Control
Room
W Particulate
Ei Feeding ^^
| System .^
Rupture ^^
Disc "~->
• (VI * t
Alternate Gas Piping
Particulate <
Sampling
, t ,
1 !
I By-Pass
*
i
r
s
V^
'
Hot Gas
Cleaning
Pressure
Vessel
Particulate
-"Sampling
Muffler
Chamber
Figure 2. Schematic of Westinghouse hot gas cleanup facility.
CURRENT AND RECENT HOT GAS CLEANING PROGRAMS IN WHICH WESTINGHOUSE HAS
PARTICIPATED.
Over the past several years we have focused our efforts on the
development of high temperature filter systems since we believe that
such systems can be designed to meet or exceed both turbine tolerance
and environmental requirements. To this end Westinghouse has
participated in the development of and or testing of six different
filter concepts over the past three years. Table 2 presents a very
brief summary of this work including some qualitative comments. In
general all of the filtration devices tested have demonstrated very high
collection efficiencies at moderate pressure drop. It is also true,
however, that all of the filter systems have unresolved questions
concerning their cleanability, be it reflected in filter life or in
gradual blinding or loss of the filter medium. All the filter systems
listed has displayed potential for successful application to PFBC, but
none have been demonstrated to date.
Development of the last three systems shown in Table 2 has most
recently been supported by DOE. Testing of the first three devices
listed has been sponsored by EPRI, and these devices are the primary
topic of this report.
285
-------
TABLE 2 - FILTER DEVICES TESTED BY WESTINGHOUSE
Tv|,K.j| Operating parameters
Test
Scale
Device m3/hr
Schumacher 85
Ceramic Candles
EPRI
Pall 85
Sintered Metal
EPRI
3M Woven 850
Ceramic Bag
EPRI
Acurex FelteO 850
Ceramic
EPRI7DOE
Ducon *5C
Granular Bed
DOE
Wertmghouse/Coors «
Ceramic Crossflo*
DOE
filter Base
Velocity. Pressure
m/mm Drop kPa Elficiency *
3 2 99 9*
3 2 99 9+
85-1 7 5 9"*
3 5 99 9 +
12 7-10 W
2 1-2 99 9*
Comments
• Mechanically sound
corrosion resistant
• High efficiency
• Potential blinding
• Mechanical sound
• High ell/ low Ap
. Potential lor blinding
• Potential corrosion
• Easy to clean/low Ap
• Bag Hie uncertain
• Very good eft/ med Ap
• Cleaning trade-oft
• Moderate efficiency
• Blowback cleaning
still to be demonstrated
• High eft/low Ap
• Compact system/good
economics
• Mounting and mechanical
properties need improvement
PILOT-SCALE WOVEN CERAMIC BAG FILTER TESTING
The testing of filter bags woven from yarns of ceramic fiber was
accomplished through the cooperative effort of the 3M Company, Buell
Envirotech, and Westinghouse under EPRI sponsorship. Westinghouse R&D
was EPRI's prime contractor and carried out the actual HTHP test program
at their pilot-scale test facility, which is operated by the Westinghouse
Synthetic Fuels Division. 3M supplied the woven ceramic bag filters and
Buell Envirotech fabricated the bag mounting system and blowback
equipment which was installed at the test site.
WOVEN CERAMIC BAG FILTER UNIT DESCRIPTION
The general arrangement drawing of the woven ceramic filter unit
is shown in Figure 3. The unit consisted of a 19-bag array that was
suspended from a canister support assembly that incorporated a top
support flange and conical section, which in turn was gasketed and
bolted between the pressure vessel flange. The conical section was
designed to accommodate differential thermal expansion and was attached
to the canister body that shrouded the bags. Gas flow was introduced at
the bottom of the pressure vessel, around and through the perforated
wall of the bag canister, up between the bags, in through the bags, up
the inside of the bag and perforated bag support cage, and out the top
of the pressure vessel. The conical section and top support flange
served to separate the clean and dusty gas sides of the pressure vessel.
286
-------
Flow
Blowback
Manifold ITyp)
0»9.7732AI9
Stainless Steel
Tube
56.00 I. D.
Existing Pressure
Vessel
New Bag Support
Can
Collecting Bag
Figure 3. General arrangement drawing of the woven
ceramic bag filter - front view.
The individual ceramic bags were fitted over perforated bag
support cages. The cages measured 15.2 cm od by 1.4 m long. The bags
were made taut against the support cage by a pocket of sand in the lower
part of the bag and clamped to the top of the support cage with a metal
compression band. The support cage perforations were 0.63 cm diameter
holes on 0.79 cm centers, providing an approximately 58% open area.
Each bag support cage was flanged at the top to seat and seal (using a
special high-temperature gasket) against the canister top flange.
The filter bags themselves were made from 3M ceramic fibers AB-
312 in an 8-harness weave. These fibers are nonoxidizing,
nonconductive, and heat resistant, and can withstand temperatures in
excess of 1096°C.
Each bag was cleaned by a 0.97 cm pulse jet nozzle located at
the top of the bag. The nozzles of the 19-bag assembly were manifolded
to two high-pressure reservoirs located externally to the test unit and
presure tank assembly. For cleaning the bags were manifolded in groups
of two and three to eight solenoid control valves programed to actuate
sequentially upon initiation of the cleaning cycle. The duration of the
back flush pulse could be set between 0.02 and 0.180 s.
287
-------
RESULTS FROM THE WOVEN CERAMIC FILTER UNIT TESTS
Hot-gas cleanup tests were conducted on the woven ceramic filter
unit at two temperatures, 427°C and 815°C. Table 3 summarizes the
overall test results for the lower temperature condition. Data included
in the table are the nominal test passage operating conditions, the
measured filter unit pressure drop characteristics, the test average
inlet and outlet dust loadings, and corresponding overall collection
efficiencies. During any one test day several (up to four) outlet
samples were obtained. From each of these an outlet dust loading from
the filter unit was determined. The overall collection efficiency
values listed in Table 3 are average values for a given day of
testing. At the lower temperature condition tests were, conducted at two
mass flow rates corresponding to filter face velocities of 0.85 m/min
and 1.71 m/min.
TABLE 3 -' SUMMARY OF RESULTS FROM WOVEN BAG FILTER
TESTS, 427<>C (800°F)
Test
Identification
12-5-79
2ml
Cyclone
Ash
12-7-79
12-11-79
12-13-7°.
12-19-79
Asn *
Limestone
12-21-79
Ash +
Limestone
Test Conditions
P°113ikPa I165pslal
T = 427*C (JOO'F)
m=0.91kg/s IZ.OIt/sl
Q=9.8m3/mln (346fl3/mlnl
Alr-to-Clolh = 0.85m/m!n (2.8ft/ mini
P-1138kPa (165(Sl«
T=427'C (SOO'FI
m = 0.91ky/s 12. OK)/ si
g = 9.8in3/mln I346ft3/mlnl
Alr-to-C loth = 0.85m/ mini 2. 8ft/ mini
P = ll38kPa (165 psla 1
T=4Z7"C 1800*0
m =0.91 kg/ s (2.0lb/sl
Q=9.8m3/mln 1 346 (I3/ mini
Alr-lo-Clolh =0. 8 m/ mln [ 2 . 8 ft/ mini
P-U38kPa 1165 psla)
T = 427"C (JOO'FI
m=0.«lk9/s (2.010/sl
Q = 9.8m3/min (346ft3/mlnl
Alr-t8-Cloth=0.85m/mln 12. 8ft/ mini
P = 1138kPa I165pslal
I=C7'C 1800'FI
111=0 91kg/s IZ.OIb/sl
Q=9.8m3/mln ( 3M It3/ mini
Air-to -C loth = 0. 85 m/min 12. 8 ft/ mini
P = 1138kPa (165 psla)
T=427'C 1800'FI
m = 1.82kg/s I4.0lb/sl
Q*19.6m3/mln (692 ft3/ mini
Alr-to-Cloth = 1.7m /minis. 6ft/ mini
NO. of
Cleaning
Cycles
I
\
3
1
3
6
BagAP. kPalln^OI
Before
Cleaning
0.47 (1.11
0.9S 13.8)
0.52 12.1)
0.60 12.4)
0.40 11.61
1.62 16. M
0.50 (2.01
0.62 12. 51
1.37 (5.51
1.12 (4.51
1.12 (4.51
1.54 (6.21
1.24 15.01
1.00 (4.01
1.3? (5.51
After
Cleaning
0.02 10.1)
0.02 10.1)
0.02 10.11
0.07 10.31
0.07 10.31
0.07 10.31
0.17 10.71
0.02 (0.11
0.10 (0.41
0.55 (2.21
0.45 11.81
0.17 10.7)
0.17 10.71
0.17 (0.7)
0.17 10.71
Oust Loading, ppmlgr/ sell
Inlet
6726
13. SI
5487
13.11
11,328
16.41
8337
14.711
8638
14.891
3844
12.21
Outlet
26.5
10.0151
11.3
10.00641
<1
(0.00051
-0
(-01
2.1
10.00121
15.9
10.0091
Overall Collection
Efficiency
( Test Average)
99.6*
99.8*
99.9*
-100*
94.9*
99.6*
The initial testing of the filter unit corresponded to a period
of "bag conditioning" where the relatively large spaces between the
fiber weave partially filled with dust particles, preparatory to
establishing an identifiable filter cake. During this bag-conditioning
period, the outlet dust loading decreased dramatically with time. After
about 6 or 7 hours of filter unit operation, an apparent steady-state
operation was indicated. During this steady-state period of time, at
the lower face velocity tests, at relatively high inlet loading, the
overall collection efficiency was measured as 99.9% or greater.
288
-------
The last test conducted in the low temperature sequence was at a
filter face velocity twice the previous runs, corresponding to 1.71
ra/min. At this condition, the overall collection efficiency was measured
at 99.6%, somewhat reduced from the lower velocity, steady-state runs.
The pressure drop characteristics of the filter unit are also
indicated in the table and show relatively low filtration Ap. In these
tests the maximum Ap was not permitted to exceed 1.6 kPa, although
system operation at higher Ap may prove to be more desirable. At the
operating pressure drop in these tests, the filtration cake thickness
was observed to be about 0.63 cm. The effectiveness of the pulse-jet
cleaning is indicated by the very low Ap measured after bag blowback.
The effectiveness of bag cleaning was found to be dependent on the
pressure ratio between the bag operating conditions and pulse-jet
reservoir pressure as opposed to the pressure difference. Unsuccessful
cleaning was identified when multiple pulses would not result in the
lowering of the bag Ap. Successful cleaning is identified for those
data where a few or single pulses would result in the bag Ap reducing to
its previous low base-line value. Effective bag cleaning was achieved
if the pulse-jet pressure was maintained at about 3 times the filter
ambient. This apparent dependence on pressure was tested at total
pressures of 1114 and 709 kPa and is consistent with conventional
ambient bag filtration wisdom where cleaning pulses are on the order of
203 to 405 kPa, At the lower filter bag Ap a somewhat lower cleaning
pressure ratio may be indicated.
The most significant dust penetration was observed to occur
during and/or immediately following the cleaning cycle. This was
supported by filter samples that included periods of blowback and
samples that did not include blowbacks. Two 815°C tests runs were made,
one at a relatively low filter face velocity, 1.0 m/min, and the second
intended to be at a higher face velocity of 2.0 m/min. During this
second high temperature test, a malfunction occurred in the passage
combustion system, forcing operation at less than desired conditions.
Additionally, during this test, oije of the filter bags failed at a sewn
seam. This resulted in our inability to build system pressure drop, and
tests were halted at this point.
Inspection of the failed bag by 3M Company indicated the
following:
• That the top layer of Nextel fabric in a double-stitched
French seam broke at the crease
• That there is no gross evidence of chemical attack, fusion
via eutectic formation, etc.
• That only one bag failed in this manner
289
-------
• That numerous warp thread crossings in the eight-harness
weave fabric showed fiber breakage in the form of tufts of fluffy
protrusions at the point of each crossover of fill yarn by warp yarn;
that some general abrasion is shown by random broken fiber protrusion
• That two to four gently curved longitudinal wrinkles are heat
set into the fabric of each bag. These are about 0.64 cm wide and about
0.64 cm high.
A judgmental interpretation of these observations suggests:
• That the seam should be made somewhat wider
• That cleaning be accomplished more gently to reduce abrasion
wear, should this prove to limit life.
CURRENT WOVEN CERAMIC BAG FILTER PROGRAM
Subsequent to the completion of this test phase, EPRI has
supported a continued effort in the development of the woven ceramic bag
filter. The current program is a parallel effort being carried out by
Westinghouse and Acurex for the primary purpose of exploring bag life
characteristics. The Acurex effort consists of high temperature but low
pressure testing using the 19-bag unit originally installed at
Westinghouse.
The testing at Westinghouse will be carried out at both high
pressure and temperature in the bench-scale test unit on a single bag
configuration. The bench-scale test facility has been modified to
accomodate a single full size filter bag, as shown in the pressure
vessel assembly drawing, Figure 4. It is our intention to carry out
testing on various cleaning strategies to determine the cleaning regimen
that optimizes bag life. The data generated will be correlated with the
low pressure test results in hopes that eventually less costly low
pressure testing can be used to predict high pressure performance. To
this end a detailed mathematical model of the cleaning process has been
developed and will hopefully serve as the basis for correlation of high
and low pressure test results.
It should be noted that 3M has pursued improvements both in the
basic ceramic fiber used in the bag cloth and in the sewing techniques
employed. We are currently considering the use of a new zirconia fiber that
has demonstrated the potential for greatly improved abrasion resistance.
HIGH PRESSURE AND TEMPERATURE TESTING OF CANDLE FILTER
As part of the current EPRI test program we have had the
opportunity to briefly examine two different types of "candle" filters
in our bench-scale HTHP test facility. One was a porous ceramic candle
290
-------
(44)
.'ii!
Figure 4. Single bag filter HTHP test vessel.
filter, while the other filter examined was a more or less conventional
sintered metal filter.
POROUS CERAMIC CANDLE FILTERS
The porous ceramic candle filters tested were manufactured by
Schumacher'sche Fabrik of Bieligheim, West Germany, and are denoted as
Schumacel HTHP. These filters are formed by incorporating very small
pockets of pure mineral fibers in a dense matrix of silicon carbide.
The interconnecting pockets of these small diameter (3 urn) fibers gives
rise to very reasonable pressure drops in spite of the rather thick and
291
-------
rugged wall of the tube. The tubular filters we tested were nominally
50 cm long with an OD of 6 can and ID of 4 cm. The active filter area
was estimated as 0.072 m^, based on the outer area.
In the course of the preliminary testing that we carried out, we
tested a relatively porous element that had an ambient air resistance to
flow of 0.25 kPa at a face velocity of 0.67 m/min and a relatively dense
element that had essentially twice that resistance to flow at the same
velocity.
TABLE 4 - SUMMARY-CERAMIC CANDLE PERFORMANCE AT 770-800°C, 11 atm
Dwg. 7772A*
Test No.
Series 1
4. 22. 82
4. 26. 82
4.2&S2
5.3.82
5.4.82
5. 6. 82A
5.6.82B
Series 2
8.18.82
8.24.82
8.26.82
8.31.82
Test
Time,
hr
Fi Iter
Flow. Velocity.
kg/min n/min
Dust
ConcentraKtn.
Inlet, pprt
Measured Dust
Collection
Efficiency, *
- Porous Element
2.0
4.0
4.0
9.8
5.0
1.0
8.5
0.72
1.44
1.44
1.44
1.44
1.44
0.72
152
5.35
5.20
5.13
5.10
5.31
165
2.707
3.811
3.329
2.807
3,010
3,151
6,302
99.90
99.98
99.97
99.99
99.99
99.99
99.99
- More Dense Element
4.0
4.0
5.5
9.0
0.83
0.83
0.83
0.83
3.35
3.35
3.35
3.35
5.206
5.149
5,499
5.468
99.97
99.%
99.35-
99.98
Leaks in gasket seal
During the first series of tests with the more porous filter
candle, the test temperature and pressure were constant at 775°C and 11
atm. During this series seven tests were conducted accumulating a total
of about 35 hours of actual filtration time and slightly more than 100
operating cycles (blow back sequences). The filter face velocities
examined ranged from 2.5 to 5.3 m/min. The test dust used was a
redispersed ash from the Curtiss-Wright PFBC and had a mass mean
diameter of approximately 10 ym. Dust concentrations were in the range
from 2700 to 3300 ppm. The results of these tests are summarized in
Table 4. The overall efficiency measurements were made using an
isokinetic outlet sampler and in all cases indicated a penetration
smaller than we were able to resolve. A typical pressure drop/time
curve for a test is presented in Figure 5. The performance shown here
is typical of the behavior observed throughout this period of testing,
292
-------
Curve 7 39137-C
100 120 140 160 180
220 240 260 280 300 320
480 520 540
Time Imms I
560
575
Figure 5. Ap vs time for ceramic candle filter.
that is, of gradual shortening of cycle time and slowly increasing base
line pressure drop. The cleaning regimen employed for these tests was
0.5 sec reverse pulse of air from a 2.1£ reservoir charged to a pressure
of 35 atm and discharged through a 0.75 cm diameter nozzle centered
above the clean side of this filter element. This pulse typically gave
rise to a 30 to 50 kPa differential pressure from the clean to dirty
side of the element. Posttest examination of the filter revealed that a
thin "cratered" deposit of dust remained on the filter surface, as shown
in Figure 6. This dust layer was not sticky or hard and was easily
removed, as shown in the photograph. We hypothesize that the
nonuniformity arises from the local areas of high porosity at the
imbedded fiber pocket surface sites.
Subsequently a series of tests were carried out with the more
dense high pressure drop element. The only system modifications were
the insertion of a venturi section in the outlet of the candle in an
effort to improve cleaning and an increase in the pulse accumulator
volume to 8.4&. The test pressure and temperature remained essentially
the same as in previous tests, but the dust concentration was somewhat
higher at 5100 to 5500 ppm. Approximately 25 hours of actual filter
test time and 95 cleaning cycles were accumulated over four test days
293
-------
Figure 6. Posttest surface condition of porous ceramic candle filter.
with the more dense filter candle. The results of these tests are
summarized in Table 4, where it can again be observed that the filter
elements behaved essentially as absolute filters. As in the previous
tests the pressure drop/time curves generated revealed a gradual
shortening of cycle time and a slowly increasing baseline pressure
drop. The inclusion of the venturi section and the larger pulse
reservoir did not seem to significantly alter the pressure rise in the
filter during blowback, and cleaning remained somewhat incomplete.
Figure 7 focuses on the observed cleaning problem. This figure
presents a measure of the cleaned filter permeability (filter velocity
divided by the freshly cleaned baseline pressure drop) as a function of
operating cycles. Both elements are gradually becoming less permeable
at an apparently constant rate of about 0.0024 (m/min-kPa-cycle) for the
more dense element and at 1.4 times that rate for the more porous
element. It is apparent that if this constant decrease in permeability
persists in spite of any operating modifications, then the filter system
would not be viable. It should be emphasized that we do not feel that
enough operating time has been accumulated nor have a wide enough range
of operating parameters been explored to conclude that the system will
not provide adequate life. Rather, we are encouraged by the filter
element's high collection efficiency, at modest pressure drop and high
operating velocity. Another positive consideration is the mechanical
strength of the elements, their resistance to thermal shock, and the
ease with which they can be mounted in a conventional filter assembly.
294
-------
4
S. 20
"E
'e
£1.5
e
£
o>
S
IZ
o>
.2
1.0
0.5
Clean
P = 11 Aim Nominal
T=815°C Nominal
V = 2 52
V=5 35m/mln
V = 2.65m/mln
'•'•*
Clean
Candle No. 2 I More Porous)
^
*
o. Candle No. 4 IMoreOensel
... —«-«.H...H.^/~.^"»H......
•••*••*••••••.•...».,
• V = l 35
I 1 1
I Removed & Cleaned Residual
Oust Cake From Candle
i i i i i i i i i i i i i i
10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110
No. of Consective Operating Cycles
Figure 7. Effective permeability of ceramic candle and residual dust
cake in HTHP PFBC simulation tests.
SINTERED METAL CANDLE FILTERS
The porous sintered metal filters tested in this program were
supplied by the Pall Trinity Micro Corporation. The elements were 7.6 cm
in diameter and had an active length of 38 cm, yielding a filter surface of
0.091 m^. The material used for fabrication of the elements was a Grade F,
20 pm absolute, sintered Hastelloy X rated for service to 900°C.
Through the entire series of sintered metal filter tests the
temperature was maintained at 815°C and the system pressure was held at
11 atm. During the first six test days a 3-element configuration was
tested at filter face velocities from 1.22 to 2.44 m/min and dust
loadings from 1600 to 4500 ppm. During the last two days of testing one
of the three filter elements was removed so that the range of velocities
could be extended to 3.7 m/min. A total of about 58 hours of test time
were accumulated, representing 62 cleaning cycles. Table 5 presents a
summary of the test experience with the sintered metal filters. As with
the ceramic candle filters, virtually no reliable, measurable
penetration was observed in any of the tests.
Blowback cleaning of the filters for the first 6 tests consisted
of 0.5 sec pulses from a 2.1)1 reservoir charged to pressures between 21
and 28 atm. The pulse nozzles were 0.75 cm diameter and centered over
the element outlet. Each element was supplied with a venturi in the
295
-------
TABLE 5 - SUMMARY OF SINTERED METAL FILTER TESTS
Date
6/10/82
6/15/82*
6/21/82"
7/1/82*'
Test Time, hr
Passage
Flow.
kg/hr
9 (13 eye) 145.28
0. 8 (2)
0.7 (1)
2 (10)
1. 8 ( 3)
3. 5 ( 5)
97.1
97.1
48.6
48.6
145. 28
Filter
Velocity.
m/min
5/26/82
5/28/82
6/3/82
6/7/82
4. 5 ( 2 cycl
11.0 (5 eye)
11.0(8cyc)
3. 5(2cyc)
3. 5(6cyc>
72.64
7Z64
7Z64
72.64
145. 28
1.22
1.22
1.22
1.22
Z44
2,44
1.5(1 eye)
2.0(2cyc)
1.5<2cyc)
72.64
108.9
145. 28
1.22
1.83
2.44
Z48
2.48
1.22
1.22
3.71
Inlet Dust
Concentration.
ppm
2962
2508
3869
4447
1827
1625
3139
2093
1569
1139
1406
3490
1186
10%
Dwg. 7772A05
Overall
Collection
Efficiency. %
99.98
99.98
99.99
99.97
99.97
99.97
99.77
99.99
99.%
99.44
99.97
99.99+
99.80
99. 99*
• On line cleaning test
•• Two filter modules only
filter outlet to enhance the blowback effectiveness. During these tests
two modifications to the blowback system were made. After the sixth
test (6/15/82) the pulse reservoir volume was increased in volume by a
factor of 4 to 8.4£. After the seventh test (6/21/82), there was some
concern that there might be some misalignment between the pulse nozzle
and the venturi throat due to thermal expansion in the blowback lines.
A new set of nozzles and method for maintaining alignment were therefore
installed for the final test. This precaution did not result in any
observable increases in internal filter pressure rise during blowback,
as this rise remained at about 45 kPa for a 28 atra pulse.
During the first several days of testing the filter velocity was
maintained at 1.22 m/min. During this time the filter appeared to
condition and settle into stable operation as shown in Figure 8. In
subsequent tests, operation at higher velocities was explored and these
seemed to result in a very gradual decrease in both cycle time and
permeability as shown in Figure 9.
As was the case with the ceramic filter, the shortened operating
cycles appear to be a consequence of a decreasing filter permeability.
296
-------
100 120 Itt 1(C
i r
Figure 8. Pressure drop vs time for sintered metal
filter conditions 6/3/82.
p ms /r n °ver the period of testin8 at a
rate of 0.015 m/(min-kPa-cycle) or nearly six times the observed rate
for dense ceramic candle filter.
During the final test the system was operated first for a few
cycles at a face velocity of 1.22 m/min and then at 3.7 m/min. During
the high velocity segment of this test, the baseline pressure drop
increased rapidly from about 17 kPa to 19 kPa in the course of 4
cycles. Posttest examination of the elements revealed the fact that the
dust deposit on one of the elements was quite different in appearance
from any observed to date. Figure 10 shows this difference. The
element on the right looks normal, while the one of the left appears to
have a much heavier deposit. Although we were unable to discover the
reason for this difference, we feel certain that it explains the
unstable behavior observed in the last test.
thai- H,^ 8Pite.of 5h? 5*lef nature of ^is program, we can conclude
that the sintered metal filters are capable of providing very high
efficiency on PFBC ash at tolerable pressure drops. Additionally, their
mechanical properties and resistance to accidental breaking is a very
297
-------
180
!oO
^.
£ 140
O>
P 12(1
JL'
o 100
TS
I
•g 60
I «
20
1. 5r
:f 1.0
Sintered Metal Filter Unit Operating
Characteristics at 815°C and 11 Atm (Nominal)
With PFBC Ash
| Signifies Shut Down/Restart
* Replaced 3-Element Unit With 2-hlement Unit -
Removed Residual Dust Cake
* Removed Unit and Cleaned Residual Dust Cake
) .
°a
°oa t
a
aa
10
_L
15
_J J L
25 30 35
i
40 45 50 55 60 65 70
No of Consectlve Operating Cycles
1
Symbol
o
a
•
V
e
e
9
•
a
a
e
JL
•
V (m/mml
1,22
1.22
1 22
1.22
1 22
1.22
1.22
? 44
2.44
2.44
2.48
1.83
3.71
AP kPa
max
3.7
37
3.7
3.7
6.0
7 0
60
6.0
8.4
8 4
11 2
7.2
22 3
C (ppml
2962
2508
3869
4447
3490
3490
1186
1827
1625
1569
1139
2093
10%
> * I
••-••-.H...V
» I
x
Measured Prior to First Cycle
After Removal of Residual Dust
\Cake (Cleaned Filter)
l i l I i i
!' 5 10 15 20 25 30 35 40 45 50 55 60 65 70
No. of Consectlve Operating Cycles
Figure 9. Effective permeability of sintered metal filter unit
and residual dust cake at 815°C and 11 atm (nominal)
with PFBC ash.
strong attribute for this type of filter, as is the ease with which they
can be mounted and configured in a real system. The question of filter
stability or blinding can not be definitively answered on the basis of
these short-duration tests, but there was indication that stable
behavior may be possible at relatively low filter velocities (1.2 m/min)
as well as the possibility of unstable behavior at higher velocities. A
final issue that was not addressed is the corrosion resistance of
sintered metal filters in actual coal-burning applications where various
sulfur and alkali metal compounds exist.
298
-------
Figure 10. Posttest condition of sintered metal filters.
299
-------
MOVING BED-CERAMIC FILTER FOR HIGH-EFFICIENCY PARTICULATE
AND ALKALI VAPOR REMOVAL AT HIGH TEMPERATURE AND PRESSURE
By: D. Stelman, A. L. Kohl, C. A. Trilling
Rockwell International Corporation
Energy Systems Group
8900 De Soto Avenue
Canoga Park, California 91304
ABSTRACT
A moving bed-ceramic filter for high-temperature gas cleanup is
described. The concept employs a high-efficiency ceramic filter that is
cleaned continuously by the slow downward motion of a thin layer of granular
material. Laboratory tests have been conducted with a variety of dusts
carried in gas at temperatures up to 1500 F and at atmospheric pressure. The
observed particle removal efficiency from a 1500 F gas containing 1 grain/scf
of 1.6 micron median diameter particles was found to exceed 99.96%. The
pressure drop across the filter was only 3.5 in. of water at a gas velocity of
13 ft/min. It remained essentially constant as a result of the continuous
removal of the filter cake from the face of the ceramic filter by the slowly
moving granular bed. In addition to particle removal, the filter offers the
potential for alkali vapor removal through the use of reactive getters in the
moving bed material.
