oEPA
United States Industrial Environmental Research EPA-600/7-80-037
Environmental Protection Laboratory March 1980
Agency Research Triangle Park NC 27711
Pilot-scale Field Tests
of High-gradient
Magnetic Filtration
Interagency
Energy/Environment
R&D Program Report
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the INTERAGENCY ENERGY-ENVIRONMENT
RESEARCH AND DEVELOPMENT series. Reports in this series result from the
effort funded under the 17-agency Federal Energy/Environment Research and
Development Program. These studies relate to EPA's mission to protect the public
health and welfare from adverse effects of pollutants associated with energy sys-
tems. The goal of the Program is to assure the rapid development of domestic
energy supplies in an environmentally-compatible manner by providing the nec-
essary environmental data and control technology. Investigations include analy-
ses of the transport of energy-related pollutants and their health and ecological
effects; assessments of, and development of, control technologies for energy
systems; and integrated assessments of a wide range of energy-related environ-
mental issues.
EPA REVIEW NOTICE
This report has been reviewed by the participating Federal Agencies, and approved
for publication. Approval does not signify that the contents necessarily reflect
the views and policies of the Government, nor does mention of trade names or
commercial products constitute endorsement or recommendation for use.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/7-80-037
March 1980
Pilot-scale Field Tests of
High-gradient Magnetic Filtration
by
Charles H. Gooding
Research Triangle Institute
P.O. Box 12194
Research Triangle Park, North Carolina 27709
Contract No. 68-02-2650
Program Element No. EHE624A
EPA Project Officer: Dennis C. Drehmel
Industrial Environmental Research Laboratory
Office of Environmental Engineering and Technology
Research Triangle Park, NC 27711
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Washington, DC 20460
-------
ABSTRACT
A 5100 m /hr mobile pilot plant was designed and built to evaluate
the effectiveness and economics of applying high-gradient magnetic
filtration (HGMF) to particulate emission control. A 4 1/2-month test
program was conducted at a Pennsylvania sintering plant to characterize
the performance of the pilot plant and to demonstrate its practicality
under long-term operation.
The pilot plant collected approximately 90 percent of the iron-
bearing particulate under practical operating conditions, but the overall
mass efficiency was lower than expected because the windbox gas contained
a high concentration of a fine alkali-chloride aerosol. To collect the
non-magnetic aerosol, a finer filter had to be used under conditions
that were conducive to plugging. Under the practical conditions the
pilot plant was operated for over 450 hours with no significant problems.
Analysis of the results indicates that high-efficiency collection can be
achieved economically if HGMF is applied to steel industry dusts that
are more homogeneous and more strongly magnetic than the sinter dust
tested.
The report describes laboratory pilot-plant work that demonstrated
collection efficiencies of greater than 99 percent with basic oxygen
furnace and electric arc furnace dusts. The development of a filter
cleaning system and the design and construction of the mobile pilot
plant are discussed. The field start-up, performance characterization,
and long-term operation are discussed in detail, and experimental data
are reported. The final section presents an analysis of the field
results and an economic evaluation of the HGMF process. The development
of a mathematical model in conjunction with the laboratory pilot plant
work is included as an Appendix.
It should be noted that the slipstream for the HGMF pilot plant was drawn
from the plant duct upstream of the plants' air pollution control devices.
No testing was conducted on the plant stack; hence the data contained in
this report should not be construed to contain any implications about the
actual plant emissions or plant compliance with relevant emission stan-
dards at the time of the test program.
-------
TABLE OF CONTENTS
Page
Abstract -Mi
List of Figures vi
List of Tables ix
Acknowledgments x
Section 1. Summary 1
Section 2. Conclusions 3
Section 3. Recommendations 4
Section 4. Background 5
Basic Concept of the HGMF Process 5
Process Development and Applications 8
Potential Applications to Particulate Emission
Control 8
Section 5. Preliminary Design and Development 12
Laboratory Pilot Testing 12
Mathematical Modeling of Preliminary Results 18
Site Selection for Field Tests 27
Development of Filter Cleaning System 30
Section 6. Detailed Design and Construction of the Mobile
Pilot Plant 34
Section 7. Field Operations 42
Description of the Sinter Plant 42
Installation and Startup of the Pilot Plant 44
Temperature Control 44
Blower Noise 45
Filter Construction and Cleaning 45
Magnet Operation 45
Flow Measurement and Control 46
Final Site Setup 47
Performance Characterization 49
Long-term Testing 62
Additional Tests with Coarse Grade Steel Wool .... 70
iv
-------
TABLE OF CONTENTS (continue)
Pac
Section 8. Discussion and Application of Results 75
Chemical and Magnetic Analysis of Sinter Dust. . . .
Transient Emissions During Filter Cleaning
Projected Applications: Economics and Effectiveness
Discussion of Potential Candidates for Applica-
tion
Full-Scale Design Considerations
Efficiency and Economic Calculations
Section 9. References
75
83
85
86
86
90
98
Appendix A Tabulation of Experimental Conditions and Results
from Laboratory Pilot Plant Tests 102
Appendix B. Mathematical Model of HGMF 114
-------
LIST OF FIGURES
Page
Figure 1. Conceptual illustration of a high gradient magnetic
collector 6
Figure 2. Schematic representation of a high gradient magnetic
filter 7
Figure 3. Schematic diagram of the HGMF laboratory pilot plant . 13
Figure 4. Magnetization curves of the two industrial dusts used
in the preliminary experiments . 14
Figure 5. Effect of filter loading on BOF dust collection
efficiency and filter pressure drop 16
Figure 6. Effect of filter loading on EAF dust collection
efficiency and filter pressure drop 17
Figure 7. Replicate data from lab pilot plant experiments
showing 95 percent confidence intervals on the true
mean efficiency and corrected model predictions. ... 19
Figure 8. Collection of BOF and EAF dusts under identical
operating conditions 21
Figure 9. Effect of applied magnetic field on collection
efficiency 22
Figure 10. Effect of filter packing density on collection
efficiency 23
Figure 11. Effect of filter depth on collection efficiency. ... 24
Figure 12. Effect of superficial gas velocity on collection
efficiency 25
Figure 13. Results of sinter dust tests conducted in the lab
pilot plant 29
Figure 14. Comparison of forces exerted on collected particles. . 31
Figure 15. Flow schematic of HGMF mobile pilot plant 35
Figure 16. View of the pilot plant from the rear interior of the
trailer • 37
-------
LIST OF FIGURES (continue)
Figure 17. View of the pilot plant from the front interior of
the trailer 38
Figure 18. Flow control diagram 40
Figure 19. Layout of the sinter strand and the HGMF pilot plant . 43
Figure 20. HGMF mobile pilot plant setup at the sintering plant . 48
Figure 21. Side entrance of mobile pilot plant 48
Figure 22. Clean side piping and stack 50
Figure 23. View from rear of pilot plant 50
Figure 24. Size distributions from test nos. 08042, 08041,
08031, and 08062 53
Figure 25. Size distributions from test nos. 08051, 08061, 08082,
and 08183. . . .' 54
Figure 26. Size distributions from test nos. 08182, 08181, 08231,
and 08222 55
Figure 27. Size distributions from test nos. 08291, 08282, 08281,
and 08292 56
Figure 28. Size distributions from test nos. 08301, 09201, 08302,
and 09013 57
Figure 29. Size distributions from test nos. 09012, 09031, 09032,
and 09033 58
Figure 30. Fractional efficiency curves from tests conducted on
filter nos. 1 and 3 59
Figure 31. Fractional efficiency curves from tests conducted on
filter nos. 2 and 4 60
Figure 32. Outlet particle size distributions obtained during
total mass sampling 69
Figure 33. Cumulative size distributions for test nos. 10231
through 10271 72
Figure 34. Fractional collection efficiency curves for test nos.
10231 through 10271 73
vn
-------
LIST OF FIGURES (continue)
Page
Figure 35. Magnetic analysis of dust samples from HGMF pilot
plant tests 78
Figure 36. Transient emission levels during a single cycle of
operation 84
Figure 37. Schematic layout of the SALA-HGMS®480 Series Carou-
sel 89
Figure 38. Predicted HGMF collection efficiency for three dust
categories at typical operating conditions 93
Figure 39. Pressure drop-flow correlation for coarse steel wool . 95
Figure Bl. Illustration of particle capture by a single wire. . . 115
Figure B2. Geometric basis of HGMF trajectory model 116
Figure B3. Contour map of collision radius as a function of
W and K (A=l, G=0) 123
Figure B4. Comparison of experimental data to the uncorrected
theoretical prediction and to two corrected models . . 125
Figure B5. Illustration of the particle bounce reentrainment
model 126
VI11
-------
LIST OF TABLES
Page
Table 1. Reports of application-oriented experimental
investigations of HGMF 9
Table 2. Expected characteristics of uncontrolled gas streams
from several processes 10
Table 3. Results of performance characterization with medium
grade steel wool 52
Table 4. Operating log of the long-term test period 63
Table 5. Efficiency testing during long-term operation 68
Table 6. Additional testing with coarse grade steel wool
(filter #8) ' 71
Table 7. Chemical analysis of dust samples from HGMF pilot
plant tests. 76
Table 8. Component mass balances based on chemical analysis. . . 80
Table 9. Illustration of filter loading times 87
Table 10. HGMF operating parameters used for economic calcula-
tions 92
Table 11. Pressure drop characterization of coarse steel wool . . 94
Table Al. Ranges of experimental parameters 102
Table A2. Experimental data on clean filter pressure drop .... 103
Table A3. Conditions and results of laboratory pilot-plant
experiments 106
Table Bl. Correlation of reentrainment correction 129
-------
ACKNOWLEDGMENTS
The author wishes to acknowledge the assistance and support of the
following individuals in various segments of this work.
Dr. Dennis C. Drehmel, EPA Project Officer, provided overall
direction and many helpful suggestions during the course of the project.
The field work described in this report was conducted at U.S. Steel
Corporation's Saxonburg Sintering Plant near Butler, Pennsylvania. The
author would like to express his sincere appreciation to U.S. Steel for
their participation in the project. In particular, Mr. John Turnage
of the Research Laboratory was most helpful in planning and coordinating
the work, and Mr. George Frye, Superintendent of the Saxonburg Plant,
provided invaluable assistance and support throughout the field program.
Dr. Herbert Hacker, Jr. of Duke University conducted the experi-
mental measurements of particle magnetization.
From RTI, Carlos Pareja assisted in the latter stages of the design
and in the construction of the mobile pilot plant and served as field
engineer during the tests at the sinter plant. Douglas VanOsdell was
particularly helpful in several portions of the engineering design and
construction of the pilot plant and in the field startup. Technicians
John Sauerbier and Daryl Smith were key participants in the construction,
field operation and testing of the pilot plant. Several speciality
items for the sampling trains were fabricated by RTI's master machinist,
Fred Schwarz. David Carter assisted with the EPA Method 5 testing, and
Peter Grohse coordinated the chemical analysis of sinter dust samples.
-------
SECTION 1
SUMMARY
High-gradient magnetic filtration (HGMF) has been gaining recog-
nition as an effective and versatile separation technique since its
commercialization in the clay industry 10 years ago. The strength of
the process lies in its ability to separate fine, weakly magnetic
particles from fluids at flow rates over 100 times the flow rates
achievable in other filtration processes. An experimental project
conducted by RTI from 1975 to 1977 under funding from the U.S. Environ-
mental Protection Agency established that HGMF could be applied to
particle-gas separations of practical interest to air pollution control.
Several sources of submicron, magnetic particles were identified among
iron and steel industry processes, and high-efficiency filtration of two
of the dusts was demonstrated in a laboratory pilot plant. The objective
of the subsequent work described in this report was to demonstrate the
practical operation of an HGMF pilot plant in a field application.
To begin the work, additional tests were run in the laboratory
pilot plant with the basic oxygen furnace (BOF) and electric arc furnace
(EAF) dusts used in the original tests. Sufficient data were obtained
to verify the achievement of high collection efficiencies and to identify
the effects of important operating variables. A mathematical model was
developed to correlate the experimental results and to aid in the design
of the field pilot plant.
A presentation was made to the American Iron and Steel Institute's
Technical Committee on Environmental Quality Control in December, 1977,
describing the prior work with HGMF and the objectives of the field
program. Following discussions with several steel companies, an agree-
ment was reached to test the pilot plant at a Pennsylvania sintering
plant. A limited series of tests was run in the laboratory pilot plant
with dust obtained from the sintering plant to obtain final data for the
design of the magnetic filter and for the design of a practical filter
cleaning system.
-------
Over the next 12 months the design details were implemented, and
the pilot plant was constructed. The cyclic system, incorporating two
parallel filters, each with a flow capacity of 5100 m3/hr (3000 cfm),
was housed in a mobile trailer. A pulsed air filter-cleaning system was
developed, and valving was installed to allow continuous filtration as
each filter was cleaned in turn. Instrumentation and test equipment
were provided to monitor the operation and efficiency of the process.
In June, 1979 the pilot plant was transported to the sinter plant
and connected to draw a slipstream from the windbox exhaust of the #1
sinter strand. After the startup and debugging procedures were completed,
a performance characterization was conducted to determine the effects of
applied magnetic field, filter density and depth, and gas throughput on
fractional collect-ion efficiency. Optimal operating conditions were
identified from the results of the performance characterization, and a
500-hour operational test was initiated. During the long-term testing,
the pilot plant was shut down only to correct malfunctions, to accom-
modate sinter plant interruptions, or to change operating conditions.
Total mass and fractional efficiency tests were conducted, and periodic
opacity observations were made. Samples of dust entering the pilot
plant, exiting the pilot plant, and collected by the pilot plant were
obtained for magnetic and chemical analyses.
The results of the field tests were used to make a technical.and
economic assessment of the application of HGMF to sinter plants as well
as to other iron and steel industry processes. The experience gained
from the field operations demonstrates' the practicality of applying HGMF
to particulate emission control while providing the basis for several
recommended design improvements.
The following sections of this report state conclusions and recom-
mendations drawn from the results of the field tests and the earlier
laboratory work. A background description of the HGMF process develop-
ment is then presented, followed by details of the laboratory pilot
plant work, the mathematical model development, the design, construction,
and field operation of the pilot plant, and an analysis of the field-
test results with respect to potential applications of HGMF.
-------
SECTION 2
CONCLUSIONS
The following conclusions were drawn from field operation of the
HGMF pilot plant and from the laboratory work that preceded the design
of the mobile unit.
(1) High gradient magnetic filtration can be applied under prac-
tical operating conditions to collect fine magnetic dust
particles with high efficiency.
(2) For HGMF to be applied successfully to an industrial source,
non-magnetic dust components that appear in significant
concentrations must be physically or chemically bound to
ferrous particles. Emissions from most basic oxygen furnaces,
electric arc furnaces, and scarfing operations are believed to
meet this criterion. More complete dust characterization data
are needed to determine whether blast furnaces and some sinter
plants might be suitable for HGMF application.
(3) A generalized economic analysis indicates that HGMF can be
economically competitive with conventional particulate control
methods in terms of primary equipment cost and energy require-
ments. The high gas velocities demonstrated in the experimental
work indicate much smaller space requirements than either
electrostatic precipitators or baghouses, which could lead to
a substantial reduction in total installed equipment costs.
Since HGMF collects the dust in a dry form, it avoids the
problems of sludge dewatering and liquid waste treatment that
a scrubber entails.
-------
SECTION 3
RECOMMENDATIONS
The development of high gradient magnetic filtration as an alter-
nate method of participate emission control should be continued. To
maximize the potential for a successful field demonstration on an
appropriate industrial source, the following points are recommended.
(1) The pilot plant should be converted from cyclic to continuous
operation so that it can be applied to continuous industrial
process that produce high dust concentrations. For lower
concentrations or intermittent industrial processes the
results will still be applicable to the cyclic approach if
that proves to be more suitable.
(2) The filter cleaning system should be modified to eliminate the
emission puffs associated with the cleaning air pulse.
(3) More complete and reliable information should be gathered on
operating and dust characteristics of the iron and steel
industry processes that are potential candidates for HGMF
application. Before the next field test, representative dust
samples should be collected at the proposed point of applica-
tion and analyzed to provide data on dust concentration, size
distribution, composition, and magnetization.
-------
SECTION 4
BACKGROUND
BASIC CONCEPT OF THE HGMF PROCESS
The basis of high-gradient magnetic filtration is the interaction
of a small paramagnetic or ferrimagnetic particle with a ferromagnetic
wire in the presence of a strong background magnetic field. The phenomenon
is illustrated in Figure 1, with the field applied perpendicular to the
axis of a cylindrical wire. The applied field magnetizes the wire and
induces a magnetic dipole in the particle. The convergence of the
applied field near the wire produces a region of highly nonuniform field
intensity that attracts the dipolar particle toward the wire, much like
a bar magnet attracts iron filings toward its poles. Depending on the
physical characteristics of the system, inertial, viscous, and gravitational
forces may also act on the particle in the vicinity of the wire.
In its simplest form, the high gradient magnetic filter consists of
a canister loosely packed with a fibrous, ferromagnetic material such as
AISI Type 430 stainless steel wool (Figure 2). The canister is placed
in a magnetic field that is customarily generated by a solenoid, and the
resulting attractive magnetic force provides high-efficiency filtration
of particles as the fluid passes through the canister. Because of the
high porosity of the filter, particle capture is actually a particle-wire
phenomenon as opposed to the cake collection mechanism that normally
dominates conventional fabric filters. Filtration may be continued
until the pressure drop through the canister becomes prohibitively high
due to the decreased size of the interstitial flow paths or until heavy
loading on the wires decreases the efficiency of particle capture. To
regenerate the filter, the magnetic field is removed and the canister is
backflushed with clean fluid. Continuous filtration may be achieved by
using a system of several parallel modules, with each module providing
filtration for a predetermined time interval. When the flow is diverted
from a module, the magnetic field of the loaded module is deenergized,
the filter is cleaned, the filter is reenergized, and the module is ready
for reuse. An alternative scheme that results in zero downtime of the
magnet is to construct the magnet and canister so that the loaded filter
5
-------
\ \ \ \ \ \ \ I I 11 1 I 11
LINES OF MAGNETIC INDUCTION
s ) PARTICLE
FERROMAGNETIC
WIRE
Figure 1. Conceptual illustration of a high gradient magnetic collector.
-------
MAGNET
COIL
PARTICLE LADEN
GAS IN
CLEAN GAS
OUT
STEEL
WOOL
FILTER
Figure 2. Schematic representation of a high gradient magnetic filter.
-------
can be continuously or intermittently removed from the magnetized region
and replaced by a clean filter without interrupting the filtration
process. Both cyclic and continuous systems are commercially available.
PROCESS DEVELOPMENT AND APPLICATIONS
With the exception of the EPA-sponsored development that began in
1975, nearly all of the experimental work in high-gradient magnetics has
involved the separation of different types of particles in a slurry
according to their magnetic susceptibility. Oberteuffer (1974), Kolm,
et al. (1975), Oder (1976), and lannicelli (1976) have published comple-
mentary reviews that provide an excellent introduction to the process
and its chronological development. The most extensive commercial
development has occurred within the last decade in the clay industry
where HGMF is used to remove fine, paramagnetic color-bodies from
kaolin. Oder (1976) estimated three years ago that installed commercial
units already had the capability to process 75 percent of the world
production of coating-quality kaolin.
The successful demonstration of HGMF in the clay industry catalyzed
experimental investigations of the new separation process for many other
applications. Well over 100 patents, reports, and technical papers have
been published in the last decade, describing projects conducted through-
out the world. Table 1 lists a representative sample of published
references in the reported areas of experimental work. In August 1978,
the Engineering Foundation sponsored an International Conference on
Industrial Applications of Magnetic Separation that was dominated by
papers and discussions of HGMF, including reports of emerging commercial
applications for boiler water polishing in West Germany and steel mill
wastewater treatment in Japan (Liu, 1979).
POTENTIAL APPLICATIONS TO PARTICULATE EMISSION CONTROL
Although most of the experimental investigations reported in the
open literature involve the separation of particles from liquid streams,
HGMF can be applied successfully in gas streams as well. Table 2 lists
waste gas characteristics of several processes that are widely used in the
iron and steel and ferroalloy industries. In the production of iron,
-------
Table 1. Reports of application-oriented experimental
investigations of HGMF.