300
-------
INTRODUCTION
Coal-fired combined gas turbine-steam turbine cycles such as a fluidized
bed combustion combined cycle or a low- or medium-Btu coal gasification com-
bined cycle offer the potential for improved thermal efficiency and reduced
cost with acceptable environmental impact. However, the combustion products
or fuel gases produced by these processes must be purified because they
contain particles and vapors that are corrosive and erosive to a gas turbine.
The main problem is that for maximum thermal efficiency the purification
should be performed at high temperature and pressure. For long-term turbine
operation, the principal impurities that have to be removed are erosive
particles and corrosive alkali compounds that are present both as particles
and vapors.
This problem has stimulated interest in a variety of potential high-
temperature filtration techniques, some directed at removing particles and
others at removing alkali vapor. In the case of particle removal by fabric
filtration, the high temperatures involved in this application dictate the use
of ceramic fiber filter media.
Removing the dust cake from ceramic bags can be a problem. Ceramic
fibers for high-temperature applications are brittle, and attempts to employ
pulse-jet cleaning have encountered problems with ceramic fiber bags rupturing
or tearing. In light of these difficulties, it is reasonable to explore
alternative methods of cleaning high-temperature ceramic filters.
The Rockwell filter system uses a continuous-duty ceramic fiber filter
that does not use a back pulse to clean the dust cake from the filter.
Rather, it uses a thin moving bed of sand-like particles to continuously and
gently remove deposited dust particles from the surface of the ceramic fiber
sheet.
DESCRIPTION OF ROCKWELL MOVING BED-CERAMIC FILTER CONCEPT
The moving bed-ceramic filter concept is shown in Figure 1. The right
side of Figure 1 shows a single filter element; the left side shows the filter
elements arranged in interlocking leaves. The filter is a combination of a
moving granular bed and a ceramic filter. Dirty gas is introduced into a
downward moving bed through a set of louvers. The gas flows perpendicularly
to the moving bed and exits through the ceramic filter sheet. A small portion
of the dust is removed by the bed, and the balance by the ceramic filter
sheet. The concept eliminates the damaging back pulse. The moving bed
continuously and gently scrubs the deposited dust particles from the surface
of the ceramic filter sheet.
Moving bed filters alone do not show adequate particle removal effi-
ciencies (1). Since the moving bed, in the Rockwell design, is not the
primary filter, it can be much thinner than "conventional" granular bed
filters. The bed material does not need to be free of fines. In fact, the
301
-------
presence of some relatively fine particles in the bed aids filtration. The
interlocking leaf design permits a large filtration area to be contained in a
relatively small pressure vessel.
The system does not require pulsing or interruptions in the flows of gas
or bed material, operations that tend to cause "spikes" of dirty gas and
require high-temperature, high-pressure valves in other systems.
A reactive or adsorbent material can be added to the bed to remove vapor-
phase impurities. The filter is thus expected to remove alkali vapors from
the gas with high efficiency by using a variety of adsorbent materials (2-5).
The filter will also remove traces of tar without plugging. The tar can be
removed from the bed by combustion or solution in light hydrocarbons.
In summary, the Rockwell filter is basically a fabric filter except that
it uses an unconventional cleaning method (i.e., a moving granular bed). The
cleanup of the gas is done almost entirely by the ceramic filter. The func-
tion of the moving bed is to clean the filter, although it does remove some of
the particles. The advantages of the moving bed-ceramic filter are:
1) High-efficiency particulate removal at high temperature-high
pressure.
2) High gas throughput with low pressure drop and no increase in
pressure drop with time.
3) Continuous filtration without pulsing or flow interruptions.
4) No need to have the bed free of fines.
5) Vapor-phase impurity (sodium, potassium) removal capability
through use of reactive or adsorbent material.
6) Removal capability for traces of tar.
EXPERIMENTAL RESULTS
The objectives of the experimental program conducted to date have been
to:
1) Demonstrate filter operation at high temperature.
2) Evaluate the effect of the moving bed on dust cake removal.
3) Establish overall filter performance, collection efficiency,
pressure drop, and pressure drop increase with time.
4) Investigate different dusts, bed materials, and filter
materials.
LABORATORY TEST APPARATUS
The laboratory test apparatus is shown in Figures 2 and 3. The test
procedure consisted of (1) injecting about 1 grain/scf of dust particles
into the inlet stream; (2) heating the gas stream to the test temperature;
302
-------
(3) measuring the pressure drop across the filter versus time with no bed,
with a stationary bed, and with a moving bed; and (4) measuring the particle
concentration in the outlet gas stream.
The materials tested are listed in Table 1.
TABLE 1. MATERIALS TESTED
Dusts
Bed Materials
Ceramic Filter
1.6-micron alumina (Glennel Corporation)
3-micron salt fume (from Rockwell Molten Salt
Test Facility)
25-micron alumina (Buehler, Ltd.)
50-micron fly ash (Ottertail Power Company)
28 by 48 mesh alumina
40 by 70 mesh silica
Saffil alumina paper (ICI America)
Figure 4 shows the size distribution of the redispersed test dusts as
measured with a cascade impactor in the inlet gas stream.
The face of the test filter was 2 in. wide and 1 ft long. The bed was
7/8 in. thick. The hopper on the dust feeder had a 2-h capacity.
The test conditions and results are summarized in Table 2. The results
are discussed in more detail below.
TABLE 2. SUMMARY OF TEST CONDITIONS AND RESULTS
Conditions
Temperature
Air-Cloth Ratio
Inlet Dust Loadings
Moving Bed Velocity
Results
Outlet Dust Loadings
Removal Efficiency
Pressure Drop
900 to 1500 F
9 to 14 acfm/ft (afpm)
0.8 to 3.8 grains/scf
3.1 to 7.5 ft/h
Below detection limit
>99.96%
1.5 to 4.7 in. water
303
-------
EXPERIMENTAL VERIFICATION OF DUST CAKE REMOVAL BY MOVING BED
In these experiments, the system was operated with and without dust
particles and with no bed, stationary bed, and moving bed. The effect of
these variables on the pressure drop across the filter was observed. (The
buildup of a dust deposit causes the pressure drop to rise. However, if the
moving sand continuously removes the dust deposit, the pressure drop remains
constant.)
Since fluctuations in the gas flow rate or temperature also affect the
pressure drop, such extraneous variables can obscure the observation of the
desired effect. Therefore, the results are presented in terms of the "flow
resistance" rather than the pressure drop. The "flow resistance" is inde-
pendent of the flow rate and temperature,* but does increase as dust deposits
accumulate. The flow resistance is defined as the ratio of the pressure drop
to the air-to-cloth ratio:
AP (in. H20)
flow resistance = —,—t ->._•./ c—=r • (1)
air/cloth (afpm) v '
The flow resistance of the system filtering fly ash with initially clean
Saffil when no bed material was present is shown in Figure 5. The plot shows
what happens when the dust feeder was turned on, off, on again, and finally
off. As expected, the flow resistance increased during the first 40 minutes
due to the buildup of a dust layer. When the dust feed was shut off, the
resistance remained constant. When the dust feed was turned on again, the
flow resistance again increased until the dust feed was turned off.
Figure 5 shows that the accumulating dust layer caused the pressure drop
to rise at the rate of 0.88 in. I^O/h. (The pressure drop is equal to the
flow resistance times the air-to-cloth ratio.) Thus, if nothing were done to
remove the dust layer, the pressure drop would continue to rise, and even-
tually, the filter would become useless.
The dust layer was left on the Saffil paper, and silica sand was added to
the system slowly over a 1-day period. Point A in Figure 6 shows the condi-
tion of the system after the sand had been added. The drop in the flow
resistance from 0.29 in. t^O/afpm at the end of the test as shown in Figure 5
to 0.21 in. H20/afpm at Point A in Figure 6 was due to part of the dust layer
having been knocked off during the addition of the sand. From A to B, the
sand was stationary, and the flow resistance was constant. At Point B, the
sand feeder was turned on. The flow resistance immediately dropped as the
dust layer was removed by the moving sand. The flow resistance leveled out at
0.123 in. H20/afpm and remained there. At Point C, the dust feeder was turned
There is actually a temperature-dependent viscosity effect, but for the small
temperature fluctuations in these experiments, the correction is negligible
and was ignored.
304
-------
on. The flow resistance rose slightly for 10 minutes and then remained con-
stant thereafter at 0.133 in. H20/afpm. Without the moving sand, in about 2 h
the pressure drop would have more than doubled (Figure 5). Instead, the
pressure remained constant to better than +0.7% at 1.46 in.
To summarize, Figure 5 showed the formation of a dust cake on the back
surface (since there was no sand in the bed at that time) . Figure 6 showed
the removal of that dust cake from the back face by the motion of sand through
the device (Point B to Point C) . From Point C to Point D, the figure shows
that when dust is being fed, the motion of the sand bed kept the pressure drop
from rising, presumably because the dust cake was being removed continuously
by the moving sand bed.
We now turn our attention to the performance of the complete filter, that
is, to experiments where there is sand in the bed from the outset.
As shown in Figure 7, pressure taps were located at the filter inlet, in
the moving bed just in front of the ceramic filter sheet, and at the outlet of
the filter. Using three differential pressure gauges as shown in the figure,
the combined pressure drop across the louvers and the sand bed, the pressure
drop across the ceramic filter sheet, and the total pressure drop across the
filter were obtained. The bed extended beyond the edges of the louvers and
the ceramic filter sheet so that the gas could not short circuit the sand bed
by a low-resistance path along the side walls.
Figure 8 shows test results obtained at 1500 F with coarse (25-Urn mass
median diameter) alumina dust using an alumina bed. Figure 8 starts with the
alumina bed flowing. Note that the main pressure drop was across the ceramic
filter sheet. When the dust feeder was turned on, there was an initial
perturbation, but after 10 min, the flow resistance returned to its initial
value and remained constant until the bed flow was stopped. With the bed
stationary, the total pressure drop instantly started to increase at an
average rate of 2.6 in. f^O/h. Note that the dust cake formed on the front
face when the sand flow was stopped. With dust still being fed, the bed feed
was restarted. There was an instant drop in the flow resistance, followed by
a steady decline toward the initial value.
When the sand flow was restarted, the dust cake on the front face broke
up. The incoming dust no longer accumulated on the front face. The incoming
dust, together with the dust from the disintegrating dust cake, was swept
through the thin moving bed (7/8 in. thick) by the gas flow (13 ft /min) . The
gas carried the dust to the back face where, instead of accumulating on the
ceramic filter sheet, the dust was continuously removed by the moving bed
(4 ft/h).
In tests with 50-ym fly ash, 25-ym alumina dust, and 3-ym salt fume at
bed velocities of about 4 ft/h, we found no increase in pressure drop with
time. With finer dust (1.6-ym alumina), however, the pressure drop did
increase continuously with time at the same bed velocity (4 ft/h) used with
the larger particle-size dusts. This indicated that finer dusts require a
higher bed velocity to clean the dust cake off the ceramic filter sheet. The
305
-------
data for 1.6-um alumina dust at two bed velocities (4.0 and 7.5 ft/h) are
shown in Figure 9.
At both velocities, the main pressure drop was across the back half. At
the 4-ft/h bed velocity, the total pressure drop increased with time. As the
figure shows, the pressure drop across the front face was relatively constant.
The increase was occurring on the back face, probably because the dust cake
on the ceramic filter was building up at a rate faster than the bed could
remove it. When the bed velocity was increased from 4 to 7.5 ft/h, there was
no effect on the front face, but the pressure drop across the back face
dropped significantly, presumably because of the increased rate of removal of
the dust cake from the ceramic sheet. The corresponding decrease in the total
pressure drop was, therefore, due entirely to the increased rate of cleaning
of the ceramic sheet.
Thus, the filtration process and the pressure drop occurred primarily at
the back face. The removal of the dust cake from the ceramic sheet and
achievement of a constant low total pressure drop were adequately controlled
by the velocity of the moving bed. Test results indicate that very fine dusts
require a higher bed velocity than coarser dusts in order to clean the dust
off the ceramic filter sheet.
To summarize, the effect of the moving bed on the performance of the
filter as illustrated in Figures 5 through 9 can be described as follows:
1) Without the sand bed or with a stationary sand bed, there was
a continuous rise in the pressure drop.
2) With the moving bed, the pressure drop remained constant.
3) The moving bed provided continuous cleanup of the ceramic
filter, even when there is an initial dust cake.
EXPERIMENTAL COLLECTION EFFICIENCY
The outlet dust loading was measured by passing the entire outlet stream
through a 6-in.-diameter Gelman Type A/E glass fiber aerosol filter. The
change in weight was measured on an analytical balance with 0.2-mg sensi-
tivity. In every case (even for test with 1.6-ym alumina dust or 3-ym salt
fume), the downstream dust concentration was less than the amount that could
be detected. Therefore, at the present time, only a lower limit to the
efficiency can be given. Based on the sensitivity of the instruments used for
all of the tests conducted so far, this lower limit on the efficiency is
conservatively established as 99.96%. For those tests which ran longer and
used higher inlet loadings such as shown in Figure 8, the lower limit on the
efficiency was 99.999%.
EROSION CONSIDERATIONS
When any two surfaces are rubbed together, they will erode and ultimately
wear out. This would not be acceptable for a practical filter.
306
-------
In the preceding experiments, we were primarily concerned with deter-
mining whether or not a constant pressure drop could be established by the
action of the moving bed. We chose to keep the pressure drop at its initial
dust-free value. Operating in this way, we can expect the filter to wear out.
Considering this fact, it becomes clear that it is not desirable to completely
remove the dust cake from the back face. Instead, a bed velocity should be
chosen such that the pressure drop is allowed to rise to some acceptable value
in order to form a protective steady-state dust cake on the ceramic filter.
By choosing a bed velocity such that the dust cake is never completely
removed, the bed material will never come in direct contact with the ceramic
filter element. The only erosion that would be expected to take place would
be steady-state erosion of the surface of the dust cake. The portion of the
cake that erodes is replaced by incoming dust. In this way, even soft filter
materials may be rendered immune to erosion.
Since the top of the filter receives the least amount of dust from the
inlet gas stream, it will be helpful to have some fines circulated with the
bed material for the purpose of augmenting the protective dust cake at the top
of the filter. Consequently, when the bed material is reused, it should only
be partially cleaned so that some fines remain in it.
BED MATERIAL RECYCLE
In the present study, the bed material was used once and discarded. In
applications where the bed material is reused, it will be necessary to sepa-
rate some, but not all, of the collected dust from the bed material. As
discussed in the previous section, it is desirable to retain some fines in the
bed material.
In traditional moving beds, the media must be cleaned completely of all
dust. This is accomplished by fluidizing the mixture (6,7). Because the dust
is much finer than the bed material, the dust is entrained in the fluidizing
gas and is carried off while the bed material remains behind.
A practical example of such a bed cleaning and recycle system shown in
Figure 10 consists of a pneumatic circulation system that withdraws the dirty
bed material from the bottom of the filter and transports the material to an
overhead deentrainment vessel (a large-diameter section of pipe or a cyclone),
where the bed material and part of the dust drops out and feeds by gravity
into a fluidized bed where the bed material has the last traces of dust
removed (6). The dust-laden transport and fluidizing gas goes to a bag filter
where the dust is collected. The cleaned bed material flows by gravity from
the bottom of the fluidized bed back into the filter. The system has no
moving parts, making it ideal for circulating and cleaning the hot bed
material.
In the present case where it is desirable to retain some of the dust in
the bed media, the same system with the fluidized bed omitted seems appro-
priate for the purpose.
307
-------
It is not always necessary or desirable to recycle the bed material. For
example, in the application of the Rockwell filter to a fluidized bed combus-
tion combined cycle, the Rockwell filter should be able to use the spent
dolomite or limestone sorbent discharged from the PFBC, as illustrated concep-
tually in Figure 11. The spent sorbent would already be hot and at pressure,
so there would be no heat losses associated with the moving bed and no addi-
tional lock hoppers required. In this application, any alkali vapor getter
added to the system would be added to the coal-sorbent mixture and injected
directly into the PFBC.
ALKALI VAPOR REMOVAL CAPABILITY
A reactive or adsorbent material can be added to the bed to remove vapor-
phase impurities. The filter is thus expected to remove alkali vapors from
the gas with high efficiency by using a variety of adsorbent materials.
The gettering of alkali vapor by a reactive bed has been extensively
studied experimentally by others (2-5). The techniques used involve the use
of materials capable of chemically or physically absorbing alkali vapors.
Suitable materials include clay minerals, silicas, aluminas, alumino-
silicates, ash from certain coals, and industrial glasses. A wide variety
of such materials has already undergone testing.
For example, Argonne National Laboratory has obtained 98% alkali vapor
removal in 3-in.-thick fixed beds of -8 -flO-mesh diatomaceous earth at gas
velocities of 50 ft/min (4a). They have also achieved 98% alkali vapor
removal with activated bauxite (4b).
Thus, it appears that it is not only technically feasible to remove
alkali vapors by using reactive bed materials in the Rockwell filter, but also
that the dimensions of the bed and the gas velocities in this filter offer
contact times similar to those stated above.
In a recent study (8), laboratory data for various getters were scaled up
to commercial size. It was calculated that a fixed bed 1 to 10 m (3 to 33 ft)
thick would be required to prevent alkali vapor breakthrough for a period of
1 year. The use of such a thick fixed bed probably represents an undesirably
large pressure drop.
If the Rockwell moving bed-ceramic filter were used for alkali vapor
removal rather than a thick fixed bed, the main difference would be that the
getter passes through the Rockwell filter in 1 to 2 h and not 1 year. For
simplicity, if we scale the values given above linearly, then taking a worst
case of 10 m (33 ft) thick with breakthrough in 6 months (4000 hours) would
correspond to the requirement of a bed 0.20 in. thick to prevent breakthrough
of alkali vapor for 2 h.
In the present experiments, the bed thickness was arbitrarily chosen at
7/8 in. That value can be increased or decreased to suit other purposes
without adversely affecting the performance of the filter for particle
removal. However, it appears that such thin beds would also be acceptable for
308
-------
alkali removal. One would expect the Rockwell filter to perform both tasks in
a single pressure vessel at a lower overall pressure drop.
COMPARISON WITH PRESENT STATE OF THE ART
In this section, the performance of the moving bed-ceramic filter is
compared to (1) conventional commercial low-temperature fabric filters and
(2) state-of-the-art high-temperature fabric filters.
COMPARISON WITH CONVENTIONAL LOW-TEMPERATURE FILTERS
The Rockwell high-temperature filter is basically a fabric filter except
that it employs an unconventional cleaning method (i.e., a moving bed).
Therefore, it is of interest to compare its efficiency, air/cloth ratio, and
pressure drop to those of conventional low-temperature commercial fabric
filters. On the basis of the experimental results discussed above, it can be
seen that the efficiency of the Rockwell high-temperature filter (>99.96%) is
about the same as that of commercial low-temperature fabric filters.
A comparison of the air/cloth ratio and pressure drop is given in
Table 3.
TABLE 3. COMPARISON OF AIR/CLOTH RATIOS AND PRESSURE DROPS
Air/Cloth Ratio Pressure Drop
Fabric Filter Type (afpm) (in. H20)
Commercial Low Temperature
Reverse Air Type 1 to 2 3 to 6
Shaker Type 2 to 4 3 to 10
Pulse Jet Type 5 to 15 6 to 10
Rockwell High-Temperature Filter 9 to 14 1.5 to 4.7
(900 to 1500°F)
Thus, the Rockwell filter operating in the high-velocity range of commer-
cial fabric filters shows a very low pressure drop. Because of this low pres-
sure drop, it is expected that the filter can be operated at higher air/cloth
ratios than were used in these tests. Therefore, one must conclude that it
will be smaller in size than the most compact commercial low-temperature
fabric filters currently available.
COMPARISON WITH HOT MOVING BED FILTERS
The Rockwell filter employs a moving bed to clean the dust cake off the
ceramic filter. Therefore, it has some similarity to the CPC moving bed
filter; however, when one considers that a 16-in.-thick CPC moving bed
309
-------
achieves 86 to 99.5% efficiency (6) while the 7/8-in.-thick Rockwell filter
achieves >99.96% efficiency, it is clear that the similarity is superficial.
The ceramic filter and not the moving bed is responsible for the high effi-
ciency of the Rockwell filter.
COMPARISON WITH HOT FABRIC FILTER
Table 4 compares the performance of the Rockwell filter to that of hot
fabric filters under development. Acurex tested filter bags made from ceramic
papers, felts, and woven cloths but found that they burst when back-pulsed
(9). For that reason, they finally selected 1-cm-thick thermal insulation
blankets made of Saffil. Three test programs using Saffil blankets were
carried out as shown in the table (10-12). The earliest test reported much
higher efficiencies than subsequent tests. The reason for this difference is
that the dust used in the first test had no fines in it. It was dust from the
second and third cyclone catches at the Exxon Miniplant and did not include
the fine particles that are not collected by the cyclones. The second Acurex
test was done on-line, downstream from the second cyclone at the Exxon Mini-
plant, and thus contained the complete particle size distribution. The third
test used the second cyclone catch with added fines -(12). These results
indicate that some of the finer particles pass through the Acurex ceramic
blanket.
TABLE 4. COMPARISON OF HOT CERAMIC FILTERS
Acurex (10)
Acurex (11)
Acurex (12)
Buell/3M (12)
Rockwell
Face
Velocity
(ft/min)
5-18
8-16
10
2-6
9-14
Mass Mean
Particle
AP
(in. H20)
<3 cyclic
1-40 cyclic
28-52 cleaned*
1-6 cyclic
1.5-4.7 steady
Temp
<°F)
1500
1300-1500
800-1500
800-1500
900-1500
Diameter
(ym)
4 & 19
4
15
15
1.6
Efficiency
(%)
99.96-99.99
96-99.5
99.5-99.8
99.2-99.9+
>99.96
Longest
Test
Duration
(h)
200
17
50
50
41
*Pressure drop after the cleaning cycle at 10 and 50 hours, respectively
The Buell/3M filter uses a woven ceramic fiber bag which has been cleaned
by shaking, reverse flow, and back-pulsing. The reported design face velocity
is 2 ft/min (13). This means that the Buell/3M filter is larger than the
other filters by a factor of approximately 4 to 10. The Buell/3M filter was
tested for EPRI at Westinghouse (12).
Of the three filters, the Rockwell filter was challenged by the finest
dust. Even so, it appears to have an efficiency equal to or greater than the
other filters.
310
-------
There are several subtle differences between a conventional fabric filter
and the Rockwell moving bed-ceramic filter. In a conventional filter, the
thickness of the dust cake increases with time, causing the pressure drop to
rise. When the pressure drop becomes excessive, a back pulse is applied.
Except as indicated in the footnote to the table, the high and low points in
the cleaning cycle of the Acurex and Buell/3M filters are given in the table.
The average pressure drop was somewhere between these limits. In the Rockwell
filter, the dust cake was continuously removed by the moving bed. For a given
dust load and sand velocity in the filter, the dust layer reached a steady-
state thickness and then remained constant. The pressure drop also remained
constant at that point. The constancy of the pressure drop and the absence of
pulsing are responsible for the difference in behavior of the two types of
filter with regard to the relationship between the gas velocity and the pres-
sure drop and collection efficiency.
If the dust cake has a constant thickness, the pressure drop varies as
the first power of the velocity. In a conventional filter, the thickness of
the cake increases with time and the rate of change of the cake thickness is
directly proportional to the gas velocity. The net result is that the pres-
sure drop across the changing dust cake in a conventional filter varies as the
square of the gas velocity. Thus, while the pressure drop of a conventional
filter varies as the square of the velocity (14,15), that of the Rockwell
filter (with its constant thickness cake) varies as the first power of the gas
velocity. For this reason, the Rockwell filter should have a lower pressure
drop than a conventional filter. This appears to be borne out in Table 4 when
one takes into account the fact that the other filters were cycling between
the high and low values and the Buell/3M filter is being operated at a much
lower velocity.
The second difference is the dependence of efficiency on gas velocity.
When a conventional filter is pulsed, it temporarily loses its dust cake.
Pulsing also causes some of the deposit to sift through the cloth; therefore,
the greater the pulse activity, the lower the average collection efficiency
(14). Since the pressure drop increases as the square of the velocity, the
corresponding pulse frequency also increases as the square of the velocity.
As a result, the efficiency of conventional filters normally decreases with
increasing gas velocity (14). The Rockwell filter does not employ pulses and,
therefore, should not suffer a loss of efficiency with increased velocity.
CONCLUSION
The Rockwell moving bed-ceramic filter has demonstrated its ability to
remove even submicron particles at high temperatures with very high effi-
ciencies. In tests at 900 to 1500 F using a variety of dusts, including
1.6-micron mass median diameter dust, the outlet dust concentration was so
low that it could not be detected with the instrumentation being used.
Therefore, only a lower limit on the removal efficiency can be given at the
present time. The instruments were capable of detecting 3 x 10~^ and possibly
311
-------
5 x 10 -* grain/scf under the conditions of the tests. Depending on the test
conditions (i.e., inlet grain loading and duration of the tests), the lower
detection limit in the various tests corresponded to removal efficiencies of
99.96 to 99.999%.
The gas velocity in these tests was 9 to 14 afpm, and the pressure drop
was 1.5 to 4.7 in. of water. Compared to commercial low-temperature fabric
filter operation, these test velocities are equal to the upper velocity limit
of commercial fabric filters, but the pressure drop of the Rockwell filter is
less than half that of commercial fabric filters. Because of its low pressure
drop, it is expected that the Rockwell filter can be operated at higher veloc-
ities than those used in these tests. Therefore, one must conclude that it
will be smaller in size than the most compact commercial low-temperature
fabric filters available.
The system provides continuous filtration without pulsing or interrup-
tions in the flows of gas or bed material, operations that tend to damage
ceramic bags, cause "spikes" of dirty gas; and require high-temperature,
high-pressure valves in other systems.
Since the moving bed in the Rockwell design is not the primary filter, it
can be much thinner than "conventional" granular bed filters. The bed mate-
rial does not need to be free of fines. In fact, the presence of some rela-
tively fine particles in the bed aids filtration and the formation of a
protective dust cake. In some applications, such as PFBC, it is not necessary
to clean or recirculate the bed material. The filter should be able to use
the hot spent dolomite sorbent discharged by the PFBC. This would also
eliminate the need for additional lock hoppers.
A hot-gas cleanup system suitable for a combined cycle must remove both
particles and alkali vapors. The Rockwell filter has the distinct advantage
of having the potential to satisfy both these needs. The use of an alkali
getter material in the moving bed is expected to make the Rockwell filter
capable of removing both particles and alkali vapor.
The work described in this paper was not funded by the
U.S. Environmental Protection Agency and therefore the
contents do not necessarily reflect the views of the Agency
and no official endorsement should be inferred.
1 —4
The value of 3 x 10 grain/scf corresponds to a weight gain five times
greater than the sensitivity of the balance on which the sampling filter
was weighed.
312
-------
REFERENCES
1. Squires, A. M., et al. Panel Bed Filters. J. Air Poll. Control Assoc.
20:534 (1970); 21:204 (1971); 21:272 (1971).
2. Phillips, K. E. Energy Conversion from Coal Utilizing CPU-400 Techno-
logy. FE-1536-30, Vol. 1, March 1977.
3. Chamberlin, R. M., et al. An Investigation of Hot Corrosion and Erosion
in Fluid Bed Combustor-Gas Turbine Cycle Using Coal as Fuel. FE-1536-30,
Vol. 2, Appendix B, May 5, 1977.
4. Johnson, L., et al. Support Studies in Fluidized-Bed Combustion.
(a) ANL/CEN/FE-78-3, March 1978; (b) ANL/CEN/FE-77-11, January 1978.
5. Chamberlin, R. M., et al. Advanced Coal Gasification System for Electric
Power Generation. FE-1514-45, January 1976.
6. Wade, G., et al. Granular Bed Filter Development Program Final Report.
FE-2579-19, April 1978.
7. Stringer, J., et al. Assessment of Hot Gas Cleanup Systems and Turbine
Erosion/Corrosion Problems in PFBC Combined Cycle Systems. ASME Gas
Turbine Conference, San Diego, California, March 12-15, 1979.
8. Mulik, P. R., et al. High-Temperature Removal of Alkali Vapors in Hot
Gas Cleaning Systems. DOE/METC Second Annual Contractors' Meeting on
Contaminant Control in Hot Coal Derived Gas Streams, February 17-19,
1982, Morgantown, W. Virginia.
9. Shackleton, M. A. and Kennedy, J. Ceramic Fabric Filtration at High
Temperatures and Pressures. In: EPA/DOE Symposium on High Temperature
High Pressure Particulate Control, Washington, D.C., September 20-21,
1977. p. 194.
10. Shackleton, M. A. and Drehmel, D. C. Barrier Filtration for HTHP
Particulate Control. In: Symposium on the Transfer and Utilization of
Particulate Control Technology, Denver, Colorado, July 24-28, 1978.
Vol. 3, p. 441.
11. Ernst, M., et al. A Regenerative Limestone Process for Fluidized Bed
Coal Combustion and Desulfurization. Monthly Report 107, EPA Con-
tract 68-02-1312, March 7, 1979.
12. Ciliberti, D. F. and Lippert, T. E. Evaluation of Ceramic Fiber Filters
for Hot Gas Cleanup in Pressurized Fluidized-Bed Combustion Power Plants.
EPRI CS-1846, May 1981.
13. Furlong, D. A. and Shevlin, T. S. Fabric Filtration at High Tempera-
tures. In; Proc. of the Sixth International Conference on Fluidized Bed
Combustion, Atlanta, Georgia, April 9-11, 1980. Vol. 2, p. 294.
313
-------
14. Lucas, R. L. Gas-Solid Systems. Chemical Engineer's Handbook,
5th edition, Perry & Chilton (ed.). McGraw-Hill, New York, 1973.
Chapter 20.
15. Billings, C. E., et al. Handbook of Fabric Filter Technology, Vol. 1,
Fabric Filter System Study. PB200648, December 1970.
314
-------
BED MATERIAL >.
MOVING
"SAND"
BED
LOUVER
PANELS *
IMPURI
GAS
INLET
^ B rf~* POROUS
v OAS CERAMIC
W FILTER
OUT SHEETS
GAS -AMOVING \
FEED BED FILTER
LEAF LEAF
DIRTY "SAND" OUT
Figure 1. Rockwell Moving Bed -
Ceramic Filter System Concept
Figure 3. High Temperature
Moving Bed - Ceramic
Filter Test System
SAND HOPPER
SAND FEEDER
SAND RECEIVER
Figure 2. Moving Bed - Ceramic
Filter Test System
2 6 10 20
CUMULATIVE PERCENT FINER THAN STATED DIAMETER (ON A MASS BASIS)
Figure 4. Size Distribution of
Re-dispersed Test Dusts as
Measured with Cascade Impactor
in the Inlet Gas Stream
315
-------
TEMPERATURE - 1210°F
AIWCLOTH RATIO • 10.8 ifpoi
DUST F6ED TURNED ON-OFF-ON-OFF
Figure 5. Filtration of Fly Ash
with Clean Saffil Alumina Paper
and No Bed
I"
102
START SAND FLOW
START DUST FEED
200
TIME (mn
Figure 6. Filtration of Fly Ash
with Initially Dirty Saffril
Paper and Moving Silica Sand
LOUVERS
MOVING BED
CERAMIC SHEET
Figure 7. Top View Showing
the Location of the Three
Pressure Taps
TEMPERATURE • 1600°F
BED VELOCITY - 4 0 ft/h
AIR/CLOTH RATIO » 130ifpm
INLET DUST CONCENTRATION - 2 6 graim/fcf
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160
TIME (mm)
Figure 8. Filtration of Coarse Alumina
Dust with Saffil Paper and Moving
Alumina Bed
316
-------
-— BED VELOCITY = 4 0 ft/h -
BED VELOCITY = 75 ft/h
MEAN DUST DIAMETER = 16^
INLET DUST CONCENTRATION = 1 0 grains/sc
TEMPERATURE = 1500°F
AIR/CLOTH RATIO = 135afpm
Figure 9. Filtration of 1.6/j.m Fine
Alumina Dust with Saffil Paper and
Moving Alumina Bed at Two Bed Velocities
COMPRESSOR/
DEENTRAINER
TRANSPORT
AIR
TRANSPORT AIR
Figure 10. Bed Recycle System
for Separating Dust from Bed
and Recycling Cleaned Material
to the System
- ROCKWELL
FILTER
V
HOT SPENT LIMESTONE
AND ASH
Figure 11. Non-Regenerative Fluidized
Bed Combustor Combined Cycle System
317
-------
TESTING AND VERIFICATION OF GRANULAR BED FILTERS FOR
REMOVAL OF PARTICULATES AND ALKALIS
T. E. Lippert, D. F. Ciliberti, R. O'Rourke
Westinghouse Electric Corporation
Pittsburgh, PA 15235
The work described has been funded by the Department of Energy (DOE)
under Contract DE-AC21-80ET17093.
ABSTRACT
The Granular Shallow Bed Filter (GBF) is proposed as a device to
clean particulates from Pressurized Fluidized Bed Combustion (PFBC) gas
streams. The GBF is a device in which the dust-laden gas passes through
a shallow granular bed, depositing the particulate matter on the surface
of the granular media. The bed medium is cleaned by a reverse flush
that gently fluidizes the bed and elutriates the collected particulate
matter from the system. Described herein are analyses and data that
reflect on the GBF concept as it would apply to a PFBC and preliminary
results of testing done on a six-element subpilot-scale GBF unit.
Results of systems analysis have shown an overall economic
incentive for the GBF in PFBC compared to all-cyclone gas cleanup.
Based on this analysis, performance goals for the GBF have been
identified. A six-element, 24-bed, subpilot-scale GBF has been built
and tested at both ambient and simulated PFBC conditions. Ambient flow
tests were used as a basis to characterize the backflush system and
evaluate candidate bed media. At simulated PFBC conditions, the test
unit has been operated over 170 hours (cumulative), through 475 cleaning
cycles in three test phases. Test variables have included bed media,
filter flow face velocity, backflush conditions, and dust loading.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the views of the Agency and no official endorsement
should be inferred.
318
-------
INTRODUCTION
The Westinghouse Electric Corporation with Ducon, Inc., and
Burns & Roe Inc. are conducting a test and evaluation program of the
Ducon Granular-Bed Filter (GBF) for gas cleaning applications in
Pressurized-Fluidized Bed Combustion (PFBC) Power Plants.
Figure 1 shows a schematic diagram of one element of the
subpilot-scale Ducon Granular-Bed Filter. The element shown consists of
four parallel operating filter compartments or beds. Each compartment
contains a granular filter bed through which the ash and dust-laden gas
pass (Figure 1-a), depositing the ash and dust particles on the surface
of the filter medium. With increasing deposits of the particulates, the
system pressure drop increases until a point is reached when it is no
longer practical or economic to operate. The GBF system is then cleaned
by sequentially backflushing each element (Figure 1-b). For this
purpose, a backflush motive air is introduced into an eductor at the
outlet (clean-air side) of a filter element. The motive air passes
through the eductor (inducing additional clean gas from the clean-air
plenum) and into the GBF element housing, up through each filter bed at
a velocity sufficient to gently fluidize each bed and dislodge the
accumulated ash and dust material. The dislodged ash is elutriated and
expelled from the element by passing back through the inlet opening
provided at the top of each compartment. Figure 1 shows a four-bed
element. Commercial-scale GBF systems would consist of multiple
elements, each comprised of ten, twelve, or more parallel operating beds,
D.g,7?3M80
- MMIvi Air For BicUiuihlnt
'Indued Air For dckfluitilng
ClwnGll
till Plpt
DltiriDutor PMi
la. GBF Element - Filtration Mode Ib. GBF Element - Cleaning Mode
Figure 1. Shallow bed GBF concept.
319
-------
The major advantage for PFBC of the GBF over other advanced
tertiary devices is its overall ruggedized construction, simplicity, and
low operating costs. Its major technical issues are its ability to be
effectively cleaned without dislodgement of filter media, and sustaining
a high collection efficiency while maintaining a stable baseline
pressure drop. Described herein are analysis and data that reflect on
the GBF concept as it would apply to a PFBC and preliminary results of
testing done on a subpilot-scale multielement GBF unit.
PFBC SYSTEMS WITH GRANULAR BED FILTERS
Gas cleanup for PFBC will require the removal of particulate to
meet both turbine erosion tolerance and new source particulate emission
standards and removal of alkali to reduce or minimize metal corrosion
potential. The assessment of gas cleanup options for a PFBC must be
based on technical feasibility and plant economics. For the GBF, a
preliminary evaluation has been made of its relative economics and
performance for two PFBC plant concepts, the PFBC Steam-Boiler and PFBC
Indirect Air Cooled Plants, and a comparison made with PFBC plant
designs that would utilize state-of-the-art cyclones for hot gas
particulate clean-up. Overall conceptual designs for these plants are
provided in References 1 through 6 and have been modified and used in
this study. Table 1 provides a summary of the design basis for each
plant concept. As indicated, there exist different assumptions between
the PFBC concepts concerning combustor and turbine inlet design
parameters. Thus comparison of PFBC concepts is not emphasized.
TABLE 1. PFBC PLANT CONCEPTUAL DESIGN BASIS
Design Parameter
Standards
Coal Type
Net Power ( MWM
Combustor Temperature
Combustor Pressure
Ratio
Turbine Inlet
Temperature I°F)
Gas Clean-Up
Particulate
Alkali
Steam Conditions
-------
In this study, plant cost comparisons are made only for the
particulate removal systems. Figure 2 shows the overall economic
incentive that would exist for the hot gas cleaning of particulates for
the GBF as compared to all-cyclone systems. In the all-cyclone cases
the gas cleanup costs include stack gas particulate removal to meet NSPS
and costs for turbine stator and rotor blade replacement due to particle
erosion. The all-cyclone case for each PFBC concept is shown for two
different turbine-life assumptions. A six-month and one-year blade life
is taken for the steam-cooled plant and a one- and two-year assumption
for the indirect air-cooled plant. These differences in turbine life
assumptions reflect differences in the expected dust loading at the
turbine inlet of the respective plants.
Curve ?3im-A
Figure 2.
Steam Cooled PFBC
COE =67 mill/kwh
0 20 40 60 80 100 120
GBF Filter Face Velocity (ft/mini
0 20 40 60 80 100 120
GBF Filter Face Velocity (ft/mini
Basis.- • Gas Clean-up Tram Includes Primary Cyclone Thru to Expander Inlet
• Costs for Stack Gas Particulate Removal Included if Required to Meet NSPS
• Costs for Turbine Stator/Rotor Reblading
Comparison of gas cleanup costs for PFBC plants employing
GBF or all-cyclones.
For the GBF case, gas cleanup costs are shown as a function of
the assumed gas filter face velocity. The design point is taken at
40 ft/min. At this condition the particulate gas cleaning cost of
electricity (COE) are shown as about 6.0 and 9.0 mill/kWh for the steam-
and air-cooled designs respectively. The higher costs associated with
the air-cooled case are the result of the larger volumetric gas flow per
kilowatt that passes through the particulate removal system. In either
case, the GBF costs (at design point) represent about 8 to 11 percent of
total plant costs. As seen from the figure, operating at lower filter
velocities significantly increases the gas cleaning costs. The
increased number of filter units at the lower velocity along with the
added cost and complexity associated with the gas piping act to nearly
double the COE at 20 ft/min over the 40 ft/min case. Moving toward
higher face velocities above 40 ft/min reduces costs, but the effects
are less dramatic. The diminishing return at the higher velocities
results from assumptions on maximum piping sizes and limiting gas
velocities. These results suggest that pursuing high filter face gas
velocity designs for GBF that impose increased technical risks may not
321
-------
be justified in view of the relatively small incremental improvement in
plant COE that is gained.
At the 40 ft/min nominal design point, the GBF cases are shown
to be lower in costs than the most optimistic all-cyclone case. Lower
turbine blade life and/or increased filter face velocities can
significantly enhance the incentive for the GBF case. Real cost
differences of 3 or 4 mills/kWh represent capital expenditures on the
order of 35 to 50 million dollars.
The PFBC plant economics summarized above are based on assumed
performance levels for GBF and its components. The choice of operating
parameters for the GBF are not arbitrary, but with consideration of
their overall impact on PFBC plant performance. Coupling the GBF with
the PFBC plant must be in consideration of such parameters as inlet dust
loading, system pressure drop, and backflush cycle and flow. A simpli-
fied PFBC systems model has been formulated that incorporates appro-
priate models for predicting the major process operations of the GBF and
the equations for describing the overall PFBC system effects (7).
Results from this modeling have been used to identify test
operating conditions for the GBF subpilot unit test program that would
be representative of operating requirements for commercial-scale
units. For this purpose, one concern has been the design and
performance level necessary for the backflush eductor in the commercial-
scale unit compared to what can be achieved on the smaller subpilot test
unit. The sensitivity of the GBF eductor performance with both
backflush time (t^j) and average pressure drop (Apa ) is shown in
Figures 3 and 4 for the steam-boiler PFBC concept. Similar analyses
have also been conducted for the air-cooled plant design (7). The
ordinate axis in Figure 3 shows the net energy penalty associated with
GBF at the particular GBF operating point compared to a case where no
GBF would be used. The quantity of motive air for the GBF backflush is
dependent on the performance level of the eductor. Eductor performance
is defined as X, the ratio of induced flow to motive flow. The total
backflush flow requirement (motive + induced) is set by the bed medium
fluidization and dust elutriation characteristics. The GBF operating
pressure drop depends on its filtration and cleaning cycle, dust
loading, and gas face velocity.
Results from the parametric analysis show (1) a rather
pronounced sensitivity to backflush time with an indicated "optimum"
between three and six seconds, (2) an "optimum" system pressure drop
(different for the different plant concepts), and (3) near the optimum
backflush time and pressure drop conditions, a general decreasing
sensitivity to the performance of the backflush eductor.
The indicated optimums in the curves showing backflush time
(Figure 3) result as a consideration of the physical design of the GBF
elments and the assumed elutriation characteristics of the deposited
322
-------
Curve No,
X
Ap, 1 psi)
1
1
1.0
2
1
2.0
3
1
i.0
4
4
1.0
5
4
2,0
6
4
5.0
0123
Ratio of Induced to Motive Flow X
2 4 6 8 10 12
Time to Backflush (sec)
Figure 3. Effect of eductor performance and backflush time on PFBC
cycle performance for boiler PFBC.
Curvi 728482-A
I3
I
S 2
I
I 1
Curve
No.
1
2
3
4
«bf
(sec)
2
2
10
10
X
1
4
1
4
Boiler PFBC
0246
Average Pressure Drop Across GBF
APAvg(psl)
Figure 4. Effect of GBF operating pressure drop on PFBC plant
performance for the boiler PFBC cycle.
323
-------
dust cake. Below approximately a one-second backflush time, no cleaning
would occur because of the finite time required for the elutriated dust
to reach the opening in the element housing at the top of the
freeboard. From about one second to the minimum point between three and
six seconds, the effectiveness in elutriating dust is most pronounced,
and the trade-off in compressor power (and the other factors) between
shorter operating cycles or further cleaning favors the latter. Since
the dust elutriation is itself dependent on the in-bed dust concentra-
tion, this trade-off shifts once sufficient dust has been removed from
the bed. It should be emphasized, therefore, that the results indicated
will be dependent on the validity of the elutriation model assumed. As
the curves suggest, should the actual time to elutriate the dust from
the individual GBF beds differ significantly from the indicated optimum,
the sensitivity to both the eductor performance level and the GBF
pressure drop on plant performance increases.
The GBF backflush cycle characteristics are dependent on the
inlet dust loading, gas velocity, and allowed pressure drop. Pressure
drop has both systems performance and element design implications. At
the high operating temperatures (1600°F) in PFBC applications, low
system pressure drop is favored to reduce mechanical design complexity
and capital cost. For a fixed dust loading, low pressure drop implies a
short filtration cycle and, as seen from the curves in Figure 4, results
in a larger reduction in plant performance. The sensitivity of plant
performance to eductor performance is aso increased. Increasing the
system pressure drop increases total cycle time and, therefore, the
fraction of time the GBF is on backflush is reduced, resulting in a
reduced system loss. At some point, depending on plant design, this
trend is reversed, and the impact of increased system pressure drop on
plant performance becomes more detrimental. Although not indicated in
the set of curves shown, it would be expected that lowering the inlet
dust loading would shift the indicated AP "optimums" towards lower
values.
The principal conclusions for the GBF drawn from the above
described analysis are as follows:
1. There exists an optimum set of GBF system operating
parameters that correspond to minimum plant impact. At the optimum
point, the overall penalty of the GBF on plant performance is
approximately one percent or less.
2. GBF cleaning cycles of between three and six seconds appear
optimum. These times represent considerably shorter backflush cycles
than previously tested or thought necessary. Should significantly
longer backflush times prove necessary to accomplish cleaning, larger
overall plant performance penalties may be incurred.
3. The optimum operating pressure cycle for the GBF system
varies with plant concept and inlet dust loading. Tolerable levels of
324
-------
in-bed dust fines and operating temperature level may also be
constraining factors.
4. The attainment of high eductor performance levels, i.e.,
ratio of induced to motive backflush flows greater than 3 or 4, would
not appear to be a necessary objective for commercial GBF application in
PFBC. The overall impact of the eductor performance appears relatively
small except at frequent cleaning cycles. Flow ratios between one and
two should be adequate if the indicated optimum backflush times are
achievable,
GBF SUBPILOT SCALE TEST RESULTS
TEST UNIT DESIGN AND INSTRUMENTATION
The six-element, shallow bed granular filter test unit is shown
in Figures 5 and 6. The test unit design is premised on a nominal flow
of 500 acfm at 1600°F (870°C) temperature and operation at 11 atm
pressure. The unit consists of six filter elements, each element
containing four filtering compartments or beds as schematically
illustrated in Figure 7. As the photograph of the test unit in Figure 6
shows, each GBF element attaches to the eductor section at flange
connections. The eductors are each fixed (welded) to the dished head
and flange support section. The dished head and flange support section
separate the dirty and clean gas sides of the filter unit. The flange
section clamps between the vessel and vessel dome section Figure 5. In
each bed of each element there exist provisions for both a thermocouple
D-5. 7757*73
Backflush Lines and
/ Nozzle
Dished
Head and
Support flange
Eductor
Eductor to Element
Flange Connection
Flow Deflector
Plate
•—Flow In let
View A-A
Sect. B-B
Figure 5. General arrangement of six-element GBF subpilot-scale test
unit.
325
-------
6a. Top view
6b. Test unit being lifted to test facility pressure vessel.
Figure 6. GBF Subpilot test unit.
326
-------
A
Bed Media
h|
B
1 I
J (.
J I
C
..~
D
.Bed
x— N I Pressure
/2^pA j
JX3/ laP
—-^-i — Bed Thermocouple
/A^\J
!~ - Bed Thermocouple
"~~Screen and Distributor
Plate
Figure 7. Schematic representation of GBF flow, pressure, and
temperature measurements.
and pressure tap. Pressure taps are also provided across the eductor
section of each element. Figure 7 also illustrates the instrumentation
provided each GBF element. The backflush lines, visible in Figure 6,
are one-inch diameter tubing sections that pass through the ring flange
and connect to one-inch diameter flexible stainless steel tubing. These
in turn are fastened to the inlet section of each eductor, positioned
and held by the spider arrangement shown. The flexible tubing is used
to accommodate thermal expansion during heat-up and backflush temperature
transients. This test unit design was used through test Phase II. For
test Phase III, modifications were made to the GBF unit to circumvent
problems experienced in the earlier test phases in uniformly distributing
the backflush flow between parallel operating filter beds. Figure 8
illustrates the test unit modifications.
HOT GAS TEST RESULTS
A series of tests on the six-element GBF subpilot test unit have
been conducted at the Westinghouse Synfuels Division's (V_ SFD) Test and
Development Center (TDC). These tests have been conducted using a test
facility that simulates PFBC operating conditions. The test passage,
schematically illustrated in Figure 9, was operated at 1600°F (850°C)
nominal gas temperature and up to 150 psig (11 atm) pressure and a mass
flow corresponding to about 500 acfm. This provides a GBF filter face
velocity of 40 ft/min. Reentrained PFBC ash is injected into the test
passage through a specially designed pressurized-brush dust feeder. The
dust in these tests was the second-stage cyclone catch obtained from the
Curtiss-Wright PFBC Technology Rig.
327
-------
-Clean Gas Side-
5/8" Ola. S. S. -Tubing Backflush Line
with Thermal Expansion Loop and 0.286 Limiting Orifice
5/8" Flow Restricting OrlUce
Section A-A
Top View
DW9.125MM
A-A
Modified Screen
and Distributor
Spot
Weld
Figure 8. Test unit modifications.
Owq. 1718826
Operating Conditions
Pressure - Up to 150 psig I capability to 220 psi)
Temperature - 200 - 1600°F
Flow Rates - Up to 12 Ib/s
Vessel - 56" Oia x 110" Length
Piping -10" Sh. 80 with 6" Inconel Liners
Airi r
r\ L
\/ t
Air Compressors I
Fuel
Alkalis II „ No. 2 Fuel 0
1 1 <_i 1
r— O1
:Preheater r;
— "\ =
Process
Air
^1 1
~2( Combustor p""1
Fuels
Blending
Tanks Atomizing Air — '
^ Particulate
3 Feeding
System
Rupture
Disc "^
r>^r\- • '
Particulate f""^.
Sampling
1 .
Control ^~f^
Room By-Pass ^
Hot Gas
Cleaning
Pressure
Vessel
Particulate
-» Sampling
Muffler
Chamber
Figure 9. Schematic of hot gas cleanup facility.
Three distinct series of hot gas testing were conducted, and
overall results and accomplishments of the test program are summarized
in Table 2. The tests were conducted in 8 to 12 hour segments and
covered a cumulative time interval of 170 hours with approximately 475
cleaning cycles attained. During the early test phases, test operations
were plagued by several mechanical problems that occurred to the test
unit. These included leaks in the gasket seal between each of the GBF
elements and their respective eductor section. Also, several of the
fittings on the backflush lines had detached or leaked. These problems
were subsequently corrected before the second test phase was initiated.
The overall operational performance of the test unit is
determined by its baseline pressure drop, overall dust collection
328
-------
TABLE 2. SUMMARY OF GBF TEST ACCOMPLISHMENTS THROUGH PHASE III
Test Phase
&
Configuration
-I-
• 6 Element/24 Bed
• 1370 /Jm Alumina
Media
-II-
• 6 Element/24 Bed
• 1370 Aim Alumina
Media
6 Element/6 Bed
620 /im Sand
Media
Test
Conditions
(Nominal)
• 1600°F/165 psi
• 50 Mrs Operation
• 103 Cycles
• 40 Ft/Min
Major Findings
Accomplishments Limitations
• Integrated GBF Operation • Gasket Seal Leaks
With On-Line Cleaning • Failed Backflush Lines
• Backflush Times As Low
As 6 Sec.
• Low Baseline Ap
(10 In. H20)
1600°F/165 psi • Confirmed Baseline Ap
SO Mrs. Operation & Cleaning On-Line
90 Cycles • Overall Dust Colletion
20 To 40 Ft/Min Efficiency = 97.3%
• Significant Bed Media
Loss
• Warping Of Distributor
Plate Assembly
1500°F/125 psi • Stable Baseline Ap But • Relatively Inflexible
71 Hrs Operation Higher (40 To 80 In. H20) Test Constraints
282 Cycles • Overall (Test Average) • Off-Line Cleaning
50 To 100 Ft/Min Dust Collection Required
Efficiency = 99.2% • Some Bed Media Loss
(18 To 30%)
efficiency, and demonstrated backflush cycle. Figure 10 shows a segment
of the GBF pressure drop measured during system operation in Phase I.
The top portion of the trace shows test segments where the duration of
the backflush cycle is altered from 30 seconds to 6 seconds. In these
tests dust was fed to the GBF unit until a prescribed system pressure
drop was achieved (about 15 to 20 in H20) at which time the dust feed
was halted. With the hot gas still flowing, each element of the test
unit was sequentially backflushed. The dust remained turned off to
enable the baseline pressure (pressure drop after backflush) to
stabilize. Several such cycles were repeated for each backflush time
indicated. From these test results it appeared that a stable baseline
could be established even for the 6-second test conditions. Thus, it
would appear that each operating bed was successfully backflushed and
the dust expelled from the filter unit would settle into the containment
vessel (as opposed to being carried into one of the other operating
elements). The 6-second backflush cycle time corresponds to the near
optimum established from the PFBC system studies (Figures 3 and 4).
The lower portion of the figure show a reproduced segment from
the pressure drop trace where dust is continuously fed in a manner
simulating actual PFBC operation. In this test segment, the backflush
cycle was set at 6 seconds. Four operating cycles are shown
corresponding to Ap's from 10 to about 30 in. of IkO, with cycle times
ranging from about 10 to 20 minutes. Conditions for these tests
correspond to 1600°F, 150 psig (870°C, 11 atm), 40 ft/min filter face
329
-------
40
|« *>
£ 20
i 10
0
30 sec/30 sec
15 sec/15 sec
Curve 731192-A
6 sec/6 sec
10
20
10 20 '
Time! mini
20
CHARACTERISTIC GBF PRESSURE DROP - CONTINUOUS DUST FEED
60
50
S 40
CSJ
I »
I 20
<
10
0
10
30
40 50
Time ( mini
60
70
80
Figure 10. Characteristic GBF pressure drop for different backflush
cycles.
velocity with an average inlet dust loading of about 7500 ppm. This
dust loading is considerably higher than would be expected in a PFBC at
the exit of a second-stage cyclone and even higher than might be
experienced at the exit of a nonrecycle primary cyclone unit. The high
dust loadings in these tests provide for achieving a large number of
operating cycles in relatively short test times.
Test data taken during the latter portion of Test Phase I and
throughout Test Phase II showed that during the backflush cleaning
cycle, the backflush flow was not distributing uniformally between the
filter compartments in each element. This was suggested during test
operations by the bed temperature measurements. Figure 11 shows an
example trace of the bed temperature measurements made in one element.
During filtration, the bed thermocouples in each filter compartment
(A,B,C, and D) show bed temperature close to the hot gas conditions.