Application
Mineral beneficiation-general
Taconite beneficiation
Coal deashing and desulfurization
Wastewater treatment-general
Steel mill wastewater treatment
Municipal wastewater treatment
Blood component separation
Catalyst recovery
References
Kelland, 1973
Murray, 1976
Kelland and Maxwell, 1975
Ergun and Bean, 1968
Trindade and Kolm, 1973
Vives, et al., 1976
Maxwell, et al., 1976
Maxwell, et al., 1977
Liu, et al., 1978
Maxwell and Kelland, 1978
Mitchell, et al., 1975
Petrakis and Ahner, 1978
Oberteuffer, et al., 1975
Harland, et al.. 1976
deLatour and Kolm, 1975
Yadidia, et al., 1977
Melville, et al., 1975
Whites ides, et al., 1976
sintering is used to combine iron ore fines with flux in the form of
limestone or dolomite and with other iron-bearing materials such as flue
dust, mill scale, turnings, and borings, to form a blast furnace feed
material of appropriate composition and size. The blast furnace then
reduces iron ore and pellets as well as the sinter. The iron is refined
to steels of various composition in basic oxygen, electric arc, or open
hearth furnaces. Scarfing is a surface improvement process in which a
thin layer of the hot steel slab or bloom is volatilized by blasting it
with oxygen. In each of these processes, the combustion air or oxygen
entrains a substantial concentration of iron-bearing dust particles that
would be emitted to the atmosphere if the emissions were not properly
controlled. Cyclones, baghouses, wet scrubbers and electrostatic
precipitators are currently employed to control the emissions with varying
degrees of success.
-------
Table 2. Expected characteristics of uncontrolled gas streams from several processes.*
o
Process
Sinter Machine
Windbox
Discharge End
Blast Furnace
Basic Oxygen Furnace
Open System
Closed System
Electric Arc Furnace
Open Hearth Furnace
Scarfing Machine
Dust
Concentration
g/m3
1-2
5-12
10-25
10-25
40-70
0.2-7
4-7
0.5-1
Mass Median
Diameter
ym
10
10
20
1
2
1
5
0.5
Iron
Composition
% Total Fe
25-50
25-50
35-50
55-70
55-70
i fi-4fl
55-70
50-70
Noteworthy Gas
Characteristics
5-15% H20, Hydrocarbons,
Fluorides, S0x, 120-180°C
120-180°C
20-40 % CO, 2-6% H2, 200-300°C
250-300°C
75% CO, 250-300°C
40-120°C
7-15% H20, 250-350°C
H20 Saturated, 50-60°C
Compiled from numerous references including Hardison and Greathouse (1972), Dulaney (1974), Steiner (1976),
Jaasund (1977), and Whitehead (1977).
-------
With current magnet technology the capital costs and power require-
ments of large solenoids make HGMF potentially competitive with other
particulate control methods. Since the filtration process is enhanced
by magnetic forces, the void volume of the collection matrix can be
much larger than in a conventional filter, allowing very high gas
velocities at relatively low pressure drops. This combination translates
into a potential reduction in energy requirements compared to conven-
tional particulate control techniques, even though production of the
magnetic field requires some energy. High operating velocities help to
reduce both the capital costs and space requirements of the equipment.
Furthermore, if air is used as the backflush fluid, the process can be
applied completely dry to avoid the water pollution problems associated
with some scrubber installations. Magnetic stainless steels of the
400 series are compatible with both high temperature and corrosive enviro-
ments and the absence of any sparking mechanism in the collection process
should allow its application in combustible gas streams. In brief, all of
the processes listed in Table 2 should be considered as potential
candidates for HGMF fine particle control subject to more complete
evaluation based on experimental testing.
11
-------
SECTION 5
PRELIMINARY DESIGN AND DEVELOPMENT
LABORATORY PILOT TESTING
In work conducted by Research Triangle Institute for EPA from 1975
to 1977, a laboratory pilot plant was constructed and operated to obtain
data sufficient to demonstrate that HGMF could be used to remove fine
magnetic dusts from an air stream with high collection efficiency. The
design and operating procedures of the lab pilot plant are described in
detail in a prior report (Gooding, et al., 1977). The same pilot plant
was used in this project to obtain verification data needed to design
the mobile pilot plant. This preliminary work is described in the
following paragraphs.
A schematic of the lab pilot plant is shown in Figure 3. The 30-cm
3
diameter filter was operated with flow rates up to 3100 m /hr from a
slipstream off an existing wind tunnel. Two types of industrial dust
were obtained from a major steel corporation for use in the experiments.
Most of the tests were run with dust collected from the hoppers of an
electrostatic precipitator that controls emissions from a basic oxygen
furnace (EOF dust). To provide further evaluation of the effects of
dust characteristics, a limited number of tests was run with dust
obtained from the hoppers of a roof-system baghouse that controls
emissions from an electric arc furnace shop (EAF dust). Experimentally
determined magnetization curves for the two dusts are shown in Figure
4. Atomic absorption measurements indicated the iron contents to be
76 percent and 41 percent in the BOF and EAF dust, respectively. Both
of these are slightly above the general ranges reported in Table 2.
The dusts were dispersed in a room-temperature air stream by means
of a fluidized-bed dust generator. Since the concentration of dust in
3
the slipstream was normally only 50 to 100 mg/Nm , the filter could be
operated without cleaning for the two hours required for sampling.
Between tests the filter was either replaced or cleaned in situ by
backflushing with pressurized air routed to a set of perforated rings
mounted in the filter canister. A vibrator attached to the filter
assembly was also operated concurrently with the backflushing opera-
tion to assist the cleaning.
-------
TO
BAGHOUSE
WIND TUNNEL
SAMPLE
HGMS
SAMPLE
CYCLONE
r
AIR
FROM
BAGHOUSE
DUST
WASTE
Figure 3. Schematic diagram of the HGMF laboratory pilot plant.
-------
50
I I I I 1 1 1 1 1 ! 1 1 [
1.0
2.0 3.0
APPLIED FIELD, kOe
4.0
Figure 4. Magnetization curves of the two industrial dusts used in the
preliminary experiments.
-------
Numerous tests were run in the laboratory pilot plant to demonstrate
the achievement of high collection efficiencies and to study the effects
of operating parameters. Variations were made in the depth of the steel
wool filter, the density of the filter, the strength of the applied
magnetic field, and the gas velocity through the filter. In each experi-
ment the particle size distribution and concentration were determined
upstream and downstream of the filter from samples obtained with MRI
Model 1502 impactors manufactured by Meteorology Research, Inc., Altadena,
CA. In accordance with EPA's recommended procedures (Harris, 1977) the
impactor substrates were coated with Apiezon L grease (James G. Biddle
Co., Plymouth Meeting, PA) prior to use. Clean filter pressure drop
tests were also obtained with several different combinations of filter
depth and density.
Fractional collection efficiencies exceeding 99 percent were
achieved with filtration velocities as high as 10 m/s, confirming the
potential practicality of applying HGMF to particulate emission control
in the iron and steel industry. The complete results of the laboratory
pilot plant tests are presented in Appendix A.
Figures 5 and 6 show the effects of filter loading on collection
efficiency and pressure drop for each of the dusts, starting with a
clean, 420-g filter at time zero. The efficiency data were obtained
with a Climet Model 208A optical particle analyzer (Climet Instruments
Co., Redlands, CA). The BOF data show no deterioration in efficiency
during the test period despite a dust accumulation of 800 g. Apparently
the highly magnetic dust particles accumulating on the wires distorted
the magnetic field sufficiently to act as new collection sites, rather
than simply filling the gradient region as an inert material. This
result implies that with a strongly magnetic dust increasing pressure
drop rather than deteriorating efficiency will most likely determine the
allowable filtration time between cleanings. In contrast, the EAF dust,
which had a lower specific magnetization than the BOF dust, showed a
deterioration in efficiency as the filter loaded. The total accumulation
was 1270 g at the end of the test. Under the latter circumstances, the
cleaning cycle would be determined by the minimum acceptable collection
efficiency.
15
-------
100
gg
98
g?
o 96
z
UJ
ti-
o ga
ai
_J
O 92
O
go
0.5-0.7 urn
- 0.7-1.0
0
100
BOF DUST
F = 0.0050
L= 0.15m
V = 8.2 m/s
Ba = 0.4 T
0.3-0.5
FILTER AP
J_
200 300
TIME, MINUTES
400
0.7-1.
0.5-0.7
500
2.5
(O
a.
2.0 n.
cc
15
'•°
1.0
0.5
cc
UJ
cc
600
Figure 5. Effect of filter loading on BOF dust collection efficiency and filter pressure drop.
-------
100
90
80
o
1
ui 60
2
O
f= 50
o
01
40
O
o
30
20
10
0
0
EAF DUST
F = 0.0050
L = 0.15m
V = 8.2 m/s
Ba = 0.4 T
FILTER AP
100
200 300
TIME, MINUTES
400
500
2.0 a:
O
cc
1.5 Q
uu
CC
LU
0.5
600
Figure 6. Effect of filter loading on EAF dust collection efficiency and filter pressure drop.
-------
MATHEMATICAL MODELING OF PRELIMINARY RESULTS
A mathematical model of the HGMF process was developed in conjunction
with the laboratory pilot plant work to give a better understanding of
the effects of various design parameters. Like the common theoretical
approach to conventional filtration, the model is based on the calculation
of particle trajectories near the ferromagnetic wires, taking into account
inertial, viscous, and magnetic forces. The development of the model is
presented in detail in Appendix B. The final predictive equation for
fractional collection efficiency is
E = 1 - exp
4FLY D
cr
A(l-F)2
(D
where E = fractional collection efficiency, dimensionless;
F = filter packing density (actual volume of steel wires/volume
occupied by filter), dimensionless;
L = filter depth, m;
s = wire radius, m;
Y = single wire collision radius, dimensionless; and
p = probability of particle adhesion, dimensionless.
The trajectory calculations lead to a theoretical correlation of YC
in terms of two dimensionless groups, the familiar Stokes number and an
unnamed magnetic parameter (see Appendix B). Each of these may be
calculated from operating parameters of the system and physical properties
of the dust and the gas stream. Comparison of the theoretical predictions
of collection efficiency to experimental data revealed that the model
agreed well with small particle data but underestimated the penetration of
large particles through the filter. Two theoretical reentrainment models
were investigated but neither proved to be satisfactory. To reconcile the
discrepancy between the uncorrected theory and the experimental data, an
empirical correlation was developed for the probability of particle
adhesion in terms of the same two dimensionless groups used to predict Y .
\*
The agreement between the corrected model and the data is illustrated in
Figure 7, which also demonstrates the high collection efficiency attain-
able with BOF dust.
18
-------
o
UJ
O
LJL
LL
LLI
O
UJ
_J
O
o
99.9
99.8
99.5
99
98
95
90
80
70
60
50
40
30
20
10
CORRECTED
THEORY
0.2
Figure 7.
BOF DUST
O
A
V
I
F
L
V
B
a
Run No. 02091
Run No. 02101
Run No. 02102
95% Confidence Interval
0.0080
0.225 m
10.0 m/s
0.30 T
j I
0.5 1.0 2
PARTICLE DIAMETER,
10
Replicate data from lab pilot plant experiments showing
95 percent confidence intervals on the true mean
efficiency and corrected model predictions.
19
-------
The next five figures illustrate the effects of different operating
parameters on collection efficiency. The curves in each case were
generated by the corrected theoretical model [Equation (1)] with the
empirically determined adhesion probability; hence they reflect the
experimental behavior of the process in the lab pilot plant. Figure 8
compares collection of the two dusts under identical operating conditions.
The BOF dust, which had roughly twice the specific magnetization of the
EAF dust at any field, was collected with significantly higher efficiency.
The EAF dust could be collected with a peak efficiency above 99 percent,
but a higher field and deeper filter were required to accomplish this
(see Run No. 11151 in Appendix A).
Figure 9 shows the effect of applied field on the collection of BOF
dust. The improvement in efficiency provided by the increased magnetic
force is clear, but the return is diminished as the dust particles
approach magnetic saturation.
Figures 10 and 11 show the effects on collection efficiency of
filter packing density and filter depth, respectively. The effects of
these two parameters are identical. One can deduce from Equation (1)
that doubling either the packing density or the filter depth squares the
fractional particle penetration. For example, in Figure 10 the collection
2
efficiency at 6 ym goes from 95 percent to 99.75 percent [0.0025 = (0.05) ]
when the packing density is doubled from 0.005 to 0.01.
Figure 12 shows the typical effect of gas velocity on collection
efficiency. Theoretically, the role of velocity is a complicated one.
As a component of particle inertia, increasing velocity can be bene-
ficial or detrimental to particle collection depending on the value of
other physical properties and conditions. But increasing velocity
always reduces the probability of particle adherence, so the net result
is usually the relatively weak detrimental effect shown in Figure 12.
Appendix B includes a more detailed discussion of the theoretical effect
of velocity.
20
-------
o
UJ
O
u.
u.
UJ
O
O
UJ
—i
O
O
99.9
99.8
99.5
99
98
95
90
80
70
60
50
40
30
20
10
0.2
I I I I II
1 I I TT
BOF DUST
a
0.0050
0.15 m
5.0 m/s
0.20 T
i i iii j
i i i
0.5 1.0 2
PARTICLE DIAMETER,
i i i
10
Figure 8. Collection of BOF and EAF dusts under identical operating
conditions.
21
-------
I I I I I I I 1
FIELD OFF
BOF DUST
F = 0.0050
L = 0.15 m
V = 5.0 m/s
i i t i i i i i
0.5 1.0 2 5
PARTICLE DIAMETER, /mi
Figure 9. Effect of applied magnetic field on collection efficiency.
•22
-------
F = 0.0100
^
F = 0.0075
1
0.2
L = 0.15 m
V = 5.0 m/s
Ba = 0.20 T
i i i i i i
0.5 1.0 2
PARTICLE DIAMETER,
10
Figure 10. Effect of filter packing density on collection efficiency
23
-------
F = 0.0050
V = 5.0 m/s
0.5 1.0 2 5
PARTICLE DIAMETER,
Figure 11. Effect of filter depth on collection efficiency.
24
-------
o
LLJ
O
LL
LL
LU
99.9
99.8
99.5
99
98
95
90
80
70
60
O
O
LU 30
50
40
O
O
20
10
i i I i r
0.2
i i TT
= 5 m/s
BOF DUST
F = 0.0050
L = 0.15 m
Ba = 0.20 T
0.5 1.0 2
PARTICLE DIAMETER,
10
Figure 12. Effect of superficial gas velocity on collection efficiency
25
-------
It can be noted that all of the curves show a peak in collection
efficiency in the 1 to 2 um particle size range. This prediction, which
accurately reflects the observed experimental system behavior, is
clearly a consequence of the reentrainment phenomenon- The uncorrected
model predicts a monotonic increase in collection efficiency with
particle size up to at least 5 ym, above which the efficiency may level
off but does not decrease. The empirical correlation does not reveal
whether reentrainment is caused by particle bounce, drag forces on
collected particles, or a combination of these and perhaps other
phenomena, but the results clearly indicate that reentrainment is an
important factor in determining collection efficiency under the experimental
conditions of interest to air pollution control applications of HGMF.
One other factor of particular interest to air pollution control
applications is the effect of operating temperature. Theoretically,
increasing temperature exerts a detrimental effect in two ways—by
increasing the gas viscosity and by decreasing the magnetization of the
wire and the particles. Thus the drag force tending to keep the particles
entrained in the gas increases, and the magnetic force decreases, but
neither of these effects is prohibitive to operation until the Curie
point of the magnetic materials is reached. The Curie point of the
most commonly used filter material, AISI Type 430 stainless steel, is
approximately 680°C (Bozorth, 1951). Above that temperature the
material loses its ferromagnetism. Below the Curie point the saturation
magnetization obeys the relationship
M /M
MS/MQ = tanh 2. (2)
where M = saturation magnetization at T, A/m;
M = saturation magnetization at absolute zero, A/m;
T = absolute temperature, °K; and
T = Curie point, °K.
C
26
-------
Equation (2) indicates that the saturation magnetization of 430 stainless
steel at 310°C is 90 percent of its value at 25°C. At 500°C it is still
70 percent of the 25°C value. Ferromagnetic dust particles could be
expected to exhibit roughly the same behavior. Viscosity of air is
approximately proportional to absolute temperature raised to the 0.7
power. Thus, an increase in temperature from 25°C to 550°C causes a
doubling of viscosity. This theoretical analysis reveals that HGMF
should certainly be applicable at temperatures of 500 to 600°C although
the filtration of particles will be somewhat more difficult than at
lower temperatures.
The corrected theoretical model of HGMF developed in conjunction
with the lab pilot plant work provided a valuable tool for the design
of the mobile pilot plant and the field experiment. The model can also
be used in conjunction with economic estimates when one must compare
alternative operating conditions. In its present form the model has four
significant limitations:
(1) It is based on a clean wire assumption, so it is not valid under
conditions in which the filter loading significantly affects
collection efficiency;
(2) The effect of wire size has not been confirmed experimentally,
and the method of estimating wire size for model calculations
is subject to error because it ignores possible influences
of shape and surface irregularities and wire-size distributions
(see Appendix A);
(3) The effect of gas temperature has not been evaluated experi-
mentally; and
(4) The reentrainment correction contains empirical constants that
most probably vary with dust properties.
SITE SELECTION FOR FIELD TESTS
Having demonstrated the attainment of high collection efficiency
with two steel industry dusts, RTI approached the American Iron and
Steel Institute to solicit their assistance in finding a suitable site
for field tests. In December, 1977, a presentation was made before the
AISI Technical Committee on Environmental Quality Control to describe the
lab pilot plant work and to introduce the objectives of the field program.
27
-------
Subsequent discussions were held with several individual companies until
one company expressed a definite interest in testing the HGMF process at
a sintering plant. Sinter dust was not the obvious choice for a first
evaluation of HGMF because it was anticipated to have a relatively low
magnetization, but the sintering process is one of the more difficult
environmental control problems presently facing the steel industry;
hence the opportunity is greatest for the application of a new concept.
In April, 1978, a drum of dust was received from the sintering
plant that had been tentatively chosen for the field tests. The dust
was obtained from the hoppers of the electrostatic precipitators that
presently control windbox emissions. The magnetization curve of the
bulk dust was similar in curvature to those of the BOF and EAF dust
(Figure 4), but the saturation value was only 9 emu/g. The dust was
quite coarse in appearance, containing some particles as large as 3 mm.
The average particle density was determined by pycnometer measurements
to be 3.9 g/cm3.
A series of tests was initiated with the sinter dust in the labora-
tory pilot plant, but the number of runs was limited by severe problems
feeding the coarse dust into the dust generator. Because of difficulty
entraining the large dust, the particle stream entering the HGMF had a
mass median diameter of approximately 6 ym (aerodynamic) and a dust
3
concentration of only 10 to 40 mg/Nm . The pilot plant gas conditions
were adjusted to approximately 130°C and 5 percent moisture to simulate
actual windbox conditions. The experimental results of the sinter runs
are shown in Figure 13. All of the runs were made with an applied flux
density of 0.5 T. In comparison to the theoretical model, the data
actually showed somewhat better small-particle collection than was
expected and not quite as good as expected with the larger particles.
The test series was too limited to draw any conclusions about optimal
operating conditions, but the results were sufficiently promising to
move ahead with plans for the field work.
28
-------
o
2
UJ
O
u.
Li.
UJ
U
HI
O
U
99.9
99.8
99.5
99
98
95
90
80
70
60
50
40
30
20
10
2
1
0.2
TEST NO.
05041
05151
05121
06061
FILTER FILTER VELOCITY,
DEPTH, m DENSITY m/s
0.300
0.300
0.300
0.225
0.006
0.006
0.006
0.010
9.9
10.1
4.9
9.0
i i 1 I
PRESSURE DROP
cm
22
23
7.6
28
0.5 1.0 2
AERODYNAMIC DIAMETER, ,um
10
Figure 13. Results of sinter dust tests conducted in the lab pilot plant.
-------
DEVELOPMENT OF FILTER CLEANING SYSTEM
The development of an effective filter cleaning system for the
mobile pilot plant was of critical importance to the field operations.
Figure 14 shows the result of simplified theoretical calculations that
were conducted to evaluate the magnitude of competitive forces that
would be significant in a cleaning system. The magnetic force acting
on a captured particle was calculated from a simplified form of Equation
(B-9); i.e.,
X*H 2b3
Fm '
where F = magnetic force, N;
-7
y = magnetic permeability of a vacuum, 477x10 h/m;
x* = effective magnetic susceptibility, dimensionless;
H = applied magnetic field, ampere turns per meter;
a
b = particle radius, m; and
s = wire radius, m.