During backflush, relatively cold motive flow is provided that passes
through the filter beds, causing a momentary temperature transient. As
indicated by the recorded temperatures shown in Figure 11, only the A
and C beds appear to show any temperature transient during the backflush
cycle. These results (typical of all six operating elements of the GBF
unit) suggest that the B and D beds in each elenent (half the total
filter beds) were not seeing any significant backflush flow. In
addition, those beds being backflushed did not appear to receive equal
flow based on the magnitude of the recorded temperature transient.
The consequence of operating through a large number of backflush
cycles with high backflush flow and poor (or no) flow distribution
between beds is high superficial gas velocity and the possibility of
elutriating bed medium on backflush. Subsequent inspection of the test
unit at the end of Test Phase II confirmed this observation. In all six
330
-------
Curve 71*0517-6
1600
75 100
Time (sec)
Figure 11. Measured bed temperatures during GBF operation, element 4.
GBF elements, the "C" beds showed nearly complete loss of the alumina
bed medium. Likewise, most of the "A" beds showed all or substantial
bed medium loss. The poor distribution of backflush flow also results
in the inability to clean those beds that are starved of backflush
flow. This in turn can overload the cleaned beds during the filtration
cycle. In Test Phase III, the problem of the distribution of backflush
flow was circumvented by eliminating three of the four filter
compartments in each GBF element (see Figure 8).
Even though a significant maldistribution of backflush flow was
evident over the course of the Phase I and Phase II test programs,
relatively stable baseline pressure drops (i.e., operating pressure drop
over the filter unit after cleaning) were achieved. These ranged from
between 5 and 10 in. H20 for the 1370 ym alumina bed medium. In test
Phase III significantly higher operating baseline pressure drops were
experienced. These ranged from between 40 to 60 in. lUO and are the
result of utilizing the modified GBF test unit that incorporated a
double distributor plate assembly and the 620 urn sand as filter
medium. Phase III testing was also conducted at somewhat higher filter
face velocities and required the test unit to be cleaned off line, a
331
-------
result of the modification made to circumvent the distribution of
backflush flow problem experienced earlier in Test Phases I and II.
Under these conditions and slightly modified backflush parameters, a
steady operating baseline pressure drop was achieved. PFBC systems
analysis suggests that GBF operating pressure drops of 40 to 60 in. H-0
would not significantly affect system performance or economics.
Dust sampling taken during the early portion of Test Phase II
for the 1370 ym alumina medium and throughout Test Phase III for the 620
yrn sand medium have been used to evaluate filter dust collection effi-
ciency. Figure 12 shows the measured grade efficiency curves for each
medium and the test basis. Table 3 gives a preliminary comparison of
the GBF test results with PFBC requirements and includes comparison with
an all-cyclone case. Neither the all-cyclone case nor the GBF with the
1370 vim bed medium appear to provide sufficient particle collection to
meet PFBC system requirements. The GBF tests with the 620 ym sand medium
show performance levels considerably improved and more than sufficient
to meet New Source Performance Standards based on total particulate as
well as a projected turbine life exceeding two years. These test
results show considerable promise for the GBF for hot gas cleanup.
100.0
98.0
1*
g
JE
£
92.0
90.0
» 620 Mm Sand Media-3 in. (7.6cm) Deep Bed
Pressure = 120 psia (827 kPa)
Temperature =1500°F (815°C) Nominal
Filter Flow = 50 ACFM/Ft2 (15.2 m^/min m2)
Overall Collection Eff. =99.2% Test Average
• Tabular Alumia Bed Media d = 1370pm at
P = 165 psia (H38kPa)
T=1500°F1815°C) Nominal
Flow =40 ACFM/Ft2 (tt 2 m3/min m2)
Overall Collection Eff. = 97.3*
_L
10 12 14 16 18 20 22 24 26 28 30
Particle Size (gm)
Figure 12. Measured grade efficiency curve for granular bed filter.
CONCLUSIONS
Preliminary test results on the GBF and analysis show that the
GBF should be capable of performance levels sufficient to meet PFBC hot
gas particulate cleaning requirements. Testing with a six-element, six-
bed test unit arrangement, overall collection efficiencies averaged 99.2
percent under simulated PFBC conditions. Overall system pressure drops
and backflush cleaning parameters achieved in test operations appear
332
-------
TABLE 3. COMPARISON OF GBF TEST RESULTS WITH PFBC REQUIREMENTS
Case
GBF
Alumina
1370 urn
GBF
620 Mm
Sand
All
Cyclones
Ash
C.W. -2nd
Stage_Cyc.
Ash 5.. = 10 pm
Mixed Ash'11
djg = 7. 7 M m
PFBC
Combustor
71 Overall
97.3%<2'
99.2%(3)
87. 5 to
96.3*
Projected Outlet
Loading gr/scf
0.027141
0.008(3) '
0.06
NSPS=. 013 gr/scf
(.03lb/in6Btu)
No
Yes
No
Projected
Turbine Life
2Yr.
lYr.
11) 66% by VVt. C. W. Ash Mixed with 33% by Wt. Ground Second Stage Catch from Exxon Mini Plant
(2) Representative of Best Data - Phase I & II
(3) Test Average - Phase III
(4) Assumed Inlet Loading = 1.0 gr/scf
consistent with the requirements for a commercial-scale unit for econo-
mic operation. Further testing and analysis are planned to further sub-
stantiate the GBF design basis and to pursue still higher performance
levels.
REFERENCES
1. "Engineer, Design, Construct, Test and Evaluate a Pressurized
Fluidized Bed Pilot Plant Using High Sulfur Coal or Production of
Electric Power", Curtiss-Wright Corporation, Woodridge, NJ, 07075,
FE-1726-20A, March 15, 1977.
2. "Preliminary Assessment of Alternative PFBC Power Plant Systems",
Burns & Roe, Inc., Woodbury, NJ, 11797, EPRI CS-1451, Research
Project 1645-2, July 1980.
3. CFCC Development Program, DOE Commercial Plant Economic Analysis
(Task 1.6), Contract No. EX-76-C-01-2357, General Electric Co.,
Schenectady, NY, Preliminary, June 1979.
4. CFCC Development Program, DOE Commercial Plant Design Definition
(Task 1.2), Contract No. EX-76-C-01-2357, General Electric Co.
Schenectady, NY, March 1978.
5. "Design of Advanced Fossil Fuel Systems Study, Air-Cooled
Pressurized Fluidied Bed Power Plant", Draft Report, Prepared for
ANL Contract No. 37-109-38-6212 by Bechtel Group, Inc., December
1981.
333
-------
6. "Design of Advanced Fossil Fuel System Study, Steam-Cooled
Pressurized Fluidized Bed Power Plant", Draft Report, Prepared for
ANL Contract No. 37-109-38-6212, by Bechtel Group, Inc., November
1981.
7. "Testing and Verification of Granular Bed Filters for the Removal of
Particulate and Alkalies", Third Quarterly Project Report, April 1,
1981 through June 30, 1981, DOE Contract DE-AC21-80ET17093.
334
-------
BAGHOUSE OPERATION IN GEORGETOWN UNIVERSITY
COAL-FIRED, FLTJIDI EID-BED BOILER PLANT, WASHINGTON, D.C.
by: Victor Buck
Pope, Evans and Robbins Incorporated
New York, New York 10004
and
David Suhre
Georgetown University
Washington, D.C. 20057
ABSTRACT
Since 1979, Georgetown University has operated the nation's first com-
mercial sized, coal-fired, fluidized-bed boiler plant for over 10,000 hours,
utilizing a baghouse for particulate emissions control.
In plant startup, the bags are first coated with limestone dust by op-
erating the forced draft and induced draft fans to fluidize the bed. This is
followed by firing of No. 2 fuel oil to preheat the boiler and the limestone
bed. Upon achieving 100 psig boiler steam pressure, coal is introduced into
the preheated bed and ignited by the oil burner to initiate boiler operation.
The baghouse has proven to be an efficient particulate collector. How-
ever, excessive pressure drop cross the baghouse has proven to be an ongoing
problem. Various baghouse modifications have been implemented and different
bags tested. This paper presents the results of this operation.
INTRODUCTION
Georgetown University has operated its 100,000 pounds per hour, 625
psig, coal-fired, fluidized-bed boiler plant in Washington, D.C. since July
1979. This is a national demonstration plant funded by the Department of
Energy for the two-fold purpose of:
Operating an atmospheric fluidized-bed boiler burning high-
sulfur coals in an environmentally acceptable manner in an
urban institutional complex, and
. Obtaining sufficient information from the prototypical op-
eration to encourage industry to move directly into the
design and construction of commercially warranted indus-
trail size fluidized-bed boiler units.
335
-------
Table 1 indicates the design basis for the Georgetown boiler.
TABLE 1. DESIGN BASIS - GEORGETOWN UNIVERSITY FLUIDIZED BED BOILER
STEAM FLOW
OUTLET TEMP./PRESS
DESIGN COAL
Heating Value
Ash %
Moisture %
Sulfur %
DESIGN PARAMETERS
Bed Operating Temperature
Ca/S Ratio (for SO Compliance)
COAL FEED SYSTEM
COAL SIZE
COAL FLOW RATE
LIMESTONE SIZE
LIMESTONE FLOW RATE
RE-INJECTION FLOW RATE
FLUE GAS FLOW
MAXIMUM BAGHOUSE FLUE GAS
TEMPERATURE
INDUCED DRAFT FAN FLOW RATE
INLET PRESSURE AT INDUCED
DRAFT FAN
100,000 Ibs/hr
Saturated 414°F/275 psig, or
493°F/625 psig
Bituminous
12,750 Btu/lb
7.97%
5.0%
3.29%
1594F
3/1
Stoker (2) - Side Wall Mount
1-1/4" x 1/4"
9,565 Ibs/hr
1/8" x 16 Mesh (0.0469")
3,133 Ibs/hr
7,500 Ibs/hr
120,000 Ibs/hr
400°F
44,430 ACFM
(-) 18"
The boiler burns eastern bituminous coal which, in practice has had a
sulfur content averaging between 2 and 2-1/2 percent, an ash content of from
10 to 16 percent, 25 percent and greater volatiles, and up to 5 percent mois-
ture. Coal is sized 1-1/4 inch x 3/8 inch for overbad stoker feed into the
boiler. The proportion of fines (<1/4 inch) has at times reached 70 percent.
As a result, a considerable amount of unburned carbon is elutriated with the
flue gas stream by the upward flow of combustion air thorugh the boiler bed.
Limestone, which constitutes the basic bed material in the boiler for
purposes of sulfur capture, averages about 1300 microns in size. By design
intent, delivered limestone is sized to fall within the limits indicated in
Table 2. These limits were based on early developmental work and had been
found to insure optimum sulfur capture.
336
-------
TABLE 2
DELIVERED LIMESTONE SIZE (DESIGN BASIS)
1/4"
6 Mesh
8 Mesh
10 Mesh
16 Mesh
20 Mesh
30 Mesh
(0.132")
(0.094")
(0.079")
(0.047")
(0.033")
(0.023")
100 %
98-100%
85-95 %
70-80 %
20-40 %
10-20 %
3-5 %
In operation, limestone is delivered by gravity into the two boiler beds
as shown on Figure 1. In the presence of heat, the limestone is calcined and
the calcined lime in turn reacts with the sulfur dioxide (SCL) produced by
burning coal to form calcium sulfate (CaSO.) thus limiting sulfur dioxide
emissions in the manner expressed by the following reactions:
CaCO (limestone) + Heat = CaO + CO.
CaO
1/2
= CaS0
BED DRAIN COOLER
WITH AIR LOCK
FIGURE 1
PLANT CROSS SECTION
THROUGH FLUE GAS STREAM
337
-------
Combustion air entering the boiler from below serves to place the bed
in suspension, i.e. the bed is fluidized. In the process, fine particles
of bed material as well as of the coal fuel is driven off in the flue gas
stream.
Immediately downstream of the boiler, a mechanical cyclone collector re-
moves the bulk of the large entrained particles from the flue gas for rein-
jection back into the boiler bed. This serves to improve boiler efficiency
by achieving more complete coal combustion and by increasing the amount of
sulfur capture per unit of limestone used.
Smaller flue gas particles which pass the mechanical collector are
directed to the baghouse for final cleanup.
INITIAL PLANT OPERATION
The baghouse initially installed in the plant was manufactured by En-
viro-Systems and Research, Inc. It consisted of a 22 cell structure, with
each cell containing 36-5 inch diameter by 8 feet 6 inch long bags - a total
of 792. Bag cleaning was performed off-line by reverse air. The air to
cloth ratio was 4.60 with one cell cleaning; 4.39 with all cells active.
The cleaning cycle was initiated automatically whenever the pressure drop
across the baghouse exceeded 4 inches. It continued to subject the bags (in
one cell at a time in sequence) to reverse air cleaning until the pressure
drop reduced below 4 inches. In the original installation, the first six
cells were fitted with Nomex glass bags and the remaining 16 cells with
Teflon felt bags.
Boiler lightoff involves preheating the boiler and the limestone bed by
means of a No. 2 fuel oil igniter until a drum pressure of about 100 psig
is reached. Coal feed is then initiated. The overall lightoff period,
until the boiler is operational producing steam, has averaged about 4 hours.
As a precautionary measure against bag fouling due to condensing of volatile
hydrocarbons, the bags were precoated with limestone dust by operating the
forced draft and induced draft fans before beginning lightoff.
Problems developed early due to increasing pressure drop across the
baghouse. At boiler loads of less than 50 percent, this drop reached 13 to
14 inches, well in excess of the expected 4 inch drop at full load. The
Teflon felt bags, in particular, were blinding and became impossible to
clean by the reverse air method.
Early in 1980, the manufacturer modified the baghouse by the addition
of a pulse jet cleaning system which would operate coincident with the re-
verse air cleaning cycle in each cell. In theory, the pulse jet would dis-
lodge the particles while the more sustained reverse air flow would drive
the particles further from the bags and thereby assure that a greater pro-
portion of dislodged material would not be recaptured on the bag when the
cell was returned to the cleaning mode. Coincident with this modification,
the manufacturer also replaced the Teflon felt bags with Nomex glass as the
latter type, though limited to a 400°F operating temperature, had given
evidence of better cleaning characteristics.
338
-------
This modification appeared to have succeeded in reducing the pressure
drop across the baghouse but in time, as the boiler was operated at ever in-
creasing loads, the drop across the baghouse again reached unacceptably high
levels. Various attempts at improving operations by adjusting the applica-
tion of the pulse jet to various points in the reverse air cleaning period
did not appear to improve bag cleaning and reduce the pressure drop. In
addition, bag failures became a significant problem with the Nomex fabric
and replacements were made with Teflon felt bags.
During this entire operating period, the demonstration boiler plant op-
eration was also undergoing refinements, many of which interacted with the
baghouse operation. Included among those changes were flyash reinjection
improvements, and improved boiler instrumentation leading up to automatic
operation. The flyash reinjection system required modifications to deter-
mine equipment which was capable of reinjecting all fines collected in the
mechanical collector back into the boiler. Obviously, when the mechanical
collector hoppers filled up, all flyash passed through to the baghouse for
final collection. In general, this had the following e.ffect upon the flyash
entering the baghouse:
Larger average particle size,
Higher carbon content,
Higher calcined lime content.
By late 1980, the installation of an air lock below one of the five
mechanical collector hoppers resulted in full reinjection of collected fly-
ash from this point. Air locks were subsequently installed on all hoppers,
and have been operational since July 1981.
Boiler instrumentation required modifications and refinements before
automatic operation of the plant was attainable. Due to poor performance
of the gas analyzer systems at the beds and stack, operators had a tendency
to operate with a higher than design level of both excess air and also lime-
stone feed for sulfur capture. Higher excess air rates increased the flue
gas flow rate through the baghouse above design levels for a given boiler
load. Higher limestone feed rates than required for maintaining emissions
level below allowable led to higher dust loadings in the flue gas stream due
to fines elutriation. Both factors served to hamper evaluation of baghouse
pressure drops, but were not considered to be governing factors. Final in-
strumentation changes were implemented in the early summer of 1982 and the
boiler is now capable of sustained operation in the full automatic mode.
INTERIM MODIFICATIONS
In April 1981, spot modifications were made to the baghouse as follows:
,. Modified pulse jet system in one cell to inject more air
during cleaning cycle;
. Added "Staclean" diffusers with venturies to all bags in one cell
to improve distribution of the pulse during cleaning.
Added "Staclean" diffusers without venturies to one row (6 bags)
in a separate cell.
339
-------
During a subsequent 12 day operating period, the baghouse was cleaned
continuously with both pulse jets and reverse air in which a 0.1 second pulse
was imposed upon the bags 3 seconds into the reverse air cleaning cycle. The
reverse air flow continued for 2 seconds after the pulse.
Tests on the bags following this experiment indicated that bags sub-
jected only to reverse air cleaning were cleaned the least. The greatest
improvement was found in cells in which the pulse jet volume was increased.
In a follow-on test, the reverse air cleaning was then eliminated and further
improved bag cleaning was noted using only the modified pulse jet system.
Based on these tests, the baghouse was modified throughout to increase
the pulse jet air flow into each bags. However, operations following this
modification did not bear out the expectations envisioned from the above
tests as shown in Figure 2.
Late in August 1981, the decision was made to replace all remaining
Nomex glass bags with the original Teflon felt in order to reduce further
bag losses. Nomex, while responding better than Teflon to bag cleaning, was
succeptible to bag damage due to flue gas temperature excursions above 400 F,
and to excessive bag wear in this particular baghouse.
Throughout the latter part of 1981, in addition to the above bag re-
placements, the University made adjustments in the operation of the baghouse
including those listed below:
Duration of pulse,
Pulse Header Pressure,
Timing of pulse with respect to reverse air,
Increasing number of pulses during cleaning cycle from
one to two, and
Eliminating the reverse air function.
None of these changes resulted in significant improvements in the bag-
house pressure drop which continued to increase to a point where boiler load
was limited to 50% full rated capacity.
Through 1981, the bags employed at this installation were predominantly
either of Teflon felt or Nomex glass. In late 1981, all bags were replaced
with Goretex Teflon B fiberglass as a last measure to obtain acceptable pres-
sure drops across this baghouse. Subsequent operations indicated that with
the Goretex bags, the boiler could be operated at 90 percent output, but with
a drop across the baghouse in excess of 10 inches. Whereas these bags were
in service for just two months, no reliable information on bag failure could
be obtained. It was noted that in this period, a boiler tube leak caused a
plant shutdown and upon returning to operation 10 days later, the bags re-
verted to the pressure drop that existed prior, an indication that despite
high moisture content in the flue gas stream, the bags were capable of re-
sponding to the cleaning cycle. The overall results of this experiment,
however, led to the conclusion that in order to reduce the drop across the
baghouse to acceptable levels at all loads, the baghouse would require major
modification.
340
-------
MAY 1981 (BEFORE PULSE MODIFICATION)
BAGS -20% NOMEX GLASS, 80% TEFLON FELT
9 5
90
85
— 80
-------
FINAL MODIFICATION
Based upon a report prepared by Davy McKee, Consultants to DOE, the bag-
house at Georgetown University was completely rebuilt between April and July
1982. Salient features of the revised baghouse as constructed by MLkro-Pul
Corporation are:
1. Four cell construction with capability of isolating one cell
for maintenance or inspection without interrupting operation;
2. 308 bags per cell, each bag 4.5 inches in diameter by 10 feet
in length;
3. Pulse jet cleaning in on-line mode;
4. Air to cloth ratio of 3.07:1 (on-line cleaning);
5. Compressed air requirement - 80 cfm @ 100 psig (on-line
cleaning);
6. Bags - Felted Fiberglass "BWF" 25 oz/sq yd.
The new baghouse configuration is shown on Figure 3.
•ED DRAIN COOLER
WITH AIM LOCK
FIGURE 3
PLANT CROSS SECTION
AFTER 1982 BAGHOUSE RECONSTRUCTION
342
-------
Since plant operations were resumed in early August 1982, a total of
1,100 hours of steam generation have been logged up to October 1 at output
levels ranging between 55 and 100 percent of boiler rated output. The pres-
sure drop across the baghouse has remained well within operational limits.
Figure 4 depicts the pressure drops through the flue gas system for a three
week period in September 1982. It is expected that the cleaning cycle will
require some further adjustment until a stable operating point is reached
that will permit sustained operation at up to boiler full load without ex-
cessive drop.
tn
3
o
z
a
<
CD
IU. U
9 5
90
8 5
8.0
7.5
7.0
6.5<
8.0
55
5.0
"5
4.0<
3.5
3,0
2.5
2.0
I.S
1.0
0.5
-
-
® ®
[ g,
i ® ® ®
n ...
a D a D a
n D a a o o o
? 8 D D o o o ° 0 °° ° ° o
i o 0 <
I- O
1-
-
-
* CHANGED 4 MODULE
CONTINUOUS CLEANING
CYCLE FROM 9 TO 6.3
MINUTES
IMtTlATT llt^II A
— IN 1 MM 1 t nlfan Op
ALARM -6" Ap
LEGEND-'
O - 70 <80« I05
D - 80 £ 90 « I03
® - 90S 100 K I03
• - > 100 x I03
13 14 IS 16 17 18 19 20 21 22 23 24 25 26 27
DAY OF MONTH
28 29 30
FIGURE 4
BAGHOUSE AP VS TOTAL AIR FLOW (Ibs/hr)
SEPTEMBER 1982 ( WITH REBUILT BAGHOUSE)
FLY ASH COMPOSITION
In fluidized bed combustion, the fly ash composition may vary consider-
ably within a given plant, the composition largely determined by the type,
quantity, size and make-up of the coal and limestone delivered to the boiler.
At Georgetown University, samples of flyash were taken from the flyash
silo at intervals and analyzed by the DOE laboratory in Alexandria, Virginia.
The tabulation in Table 3 below summarizes the results and demonstrates the
variations that are found.
343
-------
TABLE 3
FLYASH SAMPLES ANALYSES (%)
Measured Value Range Mean
C (total) 20.55-35.61
S 1.65-2.68
CaO 14.81-20.58
CaS04 8.11-11.38
Si02 20.27-26.19
A1203 7.95-10.60
Fe203 4.47-8.96
MgO 0.13-1.06
L.O.i 29.77-37.16
HHV (Btu/lb) 3303-5163
Bulk Density 28.4-36.5
31.31
2.25
17.26
9.53
22.84
9.42
6.34
0.56
34.58
4473
32.33
(Ib/cf)
Ave. Micron 37.6-51.8 43.16
Size
A word of caution applies in interpreting the above values. The samples
were taken from the ash silo and therefore may have undergone further reac-
tion from the time that the material was collected at the baghouse. After
collection, the samples were not always analyzed promptly and hence the op-
portunity existed for further reactions to take place before the analysis
was made. They are indicative, however, of the range that may be encounter-
ed in this type of plant.
CONCLUSIONS
The baghouse at Georgetown University's fluidized bed boiler plant has
performed to limit particulate emissions well below D.C. allowable limits.
The problem of excessive pressure drop across the baghouse appears to have
been overcome with the rebuilding of the baghouse. However, the performance
evaluation of bags now in place must await further periods of operation.
The operating experience at Georgetown has been documented in a series of
quarterly reports issued by DOE and listed in the References below. The re-
sults cannot be considered typical of all fluidized bed boiler plants due to
differences from plant to plant in the type of boiler, coal and limestone;
the size and method of feeding coal and limestone; and other factors. It is
344
-------
expected, however, that with the accumulation of similar information on
other operational fluidized bed boiler plants, the selection of a baghouse
and bag material for this type of plant can be made in the future with assur-
ance that the initial selection will operate successfully over the full
range of boiler output.
ACKNOWLEDGEMENTS
Capital funding for this project was provided by the U.S. Department of
Energy.
The work described in this paper was not funded by the
U.S. Environmental Protection Agency and therefore the
contents do not necessarily reflect the views of the
Agency and no official endorsement should be inferred.
REFERENCES
1. Davy McKee Corporation, Fabric Filter Design for Application in the
Fluidized Bed Combustion of Coal, DOE/ET/10123-1171.
2. Georgetown University, Industrial Application of Fluidized Bed Com-
bustion, Quarterly Technical Progress Report, July-September 1979,
DOE/FE/2461-13.
3. Georgetown University, Industrial Application of Fluidized Bed Com-
bustion, Quarterly Technical Progress Report, October-December 1979,
HCP/T2461-13, UC-9UE.
4. Georgetown University, Industrial Application of Fluidized Bed Com-
bustion, Quarterly Technical Progress Report, January-March 1980,
HCP/T2461-13, UC-9UE.
5. Georgetown University, Industrial Application of Fluidized Bed Com-
bustion, Quarterly Technical Progress Report, April-June 1980,
HCP/T2461-13, UC-9UE.
6. Georgetown University, Industrial Application of Fluidized Bed Com-
bustion, Quarterly Technical Progress Report, July-September 1980,
METC/DOE/10381/135.
7. Georgetown University, Industrial Application of Fluidized-Bed Com-
bustion, Quarterly Technical Progress Report, October-December 1980,
HCP/T2461-13, UC-9UE.
345
-------
8. Georgetown University, Industrial Application of Fluidized Bed
Combustion, Quarterly Technical Progress Report, January-March
1981, EO/ET/10381-197 (DE 81030272).
9. Georgetown University, Industrial Application of Fluidized Bed
Combustion, Quarterly Technical Progress Report, April-June
1981, DOE/ET/10381-1109 (DE 82006241).
10. Georgetown University, Industrial Application of Fluidized Bed
Combustion, Quarterly Technical Progress Report, July-September
1981, DOE/ET 10381-1143.
11. Georgetown University, Industrial Application of Fluidized Bed
Combustion, Quarterly Technical Progress Report, October-December
1981 (In printing).
12. Georgetown University, Industrial Application of Fluidized Bed
Combustion, Quarterly Technical Progress Report, January-March
1982 (In preparation).
13. Georgetown University, Industrial Application of Fluidized Bed
Combustion, Quarterly Technical Progress Report, April-June 1982
(In preparation).
14. Pope, Evans and Robbins Incorporated, Baghouse Test Program
Final Report, April 1980 (Prepared for U.S. Department of Energy).
346
-------
PARTICLE CAPTURE MECHANISMS ON SINGLE FIBERS IN
THE PRESENCE OF ELECTROSTATIC FIELDS
by: M.B. Ranade, F-L. Chen, and D.S. Ensor
Research Triangle Institute
Research Triangle Park, NC 27709
L.S. Hovis
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
ABSTRACT
Fabric filtration, although simple mechanically, is a complex phenome-
non. As part of an effort to isolate the mechanisms significant in fabric
filtration, simple experiments have been devised to evaluate the effects of
electrostatic fields on particle capture. A series of experiments, with
charged and neutral particles with various applied fields, were conducted to
determine the location of deposits on the fiber. In particular, the location
of the attachment of the aerosol, with respect to the direction of flow, was
found to be strongly dependent on the applied field. The implications of
these data, comparison to theory, and implications when applied to fabric
filtration are described.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
347
-------
INTRODUCTION
Filtration using media composed of fibers is one of the most efficient
processes for fluid/particle separation. To predict the performance of the
media, it is necessary to understand the process by which particles are
deposited on the individual fibers.