Equation (3) represents the maximum radial magnetic force acting on a
particle at the surface of a magnetized wire. In terms of the specific
magnetization
(4)
where a = specific magnetization, emu/g; and
3
p = particle density, kg/m .
The magnetic force lines in Figure 14 were calculated for a particle density
of 4000 kg/m , a wire radius of 50 urn, and an applied field of 3.98xl05 A/m
(0.5 T). With the field off for filter cleaning, the residual force
would be much smaller than that shown.
The drag force, F^, was calculated from the expression
(5)
where C, = drag coefficient, dimensionless;
pf = fluid density, kg/m ; and
V = fluid velocity, m/s.
30
-------
LLJ
O
oc
O
10
,-6
10
,-7
10
-8
10
-9
10
-10
0.1
0.2
0.5 1 2
PARTICLE DIAMETER, M
10
Figure 14. Comparison of forces exerted on collected particles,
31
-------
The fluid was assumed to be air at 25°C and standard atmospheric pres-
sure. The flow conditions indicated are in the transitional region
above the Stokes law regime- but well, below the region of constant drag
coefficient.
The van der Waals force, F , was calculated from
a
- Ub
a
P -
F -- ~
where U = Hamaker constant, J; and
£ = equilibrium separation distance, m.
The separation distance was estimated to be 4xlO~ m after Krupp (1967),
1 n
and a Hamaker constant of 2x10 J was used, which is appropriate for
ferrous materials in air (Visser, 1972). Because of the uncertainty in
estimating these two parameters, the van der Waals force must be regarded
as a rough approximation, but the comparison shown in Figure 14 makes
two important points:
(1) The van der Waals force is in all likelihood more important
than in any residual magnetic forces holding the particles on
the wires (the van der Waals force is not important to initial
particle capture because it is an extremely short-range force).
(2) Flushing velocities of at least 50 to 100 m/s are most probably
required to remove the particles from the wires.
The first quantitative cleaning experiments were conducted by
removing a used filter from the lab pilot plant and cleaning sections of
it with a small air jet. The dust accumulation was virtually unaffected
by air velocities lower than 50 m/s. When higher jet velocities were
applied, the dust was denuded from the wires in chunks on the order of
500 ym. The denuding action and the high air velocity requirement
suggested that a pulse cleaning approach would be much more practical in
application than a flush of extended duration.
A series of bench-scale filter cleaning tests was then run in the
following manner. Forty samples of steel wool were placed in 10-mm
diameter glass tubes, and the tubes were imbedded in a larger filter of
the same length and packing density as the small samples. A collection
run was then made in the laboratory pilot plant using the sinter dust.
32
-------
The tubes were carefully removed and cleaned individually in a labora-
tory apparatus. This cleaning was accomplished by pulsing each small
filter with a burst of air released from a pressurized chamber. Vari-
ations were made in the volume and pressure of the air chamber, the
surge volume between the air chamber and the filter, and the orifice
diameter of the quick-release ball valve. Both a baffled settling
chamber and a cyclone were tested to re-collect the dust blown off the
filters. The following conclusions were drawn from the tests:
(1) If a scaled-up system could be developed to release the pres-
surized air with dynamic pulse characteristics similar to the
lab system and the pressure drop characteristics of the flow
paths were the same, then the ratio of chamber volume to
filter face area should be a measure of the maximum pulse
velocity and thus of the cleaning potential. The lab experi-
ments indicated that a ratio of 2 m was needed for efficient
cleaning.
(2) The surge volume between the air chamber and the filter should
be kept to a minimum to avoid dissipation of the air pressure.
In the lab tests a surge volume of 1/3 the air-chamber volume
was acceptable.
(3) A chamber pressure of at least 140 kPa gauge (20 psig) was
needed for acceptable cleaning. Higher pressures gave even
better results.
(4) Dynamic flow calculations led to an estimate that the test
cleaning system developed a pulse velocity through the filter
of 100 m/s within 0.05 seconds of the initial valve opening.
To achieve this in the mobile unit would require that the
quick-opening valve develop a flow coefficient of 750 in 0.05
seconds. (Valve flow coefficient is a commerical measure of
the rated valve capacity. In English units of measure it has
a direct physical significance in that the flow coefficient
represents the number of gallons per minute of water that the
valve will pass at 60°F with a pressure drop of 1 psig. For
example, most 4-inch nominal butterfly valves are rated with a
flow coefficient in the vicinity of 750 when in their fully-
open positions, but standard pneumatic and electric actuators
are not capable of opening a 4" valve in 0.05 seconds. Hence,
meeting the valve requirement represented a significant design
challenge.)
(5) The cyclone consistently caught over 99 percent of the material
blown off the filter while the baffled settling chamber aver-
aged 87 percent. Since a scaled-up cyclone would also require
a minimal amount of space in the mobile unit, the cyclone
approach was adopted.
33
-------
SECTION 6
DETAILED DESIGN AND CONSTRUCTION OF THE MOBILE PILOT PLANT
The mobile pilot plant is housed in a 12.8 m (42 ft) freight van.
O
It was designed for a nominal flow capacity of 5100 m /hr (3000 CFM)
based on the previous experimental work. Figure 15 is a flow schematic
of the portion of the system that is inside the trailer.
The dirty gas enters the trailer via a 0.317 m ID, 316 stainless
steel pipe (12", Schedule 5) and passes by test ports through which
samples can be drawn to determine the size distribution and concentration
of the inlet dust. The gas is then directed to one of two functionally
identical HGMF devices. Two filtration paths are provided so that one can
be cleaned while the other is in operation. Magnet A was constructed by
Magnetic Corporation of America (MCA), Waltham, MA, and Magnet B was
originally constructed by Sala Magnetics, Inc., Cambridge, MA for use in
the laboratory pilot plant. Magnet B was later modified by the addition
of new pole pieces and stand purchased from MCA. Each of the magnetic
filters consists of the iron-bound solenoid surrounding a canister that
measures 0.432 m ID by 0.305 m long. The canisters are filled with an
appropriate amount of magnetic stainless steel wool as dictated by the
test program. Each of the magnets can be energized to provide an
applied field of up to 0.5 tesla throughout the canister volume. The
magnets are energized by DC modular power supplies manufactured by
Controlled Power Company (Troy, MI) and purchased from MCA.
After passing through the filter, the cleaned gas travels past
another set of test ports and exits the trailer. The exterior pipe size
is reduced to 0.266 m ID (10", Schedule 5). The gas passes through an
orifice and an induced draft blower and is then exhausted to the atmosphere
through an 11 m high stack.
The filters are cleaned by backflushing with compressed air provided
by a Worthington Model 7 1/2 EDBR two-stage compressor (Worthington
Compressors, Inc., Holyhoke, MA) that is mounted to the underside of the
trailer. The compressed air tank associated with each filter has a volume
3
of approximately 0.28 m . To pulse the filter, the compressed air is
34
-------
CO
en
V8A
1
CLEAN
— * GAS " i
CYCLONE
Figure 15. Flow schematic of HGMF mobile pilot plant.
-------
released through an 8" (nominal pipe size) Galigher Delta valve (Galigher
Company, Salt Lake City, UT). The Galigher valves are pneumatically
actuated pinch valves with an equivalent throat diameter of approximately
20 cm and a fully-open Cy rating of 1300. Each valve consists of a pair
of identical elastomeric diaphragms contained within a cast aluminum,
split housing. The diaphragms are closed by introducing compressed
actuator air into the chamber between the housing and the diaphragms.
The valve can then be opened very rapidly by exhausting the actuator air
through two large ports provided in the housing. To obtain a tight seal
on the air chambers the actuator air is controlled to a pressure approxi-
mately 140 kPa (20 psi) higher than the cleaning air and is then released
through 1 1/2" (nominal) Model 168S poppet valves manufactured by Kay
Pneumatics (Commack, NY). Dynamic flow calculations conducted during
the design phase indicated that the Galigher valves should be able to
achieve the required C of 750 within 0.05 seconds of the initial diaphragm
separation and thus release-the cleaning air in a pulse sufficient to
clean the filters.
The agglomerated dust that is flushed off the filer is removed from
the cleaning air by a Kirk and Blum Size 6, Type C7 cyclone (Kirk and
Blum Manufacturing Company, Cincinnati, OH). The exhaust from the top
of the cyclone is routed back to the dirty gas stream and into the
operating filter. Dust can be removed from the cyclone during operation
through the double seal formed by two 6" (nominal) Norris butterfly
valves (Dover Corporation/Morris Division, Tulsa, OK) that are mounted
in line at the bottom of the cyclone hopper. Pneumatic vibrators are
mounted on the walls of the dust hopper to aid in dust discharge.
Figures 16 and 17 are photographs taken of the interior of the trailer
during the construction phase. Figure 16, taken from the rear door, shows
the Galigher valves on either side with the air tanks mounted above. In
the foreground is the tee that receives the outlet flow from the magnetic
filters and channels it back toward the front of the trailer. The black
magnets are visible behind the pipe. Figure 17 was taken from the front
of the process area before the cyclone and the inlet and outlet test
sections were installed. The open flange in the upper portion of the
photograph connects directly to a horizontal pipe carrying the inlet gas.
36
-------
Figure 16. View of the pilot plant from the rear interior of the trailer.
-------
CJ
Figure 17. View of the pilot plant from the front interior of the trailer.
-------
The inlet of the cyclone bolts to the flange below via the transition
piece standing on the floor under the flange. The top outlet of the
cyclone exhausts to the inlet gas pipe above. The open pipe in the
lower center carries the clean outlet gas and connects to an S-shaped
pipe that rises to join the outlet test section. The magnet power
supplies are visible at the right.
The blower that moves the gas through the pilot plant is a Centri-
fan Model RB50-2 (Centrifan Company, Greenville, SC). The blower will
exhaust the required 5100 m /hr at a suction.pressure of -18.7 kPa (-75
inches H20) and a temperature of 150°C. With the exception of the
Galigher valves, pneumatically or manually actuated Norris butterfly
valves are used throughout the system. The entire system is designed to
allow continuous operation at up to 200°C. The interior pipe is insulated
with calcium silicate and the exterior pipe with jacketed fiber glass.
The front quarter of the trailer contains an enclosed, air-conditioned
laboratory and control room. Two automated devices are incorporated to
simplify operation of the pilot plant. A Xanadu Model UPT100-10-10
solid state programmable timer. (Xanadu Controls, Springfield, NJ)
sequences the operation of the butterfly valves, the pinch valves, and
the magnets as the system cycles from one flow path to the other. The
total cycle duration and the sequence of events may be changed easily
in a few seconds by inserting a pencil-coded computer card and adjusting
a thumbwheel switch. A Robertshaw DCM-1000 controller (Robertshaw
Controls Company, Anaheim, CA) maintains constant gas flow through the
pilot plant. The controller receives its signal from an orifice located
in the clean gas pipe via a Robertshaw Model 117 differential pressure
transmitter and adjusts a butterfly valve located on the blower exhaust.
The control diagram is shown in Figure 18. The orifice pressure drop
and the pressure drop across the magnetic filters are displayed and
stored on two Robertshaw Model 225 strip chart recorders located in the
laboratory/control room. An Omega Model 199KC digital temperature
indicator (Omega Engineering, Inc., Stamford, CT) is also mounted in the
control panel to display the signal from chrome!/alumel thermocouples
that are located in the inlet and outlet piping of the pilot plant.
39
-------
SET POINT
CONTROLLER
DIFFERENTIAL
PRESSURE
TRANSMITTER
ORIFICE
PLATE
CLEAN
GAS
CENTRIFAN
BLOWER
CONTROL
VALVE
Figure 18. Flow control diagram.
-------
The laboratory/control room also contains bench space, a wet sink,
a lab oven, a solvent sink and a lab hood. MRI Model 1502 cascade
impactors (Meteorology Research, Inc., Altadena, CA) are used to deter-
mine the particle size distribution and concentration. A Perkin-Elmer
Model AD-2Z microbalance (Perkin-Elmer Corp., Norwalk, CT) is used to
weigh the impactor substrates. A Climet Model 208A particle size
analyzer with a Model 210 multichannel monitor (Climet Instruments Co.,
Redlands, CA) is also available to monitor transient conditions in the
particle size distribution and concentration on the clean side of the
magnetic filters.
The utility requirements of the pilot plant consist of electricity
and water. The main power panel is breakered for 400 amperes of 480
volt AC input. The total connected load is 280 amperes, and the typical
current draw is about 150 amperes. The major equipment operates off 480
VAC, and a transformer is provided to step down to 240 VAC and 120 VAC
for lighting and smaller loads. Water consumption is approximately 5
m /hr (22 GPM) for magnet cooling plus minor usage for the compressor
aftercooler and the lab sink. As illustrated later in Section 8, these
utility requirements are much larger in proportion to gas flow capacity
than the requirements of a full-scale HGMF system.
41
-------
SECTION 7
FIELD OPERATIONS
DESCRIPTION OF THE SINTER PLANT
The sintering process is basically a scrap recovery operation
developed to make the integrated steel mill more efficient. Blast
furnaces require a feed material that is relatively uniform in size. . In
the sintering process fine, iron-bearing materials from a variety of
sources are mixed and fused together to make a suitable component of the
blast furnace feed.
Figure 19 illustrates the operation of the sintering process on the
strand where the HGMF field tests were conducted. Ore fines, blast
furnace flue dust, BOF slag, mill scale, limestone,- dolomite, and coke
breeze are fed to a traveling grate. The upper surface of the bed is
then fired with natural gas burners, and air is drawn through the bed
into a series of distribution chambers called windboxes. As the bed
travels down the strand, the combustion zone moves downward through the
bed, igniting, drying, heating, and fusing the mixture into a sinter.
At the end of the strand, the sinter is crushed, screened, and cooled
for transport to the blast furnace. Screened fines are returned to the
feed.
Part of the gas drawn through the bed into the windboxes is recycled
to the strand for combustion air. The remainder passes first through an
inertia! separator called a Lurgi Policeman where larger particles are
removed. The waste gas then passes into an electrostatic precipitator
for final cleanup before being exhausted through the plant stack. It
should be noted that the slipstream for the HGMF pilot plant was drawn
from the plant duct upstream of the plants' air pollution control
devices. No testing was conducted on the plant stack; hence the data
contained in this report should not be construed to contain any impli-
cations about the actual plant emissions or plant compliance with
relevant emission standards at the time of the test program.
42
-------
IGNITION
\J
FEED
SINTER
1C
WIND BOXES
RECYCLE LINE
LURGI
POLICEMAN
HGMF
w
E.S.P.
PLANT
STACK
HGMF
STACK
Figure 19. Layout of the sinter strand and the HGMF pilot plant.
-------
INSTALLATION AND STARTUP OF THE PILOT PLANT
The mobile pilot plant arrived at the sinter plant on June 18,
1979, and installation of the exterior piping began. Approximately 40 m
of 0.266 m ID (10" Schedule 5) pipe was run from the entrance of the
pilot plant to the windbox exhaust, terminating with a 0.266 m nozzle
positioned in the center of the 3.6 m diameter plant tunnel. The nozzle
faced into the plant gas flow and the ratio of sizes was correct to give
isokinetic flow into the nozzle at average plant and pilot plant opera-
ting conditions. (As it turned out, the effort to sample the plant gas
isokinetically was futile because much of the large-particle concentra-
tion was lost in the long run of inlet pipe.) At the point from which
the windbox exhaust was sampled, the typical temperature and pressure
were 120°C and -9 kPa (-35 inches H20) gauge.
The pilot plant blower and stack were installed as planned, and all
of the exterior piping was insulated. Additional sampling ports were
installed in the piping leading to the pilot plant and in the pilot
plant stack to allow EPA Method 5 sampling. The interior sampling ports
were suitable for impactor work, but space was insufficient for the more
bulky total-mass sampling apparatus.
The power and water connections were made, and initial checkout and
debugging of the system began. Since the mobile system operates off a
480 VAC power source, which was not available at the fabrication site in
North Carolina, most of the equipment had not been operated prior to the
field startup. Operating procedures had to be established, and several
problems had to be corrected before the performance characterization
could begin. These activities are discussed.in the next few paragraphs
and are followed by photographs of the final site setup.
Temperature Control
A Universal Model 3500 FA forced-air construction heater (National-
Riverside Co., Rancho Cucamonga, CA) was provided with the pilot plant
to assist in cold startups. The heater burns propane gas to heat 2400
m /hr of air to a maximum temperature of 115°C. A tee and appropriate
valving were included in the inlet line to the pilot plant so that the
forced-air heater could be used to preheat the pipe coming from the
plant tunnel with the assistance of the vacuum in the plant tunnel and
then to preheat the interior pilot plant piping with the assistance of
the pilot plant blower. M
-------
The entire preheat procedure required about 45 minutes to raise all
of the piping to approximately 100°C. The flow could then be started
from the sinter plant to the pilot plant without a serious drop in the
gas temperature. After all of the insulation was installed, the steady
state temperature drop from the plant tunnel to the inlet of the pilot
plant averaged less than 10°C. From the inlet to the outlet of the
pilot plant, the gas temperature normally dropped another 10°C. At the
blower the temperature of the gas rose about 30 to 40°C.
Blower Noise
The noise level produced by the 3600 rpm pilot plant blower was
initially unacceptable. A ventilated, double-insulated house was erected
around the blower, which reduced the noise level, but the high-pitched
sound emanating from the top of the stack was still far above background
noise levels. A stack muffler was designed by the sinter plant super-
intendent, and fabricated by the plant shop. The muffler consisted of a
1-m long section of the 0.266 m diameter pipe drilled with approximately
1600 equally-spaced 1-cm holes. This pipe was then placed inside a
larger-diameter cylinder, and the annular space was filled with fiber
glass insulation. The ends of the internal pipe were flanged, and it
was installed between the top two sections of the stack. The ends of
the annular space were capped to prevent rain damage to the insulation.
The muffler dramatically reduced the blower noise to an acceptable
working level.
Filter Construction and Cleaning
Preliminary tests with the filter cleaning system demonstrated that
the fundamental design was sound. The pressurized chambers emptied
virtually instantaneously when the Galigher valves were actuated,
providing the desired pulse through the filters. Operation of the system
for a few hours indicated that a cycle time of 10 to 15 minutes and a
cleaning air pressure of 170 kPa gauge (25 psig) should be sufficient to
keep the filters clean. The cleaning pulse was so strong, in fact, that
the shape of the initial test filters was found to be distorted when the
magnet canisters were opened. The steel wool used in the initial tests
was AISI Type 430 stainless steel as used before, but it had been donated
by a new supplier for the field tests. It contained an abundance of
45
-------
very fine fibers (50 pm diameter or less) that became brittle after a
few hours of operation, contributing to the degradation of the filters'
mechanical strength. To correct these problems, a stronger set of
filter backup screens was fabricated from expanded metal, and a new
supply of steel wool was obtained from the original source that had
proven satisfactory in the lab pilot plant tests. Plans were also made
to upgrade the density of the filter for the performance characterization
since the results from the initial tests indicated an outlet dust loading
of 110 to 180 mg/Nm3, well above the desired.level of 46 mg/Nm3 (0.02
gr/DSCF).
Magnet Operation
Several minor problems were experienced initially with the magnets,
but they were all corrected easily. Improper voltage indications led to
the discovery of two errors in the field-wiring of the bus cables and
power supply instrumentation. The safety interlock system tripped the
magnets several times until this problem was traced to insufficient
coolant flow. A small booster pump was added to the cooling water line.
Later the differential pressure switch on Magnet B was replaced when it
was found to be faulty. One of the power supplies failed after the
first few hours of operation. The troubleshooting procedure in the
operating manual indicated a faulty gate card (one of several internal
circuit boards), which was replaced. The problem recurred later and a
second gate card had to be replaced. Evidently the cards were deficient
at the start or were damaged during transport of the trailer because no
further problems were experienced during the several hundred hours of
operation that followed.
Flow Measurement and Control
A particularly tenacious problem involved finding the source of a
discrepancy between the flow indicated by the orifice and that measured with
pitot tubes at the sample ports. The flow profiles at the two interior
sampling points were reasonably flat and agreed with one another within
about 5 percent, but the flow indicated by the orifice was 10 to 100
percent higher than the pitot flow, depending upon the flow conditions.
46
-------
No significant leaks could be found in the piping system or the instru-
mentation lines. Finally it was observed that the orifice flanges had
been welded only to the butt ends of the pipe and not to the outside of
the pipe at the back of the flanges. While this construction was
structurally sound, it allowed a small leakage into the flange pressure
taps, which was sufficient to produce an error in the indicated orifice
pressure drop when the static pressure within the pipe was significantly
below atmospheric. A sealant was applied to stop the leakage and the
agreement between the orifice and pitots was excellent thereafter.
Appropriate values for the proportional gain and integration constant
of the flow controller were set by adjusting these parameters with the
pilot plant in operation. The flow controller was capable of maintaining
constant flow during the cyclic changes in filter pressure drop and
withstood the filter cleaning pulse with no problems. However, occasional,
unannounced fluctuations in the static pressure of the windbox exhaust
caused the automatic control system to enter an unstable, oscillating
condition. Since the flow varied very little with the controller in the
manual model and could be corrected quickly by the pilot plant operator
if necessary, manual operation was used during most of the test program.
Final Site Setup
Figures 20 through 23 show the HGMF mobile pilot plant installed on
the sinter plant site. Figure 20 is a general elevation showing the
front of the trailer where the laboratory/control room is located. The
stack is visible at the right rear of the trailer with the muffler
installed below the top section. Figure 21 shows part of the piping
that leads from the windbox exhaust to the pilot plant. The gas enters
the trailer via the pipe passing under the steps, turns, and flows
toward the rear in the pipe visible through the open door. The clean
gas pipe inside the trailer is just behind the dirty gas pipe at the
same level above the floor. The interior sampling ports are located in
these horizontal runs of pipe. In the left of Figure 21 is the flexible
duct to which the propane heater flow was connected during startups.
The exterior sampling ports on the dirty side were installed in the
vertical section of pipe just above the entrance of the flexible duct.
47
-------
Figure 20. HGMF mobile pilot plant setup at the sintering plant.
Figure 21. Side entrance of mobile pilot plant.
-------
Figure 22 shows closer detail of the clean side pipe, the blower
enclosure, and the stack. The recycle line leading from the stack to
the blower suction was used when the lowest flow rates were desired in
the pilot plant. By establishing a controlled recycle flow the blower
could have sufficient flow to remain in the stable operating region.
Figure 22 was taken from the rear of the pilot plant and shows the dirty
side piping coming over from the plant at the right. The railing around
the top of the trailer was added to provide a safe working area at the
Method 5 sampling ports, which were located just below the muffler.
PERFORMANCE CHARACTERIZATION
The objectives of the performance characterization were to evaluate
the effects of filter density and depth, applied magnetic field strength,
and gas velocity on particle collection and to identify the optimal
conditions of operation for demonstration during the long-term testing
period. A factorial experiment was designed with two levels of filter
density, two levels of filter depth, two levels of applied field, and
three levels of gas velocity. AISI Type 430 medium grade steel wool
purchased from Brillo Purex Company (London, OH) was used to construct
all four filters used during the performance characterization. The
average equivalent cylindrical diameter of this material was calculated
to be 94 pm by measuring the total length and mass of a random sampling
of strands removed from the bulk supply.
The particulate sampling procedure utilized throughout the field
tests was adopted from EPA guidelines for the use of cascade impactors
(Harris, 1977). MRI Model 1502 cascade impactors were used with their
substrates precoated with Apiezon H grease and baked at 140°C for two
hours. Prior to each run, pi tot traverses were run on the two 12-point
perpendicular diameters along which each impactor would sample. The
nozzle sizes and sample rates were then selected to establish a velocity
isokinetic with the average gas velocity in the pipe. During the runs
the impactor temperature was controlled to within ±20°C of the stack gas
temperature, which ranged from 80 to 137°C. Two control runs conducted with
the impactors at 100°C demonstrated an average substrate weight loss of
49
-------
L -
r
Figure 22. Clean side piping and stack.
Figure 23. View from rear of pilot plant.
-------
only 0.016 mg. In contrast, the stage accumulations during the actual runs
varied from a minimum of 0.03 mg to a maximum of 31.68 mg. More typically,
the stage accumulations ranged from 0.5 to 5 mg. All stage weights were
determined on a Perkin-Elmer Model AD-2Z microbalance, and stage calibra-
tions for the MRI impactors were taken from Gushing, et al. (1976).
Since the impactors were used external to the pipes, an acetone
probe wash was used to recover any material that did not reach the
impactors. The dried probe wash normally amounted to about 10 percent of
the total catch. This mass was included in the total mass calculations
derived from the impactors, but no attempt was made to assign a size to
it. Hence, the fractional efficiency calculations ignored the material
collected in the probe.
Suprisingly the filter catch ranged from 0.79 to 23.11 mg. In most
of the runs, the filter mass dominated the total mass calculations and
shifted the mass median diameter to a much lower size than was expected
since sinter dust is generally regarded as a coarse material. As
mentioned earlier, it is quite probable that some of the larger material
dropped out in the pipe leading from the windbox exhaust to the pilot
3
plant. The inlet concentration averaged 318 mg/Nm which is about 30
percent of the value typically reported for sinter plants.
Table 3 summarizes the results obtained during the performance
characterization. For additional information, Figures 24 through 29
present the cumulative particle size distributions calculated directly
from the experimental data. The curves drawn through the data points were
produced by a computerized data reduction scheme. Basically the scheme
calculates the stage cut points and then transforms the cumulative distri-
bution versus stage cut points to log-normal space. It then fits a
natural cubic spline to the transformed data. After fitting the cumulative
distribution data, the procedure differentiates the cumulative curve to
obtain the differential distribution. The differential distributions of
corresponding inlet and outlet data are then multiplied by the total mass
concentrations (excluding the probe washes) and ratioed to obtain the
fractional efficiency curve. The fractional efficiency curves of the
performance characterization tests are presented in Figures 30 and 31.
51
-------
Table 3. Results of performance characterization with medium grade steel wool.
Filter
#1
F=0.010
L=0.20m
#2
"F=0.014
L=0.30m
#3
F=0.010
L=0.30m
#4
F=0.014
L=0.20m
Test No.
08042
08041
08031
08062
08051
08061
08082
08183
08182
08181
08231
08222
08291
08282
08281
08292
08301
08302
09021
09013
09012
09031
09032
09033
Applied
Field
tesla
0.25
0.25
0.25
0.50
0.50
0.50
0.50
0.25
0.25
0.25
0.50
0.50
0.25
0.25
0.25
0.50
0.50
0.50
0.25
0.25
0.25
0.50
0.50
0.50
Superficial
Velocity
m/s
3.6
4.4
5.7
3.8
3.9
4.9
7.9
3.0
6.4
8.8
6.5
7.9
3.8
5.9
7.7
3.9
6.0
6.9
3.7
5.8
7.4
4.0
5.8
7.9
Filter AP
cm H^O
9
11
18
9
11
17
32
89
102
109
76
114
38
44
79
31
72
114
18
41
66
25
64
76
Temperature
°C
107
114
137
121
135
120
127
95
98
109
101
99
107
100
104
128
106
93
93
87
95
98
104
110
Collection
Efficiency
%
56.6
51.9
0.0
48.7
46.6
51.1
66.1
94.5
83.4
75.7
76.1
69.9
78.2
90.5
91.2
81.5
87.2
87.7
85.6
83.9
86.2
86.7
79.6
74.4
Outlet
Concentration
mg/Nm3 (wet)
96
107
166
122
126
103
134
29
36
66
22
83
75
32
29
55
50
86
62
61
38
84
39
69
-------
en
Co
95
90
« 80
ffl
u: 70
ui
J 60
I 5°
" 40
O
4 30
CC
Ul
20
10
5
0.2
OUTLET
I
I I
INLET
I
0.5 1.0 2
AERODYNAMIC DIAMETER,
08042
10
4
95 f-
90
80
70
60
50
! 40
a
< 30
20
cc
ui
0-
10
0.2
OUTLET
08041
0.5 1.0 2
AERODYNAMIC DIAMETER, j
10
95 r
90
80
70
60
50
40
O
< 30
I 20
CC
Ul
"• 10
0.2
OUTLET
INLET
I
I
0.5 1.0 2
AERODYNAMIC DIAMETER,
08031
10
CD
95
90
80
70
ui
J 60
| 50
S 40
(3
4 30
I 20
CC
Ul
°- 10
0.2
OUTLET
INLET
08062
0.5 1.0 2 5
AERODYNAMIC DIAMETER./urn
10
Figure 24. Size distributions from test nos. 08042, 08041, 08031, and 08062.
-------
95
90
80
60
50
< 30
20
O
cc
UJ
o- 10
0.2
OUTLET
I
I
INLET
I
I
08051
, I . . i i I
0.5 1.0 2
AERODYNAMIC DIAMETER,
10
95
90
5 80
_! 60
< 50
« 40
Ui
30
O
EC
UJ
20
10
0.2
OUTLET
I
I
INLET
08061
I
0.5 1.0 2 5
AERODYNAMIC DIAMETER,Mm
10
95 <-
90
80
70
60
50
2 40
§j 30
Ju 20
O
cc
Ul
°- 10
0.2
OUTLET
I I I I ll
INLET
I
08082
I I i i i I I
0.5 1.0 2 5
AERODYNAMIC DIAMETER, um
10
95
90
80
70
60
50
40
30
20
10
5
0.2
OUTLET
0.5 1.0 2 5
AERODYNAMIC DIAMETER, iim
10
Figure 25. Size distributions from test nos. 08051, 08061, 08082, and 08183.
-------
en
m
95
90
80
70
-j 60
| 50
£ 40
30
ul 20
U
oc
UI
°- 10
0.2
OUTLET
I I I I
I I
0.5 1.0 2
AERODYNAMIC DIAMETER,
08182
I
10
95
i/j 90
>• 8°
m
cc 70
ui
j 60
£ 60
Ifi
ill 40
U
< 30
20
10
0.2
I
INLET
OUTLET
08181
i
I I
I
0.5 1.0 2 &
AERODYNAMIC DIAMETER, tan
10
o
££
95
90
80
70
60
50
40
30
20
10
0.2
OUTLET
INLET
I I I I ll
0.5 1.0 2
AERODYNAMIC DIAMETER,
08231
.... I
10
95 i-
90
BO
s
m
u: 70
111
- 60
I 50
(A
iu 40
U
< 30
2 20
£C
ill
"- 10
5
0.2
I
. .1
O INLET
AOUTLET
I
08222
. I ! . , ,1
0.5 1.0 2
AERODYNAMIC DIAMETER, /
10
Figure 26. Size distributions from test nos. 08182, 08181, 08231, and 08222.
-------
o
(C
IU
90
80
70
60
50
40
30
10
5
0.2
OUTLET
I i
ILL
_L
i
08291
i I i i i 11
0.5 1.0 2
AERODYNAMIC DIAMETER,
10
95
90
80
70
60
50
40
30
20
10
S
0.2
OUTLET
I . I II I
08282
i I i i i 11
0.5 1.0 2 5
AERODYNAMIC DIAMETER, /im
10
95 r-
0.2
0.5 1.0 2 5
AERODYNAMIC DIAMETER. ium
95
90
* 80
to
a: 70
ui
-j 60
| 50
S 40
0
< 30
I 20
K
lu
o- 10
0.2
OUTLET
_L
INLET
08292
J I It I I ll
0.5 1.0 2
AERODYNAMIC DIAMETER.
10
Figure 27. Size distributions from test nos. 08291, 08282, 08281, and 08292.
-------
0.5 1.0 2
AERODYNAMIC DIAMETER,
10
Ill
o
a:
95 r~
90
80
70
60
SO
40
30
20
10
5
0.2
OUTLET
I I I I 1 I
J I
0.5
1.0
09201
I i i i il
5 10
AERODYNAMIC DIAMETER. >tm
95 •-
0.5 1.0 2
AERODYNAMIC DIAMETER,
10
I
09013
I I i i i il
0.5 1.0 2 5
AERODYNAMIC DIAMETER. >im
10
Figure 28. Size distributions from test nos. 08301, 09201, 08302, and 09013.
-------
Ul
CO
95 i—
95 r-
90
80
£ 70
60
50
S 40
<
H 30
LU
O 20
ill
O.
10
0.2
0.5 1.0 2 5
AERODYNAMIC DIAMETER, jam
10
OUTLET
INLET
I I I I I I I I
09032
II. i i I
O.S 1.0 2
AERODYNAMIC DIAMETER, nm
10
5
>•
95 r
90
80
£ 70
60
50
UJ
O 40
<
H 30
O 20
10
5
0.2
0.5 1.0 2
AERODYNAMIC DIAMETER. «
10
OUTLET
INLET
i i i i
09033
J LI ill
0.5 1.0 2
AERODYNAMIC DIAMETER,
10
Figure 29. Size distributions from test nos. 09012, 09031, 09032, and 09033.
-------
en
V.O
99.9
99.8
99.5
99
98
95
o
2
Uj
0
70
60
H 50
UJ
-j 40
O
0 30
20
10
5
0.2
08042
08041
08031
08062
08051
08061
08082
FILTER #1 (F =0.010. L = 0.20m)
Test No. Field. T
0.25
0.25
0.25
0.50
0.50
0.50
0.50
Cone, out,
Velocity, m/s
3.6
4.4
5.7
3.8
3.9
4.9
96
107
166
122
126
103
134
Efficiency. %
56.6
51.9
0.0
48.7
46.6
51.1
I6*1. .
0.5 1.0
AERODYNAMIC DIAMETER,
10
99.9
99.8
99.5
99
98
95
* 90
O
W 80
O
£ 70
ui
Z 60
O
& 50
ul
1 40
O
0 30
20
10
5
0.2
81
08291
08282
08281
08292
08301
08302
FILTER #3 (F - 0.010, L = 0.30 m)
Cone, out.
Test No. Field. T
Velocity, m/s
0.25
0.25
0.25
0.50
0.50
0.50
3.8
5.9
7.7
3.9
6.0
6.9
75
32
29
55
50
86
Efficiency. %
78.2
90.5
91.2
81.5
87.2
87.7
J_L
0.5 1.0 2
AERODYNAMIC DIAMETER.
10
Figure 30. Fractional efficiency curves from tests conducted on filter nos. 1 and 3.
-------
CD
u
z
Ul
u
o
u
o
u
99.9
99.8
99.5
99
98
95
90
80
70
60
50
40
30
20
10
5
0.2
09021
09013
09012
09031
09032
09033
31
FILTER W4 (f = 0.014, L = 0.20 m)
Cone, out,
Tesi No. Field,! Velocity, rn/s mg/IMm
Efficiency, %
0.25
0.25
0.25
0.50
0.50
0.50
3.7
5.8
7.4
4.0
5.8
7.9
62
61
38
84
39
69
856
83.9
86.2
86.7
79.6
74.4
0.5 1.0 2
AERODYNAMIC DIAMETER, jjm
10
O
2
U
UJ
o
u
Ul
o
o
99.9
99.8
99.5
99
98
95
90
80
70
60
60
40
30
20
10
5
0.2
31
22
08183
08182
08181
08231
08222
FILTER trl (F =0.014, L = 0.30m)
Cone, out.
Test No. Field, T
Velocity, m/s
Efficiency,
0.25
0.25
0.25
0.50
0.50
3.0
6.4
8.8
6.5
7.9
29
36
66
22
83
94.5
83.4
75.7
76.1
69.9
0.5 1.0 2
AERODYNAMIC DIAMETER,
10
Figure 31. Fractional efficiency curves from tests conducted on filter nos. 2 and 4.
-------
The cubic spline procedure tends to produce multimodal efficiency curves
that can result from real effects of the collection device, from arti-
facts of the impactor, and from random error in the data. The curves
shown in Figures 30 and 31 have been smoothed for clarity to show only
the dominant behavior. A more complete explanation of the merits of the
natural cubic spline and other impactor data-fitting procedures is given
by Lawless (1978).
The goal for the outlet concentration was set at 46 mg/Nm to
correspond with typical regulatory requirements. Table 3 reveals that
this goal was achieved in several of the tests, but the lowest filter
pressure drop under which it was accomplished was 44 cm H20--a value
that is probably too high for practical application of HGMF to particu-
late control. The data also show several features with respect to
parametric effects. First, in Table 3 there is no consistent correlation
of pressure drop with gas velocity and filter characteristics because
the filters showed a tendency to plug during the test series. Thus,
most of the pressure drop figures are not clean filter values, but
correspond to some partially plugged state. A cycle duration of 16
minutes was used during most of the tests with a cleaning air pressure
of 275 kPa (40 psig). With filter #2, which was the most troublesome
with respect to plugging, the cycle time was reduced to 8 minutes, but
plugging was still a problem.
The performance characterization data also show a lack of correla-
tion of either collection efficiency or outlet concentration with gas
velocity or applied magnetic field. They do demonstrate with reasonable
consistency that increasing either filter depth or density increases the
collection efficiency and reduces the outlet concentration. For example,
direct comparison of filter nos. 1 and 3 in Figure 3 adequately demonstrates
the effect of filter depth, while comparison of filter nos. 1 and 4
shows the filter density effect. During the performance characterization
several observations were made:
(1) The dust passing through the HGMF was extremely fine as
evidenced by the size distribution curves shown in Figures 24
through 29. Furthermore, there was a marked difference in
appearance between the inlet dust and the outlet dust and
between the larger dust and the finer dust. The larger inlet
dust, which was collected with higher efficiency, was typically
of a reddish-brown or reddish-gray color indicative of iron
oxides. The finer dust that escaped was either white or light
gray. Plans were made to collect samples for magnetic and
chemical analyses.
61
-------
(2) Although the outlet concentration goal was attained, the
tendency of the filter to plug made it doubtful that this mode
of operation could be sustained over a long period of time.
In addition, the pressure drops experienced were impractically
high. The plugging also changed, at least to some extent, the
collection mechanism from HGMF to cake filtration, rendering
the mathematical model of the process invalid. (The model
also assumes homogeneous particle composition with size, which
was clearly not true in this application.) Plans were made to
try a more coarse grade of steel wool in the long-term testing
period in hopes of improving the cleanability of the filter
while achieving satisfactory collection.
(3) During most of the performance characterization the opacity of
the pilot-plant stack plume was less than 5 percent when
viewed against a blue sky in accordance with the Ringleman
procedure, but when viewed against a dark background, a blue
or white plume was always visible. Also the opacity increased
noticeably for a few seconds each time the filter cleaning
pulse was released. The Climet particle size analyzer was
used to conduct a limited study of transient emission effects.
The results of this work are reported in Section 8.
LONG-TERM TESTING
The overall objective of the long-term test program was to demon-
strate the effectiveness and reliability of the HGMF collection process
over an operating period of at least 500 hours. To have practical
significance this meant that the throughput of the pilot plant should be
high enough to lead to a reasonable projection of capital cost (>1500
3
m /hr); that the outlet concentration should be low enough to meet
2
typical emission regulations (< 46 g/Nm ); that the filter pressure drop
should stabilize at a level that would project to reasonable operating
costs (< 30 cm H20); and that all of the equipment, especially the
magnetic filters and the cleaning apparatus, should function continuously
without an excessive number of forced shutdowns. Table 4 is a condensed
log of the 500 hours of operation accumulated during the long-term test
period. The highlights are discussed in more detail below.
The long-term operations began on September 12, 1979. The first
filters loaded into the magnet canisters were made of medium-grade steel
wool because the coarse-grade material had not yet been received from
the vendor. The filters had a depth of 0.20 m and a packing density of 0.015
62
-------
Table 4. Operating log of the long-term test period.
Time, Date
Cumulative Time
of Operation,
Mrs.