Deposition of particles on a collector is the preliminary step leading
to particle collection in a fibrous or fabric filter. In general, most of
the past theoretical studies in fibrous filtration were based on the idealized
single-fiber concept and were confined to the initial filtration period;
i.e., before significant particle buildup. Thus, they do not deal with the
more important period when deposition is at an advanced stage. It is known
that deposited particles not only distribute themselves on the surface of the
fiber but also may build up chain-like agglomerates called dendrites.
In this work, the quantitative characteristics of monodisperse particles
deposited on a single metal fiber were investigated under three situations:
1) neutral particles and neutral fibers; 2) non-neutralized particles and
neutral fibers; and 3) neutral particles and neutral fibers in a variable
electric field.
BACKGROUND
The theory of filtration by fibrous filters is based on the concept of
the capture efficiency of a single fiber and supposes that the fibrous layer
is composed of such single fibers. In the past, experiments conducted to
prove the above theory assumed that the packing density was very small and
that effects from neighboring fibers could be neglected. No experiments were
applied to determine the deposition efficiencies and to describe the dendrite
formation of particles collected by a single fiber until 1966.
Billings (1) used a single glass fiber to collect neutral polystyrene
latex particles and obtained photographs of the particle dendrite formation.
He observed that particle deposition on a collector was not uniform but
varied along the angular sectors of the collector. Also, the collection
efficiency increased as the number of particles on the collector increased.
These dendrites were apparently better aerodynamic targets than the bare
fibers.
Several models have been published by different authors to explain the
particle dendrite formation on a single fiber. Payatakes and Tien (2) and
Payatakes (3, 4, 5) have developed deterministic expressions for dendrite
growth by solving successive differential equations. Payatakes and Gradon
(6, 7) extended this model to include different dominant mechanisms such as
interception (Payatakes [3, 4]), interception and inertial impaction
(Payatakes and Gradon [7]), and the Brownian diffusion effect (Payatakes and
Gradon [6]). With these models, the configuration of individual particle
dendrites and the rate of growth of these dendrites, as a function of deposi-
tion age and angular coordinates on the fiber surface, can be predicted.
348
-------
Tien et al. (8) and Wang et al. (9) have made a study of particle den-
drite growth using stochastic simulation. The number of dendrites formed on
a given length of fiber and their size and shape were determined by the loca-
tion of individual particles arriving from upstream and their order of arriv-
al. The stochastic simulation method will serve as a tool for testing the
hypothesis and will be a guide for the development of particle collection
models. Kanaoka et al. (10, 11) have developed a similar method to predict
the growing process of a dendrite and to determine the collection efficiency
of a single fiber under dust loading conditions. The Kanaoka model was
simple in structure and could generate a statistically sufficient number of
simulations in short computation time.
Several researchers have utilized experimental equipment and methods to
study the phenomena of particle deposition on a collector. Their results
have indicated that the number of particles collected by a fiber is a func-
tion of deposition age, particle concentration, particle dendrite structure,
Stokes number, and external forces. Beizaie (12), Bhutra and Payatakes (13),
and Barot et al. (14) collected particles on a single fiber by interception
and inertial impaction mechanisms. They observed that particle distribution
on a single fiber is not uniform and significantly depends on the angular
sector of the fiber. Wang et al. (15) accumulated solid particles on single
cylinders in an electric field. They observed that dendrites formed straight
chains and that collection efficiency markedly increased under the electro-
static effect. Oak and Saville (16) did an experiment to consider the depo-
sition of weakly charged particles on a collector in a strong external field.
They observed results similar to Wang's and concluded that the particle-
particle bonds were stronger than the particle-fiber bonds on a fiber.
EXPERIMENTAL WORK
EXPERIMENTAL APPARATUS
The experimental apparatus was designed and built to enable deposition
and observation of fine particles on a single fiber under well-defined and
controlled conditions. It consisted of a particle generator, a particle col-
lector, and monitoring devices. The different parts of the experimental
apparatus are shown schematically in Figure 1.
The vibrating-orifice aerosol generator, TSI Model 3050, was used to
produce well characterized monodisperse aerosols. It was supplied with a
constant feed of solution of methylene blue dissolved in 2-propanol. For the
first part of the experiment, the solid methylene blue particles generated
were neutralized effectively by passing them through a TSI Model 3054
neutralizer containing a Krypton-85 source. For the second part, charged
particles were generated by the above method except that the neutralizer was
not used. The particles were negatively charged with an average 850 electron
units charge per particle.
The aerosol was then passed into a specially designed chamber under
laminar flow conditions. The chamber was similar to the one used by Bhutra
and Payatakes (13). As shown in Figure 2, a smooth cylindrical tungsten
349
-------
156
CO
1
ff
•H
M
S
4)
1-4
U
2
o
U
i
2E2
n
O V
** a
3.
350
-------
Aerosol
In
Aerosol
Flow
Removable Plug
Screen Electrodes
(0
Figure 2. Single fiber aerosol, collection chamber—a) top view;
b) side view; and c) electrode assembly.
351
-------
metal fiber was supported by two copper tubes and could be rotated without
torsion using the gear arrangement. The chamber had an opening just above
the fiber, which could be sealed with a removable plug. To create a field
around the fiber parallel to the aerosol flow, two screens, 0.6 cm apart,
were added to the plug and a high voltage power supply was connected to them.
After the particles were collected for a predetermined time, the plug
was removed and the objective lens of the microscope camera system was lower-
ed to examine the fiber. The microscope camera system was mounted on a
stable base and could be moved with a control knob to scan the length of the
fiber. Using the gear arrangement, the fiber could be rotated and photo-
graphed at different angles.
A Climet 208 optical particle analyzer and a Climet CI-210 multi-channel
monitor were used to obtain the particle concentration. The Climet 208 was
operated at the same flowrate as that of the aerosol stream approaching the
fiber.
EXPERIMENTAL PROCEDURE
At the beginning of an experimental run, a new fiber was supported in
the fiber holder across the path of the aerosol. In the first part of the
investigation, neutral particles were collected on a fiber which was not
subjected to an external electric field. After a period of time, the aerosol
feed was stopped, the plug was removed from the fiber holder, and the objec-
tive lens of the microscope was lowered into position. Several parameters in
the generation and collection of the methylene blue particles were kept
constant during all three parts of the experiment and are listed in Table 1.
The fiber diameter, particle size, and the fluid velocity were so chosen that
a direct comparison could be made with the results reported by Bhutra and
Payatakes (13) in absence of an electric field. The velocity is considerably
higher than expected for fabric filtration, but is well within practical
filtration range such as in depth filtration (17).
Photographs of dendrite structures were taken at desired locations, and
the film and frame numbers were recorded. After all the dendrites in the
field of view had been examined, the objective lens was retracted, the plug
was set in position, and the entire procedure was repeated. The same area on
the fiber was examined to follow the growth of individual dendrites since the
previous viewing. Several runs were needed to accumulate enough information
for a given set of data.
In the second part of the investigation, non-neutralized particles were
collected on the fiber. The procedure was identical to that of the first
part, except the particle neutralizer was not used. An electrometer was used
to measure the amount of charge on the particles.
In the third part of the investigation, neutral particles were collected
on the fiber placed in an external electric field directed parallel to the
flowstream. The procedure was identical to that of the first part of the
experiment, except that two screens were attached to the removable plug, as
shown in Figure 2, and were connected to a high voltage supply with the
downstream screen grounded.
352
-------
RESULTS
Extensive data were collected from each of the three parts of this in-
vestigation: 1) neutral particles versus neutral fibers; 2) non-neutralized
particles versus neutral fibers; and 3) neutral particles versus neutral
fibers in an external electric field.
From the series of photographs which were taken of the same section of
fiber, it is apparent that particle number and dendrite length both increase
with time. However, the distribution of this growth between the various
sectors is different for each of the three parts of the experiment. Figure 3
shows the particles deposited on the same degree sector and the same exposure
time of the fiber under the different electrical conditions.
TABLE 1. PARAMETER VALUES USED IN THE EXPERIMENT
Particle Diameter : 3 |Jm
Fiber Diameter : 35 pm
Fiber Length : 700 pro
Particle Density : 1.2 g/cm3
Viscosity : 0.0000183 Pa»s
Velocity of Flow : 41 cm/sec
Reynolds Number of Fiber : 1.0116
Stokes Number : 0.8
Interception Parameter : 0.086
For the neutral particles in absence of an external electric field,
particle deposition on a fiber was entirely due to inertial impaction and
interception. All of the particle deposition occurred in the upstream sector.
For the non-neutralized particles, deposition was observed on the down-
stream side of the fiber. Their presence was due to the image forces between
the slightly charged particles and the neutral fibers. However, since the
charge on the particles was very small, the image force--significant over a
small distance of the order of particle dimension—caused only a few parti-
cles to deposit on the downstream sectors.
When the fiber was subjected to the electric field, the particle deposi-
tion was more uniform over the entire fiber surface compared to the' first two
353
-------
Jt
m
£
5
x
w
c
o
•H
O
U
n)
o
M
3
O
•H
M
tfl
3
j(
10
to
o
p.
0)
T3
R)
PL.
CO
01
V4
00
354
-------
parts of the investigation. The electric field also affected the shape and
size of the dendrites. In contrast to the branching, irregularly shaped
dendrites formed in the absence of an electric field, these dendrites were
much more slender and very straight. Contact between neighboring chains was
very limited.
In all three parts of the investigation, the number of particles and
dendrites collected on the fiber increased as the sampling time increased.
For any given sampling time, many more particles were collected on the fiber
when an electric field was present than when one was not present. Figure 4
shows the dendrite configuration in an electric field strength of 3 kV/cm.
The tendency of dendrite formation on the fiber surface was different
for each set of conditions. As shown in Figure 5, the ratio of M2 to MI
reflects the probability of dendrite formation. M2 is defined as the number
of particle chains which have a minimum of two particles; that is, the number
of dendrites. MJ represents the number of particles which contact the fiber
surface, regardless of whether or not they form the base of a dendrite. The
neutralized particles show a greater tendency for dendrite formation than the
non-neutralized particles. This may be caused by repulsion between the
similarly charged non-neutralized particles, thus preventing particle link-
ages.
The probability of dendrite formation increased as the electric field
strength was increased. In the presence of an electric field, the particles
in the flowstream were polarized and were attracted by dielectrophoresis to
the metal fiber with a high electric field gradient near it.
The overall collection efficiency was calculated for each experimental
run. The efficiency was defined as the ratio between the number of particles
collected per unit length of the particles, and the total particle flux
across the cross section of the fiber, as shown in Figure 6.
The measured efficiency at zero field agreed very well with the predic-
ted value using the Langmuir-Blodgett theory (18) based on inertial impaction.
When the electric field strength was increased, the interaction force between
the particle and the fiber increased and the collection efficiency of the
fiber increased. Figure 6 indicates that the collection efficiency monoton-
ically increased with electric field strength. The collection efficiency was
also calculated for each degree sector of the fiber and is plotted in Figure
7. For the fiber not placed in an electric field, the collection efficiency
plots for neutral and non-neutralized particles appear very similar. Both
exhibit sharp peaks in the upstream portion of the fiber, and the natural
charged particles show a little collection on the backside of the fiber. For
the fiber placed in an electric field, the collection efficiency is higher in
all degree sectors and exhibits a greater uniformity between the different
sectors.
355
-------
3/5/81, 200X, 2.5 hr, 3 kV/cm
Figure 4. Dendrite configuration deposited on the fiber
in an electric field (270° sector).
356
-------
o
I CM
O
en
o
(O
I
0>
o
'•p
I
o
u
o
CO
<£>.
o
in
o
•
-------
-no
n
0
•o o
o
M o
id o
a 0)
O
o
•H
H
cd
at
0)
O 4)
O £
vO
01
E
uoiioe||00
358
-------
i
5
1
q q q q u>
iri ^ co oi «-•
to
C
o
TJ
g
O
JJ
C
111
M-l
IM
•H
"O
t-i
-------
CONCLUSIONS
This investigation was performed to observe the quantitative characteris-
tics of particle deposition on a single fiber under three conditions. Based
on the observations, the following conclusions were reached:
1. Particle collection efficiency increases as electric field strength
increases.
2. In the presence of an external electric field, deposited particles and
particle dendrites are more uniformly distributed over the entire metal
fiber surface than without the electric field.
3. As electric field strength increases, the probability of dendrite forma-
tion increases.
4. Dendrites formed with electric field collection are much more slender
and straighter.
REFERENCES
1. Billings, C.E. Effects of particle accumulation in aerosol filtration.
Ph.D. dissertation, California Inst. Technology, Pasadena, CA, 1966.
2. Payatakes, A.C. and Tien, C. Particle deposition in fibrous media with
dendrite-like pattern: A preliminary model. J. Aerosol Sci. Vol. 7,
85, 1976.
3. Payatakes, A.C. Model of the dynamic behavior of a fibrous filter
application to case of pure interception during period of unhindered
growth. Powder Technology. 14: 267, 1976.
4. Payatakes, A.C. Model of aerosol particle deposition in fibrous media
with dendrite-like pattern: Application to pure interception during
period of unhindered growth. Filtration and Separation. 13: 602, 1976.
5. Payatakes, A.C. Model of transient aerosol particle deposition in
fibrous media with dendritic pattern. AIChE J. 23: 192, 1977.
6. Payatakes, A.C. and Gradon, L. Dendritic deposition of aerosols by
convective Brownian diffusion for small, intermediate, and high particle
Knudsen numbers. AIChE J. Vol. 26, No. 3, 443, 1980.
7. Payatakes, A.C. and Gradon, L. Dendritic deposition of aerosol parti-
cles in fibrous media by inertial impaction and interception. Chem.
Engineering Sci. Vol. 35, 1083, 1980.
8. Tien, C., Wang, C.S., and Barot, D.T. Chain-like formation of particle
deposits in fluid-particle separation. Science. 196, 983, 1977.
360
-------
9. Wang, C.S., Beizaie, M., and Tien, C. Deposition of solid particles on
a collector: Formulation of a new theory. AIChE J. Vol. 23, No. 6,
879, 1977.
10. Kanaoka, C., Emi, H., and Myojyo, T. Simulation of deposition and
growth of airborne particles on a filter. Kagako Kogako Ronbunshu,
Japan. 4, 535, 1978.
11. Kanaoka, C., Emi, H., and Myojyo, T. Simulation of the growth process
of a particle dendrite and evaluation of a single fiber collection
efficiency with dust load. J. Aerosol Sci. Vol. 11, 377, 1980.
12. Beizaie, M. Deposition of particles on a single collector. Ph.D.
dissertation, Syracuse University, Syracuse, NY, 1977.
13. Bhutra, S. and Payatakes, A.C. Experimental investigation of dendritic
disposition of aerosol particles. J. Aerosol Sci. Vol. 10, 445, 1979.
14. Barot, D.T., Tien, C., and Wang, C.S. Accumulation of solid particles
on single fibers exposed to aerosol flows. AIChE J. Vol. 26, No. 2,
289, 1980.
15. Wang, C.S., Ho, C.P., Makino, H., and linoya, K. Effect of electro-
static fields on accumulation of solid particles on single cylinders.
AIChE J. Vol. 26, No. 4, 680, 1980.
16. Oak, M.J. and Saville, D.A. The buildup of dendrite structures on
fibers in the presence of strong electrostatic fields. J. of Colloid
and Interface Sci. Vol. 76, No. 1, 259, 1980.
17. Miller, V.R. and Loeffler, F. Reinhalt Luft 40, 405, 1980.
18. Langmuir, I. and Blodgett, K. Report on Smokes and Filters. Supplement
to Section I and Section II. No. 3460, U.S. Office of Scientific
Research and Development, 1944.
361
-------
PILOT DEMONSTRATION OF PARTICULATE REMOVAL
USING A CHARGED FILTER BED
by: Paul H. Sorenson
Air Correction Divison, UOP Inc.
Norwalk, Connecticut 06856
ABSTRACT
The concept of fine particulate collection in a gas stream using a highly porous,
charged fiber bed was first developed at Battelle Pacific Northwest Laboratories while
studying the collection process by charged spray drops. Laboratory testing by Batelle
using highly resistive, charged submicron aerosols showed that extremely high collec-
tion efficiencies were possible by this process. Air Correction Division, UOP Inc. has
undertaken a program to develop the concept under field scale conditions. A
transportable 4000 cfm pilot plant was constructed and installed slipstream on a lignite-
fired utility boiler at the outlet of an existing precipitator. The collection efficiency of
the bed was monitored as a function of bed face velocity, gas temperature, and particle
charge levels. This paper reports the results of the program.
INTRODUCTION
Trends in air pollution control during the last several years have emphasized
reduction in particulate emissions without appreciably increasing operating costs.
Advances in electrostatic precipitator technology have reduced emissions, but success
has been limited by highly resistive ash. Fabric filters have successfully reduced
resistive ash emissions but only at the expense of higher operating costs. The
Electrostatic Fiber Mat (EFM) filter system invented by the Battelle Pacific Northwest
Laboratories (1) appeared to be a potential method for reducing high-resitivity
particulate emissions at the outlet of an electrostatic precipitator without an appre-
ciable increase in pressure loss.
To explore the potential of the EFM for collection of flyash, Air Correction
Division, UOP Inc. established a program involving laboratory work and field pilot tests.
UOP sponsored additional research at Battelle using their laboratory and expertise.
Laboratory tests were performed using ash samples collected at boilers burning high-
and low-sulfur coal and lignite. The results of these tests showed that the EFM is
effective in collecting highly resistive ash with low energy consumption (2). On the
basis of these tests, Air Correction Division undertook a program to verify those results
under field conditions.
362
-------
A field test program was established to demonstrate that the EFM was a practical
device for the collection of high-resistivity ash that had penetrated an operating
electrostatic precipitator. To be considered a practical device, the EFM should remove
at least 90% of all incident particulate with a pressure drop of less than one inch of
water column. It was also necessary to show that accumulated particulate could be
removed from the mat and separated from the flue gas stream for disposal and the mat
restored to service. When these criteria have been demonstrated, the EFM could be
considered a potential retrofit for nonconforming precipitators. It could also be
incorporated into an effective primary particulate emission control system for ne\v
installations.
Air Correction has conducted a program of field testing to demonstrate the
potential of the EFM. The first site chosen to study high-resistivity ash was a utility
boiler burning lignite. An EFM pilot plant was set up and a test program conducted to
demonstrate the stated objectives. This paper is a report on that program.
DESCRIPTION OF THE PILOT PLANT
The pilot plant was designed as a five-to-one scale-up of the laboratory apparatus.
The EFM system consisted of a particle charging section and fiber mat section with an
in-situ mat-cleaning mechanism and interconnecting ductwork. A schematic of the
system is shown in Figure 1.
Two particle-charging electrode configurations were tested during the program.
The first electrode configuration consisted of wire emitting electrodes and rod
grounded electrodes. The electrodes were spaced to emphasize particle charging and to
minimize particle collection. Gas velocity in the charging section equaled mat face
velocity. In the second configuration, each row of rod electrodes was replaced with a
single plate. This design permitted charger operation at lower current densities and
reduced electrical stress across the collected ash area.
The fiber mat itself was made of highly resistive fibers knit in an open mesh. The
resulting fabric was then layered and compressed to the appropriate thickness and void
fraction. Two types of fibers were tested in the field, a monofilament and a yarn. Both
fibers were capable of operation up to 500°F. The mats were held between fiberglass
grids. Fiberglass pins pierced the mat to provide support and were terminated in the
face grids.
DESCRIPTION OF TEST METHODS
Operation of the pilot plant required control and recording of temperature and
flow. Temperature was controlled by adjusting a dilution air damper on the inlet duct
and was measured with a thermocouple immediately before expansion into the test
cross-section. Flow was controlled by adjusting the outlet damper on the slipstream
fan. Flow was monitored with a pitot tube and thermocouple located in the high-
velocity duct between the test section and the fan. Velocity pressure and mat pressure
drop were monitored with an inclined manometer.
363
-------
UJ
ed
E
o
o
0)
«rf
c
rt
"5.
3
O)
364
-------
Particle concentrations were measured with conventional mass-sampling trains.
Standard buttonhook nozzles were mounted on in-line filter holders to provide in-stack
particulate sampling at the system inlet and outlet locations. Because of the low
velocity at the test locations, sampling rates were calculated from the velocities
measured in the high-velocity duct and actual static pressure and temperature readings
in the sampling area. Sampling times were calculated so that the lowest anticipated
mass loading would be at least two orders of magnitude greater than the smallest
division on the mass balance, i.e., approximately 1.0 milligrams when measuring to 0.01
milligrams. The sampling train consisted of probe, filter, desiccant, rotometer, pump,
and gas meter. The pilot plant efficiency was calculated from the mass concentrations.
DESCRIPTION OF THE TEST SITE
The test site chosen for evaluation of the EFM with flyash from a utility power
boiler burning lignite consisted of a 550 megawatt boiler supplied with lignite from a
local strip mine. The pilot plant slipstream was taken from the outlet of the
electrostatic precipitator and reinjected at the I.D. fan suction. The precipitator was
of conventional design upgraded for increased power and better gas flow distribution.
Precipitator performance was generally between 99.3% and 99.5% efficient. The plant
uses a flue gas conditioning system in conjunction with the precipitator. The gas
conditioning system was in operation during the test period.
A comparison was made between the flue gases entering the pilot plant and the
gases leaving the stack to obtain an indication of how well the slipstream represented
the plant discharge. Table 1 presents a summary of results of tests taken at the pilot
plant inlet and the stack. The close agreement in the two tests shows that the pilot
plant slipstream is a reasonable representation of the plant exhaust gases. The lower
temperatures recorded at the pilot plant probably result from pilot plant duct in-
leakage as is evident by the slight increase in oxygen content.
TABLE 1. COMPARISON OF PILOT STREAM TO STACK
PILOT STACK
Flow Rate, acfm 2800 2.1xl06
Flow Rate, dscfm 1558 1.2xl06
Temperature, °F 322 366
Moisture, % 13 13
CO2, % 13 13.3
O2, % 6 5.8
CO, % - -
N2, % 81 80.8
EA % 38.8 34.9
Grains/dscf 0.06 0.075
365
-------
Flyash resistivity readings were taken on samples of ash taken from the hoppers
on the last stage of the precipitator. The results are shown in Table 2.
TABLE 2. RESISTIVITY OF FLYASH SAMPLES
TEMPERATURE MOISTURE PERCENT VOLUME
Degrees F 11.5% 16.5%
200 3.7xlOU 1.8 xlO11
240 3.6xlOU 2.4 xlO11
280 3.7xlOH 2.2 xlO11
315 3.1xlOH 2.1 xlO11
355 1.8 xlO11 1.6 xlO11
Resistivity in ohm-cm at 3.87 KV/cm
DISCUSSION OF RESULTS
PARTICLE COLLECTION EFFICIENCY
The test program was arranged to compare the wire-rod charger with the wire-
plate charger by operating each charger with the monofilament mat and obtaining the
mass efficiencies of both combinations. The charger configuration providing higher
efficiency would then be operated with the yarn mat and the test series would be
repeated. Each combination was tested for mat face velocities from 200 to 350 feet
per minute and temperatures ranging from 200°F to 350 F.
Each charger configuration was ooerated with the monofilament mat over
temperatures ranging from 200°F to 350 F. The collecting efficiency for the EFM
increased from 81.2% for the wire-rod arrangement to 90.5 for the wire-plate
arrangement. Lower efficiencies for the wire-rod configuration can be attributed to
higher current densities and electrical field stress resulting in back corona. (3) Fiber
mat face velocity was varied from 200 to 350 feet per minute without a noticeable
change in collection efficiency.
The yarn fiber bed was then installed in the pilot plant and compared with the
monofilament performance. The particle collection efficiency increased to 94.6%. The
increase in efficiency can be attributed to the smaller diameter of the yarn fiber and a
lower sensitivity of the material in the yarn to changes in resistivity.
Laboratory tests showed that efficiency was essentially insensitive to inlet dust
loading from 0.08 to 0.12 gr/dscf. Actual inlet loadings to the pilot plant covered a
wider range. Testing during periods of precipitator maintenance and other upset
conditions resulted in inlet loadings ranging from 0.05 to 1.32 gr/dscf. Particle
collection efficiency remained constant over the entire range of inlet concentrations.
Mat dust loadings and therefore mat pressure drop increased due to the heavy
accumulation within the mat during periods of heavy loading. Once the bed was
cleaned, pressure drop returned to the desired operating range.
366
-------
PRESSURE DROP AND MAT CLEANING
Both mat configurations were tested for pressure drop before they were contami-
nated with ash. The monofilament mat construction had a pressure drop of 0.21 inches
of water column at 300 feet per minute and the yarn mat construction had a pressure
drop of 0.46 inches of v/ater column at 300 feet per minute. The increase in pressure
drop for the yearn is a result of the difference in the diameter of the fiber and the
differences in aerodynamic profile. The monofilament presents a smooth, cylindrical
obstruction to the gas flow. The yarn forms a ribbon-type obstruction with its oval
cross-section constantly changing orientation with respect to the flow path. In
addition, while the monofilament has a smooth surface, the yarn has an irregular
surface resulting from the multitude of filaments of which the yarn is composed.
Two methods of mat cleaning have been evaluated. The first method used high-
velocity water sprays to remove the ash from the mat fibers. The second method used
a compressed air and vacuum system where the concentrated dust cloud is drawn out of
the gas duct and separation occurs in an auxiliary baghouse. Both systems removed
sufficient ash to maintain a constant pressure drop across the mat of less than one inch
of water column. The water systems was used exclusively with monofilament mats and
the air system with yarn mats.
Inspection of the mats constructed of monofilament following several cleaning
cycles showed that the wash system was effective in restoring the fibers to their clean
state wherever the sprays contacted the mat. The spraying action did, however, drive
some of the ash into unwashed areas, especially around the edges. These areas
eventually collected sufficient ash to block off flow and resulted in increased pressure
drop. Once these areas were packed, no furthur build-up was noticed. Thus, the
pressure drop after cleaning increased gradually from 0.21 inches of water column to
0.28 inches of water column. All subsequent cleaning cycles returned the pressure drop
to 0.28 inches. Three sample collecting/cleaning cycles for the water system with the
monofilament mat are shown in Figure 2. It can be noted that the collecting time for
the mat is only limited by the maximum tolerable pressure drop.
Operation of the compressed air vacuum system was considerably different from
the water system. Unlike the water system, which carries all of the particulate away
with the slurry, the air system momentarily resuspends collected particulate in a
concentrated cloud, which is then drawn off through the vacuum system. Since the
vacuum cannot draw from deep within the mat, the cleaning cycle must be timed so
that the particulate is resuspended while still within reach of the vacuum. The
compressed air vacuum system must therefore be used more frquently than the water
system. The yarn presents an additional problem in cleaning. Because of its
construction from many filaments, the yarn traps particulate within the filaments,
especially when direct interception occurs. Cleaning vigorously enough to dislodge such
trapped particles would at the same time destroy the integrity of the yarn. Increases in
pressure loss due to this type of particle accumulation appear to diminish after eight to
ten hours of operation. After this initial increase, a constant range of operation can be
held. Figure 3 shows the pressure drop history for a yarn mat with compressed air
vacuum cleaning The gradually increasing pressure drop due to intrafiber capture can be
seen stabilizing after about seven hours of operation and several cleaning cycles.
367
-------
0
CO
111
X
o
z
I
Q.
O
lit
CC
CO
CO
UJ
CC
Q.
1.0
0.8
0.6
0.4
0.2
5 10 15 20 25 30
TIME-HOURS
FIGURE 2
PRESSURE DROP HISTORY FOR WATER CLEANED MATS
O
CO
UJ
X
o
z
I
Q.