Observations
1615, 9/12/79
0100, 9/13/79
0515, 9/13/79
0553, 9/13/79
0750, 9/14/79
0850, 9/14/79
1601, 9/14/79
0 Start with filter #5:
Medium-grade steel wool, F=0.015,
L=0.20 m
Applied field = 0.25 T
Superficial velocity = 4.8 m/s
(2550 m3/hr)
Initial AP = 20 cm H20
Cycle duration = 10 min
Cleaning air pressure - 275 kPa
gauge
Inlet gas temperature - 93°C
8.8 Temperature - 79°C; filter AP = 31 cm H20;
reduced cycle duration to 8 min
13.0 Temperature = 69°C; filter AP = 39 cm H20;
stopped inlet gas flow and warmed
system to 100°C with hot air while
operating cleaning system
13.0 Restarted inlet gas flow; filter AP - 41
cm H20
39.0 Filter AP has continued to be a problem,
now at 34 cm H20; temperature - 80°C;
stopped inlet gas flow and warmed
system to 100°C with hot air while
operating cleaning system
39.0 Restarted inlet gas flow; filter AP - 25
cm H20
46.1 Stopped system, compressor is not filling
--.. cleaning tanks to desired pressure
Sinter strand down for several days for annual maintenance
1515, 9/22/79
1635, 9/23/79
1030, 9/24/79
1300, 9/24/79
0552, 9/24/79
0616, 9/25/79
46.1 Start with filter #6:
Coarse-grade steel wool, F=0.015,
L=0.20 m
Applied field = 0.25 T
Superficial velocity = 4.8 m/s
(2550 m3/hr)
Initial AP = 9 cm H20
Cycle duration - 8 min
Cleaning air pressure = 275 kPa
Inlet gas temperature = 100°C
71.4 Ringlemann reading 8%
89.4 Filter AP = 14 cm H20, running well but
shut down briefly for compressor
maintenance
89.4 System backup, Ringlemann reading 5%
106.2 Valve V7B not opening fully; system shut
down to investigate problem
106.2 System back up; V7B working correctly
53
-------
Table 4. (continued)
1100, 9/26/79
1130, 9/26/79
0700, 9/27/79
1705, 9/27/79
1345, 9/28/79
1400, 9/28/79
1716, 9/28/79
0800, 9/30/79
1620, 9/30/79
10/2/79 - 10/3/79
1030, 10/3/79
1100, 10/3/79
1400, 10/3/79
1540, 10/3/79
1110, 10/4/79
1120, 10/4/79
2400, 10/4/79
1730, 10/9/79
m.O Tank A is filling to 210 kPa and Tank B
is filling to 420 kPa; shut down to
investigate; Tank B is filling via
leak in diaphram of Pinch Valve 8;
Adjusted actuator pressure and
program cycle to correct
111.0 System restarted; tanks filling to 275
kPa
154.5 Shut down for weekly maintenance on
sinter strand
154.5 System restarted as before; filter AP =
11 cm H20; collected samples of
stack particulate and dump from
cyclone hopper
175.2 Shut down to check operation of flow
controller
175.2 Back on line
178.4 Inlet gas temperature hit, highest value
of 149°C; filter AP = 11.4 cm H20;
217.2 Shut down to change filter
217.2 Start with filter #7:
Coarse-grade steel wool, F=0.0175,
L=0.20 m
Applied field - 0.25 T
Superficial velocity = 6.5 m/s
(3400 m3/hr)
Initial AP = 22 cm H20
Cycle duration = 10 min
Cleaning air pressure = 275 kPa
Inlet gas temperature = 117°C
Steel company research team is conducting
emissions tests on HGMF pilot plant;
Ringlemann reading about 10%
273.3 Valve V7B malfunctioning again; shut
down to check
273.3 System back up so that testing teams can
finish; V7B is not functioning
properly but won't affect tests
276.3 Stack testing completed; shut down to
replace V7B with spare
276.3 System back up and operating as before
295.8 Sinter strand tripped; HGMF flow
continued
295.8 Strand back on
308.5 System running well; filter AP = 20 cm
H20; shutdown to give RTI operators
time off and to accomodate weekly
maintenance on sinter strand
308.5 System restarted; same filter and opera-
ting parameters as before
64
-------
Table 4. (continued)
2250, 10/9/79 313.8
10/10/79 - 10/11/79
1120, 10/12/79 374.3
1330, 10/12/79 374.3
2340, 10/17/79 504.5
Filter aP up to 25 cm H20; decreased
cycle time to 8 min
RTI team is conducting emissions tests
on HGMF pilot plant; Ringlemann
reading about ~\Q%; collected dust
samples for analysis
Sinter strand down for unscheduled
maintenance; shut off HGMF
Sinter strand and HGMF back on line
Duration goal achieved; system shut down;
final filter AP = 21 cm H20.
65
-------
Since increasing the magnetic field strength had resulted in no signifi-
cant improvement in collection efficiency during the characterization
runs, the field was set at the lower value. A moderate flow rate of
2550 m /s was selected for initial operation.
As expected, the first filter developed problems with plugging,
particularly at night when the inlet gas temperature dropped to as low
as 70°C. On two occasions the sinter gas flow was stopped temporarily
and the filters were flushed with hot air and backflushed with pulse air
to loosen the deposits that accumulated. After 46 hours of operation,
the system was shut down due to an apparent compressor malfunction that
was not allowing the cleaning air tanks to fill properly.
The pilot plant remained down for 8 days while scheduled annual
maintenance was conducted on the sinter strand. During this time the
pressurized air problem was investigated but not completely resolved.
A new filter of Purex coarse-grade steel wool was inserted, and the
pilot plant was restarted on September 22 with the same operating
parameters used with the first filter. In contrast to the medium-
grade steel wool, which exhibited a clean filter AP of 20 cm H?0 at 2550
3
m /hr, the coarse-grade filter showed an initial AP of only 9 cm H^O at
the same flow. The equivalent diameter of the coarse fibers was deter-
mined to be 195 jam. Over 171 hours of operation the second filter
experienced no significant plugging problems. Ringlemann readings taken
on the pilot plant stack on two occasions were 8 percent and 5 percent.
A sample of the stack dust was collected in an alundum thimble for
chemical and magnetic analysis, but no quantitative stack, testing was
conducted.
During the operation of the second filter, the air pressure problem
was traced to a leak in the diaphragm of one of the pinch valves. The
leak allowed a flow of actuator air from between the housing and diaphragm
of the valve into the cleaning air tank and into the clean side of the
pilot plant flow. This malfunction created an imbalance in the tank
pressures and also resulted in an air consumption rate beyond the
capacity of the pilot plant compressor. The leakage problem was minimized
and the tank pressures were balanced by adjusting the actuator pressure
and by changing the duty cycle of the valves, so that each would be
closed (air-actuated) only when the corresponding tank needed to be filled.
66
-------
The remaining deficit in air capacity was made up by tying into the
compressed air system of the sinter plant.
The pilot plant was shut down on September 30 to change filters in
preparation for emission testing. Since the coarse steel wool had shown
no tendency to plug, the packing density of the new filters was increased
o
to 0.0175, and the flow was increased to 3400 m /hr. An increase in
efficiency was anticipated with the increased packing density. The flow
rate was increased since the performance characterization had shown
velocity to have no detrimental effect on collection, and higher flow
rates naturally lead to more favorable capital cost projections.
The system was restarted on the same day with an initial filter AP
of 22 cm HgO. With the exception of the pinch valve and a sticky butter-
fly valve, all components of the pilot plant functioned perfectly for
287 more hours of operation. Four particulate concentration tests were
conducted by a team from the steel company, and two were conducted by RTI
personnel. The results of the tests are summarized in Table 5, which shows
the mean of each measured parameter plus or minus one standard deviation.
Both testing groups used EPA Method 5 on the outlet. The concentrations
reported represent material collected in the probe and on the filter.
The impinger filtrate proved to be negligible. The steel company used
alundum thimbles for inlet sampling, and the RTI team ran two additional
outlet tests with MRI impactors for comparison to the Method 5 results.
The outlet concentrations determined by the RTI sampling team are
somewhat lower than those measured by the steel company team; but unfor-
tunately even the lower values are well above the goal of 46 mg/Nm .
Ringlemann readings on the light blue plume averaged less than 10 percent
during both test periods although higher-opacity puffs occurred during
each filter cleaning. The outlet size distributions determined by RTI's
impactor samples are shown in Figure 32. They demonstrate that the
outlet dust was extremely fine, having a mass median aerodynamic diameter
of 0.2 to 0.3 ym. Comparison of the total mass results obtained from
the impactors with the simultaneous Method 5 results indicate good
agreement. Hence the outlet total mass concentrations obtained during
the performance characterization should be valid.
67
-------
Table 5. Efficiency testing during long-term operation (filter #7).
Test Team
Steel
g Steel
RTI
RTI
Location
Inlet
Outlet
Outlet
Outlet (MRI)
Temperature Water Content
C %
96 ± 13 11.3 ± 1.0
136 ± 6 10.5 ± 0.3
142 ±9 8.3 ± 3.8
101+0
Dust Concentration
mg/Nm3 (dry)
550 ± 25
260 ± 66
185 ± 45
190 ± 11
Collection
Efficiency
%
-
53*
66*
65*
_
Based on ratio of indicated mean outlet concentration to the mean inlet concentration measured by the
steel company sampling team.
-------
en
DC
V)
m
CD
95
90
80
70
60
50
Z 30
LU
O
DC
20
10
5
0.2
I I I
.I.
I
I
o 10101
A 10111
I
I 111! | I
0.5 1.0 2
AERODYNAMIC DIAMETER, jum
10
Figure 32. Outlet particle size distributions obtained during total mass sampling.
-------
In summary, after the coarse-grade steel wool was inserted as a
filter material, the pilot plant performed in a relatively trouble-free
manner for over 450 hours. The pressure drop remained constant at a
level lower than the pre-operational goal with a flow rate higher than
the goal. But the particulate concentration tests showed that the
magnetic filter could not remove a sufficient percentage of the very
fine fraction of the dust while operating under these practical condi-
tions. The reasons for the unsatisfactory collection are discussed in
Section 8.
ADDITIONAL TESTS WITH COARSE-GRADE STEEL WOOL
Since a few days remained for field work at the end of the long-
term operation, a brief series of additional tests was conducted with
coarse steel wool. The objectives were to make a final effort at
achieving low outlet concentrations under practical operating conditions
and to take a second look at the effects of magnetic field strength and
velocity on collection without the interference of filter plugging. The
new filters were made with the same packing density as the last filter
used in the long-term testing (F=0.0175) and the full available depth of
the canisters (L=0.30 m). Two velocities and two fields were used and
one of the four operating conditions was duplicated. Both EPA Method 5
and impactor sampling were conducted. In addition, one inlet and one
outlet test were run with an alundum thimble to collect sufficient
material for chemical and magnetic analysis. The results of the efficiency
tests are summarized in Table 6. Cumulative particle size distributions
are presented in Figure 33, and the fractional efficiency curves are
shown in Figure 34.
The duplicate runs (10231 and 10271) exhibited a significant dif-
ference in the outlet concentration that was directly attributable to a
similar difference in the inlet concentration. The collection efficiency
of the two runs was approximately the same. Comparison of the corre-
sponding Method 5 and outlet impactor results showed good agreement on
total concentration with the exception of test no. 10241. The impactor
sample of test no. 10241 showed an outlet concentration close to the
established goal, but this low concentration was suspect because it was
not confirmed by the corresponding'Method 5 test.
70
-------
Table 6. Additional testing with coarse grade steel wool (filter #8).
Impactor Te
Test No.
10231
10271
10251
10261
10241
Total Mass
Test No.
10231
ir>S>71
\\jt-l 1
10251
10251
10261
10261
10241
ists
Applied Field
tesla
0.25
0.25
0.25
0.50
0.50
Tests
Location
Outlet
fin-Hat-
UK L 1 c L
Outlet
Outlet*
Inlet*
Outlet
Outlet
Superficial
Velocity
m/s
4.0
4.4
8.8
3.9
8.5
Temperature
°C
133
.. _. — _ _ — ^amnlinn
134
134
92
129
135
Filter AP Temperature
cm H20 °C
15 97
14 80
62 99
13 86
51 98
Water
Content Flow
% Nm3/hr (dry)
6.1 1270
8.4 2830
6.6
7.0
10.8 1370
9.1 2900
Collection
Efficiency
%
58.3
54.1
51.2
41.9
89.8
Dust
Concentration
mg/flm3 (dry)
227
187
172
488
222
192
Outlet
Concentration
mg/Nm3 (wet)
218
125
149
185
62
Collection
Efficiency
%
-
_
54.5
-
Using alundum thimble in lieu of Method 5 filter.
-------
ro
95 r
0.2
o A 10231
• * 10271
0.5 1.0 2
AERODYNAMIC DIAMETER,
10
0.5
2
>
UJ
O
<
z
UJ
O
DC
95 I-
90
80
70
60
50
30
20
10
5
0.2
95
90
80
70
60
50
S 40
30
20
10
0.2
OUTLET
INLET
0.5 1.0 2
AERODYNAMIC DIAMETER,
10251
i I i ... I
5 10
OUTLET
INLET
10241
I
I
AERODYNAMIC DIAMETER. (Jm
0.5 1.0 2
AERODYNAMIC DIAMETER,
10
Figure 33. Cumulative size distributions for test nos. 10231 through 10271.
-------
o
UJ
u
!T
99.9
99.8
99.5
99
98
95
90
80
£ 70
Z
O
60
50
O 40
o
30
20
10
5
0.2
51
Test No.
10231
10271
10251
10261
10241
Field, T
0.25
0.25
0.25
0.50
0.50
j I
Cone, out,
Velocity, m/s
4.0
4.4
8.8
3.9
8.5
218
125
149
185
62
0.5 1.0 2
AERODYNAMIC DIAMETER,/
Efficiency, %
58.3
54.1
51.2
41.9
89.8
41
51
I I
10
Figure 34. Fractional collection efficiency curves for test nos. 10231
through 10271.
73'
-------
Once again the results were inconclusive on the effects of magnetic
field and gas velocity. The low-field tests (10231, 10271, 10251)
showed little difference in overall collection efficiency, although the
fractional efficiency plots imply that increasing the velocity improved
small particle collection and degraded large particle collection. This
observation would be in qualitative agreement with the combined inertial
impaction and reentrainment phenomena discussed in conjunction with the
mathematical model. The high-field impactor data indicate that high
velocity is beneficial to particle collection, but this conclusion is
dependent on the questionable results of test no. 10241. The Method 5
tests showed little if any significant difference in the outlet concen-
trations of the four runs.
The salient conclusion from these additonal tests was similar to
that of the earlier efficiency tests—that the emission rate of very
fine particles was too large for the HGMF system to be applied success-
fully to this dust source. The reason for the high emission rate of
fine particles was discovered through chemical and magnetic analyses of
dust samples that are discussed in the following section.
74
-------
SECTION 8
DISCUSSION AND APPLICATION OF RESULTS
This section discusses in further detail experimental data and
observations that indicate why the magnetic filtration pilot plant was
unable to collect a satisfactory amount of the sinter dust under practical
operating conditions. Following that discussion an analysis is presented
to define the criteria for successful application of HGMF to particulate
emission control and to make economic estimates for applications to
specific industrial processes.
CHEMICAL AND MAGNETIC ANALYSIS OF SINTER DUST
Durin-g the long-term operational period and the additional tests that
were conducted with the coarse steel wool filter, samples of dust were
collected for analysis from three points: (1) from the sample ports
installed in the vertical section of pipe leading into the pilot plant
(inlet); (2) from the sample ports installed in the pilot plant stack
(outlet); and (3) from the hopper of the cyclone that acts as a secondary
collector in the filter cleaning system (collected). The inlet and outlet
samples were obtained by sampling isokinetically at the centerline of the
respective pipes with an in-stack alundum thimble. The cyclone samples
were grab samples obtained by dumping the hopper contents.at designated
times. Table 7 reports the results of the chemical analyses conducted
on the samples. All of the metal determinations were made by flame
atomic absorption. The chloride and sulfate concentrations were determined
by ion chromatography.
The bottom row of Table 7 reports data on a wet, green crystalline
material that was found dripping from the stack muffler. A sample was
collected, and the analysis indicates that it was primarily ferrous
sulfate and ferrous chloride. Both of these compounds exist in a variety
of hydrated forms at the stack temperature. The material most probably
originated from condensation of acid mists in the stack and muffler followed
by chemical attack on the metal. The outer cylinder of the muffler was
constructed of carbon steel.
75
-------
Table 7. Chemical analysis of dust samples from HGMF pilot plant tests.
Sampler Description
Inlet, 10/10/79
Inlet, 10/26/79
Outlet, 9/27/79
Outlet, 10/25/79
ov
Collected, 9/27/79
Collected, 10/9/79
Collected, 10/25/79
Chemical Analysis, Reported in
7
10
1
1
12
14
12
Fe
.55
.3
.24
.52
.7
.0
.3
Ca
4.80
5.60
1.76
2.88
6.40
8.80
8.00
Mg
1.45
1.76
0.44
0.76
1.84
2.44
2.32
Al
8.10
8.08
2.56
3.68
10.0
12.4
11.8
Si
1.9
1.4
0.5
0.6
2.2
2.4
2.6
Na
0.64
0.75
1.12
1.36
0.42
0.36
0.42
% by Mass
K
16.0
17.4
35.4
29.9
11.3
7.2
9.4
Cl"
23.5
22.7
37.7
31.2
14.6
8.40
15.8
so4=
1
1
2
1
1
2
2
.62
.61
.26
.33
.98
.04
.96
23'3 <] °'18 <] °'2 °'22 <] 4'30 32'5
-------
The inlet data in Table 7 show an average iron content of only 8.9
percent compared to the 25 to 50 percent reported in Table 2 as a
generally accepted range for sinter dust. In addition, the inlet
•3
concentration of the dust averaged less than 0.5 g/Nm with a mass
median aerodynamic diameter of less than 2 ym compared to the generalized
o
values of 1 to 2 g/m and 10 ym. When combined, these figures indicate
that a substantial quantity of larger dust particles fell out of the gas
stream before reaching the pilot plant, and that the larger particles
were most likely higher in iron content than the dust that reached the
pilot plant. The latter contention is further substantiated by the
magnetic analysis of the same seven samples that is shown in Figure 35.
The specific magnetization curves of the two inlet samples are coinci-
dent and show a saturation magnetization at 3 kOe of 2.4 emu/g. In
contrast, the coarser sinter dust that was obtained from the plant
precipitator hoppers and tested in the laboratory pilot plant had a
saturation magnetization of 9 emu/g. This partially explains the
disappointing collection efficiency experienced in the field tests. The
inlet dust was simply lower in magnetic susceptibility than expected.
Further analysis of Table 7 and Figure 35 reveals additional infor-
mation of interest. The magnetization curves of the outlet dust are
quite low because the outlet dust contained a very low percentage of
iron and a high percentage of alkali chlorides. This finding is consis-
tent with the qualitative observations made in Section 7 that the finer
outlet dust was white or light gray in comparison to the inlet dust,
which had a reddish-brown or reddish-gray color. In contrast to the
inlet dust the collected dust had a higher iron content, a higher
magnetization, and was a dark red color. Comparison of Table 7 and
Figure 35 reveals, as expected, that the magnetization of each of the
samples is roughly proportional to its iron content. None of the other
elements or their compounds contributes significantly to the magnetic
susceptibility.
77
-------
5.0 r
COLLECTED
10/9
COLLECTED
9/27 & 10/25
INLET
10/10 & 10/26
OUTLET 10/25
OUTLET 9/27
1.0
2.0 3.0
APPLIED FIELD, kOe
4.0
5.0
Figure 35. Magnetic analysis of dust samples from HGMF pilot plant tests
78.
-------
The data in Table 7 can be utilized to construct a component mass
balance that yields still more information. Although the seven samples
were not collected simultaneously, the compositions of different samples
from the same source are quite similar. Thus, for the sake of argument
one can treat the samples from each source as if they were replicates
taken during steady-state operations. The corresponding chemical
analyses can then be averaged as if to minimize random error. The
average analyses of the inlet, outlet, and collected dusts are reported
in the first three rows of Table 8. If the three compositions of any
component are treated as exactly correct, one can calculate the overall
mass collection efficiency of the HGMF by constructing a mass balance
on that component.
Mass In = Mass Out + Mass Collected; (7)
Mtci ^t^WWc '
where M. = total mass flow rate of dust into system;
C- = mass percentage of component in inlet;
C = mass percentage of component in outlet;
C = mass percentage of component collected; and
E. = overall mass collection efficiency.
Solving Equation (8) for the overall collection efficiency yields
C.-C
C = _J 9_
Lt C -C '
r Lc Lo
The estimates of E. obtained from each of the components are reported in
the sixth row of Table 8. All of the estimates are in the range from
0.52 to 0.69 with the exception of the sulfate data, which are inconsis-
tent. These data indicate a production of sulfate in the pilot plant,
most probably due to the condensation of acid mist in the cyclone.
7.9
-------
Table 8. Component mass balances based on chemical analysis.
co
o
Analysis
Avg. Inlet (C.. ), mass %
Avg. Outlet (CQ), mass %
Avg. Collected (C ), mass
r r °/
L • - L , a
1 0
r r °i
\j-\j , h
C 0
Et
Component Balance (based on
Inlet Mass, kg
Outlet Mass, kg
Collected Mass, kg
Component Closure, %
Component Efficiency, %
Fe
8.92
1.38
% 13.0
7.54
11.62
0.649
Et = 0.61
8.92
0.54
7.93
95
89
5
2
7
2
5
0
and
5
0
4
Ca
.20
.32
.73
.88
.41
.532
Mg
1.60
0.60
2.20
1.00
1.60
0.625
Al
8.09
3.12
11.4
4.97
8.28
0.600
Si
1.65
0.55
2.4
1.10
1.85
0.595
Na
0.69
1.24.