O
CC
a
UJ
CC
D
CO
CO
UJ
CC
0.
1.0
0.8
0.6
0.4
0.2
10
15 20
-HOURS
25
30
FIGURE 3
PRESSURE DROP HISTORY FOR AIR CLEANED MATS
368
-------
CONCLUSIONS
The test program was designed to demonstrate that the Electrostatic Fiber Mat is
a practical method of reducing highly resistive flyash emissions. The pilot plant was
operated to show that practicality in efficiency, operating pressure drop, and mat
cleaning techniques.
Mass efficiencies of 94% for the wire-plate charger and yarn mat system
exceeded the objective of 90%. Projecting this efficiency to a full-size installation
such as the host site would reduce the current average emission of 0.075 gr/dscf to less
than 0.007 gr/dscf. The EFM efficiency has also been shown to apply to a range of inlet
concentrations from 0.05 to 1.3 gr/dscf. The ability to retain collection efficiency
during inlet upsets is especially advantageous when emissions drift out of compliance
during plate rapping, soot blowing cycles, or load changes.
The practicality of this device has been further shov/n in the operation of the
compressed air vacuum mat cleaning system. The water wash system is extremely
effective in regenerating the mat. The quality of cleaning does not offset the
drawbacks of sectionalization and isolation required to cool, clean, and reheat the mat
assembly. The development of the air system overcomes the difficulties of the water
system at the expense of reduced time between cycles. Thus, while the monofilament
mat with water washing was regnerated every eight to ten hours, the yarn mat was air
cleaned every two hours. The combined system of a yarn mat with compressed air
vacuum cleaning therefore presents a practical method of reducing the emissions from
an operating electrostatic precipitator.
The work described in this paper was not funded by the U. S. Environmental
Protection Agency and therefore the contents do not necessarily reflect the views of
the Agency and no official endorsement should be inferred.
REFERENCES
1. Reid, D.L. and Browne, L.M., "Electrostatic Capture of Fine Particles in Fiber
Beds," EPR Report 600/2-76-132, National Technical Information Service, Spring-
field, VA (976)
2. Bamberger, J.R. and Winegardner, W.K», "Fiber Bed Filter System Control of Ash
Particualtes," ASME paper 81-WA/APC-l, ASME Publication, New York, NY
(1981)
3. White, H.3., "Resistivity Problem in Electrostatic Precipiator," APCA 24(4): 314-
338 (April 1974)
4. Yu, H.S. and Teague, R.K., "Performance of Electrostatic Fiberbed," presented at
the First Annual Conference of the Aerosol Research Association, February 17-
19, 1982, Santa Monica, CA
369
-------
PILOT DEMONSTRATION OF MAGNETIC FILTRATION WITH CONTINUOUS MEDIA REGENERATION
by: Carroll E. Ball and David W. Coy
Research Triangle Institute
Research Triangle Park, NC 27709
ABSTRACT
A mobile pilot plant with a nominal flow capacity of 3,060 m3/hr (1,800
cfm) was designed and built to evaluate the use of high gradient magnetic
filtration (HGMF) for particulate emission control on an electric arc furnace
(EAF). A five-month test program was conducted at Georgetown Steel Corpora-
tion1 s plant in Georgetown, South Carolina, to test the performance of the
HGMF. A 500-hour long-term test was scheduled and later changed in order to
perform additional characterization studies.
The pilot-plant collection efficiency was less than expected for the
stainless steel wool matrix packed to a density of 1.5 percent by volume.
The matrix was then changed to an expanded metal, packed to a density of 3.5
percent by volume, which resulted in much lower pressure drop measurements,
but even lower collection efficiencies. The expanded metal matrix was then
packed to a density of 6.0 percent by volume, which gave higher collection
efficiencies than the steel wool and a slightly lower pressure drop.
During the field test operations, there were no significant problems
with the HGMF mobile pilot-plant equipment.
The report describes the design and construction of the continuous HGMF
mobile pilot plant, as well as some of the background work in high gradient
magnetic filtration done at RTI. The field start-up and performance charac-
terization of the mobile pilot plant are discussed in detail. The experi-
mental data and data analysis are given, as well as an economic evaluation and
comparison of the HGMF with other particulate control devices.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
370
-------
INTRODUCTION
Since the commercialization of high gradient magnetic filtration (HGMF)
in the clay industry 10 years ago, the applications of magnetic separation of
process streams have been steadily increasing. Some of the applications are
mineral beneficiation, coal deashing, coal desulfurization, wastewater treat-
ment, and blood component separation. From 1975 to 1977 Research Triangle
Institute (RTI) conducted experimental work, funded by the U.S. Environmental
Protection Agency (EPA), in the application of HGMF to air pollution control.
Magnetic separation was tested in the laboratory on several dusts from the
iron and steel industry. There were promising results from the following
iron and steel industry sources: basic oxygen furnace (BOF); electric arc
furnaces (EAF); open hearth furnace; scarfing machine; and the sinter
machine.
An earlier pilot plant was designed and built by RTI and tested on a
Pennsylvania sintering plant. The overall efficiency data were low for these
tests, however, due to the low specific magnetization of the sinter plant
dust. It was then decided to design and build a pilot plant with continuous
media regeneration and test it on a dust with higher specific magnetization
such as BOF or EAF dust.
In June, 1981 the pilot plant was moved to Georgetown, South Carolina
and connected to a slipstream from the exhaust of three EAF's just upstream
of a baghouse. After startup and debugging, the test program was begun to
obtain results on the effects of filter density, applied magnetic field, and
gas velocity on overall and fractional collection efficiency. Total mass and
fractional collection efficiency tests were conducted, and samples of the
dust entering, exiting, and captured by the pilot plant were collected for
magnetic and chemical analyses.
The results of the field tests were used to make technical and economic
assessments of the application of HGMF to EAF's, and to compare HGMF to other
types of pollution control devices.
BACKGROUND DEVELOPMENT
BASIC CONCEPT
The fundamental concept of the HGMF process is the interaction between
paramagnetic or ferromagnetic particles and ferromagnetic fibers while in the
presence of an applied background magnetic field. The applied magnetic field
induces a magnetic dipole in the particle and magnetizes the wire. This
creates a convergence of the field near the wire resulting in a net force
being applied to the particle. The magnetic force, in competition with the
viscous, inertial, and gravitational forces, causes the particle to be
attracted to the wire and held there until the applied field is removed.
The high gradient magnetic filter consists of several cassettes packed
with ferromagnetic fibers (such as stainless steel wool or expanded metal)
371
-------
which are moved into a magnetic field as the particle-laden gas is being
passed through. The particle-laden gas is cleansed as the particles are
attracted to and held by the fibers. When the matrix is loaded, the cassette
is then moved out of the magnetic field and the particles are flushed from
the fibers.
HGMF DEVELOPMENT AND APPLICATIONS
The experimental work in high gradient magnetic filtration, with the
exception of the EPA sponsored development begun in 1975, has been most con-
cerned with the magnetic separation of particles in a slurry. Oberteuffer
(1), Kolm et al (2), Oder (3), and lannicelli (4) have published excellent
reviews of the process and its chronological development. A brief review of
HGMF development can also be found in the EPA report "Application of High
Gradient Magnetic Separation to Fine Particle Control" by Gooding et al (5).
The most extensive development of HGMF has been within the last decade
in the clay industry. Here, the HGMF is used to separate small paramagnetic
color bodies from kaolin clay. The successful demonstration of this process
in the clay industry has sparked investigations of HGMF for many different
applications.
DETAILED DESIGN AND CONSTRUCTION OF THE HGMF MOBILE PILOT PLANT
The mobile pilot plant is a continuous HGMF system housed in a 12.8 m
(42 ft) freight van. The system is designed for a nominal flow capacity of
3,060 m3/hr (1,800 cfm). Figure 1 is a flow schematic of the continuous HGMF
system.
The dirty gas enters the pilot plant through a 0.267 m ID stainless
steel pipe (10" schedule 5). The gas passes by test ports through which
samples can be drawn to determine inlet dust concentration, chemical composi-
tion, and size distribution, and then is directed to the HGMF device. The
magnetic filter is a Sala-HGMS® Carousel Model 120-05-00 (Sala Magnetics,
Inc., Cambridge, MA) incorporating a magnet head and a cleaning station
mounted 180° apart on a rotating carousel. The magnet coils are split into a
saddle configuration to allow the carousel to be rotated through the magne-
tized zone by a variable speed drive. The carousel contains 48 removable
cassettes which can be loaded with filter material to a depth of 0.15 m (5.8
in.). The magnet head encloses an active face area of 0.085 m2 (133 in.2) in
the direction of fluid flow. The magnet head is designed to provide an
applied field from 0.0 to 5.0 kG. In the range of gas velocities tested, 2
to 10 m/s, the gas residence time in the filter varied from 0.015 to 0.075s.
After passing through the magnet, the gas then passes by another set of
test ports and exits the pilot plant. Once leaving the pilot plant, the gas
is directed through an orifice, for velocity determination, through an in-
duced draft blower and then is exhausted to the atmosphere through an 8 m (26
ft) high stack.
After the filter matrix has passed through the magnetized zone and
collected dust from the gas stream, it then passes through the cleaning
station. The filter is cleaned by backflushing with compressed air.
372
-------
O)
13
O
-P
c
O
-------
The agglomerated dust that is cleaned from the filter matrix with the
cleaning air pulse is sent to a cyclone. Exhaust from the top of the cyclone
is recycled into the dirty gas stream. Dust can \>e removed from the cyclone
while the pilot plant is in operation through a double-sealed valve.
The induced draft blower which moves the gas through the pilot plant is
rated at 3,060 m3/hr at a suction pressure of -13.7 kPa (-55 inches 1^0) and
a temperature of 150°C. The entire system is designed to allow continuous
operation at temperatures of up to 200°C. All interior and exterior pipe is
insulated with jacketed fiberglass.
The utility requirements of the pilot plant are electricity and water.
The main power panel has 400 ampere service at 440 ac volt input. The total
connected load is 300 amperes. The major equipment operates off 440 vac and
a transformer is provided to step down to 240 vac and 120 vac. Water con-
sumption is approximately 2.3 m3/hr (10 gpm) for magnet cooling, compressor
aftercooler, and occasional use of the lab sink.
FIELD OPERATIONS
DESCRIPTION OF THE ELECTRIC ARC FURNACE AT GEORGETOWN, SOUTH CAROLINA
The dust source was an EAF shop utilizing three arc furnaces operating
continuously in a staggered batch operation. The Georgetown Steel raw steel
production facilities are composed of three 68 Mg (75 ton) per cycle DeMag
electric arc furnaces. The charge to the furnaces consists of scrap and pre-
reduced iron pellets. The scrap charged is obtained primarily from external
sources; about 5 to 10 percent is reclaimed scrap. Prereduced pellets are
produced on-site from South American iron ores. Other materials added to the
furnaces during the course of the production cycle include limestone, coke,
ferromanganese, and ferrosilicon.
Gas cleaning for the furnaces is provided by a positive pressure bag-
house supplied by American Air Filter. A slipstream of gas was taken from
the duct. The gas stream conditions at the extraction point are listed
below:
Pressure -1.7 kPa (-7 inches H20)
Temperature 71°C (160°F)
Velocity 17 m/s (55.65 ft/s)
Reynolds Number 4.1 x 106
PERFORMANCE CHARACTERIZATION
The test program was designed to test the effects of four parameters on
collection efficiency, and the reliability of the equipment during long-term
operation. The four parameters to be varied were applied field, gas veloc-
ity, filter type, and filter packing density.
374
-------
Selection of an optimum set of operating conditions was the goal of the
performance characterization. The selection of optimum conditions was to be
based on statistical analysis of the performance data obtained under varied
operating conditions. Performance under each set of conditions was to be
measured by sampling the HGMF inlet and outlet particulate concentrations.
Overall efficiency would be determined on a mass basis, and fractional effi-
ciency would be determined by measuring inlet and outlet particle size dis-
tribution. Periodically during the characterization, bulk samples of the
particulate were to be obtained from the inlet, the outlet, and the cyclone
catch in order to obtain the chemical composition and its effect on perfor-
mance and vice versa.
The applied magnetic fields were varied between 0 and 5 kG for each of
three test series. Gas velocities through the filter were varied between 2
and 10 m/s.
The initial filter medium was American Iron and Steel Institute Type 430
medium grade stainless steel wool packed to a density of 0.015 (1.5 percent
by volume) and came packed in the carousel from Sala Magnetics. The average
fiber diameter of this material is 120 pm.
The second and third test series were run with expanded metal matrices
(layers of wide mesh screen separated by spacers). The expanded metal matrix
can be packed with all of the fibers perpendicular to the magnetic field and
in the optimum position for particle capture. This we had hoped would allow
us to obtain a higher collection efficiency for the same pressure drop as was
attained with a stainless steel wool matrix.
The average fiber diameter of the expanded metal matrix is 300 pm. In
the second test series, the approximate packing density of the expanded metal
matrix was calculated to be 0.035. In the third test series the packing
density was calculated as 0.06.
Multiple linear (slope-intercept form) regression techniques were used
to analyze the overall mass efficiency data. The groups of data for each
matrix type and packing density were obtained in sequence, rather than ran-
domly, owing to the difficulty of changing the carousel cassettes. Also
because of this factor, statistical analyses were performed on the data
grouped by packing type and packing density. Using statistical techniques,
the model found to produce the best data fit was an exponential model.
The comparative performance of the filter system for both matrices and
packing densities can be seen in Figure 2. The performance data are plotted
as penetration (1-efficiency) versus regression function divided by inlet
mass rate. The line through each data set is a regression curve for that
data set. The regression coefficients a, b, c, and intercepts for each set
of data are given in Table 1. To compare the performance one must examine
the penetration at low values of the regression function for each data set.
looking at the functional relation of each variable in the regression
function, it is expected that best performance would be measured for low
regression function values, i.e., low inlet mass, low velocity, and high
field strength. Differing absolute values of the regression function occur
between sets because of the different regression coefficients for each data
375
-------
100
Expanded
Metal
0.035
B
I 10
s.
- 0
Steel Wool
0.015
Expanded
Metal
A 0.06
10 100
Regression Function (Inlet Mass)8 (Velocity)11
Inlet Mass " (Field Strength)e(lnlet Mass)
Figure 2. Georgetown Steel HGMF test—Penetration vs. regression function
(all matrices and packing densities).
376
-------
TABLE 1. REGRESSION COEFFICIENTS FOR EACH FILTER MATRIX AND PACKING DENSITY
Model: Outlet mass rate = Intercept x (In1et mass ^te)a(Ve1ocity)b
(Field Strength)0
Matrix and
packing density
Regression Coefficients
Intercept
Steel wool, 0.015
Expanded
Expanded
metal ,
metal ,
0.
0.
035
06
0.
0.
0.
034
123
005
0.
0.
0.
482
647
364
0.
0.
1.
801
351
39
0.
0.
0.
315
139
111
set. At low values of the regression function, poorest performance was
measured for expanded metal at 3.5 percent packing density, with clearly best
performance from expanded metal at 6 percent packing density.
Table 2 presents the predicted penetration based on the model and
coefficients in Table 1 for both matrices and packing densities at fixed
system operating conditions. Predicted penetrations are lower for expanded
metal at 6 percent packing density and the observed pressure drop was also
lower for this matrix than for steel wool. Pressure drop was significantly
lower for expanded metal at 3.5 percent packing density, but its predicted
penetration is about twice as high as that for 6 percent packing density at
the higher field strengths. Based on Table 2, expanded metal at 6 percent
packing density would be the preferred matrix.
TABLE 2. PREDICTED FILTER PERFORMANCE FOR
EACH FILTER MATRIX AND PACKING DENSITY
Matrix and Packing Density
Operating
conditions*
Steel wool
0.015
Penet.% AP**
Expanded metal
0.035 0.06
Penet.% AP** Penet.% AP**
0.1 kilogauss
7m/s
2.5 kilogauss
7m/s
5.0 kilogauss
7m/s
26.2 38.1
9.5 38.1
7.6 38.1
22.5 25.4
14.4 25.4
13.1 25.4
15.2 30.5
7.3 30.5
6.2 30.5
* Inlet concentration =1.0 gram/m3.
**Actual pilot-plant AP, cm H20.
377
-------
Performance By Particle Size
The MRI cascade impactor size data were used to generate particle size
penetration curves. Simultaneously obtained inlet and outlet distribution
curves were used, and in some cases data taken under duplicate conditions
were combined to provide composite fractional efficiency curves.
Figure 3 shows typical fractional penetration curves for each matrix and
packing density. It is evident that the filter is relatively inefficient for
particle sizes below 1 |Jm in diameter. This leads to the question of whether
inefficiency in the smaller particle size range is inherent to the design and
magnet operating conditions chosen for these tests, or attributable to varia-
tion in size-related particle characteristics.
Chemical And Magnetic Characteristics Of The Dust
The chemical and magnetic characteristics of suspended dust entering and
leaving the pilot plant were analyzed. The outlet dust samples contained a
relatively significant amount of magnetic material. For the low field test
case (0.1 kilogauss), the outlet dust specific magnetization was about 80
percent of the inlet dust value. For the 2.5 kilogauss field cases, the
outlet dust specific magnetization was about 65 percent of the inlet values.
In the high field case (5.0 kilogauss), the outlet value was about 50 percent
of the inlet value.
Elemental chemical analyses were performed on collected inlet and outlet
dust samples by atomic adsorption. In each set of samples the percentage of
iron in the outlet sample was lower than in the the inlet sample, as ex-
pected. However, even in the case of the maximum magnetic field strength
tests there was 6 percent by weight iron penetration. The fact that incom-
plete elemental iron separation occurred in the magnetic filtering process
suggests complex particle chemistry.
Other studies (6) of EAF dust have shown zinc (a diamagnetic element) to
be associated with iron in particles labelled mixed ferrite. Some of these
particles were found in both magnetic and non-magnetic fractions. Since some
of the iron occurs in ferrite particles that are non-magnetic, this offers
one explanation for iron penetration not approaching zero in these pilot-plant
tests.
Discussion of Performance Characterization Results
This pilot program was the first attempt at using a continuously cleaned
magnetic filter unit on a gas stream as opposed to a liquid stream.
The performance data discussed, showed the best performance (efficiency),
was achieved with an expanded metal matrix at 6 percent packing density.
This best performance was achieved at a pressure drop lower than the second
best performing matrix, steel wool. This test program was the first in the
HGMF development program attempting to use an expanded metal matrix instead
of steel wool. Given the better performance at lower pressure drop observed
in the initial experiments, expanded metal matrices deserve further study
toward additional optimization.
378
-------
100
10
O Steel wool matrix 0.015 packing density
D Expanded metal matrix 0.035 packing density
A Expanded metal matrix 0.06 packing density
7.0 m/s Velocity
2.5 kG Field strength
.1
I
I
.1
1 10
Particle Diameter (ptm)
100
Figure 3. Georgetown Steel HGMF test—penetration vs. particle size.
379
-------
For the given conditions of applied field, velocity (residence time),
and inlet concentrations, the best overall efficiencies achieved were in the
range of 94 to 96 percent with outlet concentrations in the range of 20 to 70
mg/m3. The efficiency levels are much improved over those achieved in the
previous pilot plant work on sinter plant emissions. The performance levels,
however, are not competitive with conventional high efficiency control de-
vices applied to electric arc furnaces. The New Source Performance Standard
(NSPS) for electric arc furnaces limits particulate emissions to 12 mg/dsm3.
The HGMF outlet concentrations in these tests were 2 to 6 times the required
level. State standards for existing sources vary considerably e.g.
Pennsylvania equivalent to 18 mg/dsm3, Michigan equivalent to 130 mg/dsm3.
On a performance basis, HGMF might have some limited retrofit potential.
The fractional penetration data show the HGMF was not as effective on
particles below 1 pm in diameter as on those above 1 (Jm. In terms of frac-
tional particle size penetration, it is not evident that any significant
qualitative differences exist between HGMF and conventional control devices.
The magnetic analyses data reveal that the HGMF did not remove all of
the magnetic material. The magnetic material penetrating the collector may
do so because of insufficient residence time or reentrainment. The chemical
analyses data reveal a significant amount of iron penetrated the filter,
especially in the small particle size (below 1 (Jm) fraction. Penetration of
the iron may be due to insufficient residence time and reentrainment. How-
ever, the recent report (6) indicating some iron in electric arc furnace dust
to be present in a "non-magnetic" form (probably meaning not ferromagnetic)
suggests a third mechanism for penetration. The HGMF's sensitivity to chem-
ical composition of particles and their resulting magnetic susceptibility is
analagous to the effects of chemical composition on particle resistivity and
electrostatic precipitator performance.
An alternative use for HGMF not explored in this study is to separate
non-magnetic components of waste EAF dust from magnetic components, i.e.
ferrous and non-ferrous. At present, EAF dust is classified as hazardous
waste as a result of heavy metals contamination. Separation of the ferrous
portion with minor contamination by zinc might permit its recycle to steel-
making, reducing the residue for disposal. With sufficient concentration of
zinc, the non-ferrous portion might be sold to zinc refiners. The association
of iron and zinc in non-magnetic particles identified in the study discussed
above (6) suggests that this potential application of HGMF needs further
study to determine the degree'of separation achievable.
ECONOMICS
Approximate costs have been developed for four different options for
particulate emission control of EAF dust. The accuracy of these costs cor-
responds roughly to that of a study grade estimate (±30 percent) and are
shown in.Table 3. Although a best estimate is presented for both capital and
annual expenses, it should be kept in mind that the absolute costs of the
four options may depend on special process details -- plant location, and
380
-------
material of construction. Consideration of these details was not within the
scope of the estimate. However, the estimates were made for the following
common base case:
Volumetric flow rate:
Inlet dust loading:
Outlet dust concentration:
Annual operation:
8,500 m3/min @ 66°C
1,050 mg/sm3
12 mg/sm3
8,500 hr/yr
For purposes of the economic comparisons, it was assumed that HGMF performance
could attain the outlet concentration needed to comply with the EAF NSPS, or
12 mg/sm3.
TABLE 3. COSTS OF VARIOUS CONTROL OPTIONS FOR EAF PARTICIPATE EMISSION
$/m3/s
Total capital costs
Total annualized costs
10 yr1 20 yr1 Direct operating costs2
HGMF3
2 m/s
5 m/s
7 m/s
ESP
Fabric Filter
Venturi scrubber
31.31
20.99
14.16
11. 944
11.11
17.58
7.06
5.21
4.07
3.43
3.64
8.11
5.64
4.26
3.43
2.89
3.14
7.31
0.71
0.95
1.20
1.01
1.09
4.55
1Total annual costs are computed for both 10 and 20 year capital recovery
periods. The capital recovery factor for 10 years is .16275 and for 20
years is .11746.
2Direct operating costs include operation and maintenance labor, supervisory
overhead, and utility costs. It does not include capital recovery charges or
taxes, insurance, and administrative charges (taxes, insurance, and adminis-
tration are computed at 4 percent of the total capital costs).
3Values given for three separate face velocities.
4Capital costs for the ESP were calculated in four ways.
a)Ratio from reference (7) using empirical factors- Cost = $11.50/am3/s
b)Itemized major equipment and cost factors Cost =
c)Escalate from reference (8) Cost =
d)Ratio from reference (7) using ".6 rule" Cost =
10.47/am3/s
13.11/am3/s
13.13/am3/s
381
-------
In addition, the following system parameters specific to each option
were specified based on engineering judgment:
HGMF
Superficial face velocity
Total fan static AP, design
operating
Migration Velocity
Specific collection area
7 m/s
33.0
63.5
50.8
5 m/s
17.8
50.8
38.1
2 m/s
2.5
38.1
25.4
ESP
Total fan static AP, design
operating
4.6 cm/s
100 m2/m3/s
2.5 cm H20
5.9 cm H20
3.9 cm"H20
FABRIC FILTER
Air/cloth
Total fan static AP, design
operating
1.0 cm/s
20.3 cm H20
50.8 cm H20
38.1 cm H20
VENTURI SCRUBBER
L/G
Total fan static AP, design
operating
Water recycle ratio
93 II1,000 m3
152 cm H20
203 cm H20
191 cm H20
0.9
The scope of the cost estimates includes flange-to-flange costs from the
confluence of the particulate collection hoods (e.g., shell and canopy) to
the discharge of the clean air from the control device. For the venturi
scrubber, costs of sludge treatment equipment are also included. It is assumed
that utilities are available at the plant site at the following rates:
Electricity
Plant water
Cooling water
Compressed air
$.05/kWh
$.066/1,000 H
$.026/1,000 SL
$.706/1,000 m3
382
-------
CAPITAL COSTS
The total capital costs (TCC) were calculated using a modified Lang
method, i.e., applying factors to the purchased equipment costs to account
for direct and indirect installation costs.
A substantial amount of engineering judgment is used in formulating
these factors. However, the relative order of these factors for the ESP, VS,
and FF is consistent with other data (9). These factors reflect the expense
necessary to install and put into operation each control option. The ratio
between installation costs and purchased equipment costs for HGMF was judged
to be lower than for either the ESP or FF.
Table 3 shows that HGMF is more capital intensive than either the ESP or
FF. Looking at annualized costs, the venturi scrubber is not competitive
with any of the other three options at the given conditions. Direct opera-
ting costs, which do not include the cost of capital over the life of the
unit, are slightly higher for the HGMF at 7 m/s than for either the ESP or
FF. However, the difference in both capital and direct operating costs among
the HGMF at 7 m/s, ESP, and FF is well within the probable error of the
estimate (±30 percent).
It is important to note that tax considerations are not part of this
estimate. Investment tax credits and other tax incentives could offset some
of the initially higher HGMF capital costs by reducing the total annualized
costs of Table 3.
383
-------
CONCLUSIONS
The following conclusions were drawn from the field operation of the
HGMF mobile pilot plant:
1. Test series were performed on two types of matrices at three levels of
matrix packing density. The best overall performance of the HGMF unit
was achieved with the expanded metal matrix at a packing density of 6
percent. The highest efficiency level achieved was 96.4 percent with
five of nine tests (excluding zero applied magnetic field tests) in the
range of 93.9 to 96.4 percent.
2. In the velocity range of 4 to 8 ra/s, the expanded metal matrix at 6
percent had lower pressure drops (10 cm 1^0 to 46 cm 1^0) than the steel
wool matrix (16 cm H20 to 50 cm H20). Given the better overall per-
formance (both efficiency and pressure drop) with the expanded metal at
6 percent packing density, it was the preferred filter matrix.
3. Particle size measurements with cascade impactors were used to measure
fractional size penetration. Fractional penetration curves show per-
formance of the HGMF to be relatively poor (85 percent efficiency or
less) in the particle size range below 1 |Jm .
4. Elemental chemical analyses show iron removal efficiencies are higher
than overall mass efficiencies as determined from thimble dust samples.
However, the data show iron penetration to be as much as 6 percent when
overall mass penetration is 9 percent.
5. Potential explanations for inadequate capture of iron-bearing particles
include the following:
a. magnetic forces acting on the fine particles are not sufficent to
effect capture as the gas passes through the filter due to in-
sufficient residence time,
b. some of the iron occurs in complex compounds with zinc that is not
sufficiently magnetic to be captured; this is supported by work
done at Lehigh University on waste dusts from electric arc fur-
naces, and
c. reentrainment.