0.40
-0.55
-0.84
0.655
K
16.7
32.6
9.3
-15.9
-23.3
0.682
Cl
23.
34.
12.
-11.
-21.
0.
1
4
9
3
5
526
so4=
1.62
1.80
2.33
-0.185
0.535
—
average measured composition)
.20
.90
.70
108
90
1.60
0.23
1.34
98
84
8.09
1.22
6.95
101
86
1.65
0.21
1.46
101
88
0.69
0.48
0.24
104
35
16.7
12.7
5.7
110
34
23.
13.
7.
92
34
1
4
9
-
-
-
-
-
-------
When the eight estimates of total mass efficiency are averaged, a
value of 0.61 is obtained. This average is quite reasonable with respect
to the total mass efficiency measurements made during the same time
period (see Tables 5 and 6).- One can then use the estimated average
total mass efficiency and the measured average component compositions to
construct a mass balance for each component. For example, for iron
Inlet: (100 kg)(0.0892) = 8.92 kg (10)
Outlet: (39 kg)(0.0138) = 0.54 kg (11)
Collected: (61 kg)(0.130) = 7.93 kg (12)
The closure on the component mass balance is defined as
n _ Outlet Mass + Collected Mass 1nf)0/ ,,-*
Closure -- Inlet Mass 100/0 ( Uj
Note in Table 8 that the closure on the components is distributed about
100% as expected because of the averaging procedure used to determine the
total mass efficiency. It is remarkable, however, considering the time
spanned by the sample collection, that the range of closure values is
so small. The results show once again that the HGMF collection
process was not strongly influenced by changes in the applied magnetic
field or gas velocity or by minor changes in the filter density or
depth.
The component collection efficiencies reported in the last row of
Table 8 were calculated from the expression
Component Efficiency = • 100* . (14)
The results indicate that the collection efficiency of iron was actually
nearly 90 percent. A similar conclusion could be reached by conducting a
"magnetics balance" on the curves reported in Figure 35. The fact that
81
-------
the calculated collection efficiencies of calcium, magnesium, aluminum,
and silicon are all nearly equal to that of iron suggests that these
components are tightly bound together with the iron in the sintering
process. In contrast the much lower calculated efficiencies of sodium,
potassium, and chlorides suggest that sodium and potassium chloride
appear in discrete fine particles that are collected in the HGMF primarily
by inertial impaction. A credible explanation for the appearance of
discrete, fine particles of sodium and potassium chloride is that these
compounds are vaporized during the sintering process and reformed as a
condensation aerosol in the windbox. The very fine size of the outlet
dust supports this explanation. Furthermore, the fine alkali-chloride
aerosol, which was emitted from the HGMF stack in a bluish-white plume,
could be a major contributor to the "blue haze" that plagues many sinter
plant stacks and is most often attributed to the condensation of hydro-
carbons.
The component mass blances and the magnetization curves of the
dusts also provide insight to the observation that increasing the applied
magnetic field had no appreciable effect on dust collection. Since the
magnetics are nearly saturated at 0.25 T (2.5 kOe), increasing the field
would not substantially improve collection of magnetics and would have
no effect on the collection of alkali chlorides.
Table 8 indicates that the collection efficiency of magnetic
components was actually much higher than the total mass efficiency test
results. If the larger iron-bearing particles that are characteristic
of most sinter dusts had been included in the inlet, the collection
efficiency surely would have been even higher. But the fact remains
that the dust stream contained a concentration of non-magnetic dust
sufficient to prevent successful application of HGMF. The collection of
this fraction was improved only by operating under conditions conducive
to filter plugging, which would not be practical for commerical applica-
tions.
-------
TRANSIENT EMISSIONS DURING FILTER CLEANING
The Climet particle size analyzer was used to make one very reveal-
ing study of short-term transient emissions. With the pilot plant
operating on a 16-minute cycle on August 29, the Climet was set up to
sample from the outlet line with a 24-second counting interval synchro-
nized to begin when the flow switched from Magnet B to Magnet A. The
counting then continued through a complete cycle of operation.
The results are shown in Figure 36 for four particle size increments
ranging from the smallest to the largest sizes for which valid counts
could be obtained. The results show relatively constant emission rates
over most of the cycle except for sharp peaks corresponding to two
events—filter cleaning and flow switching. When the pressurized air is
released to clean one filter, it surges through the cyclone and exhausts
into the inlet line leading to the other filter. Figure 36 shows that
the surge of air either knocks large particles and agglomerates off the
active filter or carries large particles or agglomerates through the
cyclone and through the active filter. The former explanation is more
reasonable given the operating characteristics of cyclones and the fact
that fine particle emissions do not rise immediately after the pulse.
The emission puff subsides but is followed by a second puff of fine
particles when the flow is switched from one path to another. The
explanation of the second puff is more obscure. It is plausible that
the cleaning pulse leaves a cloud of fine particles in the idle flow
path that is then swept out when the flow is switched.
An estimate can be made of the importance of the two emission puffs
to total emissions by integrating the area under each histogram. (Note
that the relative concentration is plotted on a log scale so the peaks
are as much as ten times the normal emission rates.) This calculation
indicates that the puffs account for 16 percent of total emissions in
the 0.3-0.5 ym range, 47 percent in the 0.7-1 ym range, 43 percent in
the 2-3 ym range, and 51 percent in the 3-5 ym range.
83
-------
10<5
5
2
104
s
1 2
103
104
5
2
103
oc
H
Z
LU
o
o
o
LU
_J
o
H
cc
<
a.
LU
LU
a:
2
102
2
103
102
5
2
101
0.7-1.0 /
i j i i
i i i i
MAGNET A FILTERING
MAGNET B IDLE
_J L I I I I
D -J
O. LL
I I I I I
MAGNETS FILTERING
MAGNET A IDLE
_l l_
O. LL,
96 192 288 384 480 576 672
TIME IN CYCLE, SEC
768
864
960
Figure 36. Transient emission levels during a single cycle of operation,
84
-------
Further attempts to quantitatively evaluate the emission puffs were
stymied by operating problems with the Climet. The puffs were observed
qualitatively throughout the field program. An attempt was made to
reduce the magnitude of the problem by changing the cycle so that the
cleaning pulse occurred just after flow switching when the active
filter had little dust accumulation. This would also have given the
cloud in the idle path more time to settle, but the puffs persisted
despite these adjustments.
Had the problem been eliminated it would not have changed the
results at the sinter plant enough to make them acceptable, but the
design of the cleaning system should be changed before future work is
undertaken. The puff associated with the cleaning pulse could be
eliminated by replacing the cyclone with a surge chamber that is vented
through a single bag filter to the atmosphere or to the inlet pipe. The
size of the surge chamber would not be unreasonable. For example,
neglecting the surge capacity of the inter-connecting pipe, a surge
chamber nine times the volume of the pressurized air tank would dissi-
pate the back pressure to 10 percent of the original gauge pressure. In
the pilot plant this would have required a surge chamber with a volume
•5
of 2.5 m . If the explanation of the second pulse is correct, it could
be eliminated only by cleaning the filter outside of the normal flow
path. This suggests the use of a continuously cleaned HGMF, which is
discussed in the next subsection.
PROJECTED APPLICATIONS: ECONOMICS AND EFFECTIVENESS
The ultimate objective of EPA's support for the development of high-
gradient magnetic filtration is to determine whether the process can be
applied to one or more particulate emission control problems with result-
ing advantages over conventional control technology either in terms of
economics or collection efficiency. The work presented in this report
has increased the understanding of HGMF from a theoretical standpoint
and has demonstrated that the process can function reliably under
industrial conditions. What remains is to identify the most appropriate
candidate for application and to design the proper configuration of HGMF
equipment tailored to that application.
85
-------
Discussion of Potential Candidates for Application
Although the collection efficiency of the mobile pilot plant was
not satisfactory in the field tests with sinter dust, the previous
laboratory work and the chemical analysis of the dusts from the sinter
plant suggest that there are more promising opportunities for applica-
tion. For example, the magnetization of basic oxygen furnace dust is
roughly 20 times that of the inlet dust encountered in the field work,
and the gas stream reportedly does not contain, the undesirable alkali
chlorides that cannot be adequately controlled by HGMF. Of the other
processes listed in Table 2, the electric arc furnace has been shown in
laboratory tests to be a potential candidate. One sample of scarfer
fume has been obtained from an industrial source and determined to have
a specific magnetization of 32 emu/g at 3000 Oe, slightly higher than
the electric arc furnace dust tested. Blast furnace dust should have a
very high magnetization judging from the reduced iron content, but it is
also possible that the presence of discrete carbon particles could
create the same problem experienced in the sinter plant field work.
More information is needed on the size and composition of the dust
downstream of the inertial or gravitational dust catcher that is used
with most blast furnaces. Indeed, it is still possible that HGMF may be
applicable to some sinter plants. Steiner (1976) reports that the
characteristics of sinter dust vary widely from plant to plant. The
results of the field tests indicate that HGMF can collect a large per-
centage of the sintered metals and flux materials. If the concentration
of alkali chlorides in the gas stream is not significantly above the
allowable emission level for total mass, then HGMF may be able to meet
compliance codes.
Full-Scale Design Considerations
The cyclic HGMF design employed in the pilot plant must include a
redundancy in magnets to achieve continuous operation. This results in
a cost penalty. With dust concentrations significantly higher than the
0.5 g/Nm encountered in the field tests, it is also doubtful that the
cyclic version of the HGMF process can achieve operating practicality.
86
-------
In the lab pilot plant work it was demonstrated that with a highly
magnetic dust the HGMF filter could accumulate nearly twice its mass in
dust without a decrease in collection efficiency. But pressure drop
constraints and the need to insure effective filter cleaning would
probably limit the accumumulation to a more conservative 10 to 20 per-
cent of filter mass in most applications. In the latter stages of the
field work the accumulation during each cycle was limited to 1 to 2
percent of filter mass to insure that plugging did not occur. Table 9
illustrates the calculated range of allowable loading times for a filter
operating under typical conditions if the accumulation were limited to
10 percent of the filter mass. With the high dust concentrations that
are typical to some of the processes listed in Table 2, the loading time
is too short to apply a cyclic HGMF system practically.
Table 9. Illustration of filter loading times.
(F = 0.015, L = 0.20 m, and VQ = 8 m/s)
Dust Concentration, g/m Loading Time, seconds
0.5 580
1 290
2 145
5 58
10 29
In addition to high dust concentrations, most full-scale steel industry
processes involve very large gas flows. To take advantage of the economy
of scale, a cyclic HGMF system would employ a number of parallel filters
each much larger than that of the pilot plant. Unless a manifold system
were introduced inside each flow path to sub-divide the filter into more
87
-------
easily cleanable sections, the volume of the required cleaning air pulses
would be proportionately larger. Not only would the tank sizes increase,
the size of the release valves would increase making it much more diffi-
cult to achieve the required pulses of air.
Obviously an HGMF design that eliminates the magnet redundancy and
overcomes the loading and cleaning problems discussed above would be
advantageous in most applications. Several possible designs have been
suggested, and at least one is commerically available for liquid appli-
cations. The SALA-HGMS® Carousel Series 480 (SMI Bulletin No. D052119-
7610GB, Sala Magnetics, Inc., Cambridge, MA) is a continuous device that
incorporates four magnet heads and four cleaning stations mounted on a
rotating carousel (Figure 37). The magnet coils (not shown in the
figure) are split into a mirrored-saddle configuration to allow the
carousel to be rotated through the magnetized zone by a variable speed
drive. The carousel can be loaded with filter material to a depth of
0.22 m and an active radial width of 1.5 m. Each of the magnet heads
2
encloses an active face area of 3.4 m in the direction of fluid flow.
At superficial velocities of 5 to 10 m/s the total capacity of the unit
would be 245,000 to 490,000 m3/hr. With each of the magnet heads designed
to provide an applied field of 0.3 T, the power dissipation is 36 kW per
magnet plus a nominal amount for the carousel drive. The estimated
first-quarter 1978 price of the device was $900,000 including power
supplies and heat exchangers for indirect cooling of the magnets. The
total cooling water requirement is 7.2 m/hr (32 GPM).
The carousel design is one suitable solution to the problem of
scaling HGMF up to full-scale application on a high-concentration dust
stream. The magnets are used continuously. A rotational speed of 1/3
rpm would satisfy the shortest loading time requirement listed in Table
9. The cleaning stations could be set up to provide pulses of air to
relatively small increments of the filter at regular intervals that
would be timed so that the entire filter would be cleaned after passing
through each magnet. In practice, the carousel is already designed so
that the filter is subdivided radially into smaller increments by solid
vanes. Elastomer seals are attached to the top and bottom of each vane
ho
uu
-------
MAGNET
CO
DRIVE
CLEANING STATION
Figure 37. Schematic layout of the SALA-HGMS^480 Series Carousel.
-------
to provide a radial seal against leakage as the carousel passes through
the magnets. Similar seals are mounted to the inner and outer circum-
ferences of the carousel. The sealing system may require some modifica-
tion (or at least careful selection of the elastomer material) before
applications are proven on hot, dusty gas streams, but this is not
foreseen as a serious problem. Since most particulate emission control
equipment is installed at points where the process gas is under a
partial vacuum, any leakage would be inward and would help the seals to
be self-cleaning.
Efficiency and Economic Calculations
The design and cost of particulate emission control equipment must
be evaluated carefully for each individual application, but it is instruc-
tive to consider generalized capital and operating cost figures as a
basis for the evaluation of new technology. The capital investment for
an electrostatic precipitator or baghouse normally runs about $2.35 per
m /hr of gas flow ($4/ACFM), for the basic device itself. Auxiliary
equipment such as ducts and fans, structures, engineering, and construc-
tion will escalate the final cost considerably. For example, a news
item in the December 1978 edition of "Environmental Science and Technology"
reported the turnkey installation of four electrostatic precipitators to
control emissions from a basic oxygen furnace. The 510,000 m /hr (300,000
ACFM) system cost $5.2 million including equipment to pelletize the
collected dust, which translates to $10.20 per m3/hr ($17.33/ACFM). The
initial cost of scrubbers is lower than that of precipitators or baghouses,
but they usually require the addition of water pollution control equipment
that drives the installed cost up to a comparable level.
Energy requirements of conventional equipment also vary. Baghouses
on metallurgical processes normally operate with a pressure drop in the
range of 10 to 20 cm H20. This translates into an energy expenditure
of 1.7 to 3.4 kJ/m3 (1.1 to 2.2 hp/1000 ACFM), assuming 60 percent fan
efficiency. Bag cleaning requirements can add a significant increment.
Precipitators have much lower pressure drops (typically 1 to 2 cm FLO)
but require considerable power for energization of the electrodes.
90
-------
Their total energy requirement is on the order of 1.6 to 3.2 kJ/m (1 to
2 hp/1000 ACFM). Scrubbers normally require much greater energy input
to achieve acceptable collection efficiencies. Pressure drops as high
as 150 cm 1^0 are common, resulting in energy requirements of more than
25 kJ/m3 (15.8 hp/1000 ACFM).
The projected cost of particulate emission control by HGMF is depend-
ent on characteristics of the dust source, most notably dust size and
magnetization. In applications where the effects of gas velocity and
magnetic field are more readily apparent than they were in the field
work, trade-offs may be appropriate between these two variables as well
as filter depth and density in order to minimize costs. For example,
theoretically one can keep collection efficiency constant while reducing
filter density by increasing the applied field. This results in a
trade-off of fan power for magnet power, which may be desirable. If gas
velocity has no significant, effect on collection, it can be increased at
the expense of fan power in order to reduce capital costs.
The approximate capital cost and energy requirements of HGMF can be
illustrated by considering one operating condition that could be suitable
in a variety of applications. To take advantage of the continuous
design for which cost and power requirements are available, the operating
parameters listed in Table 10 are appropriate. The dust magnetizations
are based on the experience gained to date and should be regarded as
approximate. The predicted fractional collection efficiency for each of
the three dust categories is shown in Figure 38. The curves are based
on the mathematical model discussed in Section 5 and Appendix B. The
model was validated for the basic oxygen and electric arc furnace dusts
in the laboratory pilot plant, at least to the extent that the dusts
tested were representative of emissions from their source category. The
curve for a = 2.5 emu/g corresponds roughly to sinter dust, but the
model assumes the dust particles to be homogeneous in composition. In
comparison to the predicted curve, sinter field tests corresponding most
nearly to the conditions of Table 10 demonstrated a flatter curve with a
collection efficiency of about 80 to 90 percent for larger particles and
20 to 40 percent for the smallest particles. Note that the efficiency
91
-------
Table 10. HGMF operating parameters used for economic calculations.
Filter Material: AISI Type 430 coarse-grade steel wool
(equivalent cylindrical diameter =
200 ym)
Filter Depth: 0.20 m
Filter Density: 0.015
Applied Magnetic Field: 0.30 T
Superficial Gas Velocity: 8 m/s
Gas Temperature: 170°C 3
(density =0.8 kg/m ; viscosity =
2.4xlO"5 Pa-s)
Dust Magnetization: 50 emu/g (basic oxygen furnace, blast
furnace (?))
25 emu/g (electric arc furnace, scarfer^
2.5 emu/g (sinter machine)
Particle Size: 0.2 - 10 ym Stokes diameter
(as 0.4 - 20 ym aerodynamic diameter)
is very high for the strongly magnetic dust even down into the submicron
range. For lower magnetizations, lower efficiencies are indicated, but
not all applications require 99 percent collection to meet emission
regulations. The efficiency can be adjusted, if necessary, by varying
the operating parameters as discussed in Section 5.
Filter pressure drop can be estimated from the field test results.
Two valid pressure drop characterizations were run with clean, coarse
filters. Since the pressure taps enclosed a considerable part of the
piping network, the pressure drop was also characterized with no filter
in the canister. The results are reported in Table 11. As shown in
Figure 39 the AP data are correlated well with velocity squared as
suggested by the Burke-Plummer equation for turbulent flow in packed
columns (Bird, et al., 1960)
92
-------
35
U
2
LU
LLJ
g
o
01
o
u
99.9
99.8
99.5
99
98
95
90
80
70
60
50
40
30
20
10
5
i i i i
i i i i
0.2
0.5 1.0 2
STOKES DIAMETER, urn
10
Figure 38. Predicted HGMF collection efficiency for three dust
categories at typical operating conditions.
93
-------
Table 11. Pressure drop characterization of coarse steel wool.
Upstream
Pressure
kPa, absolute
No Filter 95.3
94.1
92.1
90.4
Filter # 6 95.4
(F = 0.015, L = 0.20 m) g4>4
93.6
92.1
90.4
Filter #8 97.1
(F = 0.0175, L = 0.30 mj gg 2
95.0
93.3
92.0
Upstream
Temoerature Flow* Pressure Drop
' °C m3/hr cm H20
103
101
93
85
101
99
99
95
100
101
94
91
86
72
2380
3090
4050
4880
2360
3090
3790
4590
5390
2480
3210
3770
4470
4950
2.1
3.7
5.6
8.5
7.5
16
20
30
42
18
27
41
61
75
Flow corrected from orifice temperature and pressure to upstream filter
temperature and pressure.
FL
AP =
s(l-F)'
where AP = pressure drop, Pa;
k = dimension!ess constant;
PC = fluid density, kg/m ;
V = superficial velocity, m/s;
F = filter density, dimensionless;
L = filter depth, m; and
s = wire radius, m.
94
-------
80,-
10 20 30 40 50 60 70 80 90
fFV0, kg/m-s
Figure 39. Pressure drop flow correlation for coarse steel wool
95
-------
In contrast, the extensive AP characterization conducted on medium-
grade steel wool in the laboratory pilot plant showed AP to be propor-
1 CQ
tional to v although the Reynolds number calculations indicated
turbulent flow conditions (see Appendix A). The data in Table 10 sug-
gest a stronger dependence on F and/or L than Equation (15) indicates
but the indication is uncertain because of the relatively small ranges
over which these parameters were varied. In the laboratory work, AP was
correlated with p and L. An observation from the field work is
consistent with the 1/s dependence shown in Equation (15). When coarse
steel wool (equivalent diameter = 195 ym) was substituted for medium
steel wool (equivalent diameter = 94 urn) in the long-term test period,
the clean-filter pressure drop decreased from 20 cm H20 to 9 cm H20 at the
same gas velocity.
Additional work is needed to better establish the pressure drop
correlation across the appropriate parametric ranges. For the conditions
illustrated in Table 10, however, the estimated clean-filter pressure
drop can be taken directly from the data on Filter #6 shown in Figure
39. Subtracting off the contribution of .the pilot plant piping, the
estimated flange-to-flange filter AP is 19 cm hLO. Allowing for a 10
percent increase at steady-state operating conditions, the fan energy
requirement would be 3.5 kj/m3 (2.2 hp/1000 ACFM). With the 8 m/s
velocity the Series 480 Carousel would accomodate a flow rate of slightly
•3
more than 390,000 m /hr (230,000 ACFM), yielding a magnet energy require-
ment of 1.3 kJ/m3 (0.8 hp/1000 ACFM). Based on the conditions used in
the field tests, a liberal estimate for the energy required by the
cleaning system is 0.8 kJ/m (0.5 hp/1000 ACFM). The combined total
energy requirement is above the generalized estimates for baghouses and
precipitators but well below typical scrubber energy requirements. At 8
m/s the estimated capital investment is $2.30 per m3/hr ($3.90/ACFM),
which is competitive with the conventional technologies.
It should be re-emphasized that the cost and energy estimates are
based on a single HGMF design and are compared to generalized figures on
competitive technologies. For specific applications other factors
should be taken into account. The coarse stainless steel wool is relatively
inexpensive and is quite durable in corrosive and hot environments.
96
-------
The potential for application of HGMF in hot, combustible gas streams
could lead to improved energy recovery from waste gases. Unlike many
wet scrubber systems the HGMF process will not turn an air pollution
control problem into a water pollution control problem. The high gas
velocities and compact equipment that are characteristic of HGMF greatly
reduce space requirements in comparison to baghouses and precipitators,
which should produce a proportionate reduction in installed costs
related to duct work, structural supports, and erection. These attrac-
tive features strongly support the continued development of the HGMF
process with emphasis on field pilot plant tests to identify the most
appropriate points of application and to demonstrate the application
of a continuous filtration-continuous cleaning design.
97
-------
SECTION 9
REFERENCES
Bird, R.B., W.E. Stewart, and E.N. Lightfoot, Transport Phenomena, John
Wiley & Sons, Inc., New York (1960).
Bozorth, R.M., Ferromagnetism, D. Van Nostrand Company, Inc., New York
(1951).
Clarkson, C.J., D.R. Kelland, and T.B. King, "Model for Calculation of
Capture Radii of High Gradient Magnetic-Separator at Moderate
Reynolds Numbers," IEEE Trans. Magn., Mag-12, 901 (1976).
Gushing, K.M., 6.E. Lacey, J.D. McCain, and W.B. Smith, Particulate
Sizing Techniques for Control Device Evaluation: Cascade Impactor
Calibrations, PB 262-849/AS, National Technical Information Service,
Springfield, VA (1976).
DeLatour, C., and H.H. Kolm, "Magnetic Separation in Water Pollution
Control-II," IEEE Trans. Magn., Mag-11, 1570 (1975).
Dulaney, E.L., Development Document for Effluent Limitations Guidelines
and New Source Performance Standards for the Steel Making Segment
of the Iron and Steel Manufacturing Point Source Category, EPA-440/
l-74-024a, National Technical Information Service, Springfield, VA
(1974).
Ergun, S., and E.H. Bean, "Magnetic Separation of Pyrite from Coals,"
Report of Investigation No. 2718, U.S. Bureau of Mines, Pittsburgh,
PA (1968).
Fuchs, N.A., The Mechanics of Aerosols, Pergamon Press, New York (1964).
Gillespie, T., "On the Adhesion of Drops and Particles on Impact at
Solid Surfaces. I.," J. Colloid Sci., 1_0, 266 (1955).
Gooding, C.H. , "High Gradient Magnetic Filtration of Fine Particles from
a Gas Stream," Ph.D. Thesis, North Carolina State University,
Raleigh, NC (1979).
Gooding, C.H., T.W. Sigmon and L.K. Monteith, Application of High-Gradient
Magnetic Separation to Fine Particle Control, PB 276-633/AS,
National Technical Information Service, Springfield, VA (1977).
Hardison, L.C. , and C.A. Greathouse, Air Pollution Control Technology and
Costs in Nine Selected Areas, PB 222-746, National Technical Informa-
tion Service, Springfield, VA (1972).
90
-------
Harland, J.R., L. Nilsson, and M. Wallin, "Pilot-Scale High Gradient
Magnetic Filtration of Steel Mill Wastewater," IEEE Trans. Magn.,
Mag-12. 904 (1976).
Harris, D.B. , Procedures for Cascade Impactor Calibration and Operation
in Process Streams, PB 263-623/AS, National Technical Information
Service, Springfield, VA (1977).
lannicelli, J. , "New Developments in Magnetic Separation," IEEE Trans.
Magn.. Mag-12. 436 (1976).
IBM Application Program System/360 Scientific Subroutine Package,
Version III Programmer's Manual, Program No. 360A-CM-03X, IBM
Corp., White Plains, NY, 5th ed. (1970).
Jaasund, S.A., and M.R. Mazer, "The Application of Wet Electrostatic
Precipitators for the Control of Emissions from Three Metallurgi-
cal Processes," Proceedings: Particulate Collection Problems
Using ESP's in the Metallurgical Industry. PB 274-017/AS, National
Technical Information Service, Springfield, VA (1977).
Jackson, J.D., Classical Electrodynamics. John Wiley & Sons, Inc.,
New York (1975).
Kelland, D.R., "High Gradient Magnetic Separation Applied to Mineral
Beneficiation," IEEE Trans. Magn.. Mag-9. 307 (1973).
Kelland, D.R., and E. Maxwell, "Oxidized Taconite Beneficiation by
Continuous High Gradient Magnetic Separation," IEEE Trans. Magn.,
Mag-11, 1582 (1975).
Kolm, H.H., "The Large-Scale Manipulation of Small Particles," IEEE
Trans. Magn.. Mag-11. 1567 (1975).
Kolm, H.H., J.A. Oberteuffer, and D.R. Kelland, "High Gradient Magnetic
Separation," Sci. Am., 223. 47 (Nov. 1975).
Krupp, H., "Particle Adhesion," Advances in Colloid and Interface
Science," 1, 111 (1967).
Lawless, P.A., Analysis of Cascade Impactor Data for Calculating
Particle Penetration. PB 288-649/AS, National Technical Information
Service, Springfield, VA (1978).
Lawson, W.F., "The Dynamics of Paramagnetic Particles Near a Magnetized
Wire," M.S. Thesis, West Virginia University, Morgantown, WV
(1976).
Lawson, W.F., W.H. Simons, and R.P. Treat, "The Dynamics of a Particle
Attracted by a Magnetized Wire," J. Appl. Phys., 48_, 3213 (1977).
99
-------
Liu, Y.A., G.E. Crow, C.J. Lin, and D.L. Vives, "A Pilot-Scale Study of
High Gradient Magnetic Desulfurization of Solvent Refined Coal
(SRC)," IEEE Trans. Magn.. Mag-14, (1978).
Liu, Y.A., Editor, Industrial Applications of Magnetic Separation,
Proceedings of Engineering Foundation Conference, Published by
IEEE, Inc., 347 E. 47th St., New York (1979).
Loeffler, F., "Adhesion Probability in Fibre Filters," Clean Air, 8(4),
75 (1974).
Lyman, T., Editor, Metals Handbook, Vol. I Properties and Selection of
Materials, American Society for Metals, Metals Park, OH, 8th ed.
(1961).
Maxwell, E., I.S. Jacobs, and L.M. Levinson, Magnetic Separation of
Mineral Matter from Coal Liquids, EPRI AF-508, Electric Power
Research Institute, Palo Alto, CA (1977).
Maxwell, E., and D.R. Kelland, "High Gradient Magnetic Separation in
Coal Desulfurization," IEEE Trans. Magn., Mag-14, 482 (1978).
Maxwell, E., D.R. Kelland, and I.Y. Akoto, "High Gradient Magnetic
Separation of Mineral Particulates from Solvent Refined Coal,"
IEEE Trans. Magn., Mag-12, 507 (1976).
Melville, D., F. Paul, and S. Roath, "High Gradient Magnetic Separation
of Red Cells from Whole Blood," IEEE Trans. Magn., Mag-11, 1701
(1975).
Mitchell, R., G. Bitton, and J.A. Oberteuffer, "High Gradient Magnetic
Filtration of Magnetic and Non-Magnetic Contaminants from Water,"
Separation and Purification Methods, 4_, 267 (1975).
Murray, H.H., "Beneficiation of Selected Industrial Minerals and Coal
by High Intensity Magnetic Separation," IEEE Trans. Maqn., Mag-12,
498 (1976).
Oberteuffer, J.A., "Magnetic Separation: A Review of Principles,
Devices, and Applications," IEEE Trans. Magn., Mag-10, 223 (1974).
Oberteuffer, J.A., I. Wechsler, P.G. Marston, and M.J. McNallan, "High
Gradient Magnetic Filtration of Steel Mill Process and Waste
Waters," IEEE Trans. Magn., Mag-11, 1591 (1975).
Oder, R.R., "High Gradient Magnetic Separation Theory and Applications,'
IEEE Trans. Magn., Mag-12, 428 (1976).
100
-------
Parson, T., Editor, Industrial Process Profiles for Environmental Use:
Chapter 24. The Iron and Steel Industry, EPA-600/2-74-023x,
National Technical Information Service, Springfield, VA (1977).
Petrakis, L., and P.P. Ahner, "High Gradient Magnetic Separations in
Water Effluents," IEEE Trans. Magn., Mag-14, 491 (1978).
Steiner, B.A., "Air Pollution Control in the Iron and Steel Industry,"
International Metals Review, No. 209, 171 (1976).
Stenhouse, J.I.T., and D.C. Freshwater, "Particle Adhesion in Fibrous
Air Filters," Trans. Inst. Chem. Engrs., 5±, 95 (1976).
Trindale, S.C., and H.H. Kolm, "Magnetic Desulfurization of Coal,"
IEEE Trans. Magn.. Mag-9. 310 (1973).
Visser, J., "on Hamaker Constants: A Comparison Between Hamaker Constants
and Lifshitz - Van der Waals Constants," Advances in Colloid and
Interface Science, 3., 331 (1972).
Vives, D.L., L.J. Hirth, and W.H. Summerlin, "Direct Reduction and
Magnetic Beneficiation of Alabama Brown Ore with Lignite," IEEE
Trans. Magn., Mag-12, 490 (1976).
Watson, J.H.P., "Magnetic Filtration," J. Appl. Phys., 44_, 4209 (1973).
Whitehead, C. , "Design and Operating Experience with Electrostatic
Precipitators on Electric Arc Furnaces," Proceedings: Particulate
Collection Problems Using ESP's in the Metallurgical Industry,
PB 274-017/AS, National Technical Information Service, Springfield,
VA (1977).
Whitesides, G.M. , C.L. Hill, and J. Brunie, "Magnetic Filtration of
Small Heterogeneous Catalyst Particles. Preparation of Ferri-
magnetic Catalyst Supports," Ind. Eng. Chem.. Process Des. Dev..
15, 226 (1976).
Yadidia, R. , A. Abeliovich, and G. Belfort, "Algae Removal by High
Gradient Magnetic Filtration," Envr. Sci. and Tech., 11, 913
(1977).
101
-------
APPENDIX A
TABULATION OF EXPERIMENTAL CONDITIONS AND RESULTS FROM
LABORATORY PILOT PLANT TESTS
The ranges of experimental parameters investigated during the
laboratory pilot plant work are given in Table A-l.
Table A-l. Ranges of experimental parameters.
Dust Source:
Filter Material:
Filter Depth:
Filter Packing Density:
Applied Magnetic Field:
Superficial Air Velocity:
Air Temperature:
Filter Inlet Pressure:
Particle Diameter (Stokes):
basic oxygen furnace, electric arc furnace
medium-grade type 430 stainless steel wool
0.075 - 0.30 m
0.0050 - 0.0100
0 - 0.6 T
4.9 - 11.9 m/s
24 - 46°C
747 - 764 mm Hg
0.22 - 8.7 ym
The equivalent cylindrical diameter of the steel wool wires was calcu-
lated to be 108 ym by separating 50 randomly chosen strands of the material
and measuring the total length and mass of the strands. Microscopic
examination of the strands showed some to be roughly cylindrical while
others had a more prismatic or ribbon-like appearance. Before each
differently constructed filter was used, flow tests were conducted to
determine the clean filter pressure drop at several velocities. The
data are reported in Table A-2. A multiple regression analysis was run
of the 41 observations to obtain the correlation
102
-------
Table A-2. Experimental data on clean filter pressure drop.
Filter depth,
m
0.15
0.15
0.15
0.15
0.15
0.30
0.30
0.30
0.15
0.15
0.15
0.15
0.15
0.15
0.225
0.225
0.225
0.225
0.225
0.225
0.225
0.225
0.225
0.225
0.225
0.225
0.15
0.15
0.15
0.15
0.15
0.15
0.15
0.15
0.15
0.15
0.30
0.30
0.30
0.30
0.30
Packing
density
0.0100
0.0100
0.0100
0.0100
0.0100
0.0050
0.0050
0.0050
0.0050
0.0050
0.0050
0.0050
0.0050
0.0050
0.0080
0.0080
0.0080
0.0080
0.0080
0.0080
0.0064
0.0064
0.0064
0.0064
0.0064
0.0064
0.0080
0.0080
0.0080
0.0080
0.0080
0.0080
0.0080
0.0080
0.0080
0.0080
0.0060
0.0060
0.0060
0.0060
0.0060
Air velocity,
m/s
5.0
7.6
8.3
10.1
10.4
5.0
8.1
10.5
4.9
8.1
10.7
4.8
7.7
10.7
4.6
6.5
8.1
9.2
10.3
10.9
4.5
6.6
7.8
9.1
10.2
10.8
5.1
7.5
9.1
10.5
11.1
6.8
8.0
9.0
10.1
10.7
4.3
6.2
7.6
9.0
9.8
Pressure drop,
kPa
0.80
1.22
1.72
2.14
2.22
0.62
1.44
2.17
0.30
0.67
1.10
0.27
0.62
1.12
0.67
1.22
1.69
2.14
2.47
2.74
0.52
1.00
1.39
1.84
2.24
2.39
0.62
1.15
1.74
2.19
2.37
0.85
1.22
1.54
1.92
2.14
0.70
1.25
1.74
2.22
2.72
103
-------
AP=62 F1.15V1.68 f (A.1}
where AP = filter pressure drop, kPa;
(1 kPa = 10 cm H20)
L = filter depth, m;
F = packing density, dimension!ess; and
V = air velocity, m/s.
The multiple regression coefficient for Equation (A-l) was 0.990.
In contrast to Equation (A-l) the Blake-Kozeny equation for laminar
flow in packed columns (Bird, et al., 1960) has the form
^ a F2V (A-2)
and the Burke-Plummer equation for highly turbulent flow is packed columns
(Bird, et al., 1960) has the form
^[ a FV2 (A-3)
All of the experimental data in Table A-2 meet the Reynolds number criterion
of the Burke-Plummer equation,
3sPfV
—~ > 1000 , (A-4)
nr
where s = wire radius, m;
P.C = air density, kg/m3; and
n = air viscosity, Pa-s;
but Equation (A-l) indicates that the flow was actually in the transition
region between laminar and fully-developed turbulent flow. Confidence
intervals were calculated on the two exponents obtained from the
regression. A 95 percent confidence interval about the F exponent
excluded 1.0, and a 95 percent confidence interval about the V exponent
excluded 2.0. A regression analysis was also run on the data allowing
the exponent on L to vary from unity. The regression exponent was 1.03,
but it could not meet even an 80 percent confidence interval test, and the
multiple regression coefficient was not improved.
104
-------
Table A-3 lists the experimental conditions and results of the indivi-
dual runs made in the lab pilot plant with BOF and EAF dusts. The
experiments conducted with BOF dust are tabulated first, followed by the
EAF dust experiments. Within each dust grouping the experiments are
tabulated in chronological order. The data from run numbers 04191
through 07181 were reported previously (Gooding, et al., 1977), but are
included here because they were used in the mathematical correlations
developed as a part of this work. The notation of the operating
parameters and results is as follows:
B = applied magnetic flux density, T
V = superficial air velocity, m/s
F = volumetric filter packing density, dimension!ess
L = filter depth, m
T = air temperature, °C
D = particle diameter, ym
Eff = collection efficiency, %.
The particle diameter reported in Table A-3 is the Stokes diameter,
which is defined as the diameter of a sphere (with the same density as the
actual particle) that would behave in an impactor the same as does the
actual particle. The mass median Stokes diameter of the inlet dust ranged
from approximately 2 to 4 ym. Photomicrographs of each of the dusts showed
that the primary particles were predominantly spheres ranging from 0.1 to
1 ym in diameter (Gooding, 1979), and larger "particles" were actually
made up of agglomerates. The particle densities used in the size calcu-
lations were determined by pressurized-air pycnometer measurements to be
o 3
4.47 g/cm for the BOF dust and 4.61 g/cm for the EAF dust. To convert
the Stokes diameter to aerodynamic diameter, one must multiply by the
square root of the particle density.
In all of the lab pilot plant experiments the impactor sample rates
were controlled carefully to provide equal stage cut points on the inlet
and outlet sides of the HGMF filter. This allowed calculation of the
fractional efficiency from a direct ratio of the fractional particle
concentrations calculated from corresponding stage weights.
105
-------
Table A-3. Conditions and results of laboratory pilot-plant
experiments.
Run no. 04191
BOF dust
B = 0 T
V = 5.6 m/s
F = 0.0050
L = 0.150 m
T = 31°C
Run no. 04272
BOF dust
B = 0.050T
V = 8.1 m/s
F = 0.0050
L = 0.150 m
T = 32°C
Run no. 04281
BOF dust
B = 0.050 T
V = 6.7 m/s
F = 0.0050
L = 0.150 m
T = 29°C
Run no. 05031
BOF dust
B = 0.100 T
V = 8.5 m/s
F = 0.0050
L = 0.150 m
T = 31°C
Run no. 05032
BOF dust
B = 0.100 T
V = 5.6 m/s
F = 0.0050
L = 0.150
T = 35°C
D, ym
0.23
0.37
0.74
1.7
3.3
7.9
D, ym
0.23
0.37
0.74
1.7
3.3
7.9
D, ym
0.41
0.82
1.8
3.6
8.3
— — —
D, ym
0.36
0.72
1.6
7.9
—
-— —
D, ym
0.23
0.36
0.72
1.6
3.3
7.9
Eff, %
7.2
39.6
80.4
93.3
96.6
91.1
Eff, %
7.3
35.6
76.9
74.1
32.5
47.8
Eff, %
33.1
82.2
89.9
57.8
50.0
---
Eff, %
62.3
84.0
87.1
81.1
—
-— -
Eff, %
42.8
51.6
86.6
92.6
85.4
62.4
Run no. 05041
BOF dust
B = 0.100 T
V = 6.9 m/s
F = 0.0050
L = 0.150 m
T = 32°C
Run no. 05042
BOF dust
B = 0.100 T
V = 10.2 m/s
F = 0.0050
L = 0.15 m
T = 37°C
Run no. 05121
BOF dust
B = 0.200 T
V = 8.0 m/s
F = 0.005
L = 0.150 m
T = 33°C
Run no. 05181
BOF dust
B = 0 T
V = 9.7 m/s
F = 0.0100
L = 0.150 m
T = 30°C
Run no. 05182
BOF dust
B = 0 T
V = 5.9 m/s
F = 0.0010
L = 0.150 m
T = 39°C
D, ym
0.25
0.40
0.79
1.8
3.6
8.2
D, ym
0.43
0.85
1.9
3.8
8.5
—
D, ym
0.23
0.37
0.73
1.7
3.3
7.9
D, ym
0.22
0.36
0.71
1.6
3.2
—
D, ym
0.23
0.36
0.73
1.6
3.3
7.9
Eff, %
22.5
61.4
81.7
83.8
51.5
46.1
Eff, %
76.4
86.6
75.5
74.6
87.5
—
Eff, %
49.0
82.5
96.3
93.5
96.9
96.3
Eff, %
59.8
84.0
95.2
94.8
94.2
—
Eff, %
16.9
39.2
87.5
94.9
92.1
93.1
106-
-------
Table A-3 (continued).