6. The overall penetration of electric arc furnace dust through HGMF
measured in this program must be reduced by a factor of 2 to 6 to compete
with the performance of conventional particulate control devices applied
to new sources. Standards for existing sources in some states might
permit the retrofit of an HGMF.
384
-------
Assuming HGMF performance can reach a competitive level in the config-
uration and operating mode tested in this program (e.g., 99 percent
efficiency), comparative annualized costs for HGMF, fabric filters,
ESPs, and venturi scrubbers show that HGMF can compete economically with
venturi scrubbers, but is more expensive than fabric filters and ESPs.
Since to achieve that level of performance on EAF dust, it would be
necessary to reduce gas velocity and/or increase filter length, it will
be difficult for HGMF to compete as a control device for EAF's.
REFERENCES
1. Obertueffer, J.A. Magnetic separation: A review of principles, devices,
and applications. IEEE Trans. Mag. Mag-110: 223, 1974.
2. Kolm, H.H., Oberteuffer, J.A., and Kelland, D.R. High gradient magnetic
separation. Sci. Am. 223: 47, Nov. 1875.
3. Oder, R.R. High gradient magnetic separation theory and applications.
IEEE Trans. Mag. Mag-12: 428, 1976.
4. lannicelli, J. New developments in magnetic separation. IEEE Trans.
Mag. Mag-12: 436, 1976.
5. Gooding, C.H., Sigmon, T.W., and Monteith, L.K. Application of high-
gradient magnetic separation to fine particle control. EPA-600/2-77-230
(NTIS PB 276-633), 1977.
6. Keyser, N.H., et al. Characterization, recovery and recycling of elec-
tric arc furnace dust. Paper presented at the Symposium on Iron and
Steel Pollution Abatement Technology for 1981, Chicago, Illinois.
October 6-8, 1981.
7. Severson, S.D., Horney, F.A., Ensor, D.S., and Markowski, G.R. Economic
evaluation of fabric filtration versus electrostatic precipitation for
ultrahigh particulate collection efficiency. FP-775, Research Project
834-1, prepared for EPRI by Steams-Roger, Inc., 1978.
8. ES&T currents. Environmental Science & Technology. Vol. 12, No. 13,
December 1978.
9. Neveril, R.B. Capital and operating costs of selected air pollution
control systems. EPA-450/5-80-002, U.S. Environmental Protection Agency,
Research Triangle Park, NC, 1978.
385
-------
NOVEL PARTICULATE CONTROL TECHNOLOGY
by: Senichi Masuda
Department of Electrical Engineering,
Faculty of Engineering, University of Tokyo
7-3-1, Kongo, Bunkyo-ku Tokyo, Japan 113
ABSTRACT
A review of pulse energization and precharging is attempted in view of
their inherently great potentials for particulate control and the current
controversies around these two technologies. They may provide three major
advantages when properly designed and applied. These are "energy-saving",
"back corona correction for high resistivity dusts", and "performance
enhancement for medium resistivity dusts. The physical backgrounds of these
technologes are examined with a special attention on the technical potential
of nanoseond pulse. Different designs and operation modes of these technolog-
ies are discussed in consideration of various application areas and dust
resistivity levels, with an intention to provide a guide-line for correct
use of these technologies.
INTRODUCTION
The current interests world-wide in the field of particulate pollution
control may be three-fold: "Cost-Effectiveness", "Energy-Saving", and "Sub-
micrometer Particules". To meet with these interests a number of novel
technologies have been proposed. Among these are discussed in this review
only on "Pulse Energization" and "Precharging", as these are currently under
great arguments and controversies, yet are quite certain to provide great
impacts when fully developed and properly applied.
PULSE ENERGIZATION
The advantages of "Pulse Energization" currently being expected are
three-fold: "Energy-Saving", "Correction of Back Corona" and "Performance—
Ehancement at Medium Dust Resistivity". These are based on its inherent
feature capable of smoothly decreasing corona current level without degrad-
ing a uniformity in its distribution and a main field strength. The current
emphasis in this technology is directed to its application primarily for
retrofitting to existing precipitators. This is a factor causing various
confusions and controversies, as the essential feature of this technology is
much more involved. Therefore, a current estimation of its cost-effective-
386
-------
ness should never be taken as representing its actual potential. The pulse
energization is not a single unit operation, but a family of similar but
different groups, each having its own specific application area. The design
and operation modes of pulse energization differ greatly depending upon the
target of its applicatio and dust resistivity level encountered, producing
"Energy-Saving Mode", "Back Corona Corection Mode" and "Performance-Enhanc-
ing Mode". Another great difference arises from the pulse to be used itself,
more exactly its duration time, t . These are "Millisecond-Pulse", "Microsec-
ond-Pulse", and "Nanosecond-Pulsfe". The third difference appears in the
construction of electrode system: the conventional "Twin-Electrode System"
and "Tri-Electrode System" having an additional third electrode. The fourth
difference is whether it is used for retrofitting to an existing plant or in
a new plant.
Mode of Operation
a) Energy-Saving Mode; This design or operation mode is directed to the
greatest current concern, and applicable in most of the precipitators with
and without back corona. The "Intermittent Energization System" - MIE -
commercialized by Mitsubishi Heavy Industries (1) may represents a typical
example although it also produces back corona correction effect. By a
periodical blocking of ac primary current with thyristors the secondary
current is blocked intermittently, with a concurrent pulsation of precipitat-
or voltage (Fig. 1) . The power consumption is reduced in proportion to
decrease in duty ratio, Re = t./Ctj + t~) • Extensive tests made at a pilot
plant and 8 full-scale plants indicate that different grades of improvement
in performance can be achieved depending primarily upon the dust resistivity
level._The best result is obtained for the high resistivity dusts with 10
- 10 ohm-cm causing severe back corona (low-sulfur low-alkali coal
fly-ashes and iron ore sinter-machine dusts) with a reduction in power con-
sumption R = 10 - 30 % and performance enhancement in terms of modified
Deutsch migration velocity H = 1.1 - 1.7 at the optimum duty ratio Re = 1/5
- 1/3. A- moderate improvement is achieved for the medium resistivity dusts
with 10 - 10 ohm-cm causing slight back corona (general overseas coal
fly-ashes) with R = 30 - 50 % and H = 1.1 - 1.3 at Re = 1/3 - 1/2. However,
in the case of medium-low resistivity dusts with less than 10 ohm-cm
causing no back corona, the advantage of the MIE is greatly reduced, indicat-
ing a maximum of power reduction R = 50 % with a slight performance degrad-
ation H = 0.8.
b) Back Corona Correction Mode; This is the most fruitful application
area of pulse energization. The pulse voltage is applied intermittently to
discharge electrodes on top of a dc "Base Voltage", V, . In the case of a
tri-electrode system, the dc base voltage is applied oetween the third and
collecting electrodes, and the pulse voltage is applied across the third and
discharge electrodes. Most important, it is imperative to produce corona
discharge only at an instant when the pulse voltage is applied, but never in
a period between two successive pulses. Otherwise, a precise control of
corona current by means of pulse parameters (frequency, f ; crest voltage
V ; duration time, t ) necessary for correction of back corona is lost. Back
corona can be corrected by lowering corona current density, i,, so as to
meet the following criterion:
387
-------
i, x r, < E, (1)
d d ^ ds v
where r = dust layer resistivity. The correct selection of V, is of utmost
importance in this case. In the case when r, is not too high, v, can be set
close to corona inception voltage, V . Care must be taken, however, when r,
becomes very high, exceeding say 10 ohm-cm, to cause a severe back corona
and a great hysteresis in V-I curve with its inception voltage, V , much
higher than its extinction voltage, V . In this case the level of V^ must
jG Q.C.
be set slightly below V (2,3). The reason for this is to avoid an
uncontrollable runnaway of back corona in a form of its lateral propagation
(4). The large hysteresis in V-I curve is due to both self-stabilization of
local back corona and its lateral propagation to cause a time-dependent
current increase. This begins to occur when the field intensity in the
collection field, E, exceeds a level of streamer propagation, E (ca. 5
kV/cm in air at NTP) (2). Now, even if V, be set lower than VS or at a
QC C
value with E lower than E , stochastic fluctuations of process parameters
may cause back corona at a local point somewhere in a large collection
field, as,well as its lateral propagation. Once this happened, it can never
be extinguished unless V, be decreased below V . The constraint to maintain
dc G
V below quite a low level V produces a great difficulty as it inevitably
impairs collection performance (3). This may lead to a thought that pulse
energization may be useless for extremely high resistivity dusts. However, a
possibility of its correction arises from a fact that it takes at least
several seconds before the lateral propagation reach a detrimental level.
The author and his co-workers confirmed the effectiveness of an operation
called "Back Corona Quenching" which is an intermittent abrupt lowering of
V from a high level close to V to zero and a succeding recovery to its
original level. The entire back corona is immediately quenched, and the
current shows a fairly slow time-dependent rise although V, is as high as
V . Before the current reaches a detrimental level by lateral propagation,
the quenching operation is repeated. Its combination with pulse energization
will be tested in more detail in the author's laboratory. It is felt that
the pulse energization for better correction of back corona will require a
more sophysticated control of power supply, which in turn will neccesitate
the development of more advanced "Back Corona Sensors". The "Bipolar Current
Probe" (5,6), enabling on-line measurement of both negative and positive
ionic current desities, may meet such a requirement.
The merits of pulse energization becomes diminished .with increasing
dust resistivity, and completely lost beyond say 10 - 10 ohm-cm. First,
a shortage occurs in corona current density (ion depletion) to meet the
condition (1), lowering greatly the particle charging speed. This
necessitates the use of a "Back Corona Free Precharger" in front of a pulsed
collection field, as discussed later. Second, at an extremely high dust
resistivity particle charge in a dust layer is retained to form a space
charge, and this modifies greatly the field distribution inside to
invalidate the above criterion (7). Back corona occurs at a certain layer
thickness, called "Limiting Thickness", independent of its resistivity (even
when i, = 0 in extreme cases), so far as an externally applied field, E,
exists (7). Back corona is enhanced by increasing corona current with a
decrease in the limiting thickness. In other words, in the presence of
corona current, back corona occurs sooner or later, unavoidable even with
the pulse energization, unless dust layer be constantly cleaned. This was
388
-------
cleary confirmed by using the bipolar current probe (5). Back corona at such
a high resistivity takes a form of a large number of very feeble scattered
glow spots (glow-mode) (5,7,8), hard to detect by visual observation, yet
emitting large quantity of positive ions in total (5). The only possible
electrical control means in this case is to use a two-stage precipitator
consisting of a back corona free precharger and a parallel-plane collection
field, as described later.
A number of pulse-energized ESP's, both of twin-electrode systems
(9-12) and tri-electrode systems (13-15), indicated more or less successful
results in improving correction efficiency for high resistivity dusts.
c) Performance-Enhancing Mode: This mode is for enhancing a collection
performance at a medium dust resistivity causing no back corona. V, is set
dc
close to the sparking voltage, V , and pulse voltage is applied on its top.
Since the pulse spark voltage becomes increasingly higher with decreas-
ing its duration time, t , the overall crest voltage can be raised beyond
the dc spark voltage, V . It is expected that particle charge could be
raised concurrently to produce an enhancement of collection performance. The
enhancement up to about H = 1.2 is reported by several investigators, while
others recognize no distinct improvement. The whole story is still underly-
ing a great controversy. More time seems to be required before a definite
condition for enhancement and its extent can be more clearly identified.
Pulse Duration Time
A pulse voltage propagates along a corona transmission line, consisting
of discharge and collecting electrodes (or discharge and third electrodes in
a tri-electrode system), with a speed close to light velocity, i.e. about =
0.3 m/nanosecond (16) The exact speed in a lossless line is given by v =
1/(LC) , where L and C represent line inductance and capacity per irnit
length. Hence, the geometrical length on a line of a pulse voltage wave with
t nanosecond duration time is about 1 = 0.3 x t m. In the case when the
length of the line, LI , is longer man about 1 /3 - a condition for the
mulitiple-reflected portion of the wave to occupy, before decaying, only a
fraction of the total wave length - the pulse voltage behaves as a
travelling wave, and the line must be treated as a distributed constant
circuit. Whereas, in the case when 1 is much longer than LI , its character
as a wave is lost, and the line should be treated as a circuit with lamped
constants, L, C and R. Assuming L = 100 m in a practical plant, the
critical pulse duration time dividing these two regions is roughly t =
1,000 nanoseconds. In other words, the nanosecond pulse with t shorter
about 1,000 nanoseconds should be treated as a travelling wave, wtrereas the
microsecond and millisecond pulses necessitate a circuit handling with
lamped constants. A great difference occurs concurrently in physical phenom-
ena and engineering approach to the designs of electrode system and pulse
power supply.
a) Nanosecond-Pulse; In this case the energy is concentrated in the
localized travelling voltage wave. The necessary energy input from the pulse
power supply into the line can become lower than the other types of pulses
depending upon the level of 1 /L... The negative streamer coronas are
triggered by this pulse wave to appear uniformly along the line (-Fig. 2),
389
-------
but lasting only a very short time to make spark voltage extremely high.
During the course of propagation the pulse wave is gradually eroded by these
streamers from its front edge of the peak (Fig. 3), finally to become
impotent so as not to trigger the streamers any more. The streamers act as a
plasma ion source with ion concentration of about 5 x 10 ion pairs/cm
(17), emitting negative ions to the charging and collection zone with the
aid of dc base field. The streamers are produced in a very short time of
only several nanoseconds, but its plasma life time is as long as several
milliseconds, owing to a slow process of ion recombination. Hence, the
corona current lasts for several milliseconds, too (17,18). The pulse energy
is also consumed by ohmic loss of the line enhanced by skin-effect especial-
ly at its sharp rising front. As a result, its rising speed becomes
gradually slowed down. Hence, after losing its power of producing streamers,
the pulse wave becomes flattened after multiple-reflection, finally to
become a dc voltage covering uniformly the entire length of the line and to
be added to the dc base voltage. This part may represent a final loss as a
pulse energy, but actually it is converted into a dc energy. The greatest
advantage of the nanosecond-pulse is based on the facts that its short
duration time is still long enogh for a streamer to be fully developed, that
its energy can be fully squeezed out into corona by using novel technologies
specific to its wave nature, such as wave reflection, wave compression,
etc., and further that its pulse power supply can be made extremely simple,
efficient and cheap, when properly desinged.
Fig. 4 illustrates one of such power supplies. The capacity of the
pulse forming condenser, C , is very small depending upon nl /L-, where n =
number of parallel corona Transmission lines to be pulsed, sunce the magni-
tude of C comparable to the electrostatic capacity of total lines is large
enough in this case. Its charging is made by an ac voltage from its zero
level through a rectifier and a very small protective resistance. This "AC
Charging" mode greatly reduces charging loss compared to an ordinary "DC
Charging". The condenser voltage after charging is held by the rectifier up
to the next half cycle, when the polarity of ac voltage is reversed. Switch-
ing is made in this cycle by a rotary spark gap to the feeder cable acting
as a resistive load equal to its surge impedance Z = (L/C) . The voltage
wave is distributed from the feeder to a number of corona transmission lines
in the collection field. The rush current from the ac main source after
switching is interrupted by the rectifier, allowing the use of such a low
protective resistance. The erosion rate of the spark elements can be made
acceptably low by a correct selection of material,, requiring a replacement
about once a year. Fig. 5 shows a picture of one such pulse power supply (V
= 55 kV; rise time t - 65 nanoseconds; t ** 400 nanoseconds (half-peafc
width); f = 50 z; 7.73 kW output pulse power for 5 ohm surge impedance; 90
% efficiency) which is very compact in size and very high in efficiency.
"Coupling" of pulse power to a feeder at a high dc base voltage is a
problem to be carefully considered. This is made in a twin-electrode system
through a coupling condenser which is not only expensive but also produces a
"Coupling Loss" inversely proportional to its capacity. In a tri-electrode
system, however, a direct coupling can be used to remove this problem.
"Tapping" of pulse energy from a feeder to the lines also causes a
problem, owing to an abrupt drop of surge impedance at each tapping point
which causes a partial reflection of pulse voltage wave. A stepwise drop of
pulse voltage occurs in the feeder downstream of each tapping point,
390
-------
impairing uniform distributiom of pulse power. Fig. 6 illustrates a "Graded
Feeder" (19) solving this problem by compensating the tapping effect with a
stepwise increaese of its surge impedance.
Finally, one must also look at the "Corona Transmission Line" itself.
Its terminal must be opened in any case to recover the pulse energy by
reflection. The modification of pulse wave form due to corona loss, with its
crest lowered and duration elongated, must be corrected so that the pulse
energy can be fully squeezed out into corona formation so as to produce a
longest possible active length of the corona transmission line and to reduce
the cost and power consumtion of the pulse power supply. One of the possible
solutions is "Pulse Compression" (20) effected by a gradual or stepwise
increase of a line surge impedance. This can be made by inserting induct-
ances (ferite cores, coils, etc.) into the line, or by lowering the line
capacity with decreasing electrode width or gap. Partial reflection of
voltage wave occurs on the line either continuously or in multiple steps to
produce such a "Compression". Fig. 7 indicates examples of a single pulse
compression made by inserting a coil in series to a 100 m zig-zag corona
transmission line. A peaking produced by its open end should also be noted.
The tri-electrode pulse-energized field is being developed for practical
application in the author's laboratory in consideration of its advantage in
eliminating the coupling condenser and its greater flexibility in electrode
design and voltage control.
b) Microsecond-Pulse; In this case a precipitator behaves essentially as
a capacitive load with a very high corona resistance in parallel. The entire
inter-electrode capacity is fully charged by the "non-±ravelling" pulse
voltage to store quite a large capacitive energy (1/2)CV . The pulse wave
forming requires the removal of this energy from the system after a desired
pulse duration time. Since this capacitive energy is much larger than that
consumed by corona, it must be recovered to the tank condenser of pulse
power supply, using "L-C oscillation" through a thyrister switch and a
reverse rectifier (Fig. 8)(11,21). The recovery rate can be made quite high,
about 70 to 90 %. However, for this energy recovery scheme to be effective,
the capacity of the high voltage tank condenser should be about one order of
magnitude larger than that of an entire electrode system. In addition, the
twin-electrode system requires a pulse transformer or a coupling condenser
for a pulse coupling (Fig. 8 (a)). The direct coupling is possible in the
tri-electrode system, requiring, however, a switching element to drain the
residual charge from the load capacity after each recovering cycle (Fig. 8
(b)).
An enhancement in collection performance in the range of H = 1.2 - 1.7
can be obtained by using the microsecond-pulse in both tri-electrode system
(13,22) and twin-electrode system (9-12). Its effect becomes pronounced with
increasing dust resistivity. The advantage of microsecond-pulse is that it
can generated by using a thyrister switch, a well-proven solid-state
element completely free of erosion, and that its noise levels, both sonic
and electromagnetic, remain substantially lower than the spark-switch. An
instantanuous power flowing through the pulse transformer is quite large,
requiring a concurrently much larger current capacity than corona current
itself. Its step-up ratio is restricted, so that many series-connected
thyristors must be used in its primary to cope with a high primary voltage.
391
-------
c) Millisecond-Pulse: A great advantage of using the millisecond-pulse
is a simplicity of its power supply with its cheaper cost. This is because
the pulse formation control can be made at the primary of the main transform-
er by simply interrupting the primary current. The "MIE System" (Fig. 1) is
one of its typical examples. Of course, the circuit as shown in Fig. 8 can
also be used to produce a better pulse wave form. As the pulse-induced
saw-teeth voltage produces a kind of dc bias voltage, it can be omitted in
this case. The millisecond pulse in both twin-electrode and tri-electrode
systems, not only produces an "Energy-Saving", but also "Back Corona Correc-
tion" for high resistivity dusts (1,13-15)). However, an advantage of
increasing spark voltage, specific to a narrow pulse, is lost in the
millisecond pulse. Hence, an overall crest voltage is restricted to a dc
spark voltage, reslting in a lower effective dc base voltage. This provides
a performance limitation in the medium resistivity applications (1).
Retrofitting or New Plant Application
The current emphasis is primarily placed on a retrofitting use of pulse
energization for improving substandard performance of existing plants.
Hence, the conventional electrode construction poses a definite constraint
on its applicable mode, often requesting a modification of sound approaches
of pulse power technology. A direct pulse coupling must be excluded because
of a twin-electrode design, and the use of a cost-effective, energy-saving
nanosecond-pulse can not be considered. This situation is quite understand-
able in view of of the current urgent interests. It should not be overlooked
that noticeable advantages are being obtained by such retrofitting when pro-
perly designed and applied. Anyway, it should be emphasized that the current
data from retrofitting applications should not be taken as reflecting its
maximum potential. Fig. 9 (c) shows one of the possibilities of introducing
a nanosecond pulse technology into an existing ESP, by attaching a long
non-corona wire (dotted line) close to the convetional corona wires to
constitute a "Corona Transmission Line" in a form of a tri-electrode system.
Figs. 9 (a) and (b) illustrate the corona transmission lines of twin- and
tri-electrode systems, both applicable to a new plant.
PRECHARGING
The precharging is used in a mode of "Two-Stage Module" for "Energy—
Saving", "Back Corona Correction", or "Performance-Enhancement" of the
downstream ESP field. The module consists of a precharger and an ESP,
either conventianal (de-energized), or pulsed, or non-corona (parallel-plane)
type. In a high resistivity application, the precharger must be, first of
all, "Back Corona Free" with its absolute charging performance being the
second priority. Only at a medium dust resistivity causing no back corona,
the "High-Performance Precharger" comes to the highest rank.
Hardwares of Precharger
a) High Intensity Ionizer; The well-known "High Intensity Ionizer" (Fig.
10) (23) represents one of the typical "High-Performance Prechargers". The
gas velocity in an annular charging zone is made very high (about 30 m/s).
The thick metal rod supporting a disc-like corona electrode produces a
radial controlling field. The sparking voltage in the charging zone is
392
-------
extremely high, beyond say 10 - 12 kV/cm, with a concurrently very high
current density of 20 - 30 mA/m , so that a charge-to-mass ratio obtainable
at its outlet can be 20 - 100 micro-coulombs/g depending upon the particle
size. This may be due to the very high gas speed to produce a downstream
shift of a dense ionic and dust space charge from the critical anode region
which otherwise launches spark in the presence of such field-enhancing space
charge. An effective shift may be assisted by a radial field of the metal
rod to confine the ionic current into a narrow disc-like pattern and to
avoid its upstream divergence. The high gas speed must be slowed down at its
outlet by locating a free conjunction space to a level acceptable to the
succeeding collection field. Care must be taken so as not to produce a too
big free volume. Otherwise, a highly charged dust cloud produces an
extremely high space-charge field as to cause positive coronas at many
protruding grounded points. These coronas continue to emit copious positive
ions to the dust cloud to reduce its charge, and never stop till the
space-charge induced local field at each point drops below a corona
inception level. Thus, particle charge is self-limited to a level inversely
proportional to the linear dimension of the conjunction space. This problem
can only be solved by dividing the space into a number of smaller regions
with metal members. High resistivity dusts can easily cause back corona at
the throat surface owing to a high current density. This is corrected by
injection of steam to the critical surface region of the throat to reduce
dust resistivity, which however resluts in an additional cost. The High
Intensity Ionizer certainly provides an enormas potential when properly
applied.
b) Tri-Electrode Precharger (EPA/SoRI Precharger): Fig. 11 illustrates
one of the "Back Corona Free Prechargers", called "EPA/SoRI Precharger" (24)
comprizing a grid electrode close to a plate of a twin-electrode corona
system. The grid is fed with a negative bias-voltage so as to avoid arrival
of negative ions to cause back corona. The negative ions from the discharge
electrode is allowed to pass through the grid openings to reach the plate.
Back coronas on the plate are expected from the first, but positive ions are
to be collected by the grid completely. When the potentials of discharge and
grid electrodes are properly tuned to each other, a very good charging
performance can be obtained (24). The grid bias-voltage must be high enough
to avoid a back corona on it, but never exceeds a critical level to produce
"Streamer-Mode Back Corona" (8) in the grid/plate interspace. Otherwise, a
"Lateral Propagation" of an original corona may start, finally to produce a
uniformly bi-ionized atmosphere in the entire interspace, emitting copious
positive ions to the charging space so that the control of grid is
completely lost. "Back Corona Quenching" control described previously, by an
intermittent abrupt zero-setting of the grid voltage, proved to be effective
for its correction. A micro-processor control may also provide a reliable
tuning. Large scale pilot plant tests of this precharger are going at a Bull
Run Power Station (25). These will bring this technology to a mature level
to make its inherent potential available to a public.
c) Water-Cooled Prechargr (EPA/DRI Precharger; Fig. 12 illustrates
another "Back Corona Free Precharger", called "EPA/DRI Precharger" (26)
using water-cooled grounded pipes for the anodes. The resistivity of dust
layer covering the pipe surface can be reduced by lowering its temperature,
393
-------
so that back corona can be eliminated. A very satisfactory charging
performance can be obtained, so far as the water temperature in the pipes be
kept below a certain critical level. The largest attraction of this
precharger is its simple construction and, possibly a lower cost. The most
essential factor to be carefully considered in this precharger is an
effective removal of dust deposit from the pipe surface. Otherwise, the
poor heat conductivity of porous dust layer should produce, in combination
with its increased thickness, a temperature rise and local increase in dust
resistivity at the dust layer surface being constantly subjected to a hot
gas flow, finally to produce back corona. The dust removal is made by
mechanical rapping of pipes, but when applied in a large ESP, the use of
mechanical scrapers may become imperative. Concentrations of S0_ and H-0 in
the flue gas are another critical factors deciding the effectiveness of this
precharger.
d) Boxer-Charger; Fig. 13 illustrates the third example of the "Back
Corona Free Prechargers", called "BOXER-CHARGER" (18,27-29). Unlike all
other prechargers, BOXER-Charger uses an alternating field in its charging
zone between two double-helix electrode units (Fig. 14 (a)). When one of
them takes a negative peak of its ac voltage, a nanosecond pulse voltage is
applied across its two herical wires. This proceeds as a travelling wave
along a corona transmission line of the two helical wires, producing uniform
streamer coronas along the line (Figs. 14 (b), 15 (b) and (c)). At first
negative streamers (Fig. 15 (b)) are launched from the negative wire. Then,
positive streamers (Fig. 15 (c)) are emitted from the positive wire when the
pulse voltage is high enough. This is very well indicated in the streak-phot-
ograph of these streamers (Fig. 15 (a)) (18). The negative streamers produce
only a small erosion in the pulse wave form ahead of its peak. Whereas the
positive streamers causes a large dip downstream of its peak with a
concurrently great energy loss (Fig. 16) (16). Negative ions are extracted
from the streamer plasma to travel across the charging zone. When the
polarity of the ac main voltage is reversed, the opposite double-helix is
energized by the nanosecond pulse voltage to emit negative ions to the
oposite direction. The dust particles are bombarded by these ions from both
sides to be charged rapidly. A great advantage of this charging scheme lies
in a possibility of periodical "Charge-Elimination" from the dust-contaminat-
ed wires of the double-helix units. The negative ions from a double-helix
unit arrive at its oposite unit and accumulate on the surface of dust
deposit on its two wires. The dust surface potential rises with time, and
would cause breakdown (back corona) if nothing happens. Before that the
polarity of the ac field is reveresed so that charge accumulation interrupt-
ed. Next, this opposite unit is energized so that plasma is produced very
close to the accumulated charge. Hence, it is immediately neutralized by
positive ions from the plasma, and the cycle repeats itself. Thus,BOXER-CHAR-
GER possesses a built-in mechanism of back corona correction, when properly
designed. After successful tests in a both laboratory and pilot plant, its
demonstration model for testing at a larger plant will be completed by
March, 1983. Fig. 14 indicates a small BOXER-CHARGER system to be tested at
a pilot at an incimerator, and its power supply. It is discovered that a
great-energy saving is possible in the pulse power supplies of BOXER-CHARGER
by selecting a correct pulse voltage to produce only negative streamers for
the plasma ion source. The negative streamers are much less energy-consuming
394
-------
than positive streamers (16,18), yet can emit almost the same amount of
negative ions (17). The nanosecond-pulse power supply, direct pulse
coupling, graded feeders, and pulse compression technology described
previously are also used in BOXER-CHARGER technology. A pilot plant test of
precharging in front of a bag filter was also performed using BOXER-CHARGER
with a satisfactory result.
Mode of Operation
a)Energy-Saving Mode; The combination of a precharger with a energy-sav-
ing collection field, such as a low-corona conventional field, pulsed-field,
or a parallel-plane field, is getting an increased attention in view of an
overall energy-saving possibility. Provided dust particles be fully charged
by a precharger, only an electrostatic field (parallel-plane system) should
be required for their collection. At a very high dust resistivity, beyond
say 10 ohm-cm, this approach is possible (30), as the particle charge is
preserved in the dust layer on a collecting electrode to produce enough
electrical adhesion force. At a medium resitivity, say below 10 ohm-cm,
the parallel-plane collection field indicates a very good performance at an
initial stage of operation. However, with the growth of dust deposit on the
planes a time-dependent performance degradation by dust reentrainment occurs
(30). This is because the particles arriving at the dust deposit surface
quickly lose their original charge at this resistivity level, and are
immediately induction-charged to the opposite polarity to be strongly
subjected to an electrostatic attraction force. Whereas the surface of dust
deposit becomes increasingly rough with its growth, producing a concurrent
reduction of its physical adhesion force and increased dust reentrainment.
The use of a short water-irrigated field could provide a solution for this
problem. In a "Corona Collection Field", either dc- or pulse-energized, the
dust reentrainment is greatly reduced as the negative ions from corona
produce an adequate electrical adhesion force.
b) Back Corona Correction Mode;
(i) Conventional Twin-Electrode Collection Field: First, let us consider
a conventional twin-electrode collection field subjected to back corona
activity with bi-ionized atmosphere. The highly charged particles from a
precharger lose their charge with time by attachment of positive ions, and
their charge finally reach a saturation level specific to the "Back Corona
Severity" of this bi-ionized field, defined as the ratio of positive to
negative ionic current density, i /i (5). In other words, the average
charge of particles in the back corona field can be always higher than that
without precharger, depending upon the back corona severity, and its space
distribution in the field which determine their charge decaying velocity and
the final level of charge (5). A great difference is produced by whether
back corona is localized in a form of channels (lower resistivity), or it is
uniformly distributed (very high resistivity). In the former case, the
overall charge decaying speed becomes fairly low to produce a high average
particle charge, even if a local severity of back corona be quite high.
Whereas the latter type of back corona is highly detrimental. Anyway, it is
highly preferable to use in series a multiple number of "Precharger/Collect-
ion Field Modules" to compensate the back corona induced degradation of
395
-------
charge repeatedly. Fig. 17 (a) shows laboratory data indicating a great
advantage of such a multi-module design (dotted line). This produced an
enhancement of Deutsch migration velocity (not modified) as high as 2.44
(Table 1) (29).
(ii) Pulsed Collection Field: Perhaps the module out of "Precharger/Pulsed
Collection Field" would represent the most favorable combination. The highly
charged particles from a precharger can preserve their charge in the pulsed
collection field operated in the back corona correction mode, as no positive
ions exist. A slight quantity of negative ionic current can provide adequate
electrical adhesion force to the dust layer on the collecting electrode to
avoid dust reentrainment. Fig. 17 (b) indicates the laboratory data of this
multiple-module design (dotted line), with the enhancement factor as high as
2.9 (Table 1) (29). This performance level is exactly the same as that
obtained for the medium resistivity dust at normal temperature, completely
free of back corona (29).
(iii) Parallel-Plane Field: ,. In the case when dust resistivity becomes
extremely high, beyond say 10 - 10 ohm-cm, neither a dc- nor pulse-ener-
gized collection field can perform at all after a precharger, indicating
practically zero electrical collection performance. Only possible way of
electrical collection is to use the parallel-plane field (30). Even Boxer—
Charger loses its charging performance when operated at negative charging
mode, and its normal charging performance is obtained only at positive
charging mode (30).
c) Performance-Enhancing Mode: The primary interest of this mode is a
retrofitting use to improve a substandard performance of an existing
precipitator operating at a medium dust resistivity causeing no back corona.
The second interest is on enhancing overall cost-effectiveness of a new
plant for a medium resistivity dust. Emphasis in the precharger is placed
primarily on its absolute charging performance, i.e. the levels of field
intensity and ionic current density obtainable in its charging field. In
retrofitting application, however, care must be taken in advance so as to
clearly identify whether the charging is really the cause of a trouble. This
is because the particle charging in a precipitator at medium dust resistivity
is usually quite good, showing enough charging speed and charging level, and
the substandard performance is mostly produced by other causes, such as
under-sizing, non-uniform gas distribution, gas sneakage, uncorrect rapping,
etc. As a result the effect of precharging may be quite small (29), or even
hardly detectable. In the case of a new plant, an ideal design of precharger
/collector module may be possible, combining a high-performance precharger
with a "Reentrainment Free Collection Field" such as with a low-current dc-
or pulse-energized field, or a parallel-plane field followed by a water-irri-
gated short corona section.
CONCLUSION
The present status of both pulse energization and precharging technolog-
ies are reviewed, with a special attention to the difference in their design
and operation modes corresponding to the difference in their application
areas and dust resistivity levels encounterd. The physical backgrounds of
396
-------
various pulse technologies are discussed from a standpoint, how to take full
advantage of their technical potentials. The pulse energization and precharg-
ing technologies have great potential advantages of energy-saving, back
corona correction, and performance-enhancement for medium resistivity dusts.
Their application areas are two-fold: retrofitting use in an existing plant,
and use in a new plant. In the former applications, the existing plant poses
a great constrain on the design and operation modes of the pulse energizat-
ion and precharging to be used. In the latter application, theoretically
most reasonable design may be possible. It should be emphasized that both
technologies are still in the course improvement and elaboration, but their
full advantages will surely become available in a near future.
REFERENCES
(1) T. Ando, N. Tachibana and Y. Matsumoto: A New Energization Method for
Electrostatic Precipitators - Mitsubishi Intermittent Energization
System, Proc. 4th EPA-Symposium on Transfer and Utilization of Particu-
late Control Technology (Oct., 1982 in Houston, Texas).
(2) S. Masuda, S. Obata and Y. Ogura: Lateral Propagation of Back-Discharge
in a Tri-Electrode System, Inst. Phys. Conf. Ser. No. 48 (The Inst. of
Phys., London), p. 9 (1979).
(3) M.D. Durham, G.A. Rinard and D.E. Rugg: Evaluation of Novel Electrostat-
tic Precipitator Technology, Proc. 75th Annual Meeting of APCA, Paper
No. 82-34.4 (June, 1982 in New Orleans, Louisiana).
(4) S. Masuda and T. Itagaki: Lateral Propagation of Back Corona in Twin-
Electrode Type Precipitators, Proc. 4th EPA-Sumposium on Transfer and
Utilizatio of Particulate Control Technology (Oct., 1982 in Houston,
Texas).
(5) S. Masuda and Y. Nonogaki: Bi-Ionized Structure of Back Discharge Field
in an Electrostatic Precipitator, Proc. IEEE/IAS 1981 Annual Conf.
p. 1111 (Oct., 1981 in Philadelphia).
(6) S. Masuda and Y. Nonogaki: Sensing of Back Discharge and Bipolar Ionic
Current, Journal of Electrostatics, 10 (1981) 73-80 (Elsevier).
(7) S. Masuda, A. Mizuno and K. Akutsu: Initiation Condition and Mode of
Back Discharge for Extremely High Resistivity Powders, Proc. IEEE/IAS
1977 Annual Conf., p. 867 (Oct., 1977 in Los Angels, California).
(8) S. Masuda and A. Mizuno: Initiation Condition and Mode of Back Discharge
Journal of Electrostatics, 4 (1977/1978) 35-52 (Elsevier).
(9) H.I. Milde and P.L. Feldman: Pulse Energization^of Electrostatic Precipi
tators, Proc. IEEE/IAS 1978 Annual Conf., p. 66 (Oct., 1978 in Tronto,
Canada).
(10) H.I. Milde and H.E. VanHoesen: Application of Fast Rising Pulses to
Electrostatic Precipitators, Proc. IEEE/IAS 1979 Annual Conf., p. 158
(Oct., 1979 in Cleveland, Ohio).
(11) P. Lausen, H. Hendriksen and H.H. Petersen: Energy Conserving Pulse
Energization of Precipitators, Proc. IEEE/IAS 1979 Annual Conf., p. 163
(Oct., 1979 in Cleveland, Ohio).
(12) H.H. Petersen: Application of Energy Conserving Pulse Energization
for Electrostatic Precipitators - Practical and Economic Aspects, Proc.
3rd EPA-Symposium on Transfer and Utilization of Particulate Control
Technology (March, 1981).
(13) S. Masuda, I. Doi, M. Aoyama and A. Shibuya: Bias-Controlled Pulse
397
-------
Charging System for Electrostatic Precipitator, Staub-Reinhalt. Luft 36
(1976) No. 1, p. 19.
(14) S. Masuda, I. Doi, I. Hattori and A. Shibuya: Back Discharge Phenomena
in Bias-Controlled Pulse Energization System, Proc. 4th Int. Clean Air
Congress, Paper No. V-52 (May, 1977 in Tokyo).
(15) S. Masuda: Novel Electrode Construction for Pulse Charging, Proc. 1st
EPA-Symposium for Transfer and Utilization of Particulate Control Tech-
nology (July, 1978 in Denver, Colorado), Vol.1 (EPA-600/7-79-044a, Feb.
1979).
(16) S. Masuda and H. Nakatani: Distorsion of Pulse Voltage Wave Form on
Corona Wires Due To Corona Discharge, Proc. 4th EPA-Symposium on Trans-
fer and Utilization of Particulate Control Technology (Oct., 1982 in
Houston, Texas).
(17) S. Masuda and Y. Shishikui: Pulse Corona As Ion Source and Its
Behavios in Monopolar Current Emission, ibid.
(18) S. Masuda, H. Nakatani, K. Yamada, M. Arikawa and A. Mizuno: Production
of Monopolar Ions by Travelling Wave Corona Discharge, Proc. IEEE/IAS
1981 Annual Conf., p. 1066 (Oct., 1981 in Philadelphia).
(19) S. Masuda, H. Nakatani and T. Kaji: Graded Feeder for Uniform Distri-
bution of Pulse Power to Many Loads, to be published.
(20) S. Masuda and S. Hosokawa: Pulse Compression for Regeneration of Its
Energy, to be published.
(21) S. Masuda, S. Obata and J. Hirai: A Pulse Voltage Source for Electro-
static Precipitators, Proc. IEEE/IAS 1978 Annual Conf., p. 23 (Oct.,
1979 in Tronto, Canada)
(22) G.W. Penny and P.C. Gelfand: The Trielectrode Electrostatic Precipitat-
or for Collecting High Resistivity Dust, JAPCA, Vol. 28, No. 1, p. 53
(Jan., 1978).
(23) 0. Tassicker and J. Schwab: EPRI Journal, June/July (1977) 56-61.
(24) D,H, Pontius, P.V. Bush and W.B. Smith: Electrostatic Precipitator for
Collection of High Resistivity Ash, EPA-Report EPA-600/7-79-189 (Aug.,
1979).
(25) P.V. Bush, D.H. Potius and L.E. Sparks: Pilot Demonstration of The
Precharger/Collector System, Proc. 3rd. EPA-Symposium on Transfer and
Utilization of Particulate Control Techanology (March, 1981)
(26) G. Rinard, M. Durham, D. Rugg and L.E. Spark: Development of A Charging
Device for High Resistivity Dust Using Heated and Cooled Electrodes,
ibid.
(27) S. Masuda, M. Washizu, A. Mizuno and K. Akutsu: Boxer-Charger - A Novel
Charging Device for High Resistivity Powders, Proc. IEEE/IAS 1978 Annual
Conf., p. 16 (Oct., 1978 in Tronto, Canada).
(28) S. Masuda, A. Mizuno and H. Nakatani: Application of Boxer Charger in
Electrostatic Precipitators, Proc. IEEE/IAS 1979 Annual Conf. p. 131
(Oct., 1979 in Cleveland, Ohio)
(29) S. Masuda, A. Mizuno, H. Nakatani and H. Kawahara: Application of
Boxer-Charger in Pulsed Electrostatic Precipitator, Proc. IEEE/IAS 1980
Annual Conf., p. 904 (Oct., 1980 in Cincinnati, Ohio).
(30) S. Masuda and S. Hosokawa: Performance of Two-Stage Type Electrostatic
Precipitators, Proc. IEEE/IAS 1982 Annual Conf. (Oct., 1982 in San
Francisco, California).
398
-------
kV
60
40
corona transmission lines
—- t (20msec/div)
Fig. 1 Voltage and Current Wave Forms Fig. 2 Comparison between Pulse and
of MIE System (Re = 1/3) DC Negative Coronas (l_i=200m)
-30
200
600 t (ns)
-50
-70
Fig. 3 J Erosion of Pulse Voltage Wave Fig. 4 Nanosecond Pulse Power Supply
by Negative Streamers with AC Charging and Rotary
(Vdc = - 30 kV; V = - 40 kV) Spark Gap (see Fig. 5)
Fig. 5
Nanosecond Pulse Power Supply with AC Charging and Rotary Spatk-
Switch (total length = 2173 nun; diameter = 360 mm ; Vp = 55 kV;
tf = 65 ns; t = 400 ns; f = 50 Hz; Wp = 7.75 kW; eff. = 90 %)
399
-------
Z° Z°/2 ,.
J * f
Zo/(N/2 -1)
BRANCH Zo/ Zo/ Zo/ Z
/PEEK. „* /2
HV r ^4
Z0 Z0 Z0 Z0 Z
0 ^-0
Z0 Z0 Z0 Z0 Z0 Z0 Zc
Fig.6 Graded Feeder
0 200 400 600 800 (ns)
-20
-40
(kV)
Caused by the
Inductance
Intermediate Inductance: L = 21 yH
24 m Ahead From the Ob-served Point
(mA/m"
jj
in
£ 0.06
£
jj
c
-------
^_r
J-
i
1
j
J1
IT
r
-L
\
J
n
u
r
I.
1
J
r
(
i—
(a) Twin-Electrode System (b) Tri-Electrode System (c) Modified System
Fig. 9 Corona Transmission Line for Nanosecond Pulse Energization
a: discharge electrode(disc) b: anode
c,d: supporting rod e: inlet bellmouth
f: outlet cone g: partition wall
Fig. 10 High Intensity Ionizer
a: discharge electrode
b: arid electrode
c: collecting electrode
Fig. 11 EPA-SoRI Tri-Electrode
Precharger
401
-------
discharge wire
cooling water out
/* //» !• _ ? *
1 ]
c
1
i
+
'1
7
I
t
c
c
>
1)
1
t
e
<
5
)
1
t
+
C
>
I
•(
\ / v. \ \ \
• cooled pii
water cooled pipe electrode cooling water in
Fig. 12 EPA-DRI Water Cooled Precharger with Heat-Exchanger
Spark
D K) Cable-1 Gap Cable-2
•«1 Vln
Spark
D Rl Cable-1 Cap Cable-2
three double-halical
electrode*
V-puls«
DOOOOOOG'
LDOOOOOOG1
(a) Pulse and Main Voltage Sources
(b) Electrode Element
Fig. 13 BOXER-CHARGER and Its Power Supply (Principle)
402
-------
(a) Double-Helix Units
(b) Excited by Nanosecond- (c) Power Supply
Pulse Voltage with AC
Main Voltage On
Fig. 14 BOXER-CHARGER and Its Power Supply for A Pilot Plant Test
time (na)
100 200 300
E » 7 kV/cm
(a) Streak Photograph of Nega- (b) Negative Streamers (c) Negative and Po-
tive and Positive Streamers only sitive Streamers
Fig. 15 Streamer Coronas on A Double-Helix Transmission Line Induced by
Nanosecond Pulse Voltage
403
-------
100 200 300 400
-30
-40
(a) Negative Streamers only (b) Negative and Positive Streamers
(Vp = - 28 kV at inlet; E = - 5 kV/cm) (Vp = - 35 kV at inle; E = - 5
kV/cm)
Fig. 16 Modification of Travelling Pulse Wave Form by Streamer Coronas
Table 1 PERFORMANCE ENHANCEMENT BY MULTI-MODULES OF
TWO-STAGE ESP AT HIGH RESISTIVIT FLY-ASH
Operation Mode of
Collection Fields
DC Operation Mode
under Back Disch-
arge
DC-Puls-Pulse
Operation Mode
at high pd
DC operation Mode
under No Back
Discharge
Tg
(°C)
100
100
20
Pd
(ohm- cm)
lo11
-io12
lo11
-io12
-------
4-> E
00 4- O 3
> O 0» v-
•— "o
C -C r— CU
o 4-> o s:
•r- O1O
4-> C S-
CO QJ T3 O
S 1 QJ 4-
•(-> N
QJ QJ -I— "O
C > Olr—
QJ T- S- QJ
a. +-> Q; -r-
U E Lu
-P QJ l-U
OO 4- I c
3 4- C_> O
Q LU Q T-
>> (0
r— S-
Lu CU
Q
>, £
-p QJ
•r- r—
>
••- E
•p o
00 O
•r- o:
00
QJ -P
O£ ro
C I
O QJ -—
E E
00 •!- (J
S 1 I
CU E
01 QJ -C
S- OO O
•O 03
..e OQO
QJ O
S- 4- r-
CL 4-
O 2
-c o
•P T5 r—
•r- E QJ
2 10 -Q
S-
CD
CT>
S-
en
to
o
O)
a.
s-
o
• I >, 00 O
oo 4- r- r— S- o
> O O U- QJ I
CJ> S- Ol O
C -E O >> i-r— O
O 4-> "O 4- -M E N T3 > O r— •.
rb CU T- i— T- Qj
S 1 Ol Ol -l-> S- E
4-> S- -i- 00 Q. 4- U
Qj Qj CU Lu T- 4-1
E > C to -E O E
CU •<— UJ E CU 4-* r-
Q- 4J i o O£ •<- T3 O
O QJ -r- 2 E
+-> CU oo •»-> JZ OJCM
00 4- r— U O1J= i—
34- 301-r-oOEO
QLUQ-i— HIcC Oi—
Q.
•i^
O
D.
QJ
(O
•P
00
I
O
TD
"ai
4-
O
(U
.
%
s
"
" 3
<
ri.
0
U -
§0
^
•a
3 S
T
w
•s
M »-*
- y
2 /
s /
J /
1 1
/ '/
/
/
/
n
I c
<
/'"'
l-l »
1
/
/
I .
//
J/
'/
D t
VJ '
«
M
>4 M
/
/
7
2 "
o
/ /
//
/
/'
s
!•;
* i
U '
t i
a. '
s I
< a
o c
/
/
c
?
^
> g,
\l T
I
C
"
. .
0 C\
3 C
00 4— O -C C
>OOI^:OOOQJ E
™ » <— oio:r>,i-_joo -|±
T— O1C_> r— QJ i — o t_
§ 4->E S.U-O1QJ ^
... tj fOCUTJO S-tOlO CU
^jj s-_icu4->v(an3 o c:
^ +J M 4-JjECQr— r- -?1
O QJ QJ •!— "O T- (J • 'r—
•S E >• Olf— > Q> O •" +->
" ajT-s_cu 'i-i-4-r- to
O D-+->Q)T--I->Q.4— E 3
^ -g OEU.OO O--0 Q
§ -M QJ LU T- .C 1
Z to4- c oo -*J -a i— < E r-
34-OOQJ-i-C JZ i —
QLuQ't-Q:2re 10
0 01
; (o LU
405
-------
AUTHOR INDEX
AUTHOR NAME PAGE
ADAIR, L 1-460
ADAMS, R.L 11-35
ANDO, T 11-474
ARMSTRONG, J III-241
ARSTIKAITIS, A.A 11-194
BALL, C.E III-370
BANKS, R.R 1-37, 1-62
BARRANGER, C.B 1-132
BAYLIS, A.P ,, 11-384
BELTRAN, M.R 11-51
BENSON, S.A 111-97
BERGMAN, F III-154
BIESE, R.J 1-446
BOSCAK, V 111-66
BRADBURN, K.M 11-499
BRADLEY, L.H 11-369
BRINKMANN, A III-211
BUCK, V III-335
BUMP, R.L 11-17
CAPPS, D.D 1-121
CARR, R.C 1-148
CHAMBERS, R 1-226, 1-239
CHANG, R III-271
CHEN, F.L III-347
CHEN, Y.J 1-506
CHIANG, T 11-184
CHRISTENSEN, E.M . 11-243
CHRISTIANSEN, J.V 11-243
CILIBERTI, D.F III-282, III-318
CLEMENTS, J.S 11-96
COE,JR, E.L 11-416
COLE, W.H III-l
406
-------
COOK, D.R 11-349
COWHERDfJR, C III-183
COY, D.W III-370
CRYNACK, R.R II-l
CUSCINO, T III-154
GUSHING, K.M 1-148
DAHLIN, R.S 1-192
DARBY, K 11-499
DAVIS, R.H 11-96
DAVIS, W.T 1-521
DAVISON, J.W III-166
DELANEY, S 1-357
DEMIAN, A 111-66
DENNIS, R 1-22, 111-81
DIRQO, J.A 111-26, 111-81
DISMUKES, E.B 11-444
DONOVAN, R.P 1-77, 1-107, 1-316, 1-327, 1-342
DORCHAK, T.P III-114
DRENKER, S III-271, III-282
DRIQGERS, G.W 11-194
DUBARD, J.L 11-337
DUFFY, M.J 11-489
DURHAM, M 11-84, III-241
EBREY, J.M 11-349
ENGLEBART, P.J III-183
ENSOR, D.S III-347
FAULKNER, M.B 11-204, 11-337
FINNEY, W.C 11-96
FORTUNE, O.F 1-482, 1-494
FOSTER, J.T 1-37, 1-91
FREDERICK, E.R 1-536
FRISCH, N.W III-114
FURLONG, D.A 1-287, 1-342
GARDNER, R.P 1-77, 1-107
GAWRELUK, G.R 11-17
GELFAND, P 11-35
407
-------
GIBBS, J.L 11-430
GILES, W.B 111-41, 111-53
GOLAN, L.P III-226
GOLCBRUNNER, P.R 11-401
GOLIGHTLEY, R.M 1-164
GOOCH, J.P 11-444
GOODWIN, J.L III-226
GRANT, M.A 111-81
GREEN, G.P * . . . . 1-192
GREINER, G.P 1-287, 1-357
GRONBERG, S III-141
GRUBB, W.T 1-62, 1-91, 1-179
HALL, H.J 11-459
HALOW, J.S 11-96
HANSON , P 1-460
HARMON, D 1-226, 1-239, III-131
HAWKINS, L.A. 11-194
HERCEG, Z 11-489
HOVIS, L.S 1-22, 1-77, 1-107, 1-287, 1-316, 1-327,
1-342, 1-357, 111-81, III-347
HOWARD, J.R 1-164
INGRAM, T.J 1-446
ISAHAYA, F 11-154
ITAGAKI, T 11-322
JACOB, R.0 1-446
JENSEN, R.M 1-431
JONES, R 1-303
KASIK, L.A 11-430
KETCHUCK, M 1-482
KINSEY, J III-154
KOHL, A.L III-300
KUBY, W III-271
KUNKA, S 1-239
KUTEMEYER, P.M III-211
LAMB, G.E.R 1-303
LARSEN, P.S 11-243
408
-------
LAWLESS, P.A 11-271
LEE, W 1-303
LEITH, D 111-26
LEONARD, G.L 11-230
LEWIS, M 1-179
LIPPERT, T.E III-280, III-318
LUGAR, T.W 11-184
MARCHANT,JR, G.H 11-444
MASON, D.M III-256
MASUDA, S 11-139, 11-169, 11-322, III-386
MATSUMOTO, Y 11-474
MATULEVICIUS, E.S III-226
MCCAIN, J.D III-198
MCCOLLOR, D.P 111-97
MCDONALD, J.R 11-204
MCKENNA, J.D 1-210
MCLEAN, K.L 11-489
MENARD, A 1-255
MILLER, M.L 1-482
MILLER, R.L 1-494
MILLER, S.J 111-97
MITCHNER, M 11-230
MOSLEHI, G.B 11-288, 11-306
MOSLEY, R.B 11-204
MDYER, R.B 1-460
MUSGROVE, J 1-382
MYCOCK, J.C 1-210
NAKATANI, H 11-169
NG, T.S 11-489
NOVOGORATZ, D 11-349
OGLESBY, S 11-534
O'ROURKE, R III-318
PEARSON, G 1-121
PETERS, H.J 1-179
PIULLE, W 11-401
PONTIUS, D.H 11-65
409
-------
PUDELEK, R.E 1-521
PUTTICK, D.G 11-126
QUACH, M.T 1-506
RAMSEY, G.H 1-316, 1-327
RANADE, M.A III-347
REED, G.D 1-521
REHMAT, A III-256
REIDER, J.P III-183
REISINGER, A.A 1-179
RICHARDS, R.M 1-255
RICHARDSON, J.W 1-210
RINARD, G 11-84, III-241
ROOT, R.N 1-460
ROSS, D.R 1-164
RUGENSTEIN, W.A 11-430
RUGG, D 11-84, III-241
RUSSELL-JONES, A 11-384
SAIBINI, J 1-132
SAMUEL, E.A 1-1, 11-218
SANDELL, M.A II-l
SAWYER, J III-271
SEARS, D.R 1-192, 111-97
SELF, S.A 11-230, 11-228, 11-306
SHACKLETON, M III-271
SHISHIKUI, Y 11-139
SMITH, W.B 1-148
SORENSON, P.H III-362
SPARKS, L.E 11-204, 11-271, 11-337
SPENCE, N 1-132
SPENCER,III, H.W 1-506
STELMAN, D III-300
STOCK, D.E 11-261
SUHRE, D III-335
SUNTER, T.C 1-48
SURATI, H 11-51
TACHIBANA, N 11-474
410
-------
TASSICKER, O.J III-271, III-282
THOMPSON, C.S 111-12
THOMSEN, H.P 11-243
TOKUNAGA, 0 11-96
TREXLER, E.C 11-96
TRILLING, C.A III-300
TSAO, K.C III-256
VANN BUSH, P 11-65
VANOSDELL, D.W 1-287, 1-342
WALSH, M.A 1-482
WEBER, E 11-111
WELLAN, W.G 1-420
WEXLER, I.M 11-521
WHITTLESEY, M 1-482
WILCOX, K III-154
WILLIAMSON, A.D III-198
YAMAMDTO, T III-241
YEAGER, K.E III-XV
f US GOVERNMENT PRINTING OFFICE 1985 - 559-H1/1G739
------- |