Run no. 05191
BOF dust
B = 0 T
V = 7.4 m/s
F = 0.0100
L = 0.150 m
T = 33°C
Run no. 05201
BOF dust
B = 0.050 T
V = 7.8 m/s
F = 0.0100
L = 0.150 m
T = 33°C
Run no. 05252
BOF dust
B = 0 T
V = 9.2 m/s
F = 0.0050
L = 0.300 m
T = 37°C
Run no. 05261
BOF dust
B = 0.050 T
V = 7.5 m/s
F = 0.0050
L = 0.300 m
T = 31°C
Run no. 05272
BOF dust
B = 0.050 T
V = 6.6 m/s
F = 0.0050
L = 0.300 m
T = 36°C
D, urn
0.41
0.81
1.8
3.6
8.3
---
D, ym
0.36
0.73
1.6
3.3
7.9
---
D, ym
0.22
0.35
0.71
1.6
3.2
7.8
D, ym
0.42
0.82
1.8
3.7
8.4
---
D, ym
0.36
0.72
1.6
3.3
7.9
—
Eff, %
84.3
97.2
98.9
97.1
96.8
---
Eff, %
75.1
99.0
99.0
98.6
97.7
—
Eff, %
16.0
44.5
73.1
67.2
32.7
13.3
Eff, %
56.2
93.1
91.2
84.4
79.4
—
Eff, %
46.5
92.6
96.7
93.6
87.5
—
Run no. 06021
BOF dust
B = 0.100 T
V = 10.3 m/s
F = 0.0050
L = 0.150 m
T = 32°C
Run no. 06022
BOF dust
B = 0.100 T
V = 6.8 m/s
F = 0.0050
L = 0.150 m
T = 37°C
Run no. 06031
BOF dust
B = 0.100 T
V = 6.8 m/s
F = 0.0050
L = 0.150 m
T = 30°C
Run no. 06032
BOF dust
B = 0.100 T
V = 7.1 m/s
F = 0.0050
L = 0.150 m
T = 37°C
Run no. 06281
BOF dust
B = 0.400 T
V = 8.2 m/s
F = 0.0050
L = 0.150 m
T = 35°C
D, ym
0.28
0.28
2.0
3.9
8.6
—
D, ym
0.41
0.81
1.8
3.6
8.3
—
D, ym
0.42
0.83
1.9
3.7
8.4
—
D, ym
0.41
0.81
1.8
3.6
8.3
---
D, pm
0.23
0.36
0.72
1.6
3.3
7.9
Eff, %
41.5
85.5
91.9
89.9
76.1
—
Eff, %
55.0
92.7
96.4-
94.0
85.7
—
Eff, %
63.3
92.3
95.8
90.0
97.3
—
Eff, %
66.5
94.6
98.1
91.3
93.6
—
Eff, %
66.6
70.1
96.4
98.9
99.2
99.4
107
-------
Table A-3 (continued).
Run no. 10261
BOF dust
B = 0.075 T
V = 5.6 m/s
F = 0.0100
L = 0.150 m
T = 26°C
Run no. 10281
BOF dust
B = 0.90 T
V = 7.7 m/s
F = 0.0100
L = 0.150 m
T = 24°C
Run no. 11011
BOF dust
B = 0.100 T
V = 10.1 m/s
F = 0.0100
L = 0.150 m
T = 27°C
Run no. 11021
BOF dust
B = 0.075 T
V = 5.0 m/s
F = 0.0050
L = 0.300 m
T = 27°C
Run no. 11032
BOF dust
B = 0.100 T
V = 10.5 m/s
F = 0.0050
L = 0.300 m
T = 34°C
D, urn
0.35
0.69
1.6
3.1
7.7
--
D, y
0.38
0.75
1.7
3.4
8.0
—
D, ym
0.30
0.46
0.90
2.0
8.7
— — —
D, um
0.36
0.71
1.6
3.2
7.8
—
D, ym
0.42
0.83
1.9
3.7
8.4
—
Eff, %
73.8
97.3
99.7
99.2
98.1
—
Eff, %
74.1
94.0
95.2
93.3
96.8
—
Eff, %
52.2
84.0
93.1
90.4
90.6
- —
Eff, %
74.5
99.0
99.9
99.8
97.3
—
Eff, %
93.0
99.3
96.4
95.0
93.8
—
Run no. 11211
BOF dust
B = 0 T
V = 4.9 m/s
F = 0.0050
L = 0.150
T = 29°C
Run no. 11221
BOF dust
B = 0 T
V = 8.1 m/s
F = 0.0050
L = 0.150 m
T = 32°C
Run no. 11231
BOF dust
B = 0.100 T
V = 4.9 m/s
F = 0.0050
L = 0.150 m
T = 30°C
Run no. 11281
BOF dust
B = 0.100 T
V = 8.1 m/s
F = 0.0050
L = 0.150
T = 28°C
Run no. 11282
BOF dust
B = 0.100 T
V = 10.9 m/s
F = 0.0050
L = 0.150 m
T = 37°C
D, ym
0.23
0.36
0.72
1.6
3.3
___
D, ym
0.24
0.38
0.75
1.7
3.4
8.0
D, ym
0.36
0.72
1.6
3.2
7.8
—
D, ym
0.70
1.6
3.2
7.7
___
—
D, ym
0.42
0.83
1.9
3.7
8.4
___
Eff, %
15.7
1.5
38.7
80.9
91.9
___
17-f-P V
L. r T , 10
1.0
4.7
8.8
38.3
79.2
92.9
Eff, %
38.8
80.2
96.4
99.5
88.1
—
Eff, %
92.3
95.7
93.1
96.1
—
Eff, %
55.9
72.5
69.6
39.7
57.9
108
-------
Table A-3 (continued).
Run no. 11291
BOF dust
B = 0.100 T
V = 10.8 m/s
F = 0.0050
L = 0.150 m
T = 33°C
Run no. 11292
BOF dust
B = 0.200 T
V = 5.4 m/s
F = 0.0050
L = 0.150 m
T = 38°C
Run no. 11301
BOF dust
B = 0.200 T
V = 8.5 m/s
F = 0.0050
L = 0.15 m
T = 34°C
Run no. 11302
BOF dust
B = 0.200 T
V = 11.6 m/s
F = 0.0050
L = 0.150 m
T = 40°C
Run no. 12011
BOF dust
B = 0.400 T
V = 5.4 m/s
F = 0.0050
L = 0.150 m
T = 34°C
D, ym
0.42
0.83
1.9
3.7
8.4
— — —
D, ym
0.36
0.72
1.6
3.2
7.8
— — —
D, ym
0.37
0.74
1.7
3.3
8.0
— — —
D, ym
0.42
0.84
1.9
3.7
8.4
— — —
D, ym
0.36
0.71
1.6
3.2
7.8
—
Eff, %
67.6
70.3
66.0
46.9
42.4
---
Eff, %
38.4
83.7
97.4
97.7
82.8
_ —
Eff, %
67.2
92.8
98.0
95.8
85.3
— -
Eff, %
66.0
92.1
91.1
77.8
91.4
— — —
Eff, %
48.4
88.7
98.0
99.0
93.2
—
Run no. 12012
BOF dust
B = 0.400 T
V = 8.9 m/s
F = 0.0050
L = 0.150 m
T = 41°C
Run no. 12021
BOF dust
B = 0.400 T
V = 8.6 m/s
F = 0.0050
L = 0.150 m
T = 31 °C
Run no. 12051
BOF dust
B = 0.400 T
V = 11.5 m/s
F = 0.0050
L = 0.150 m
T = 35°C
Run no. 12071
BOF dust
B = 0.600 T
V = 5.6 m/s
F = 0.0050
L = 0.150 m
T = 28°C
Run no. 12072
BOF dust
B = 0.600 T
V = 8.5 m/s
F = 0.0050
L = 0.150 m
T = 36°C
D, yin
0.23
0.37
0.74
1.7
3.4
8.0
D, ym
0.37
0.74
1.7
3.3
7.9
---
D, ym
0.42
0.83
1.9
3.7
8.4
-— —
D, ym
0.36
0.71
1.6
3.2
7.8
-__
D, ym
0.37
0.74
1.7
3.4
8.0
—
Eff, %
46.6
71.0
93.7
98.2
97.6
94.4
Eff, %
74.5
95.8
99.4
98.8
91.8
—
Eff, %
70.3
93.6
93.0
84.6
89.3
___
Eff, %
51.5
90.7
98.8
99.6
95.7
-— —
Eff, %
69.3
95.3
98.9
98.4
97.4
—
109
-------
Table A-3 (continued).
Run no. 12081
BOF dust
B = 0.600 T
V = 11.9 m/s
F = 0.0050
L = 0.150 m
T = 35°C
Run no. 01051
BOF dust
B = 0.200 T
V = 7.9 m/s
F = 0.0050
L = 0.075 m
T = 29°C
Run no. 01091
BOF dust
B = 0.200 T
V = 11.1 m/s
F = 0.0050
L = 0.075 m
T = 29°C
Run no. 011092
BOF dust
B = 0.200 T
V = 5.9 m/s
F = 0.0050
L = 0.075 m
T = 32°C
Run no. 01111
BOF dust
B = 0.200 T
V = 7.7 m/s
F = 0.0050
L = 0.150 m
T = 27°C
D, ym
0.43
0.84
1.9
3.8
8.4
—
D, ym
0.36
0.72
1.6
3.2
7.8
—
D, ym
0.41
0.81
1.8
3.6
8.3
—
D, ym
0.35
0.70
1.6
3.2
7.8
— __
D, ym
0.37
0.73
1.7
3.3
7.9
—
Eff, %
77.6
94.8
96.6
94.1
97.8
— -—
Eff, %
45.5
73.2
80.7
77.2
81.5
•---
Eff, %
51.4
73.0
66.0
51.2
51.6
—
Eff, %
26.8
70.6
91.3
97.2
92.5
~ — —
Eff, %
64.6
95.0
99.0
98.1
96.9
—
Run no. 01112
BOF dust
B = 0.200 T
V = 10.7 m/s
F = 0.0050
L = 0.150 m
T = 34°C
Run no. 01121
BOF dust
B = 0.200 T
V = 7.2 m/s
F = 0.0050
L = 0.150 m
T = 29°C
Run no. 02091
BOF dust
B = 0.300 T
V = 10.0 m/s
F = 0.0080
L = 0.225 m
T = 31°C
Run no. 02101
BOF dust
B = 0.300 T
V = 10.0 m/s
F = 0.0080
L = 0.225 m
T = 30°C
Run no. 02102
BOF dust
B = 0.300 T
V = 9.8 m/s
F = 0.0080
L = 0.225 m
T = 38°C
D, ym
0.42
0.82
1.8
3.7
8.4
— — «
D, ym
0.37
0.73
1.7
3.3
7.9
—
D, ym
0.27
0.42
0.83
1.9
3.7
8.4
D, ym
0.27
0.42
0.83
1.9
3.7
8.4
D, ym
0.26
0.42
0.82
1.9
3.7
8.4
Eff, %
72.5
89.6
90.7
83.8
89.0
— •«• —
Eff, %
56.6
92.0
98.1
97.7
98.4
—
Eff, %
86.1
96.8
99.8
99.4
98.4
98.3
Eff, %
86.8
98.9
99.7
99.2
99.6
99.6
Eff, %
79.4
98.9
99.8
99.5
99.3
99.0
no
-------
Table A-3 (continued).
Run no. 02211
BOF dust
B = 0.300 T
V = 7.9 m/s
F = 0.0064
L = 0.225 m
T = 30°C
Run no. 02231
BOF dust
B = 0.300 T
V = 8.0 m/s
F = 0.0064
L = 0.225 m
T = 25°C
Run no. 02232
BOF dust
B = 0.300 T
V = 8.0 m/s
F = 0.0064
L = 0.225 m
T = 28°C
Run no. 02272
BOF dust
B = 0 T
V = 8.8 m/s
F = 0.0080
L = 0.150 m
T = 34°C
Run no. 02281
BOF dust
B = 0 T
V = 11.0 m/s
F = 0.0080
L = 0.150 m
T = 32°C
D, ym
0.23
0.36
0.72
1.6
3.3
7.9
D, ym
0.23
0.36
0.72
1.6
3.3
7.9
D, ym
0.23
0.36
0.72
1.6
3.3
7.9
D, ym
0.23
0.36
0.72
1.6
3.3
7.9
D, ym
0.27
0.45
0.85
1.9
3.8
—
Eff, %
31.2
99.6
99.7
99.8
99.8
99.7
Eff, %
78.7
93.6
98.3
99.1
99.2
99.0
Eff, %
30.1
94.4
99.8
99.7
99.3
99.6
Eff, %
0.1
64.0
87.3
83.8
71.7
61.2
Eff, %
51.6
69.3
83.7
73.5
66.6
—
Run no. 02282
BOF dust
B = 0 T
V = 8.5 m/s
F = 0.0080
L = 0.150 m
T = 34°C
Run no. 03011
BOF dust
B = 0 T
V = 7.0 m/s
F = 0.0080
L = 0.150 m
T = 31°C
Run no. 03021
BOF dust
B = 0.300 T
V = 8.7 m/s
F = 0.0080
L = 0.150 m
T = 29°C
Run no. 03022
BOF dust
B = 0.300 T
V = 6.9 m/s
F = 0.0080
L = 0.150 m
T = 36°C
Run no. 03211
BOF dust
B = 0.300 T
V = 7.9 m/s
F = 0.0080
L = 0.510 m
T = 46°C
D, ym
0.37
0.73
1.7
3.3
7.9
—
D, ym
0.43
0.85
1.9
3.8
8.5
—
D, ym
0.23
0.36
0.72
1.6
3.3
7.9
D, ym
0.43
0.85
1.9
3.8
8.5
—
D, ym
0.36
0.72
1.6
3.3
7.9
___
Eff, %
80.3
84.4
81.5
71.5
59.8
—
Eff, %
47.4
84.3
85.1
76.6
59.3
—
Eff, %
60.3
90.2
98.6
99.4
98.0
97.8
Eff, %
99.3
99.7
99.5
98.1
98.2
—
Eff, %
96.6
89.9
96.1
95.9
86.0
_-_
111
-------
Table A-3 (continued).
Run no. 03221
BOF dust
B = 0.300 T
V = 7.5 m/s
F = 0.0080
L = 0.150 m
T = 29°C
Run no. 03222
BOF dust
B = 0.300 T
V = 9.5 m/s
F = 0.0080
L = 0.150 m
T = 34°C
Run no. 07141
EAF dust
B = 0.200 T
V = 7.6 m/s
F = 0.0050
L = 0.150 m
T = 36°C
Run no. 07151
EAF dust
B = 0.200 T
V = 10.8 m/s
F = 0.0050
L = 0.150 m
T = 33°C
Run no. 07181
EAF dust
B = 0.400 T
V = 8.2 m/s
F = 0.0050
L = 0.150 m
T = 37°C
D, ym
0.37
0.73
1.6
3.3
7.9
—
D, ym
0.28
0.43
0,85
1.9
3.8
8.5
D, ym
0.39
0.77
1.8
3.5
8.1
___
D, ym
0.41
0.81
1.8
3.6
8.3
_ _-.
D, ym
0.36
0.71
1.6
3.2
7.8
—
Eff, %
96.6
93.5
96.8
99.0
99.2
—
Eff, %
62.6
96.7
98.4
97.7
97.1
98.3
Eff, %
48.8
87.6
96.1
97.9
73.2
—
Eff, %
45.4
69.8
74.0
74.1
65.6
-— —
Eff, %
47.0
86.1
94.6
96.3
85.2
—
Run no. 07122
EAF dust
B = 0.100 T
V = 7.9 m/s
F = 0.0050
L = 0.150 m
T = 41 °C
Run no. 11072
EAF dust
B = 0.400 T
V = 8.1 m/s
F = 0.0050
L = 0.300 m
T = 38°C
Run no. 11081
EAF dust
B = 0.200 T
V = 5.6 m/s
F = 0.0100
L = 0.150 m
T = 31°C
Run no. 11091
EAF dust
B = 0.375 T
V = 8.9 m/s
F = 0.0100
L = 0.150 m
T = 31°C
Run no. 11092
EAF dust
B = 0.400 T
V = 11.6 m/s
F = 0.0100
L = 0.150 m
T = 37°C
D, ym
0.39
0.78
1.8
3.5
8.2
— — —
D, ym
0.23
0.36
0.72
1.6
3.3
7.9
D, ym
0.22
0.35
0.70
7.8
___
—
D, ym
0.23
0.37
0.73
1.6
7.9
—
D, ym
0.26
0.41
0.81
1.8
3.6
8.3
Eff, %
39.2
78.6
91.8
95.8
65.9
— — -
Eff, %
38.1
59.9
94.2
98.4
98.1
92.7
Eff, %
60.1
52.9
95.2
97.5
___
—
Eff, %
40.5
54.6
92.5
98.8
96.9
—
Eff, %
41.3
58.5
87.2
91.0
91.3
97.9
112
-------
Table A-3 (continued).
Run no. 11151
EAF dust
B = 0.375 T
V = 7.3 m/s
F = 0.0050
L = 0.300 m
T = 29°C
D, ym
0.23
0.37
0.74
1.7
8.0
— _
Eff, %
55.4
62.9
97.7
99.7
99.0
—
Run no. 11161
EAF dust
B = 0.400 T
V = 10.0 m/s
F = 0.0050
L = 0.300 m
T = 32°C
D, ym
0.26
0.41
0.81
1.8
3.6
8.3
Eff, %
29.9
63.6
84.6
86.2
98.0
85.0
113
-------
APPENDIX B
MATHEMATICAL MODEL OF HGMF
The approach taken to theoretical modeling of HGMF parallels the
usual treatment of conventional filtration. The particles are assumed
to be spherical, uniformly distributed in the fluid stream, and moving
at the same velocity as the fluid upstream of the collector. The basic
element of the filter is a clean, cylindrical wire of radius s, oriented
so that its axis is perpendicular to fluid flow (Figure B-l). The
objective of the theoretical calculations is to determine the portion of
the fluid stream from which particles are removed as the fluid flows
past the wire by analyzing the trajectories of particles approaching
under the influence of attendant forces. The collision radius of the
wire, y , is defined by the initial position of the particle whose
trajectory just touches the,wire, as illustrated in Figure B-l.
Watson (1973) published the first trajectory model of HGMF, taking
into account the magnetic and drag forces acting on a spherical particle
approaching a cylindrical wire. Lawson and coworkers (1976, 1977)
extended the model to include inertia! and gravitational effects. The
present work builds on the Lawson model to include the conditions appro-
priate to the magnetic filtration of particles from a gas stream.
The geometry of the system to be discussed is illustrated in Figure
B-2. The fundamental equation governing particle motion is
«*-VVfn,
-------
FLUID STREAMLINE
PARTICLE TRAJECTORY
v FLOW
Tr- ~*
cn
FLOW
Figure B-l. Illustration of particle capture by a single wire.
-------
cr>
PARTICLE
Figure B-2. Geometric basis of HGMF trajectory model.
-------
The gravitational force acting in the x-direction is
V
i
wgO ) cos e
P
r pf
- wg(l - —) sin
I P
P
Assuming potential flow about the wire and Stokes drag on the particle,
the drag force is expressed by
f
Next the particle is assumed to be small enough to be represented
magnetically as a point dipole of moment m = ft v , where v is the
particle volume. Then from the fundamental equation for the force on a
point dipole (Jackson, 1975, p. 185), the force on the sphere can be
expressed as
(B-3)
-V0(l - ^) cos 9 - ^-"1 r + 6Trnb fyi + p-) sin e - r ^-1 e, (B-4)
where b and s are the particle and wire radii, respectively, V is the
fluid velocity far upstream of the wire, and n is the fluid viscosity.
The potential flow and Stokes drag assumptions can be satisfied concur-
rently provided s»b.
The magnetization of a paramagnetic sphere immersed in a uniform
field ft is derived by Jackson (1975, p. 198) in a form equivalent to the
expression
«p -xff00 *$>-'. (B-5)
where x is the classical magnetic susceptibility, which is defined as
the ratio of magnetization of a paramagnetic body to the field inside
the body. The factor in parentheses in Equation (B-5) corrects for the
fact that ft is the field outside the sphere. For convenience, an
effective magnetic susceptibility, x*> is defined as
X*= x(l +)~1. (B-6)
117
-------
(B-7)
In an HGMF system the field surrounding the particle at any point
is determined by the uniformly applied field, H, and the contribution
a
of the magnetized wire. The determination of the field surrounding a
ferromagnetic cylinder immersed in a uniformly applied background field
is a classical magnetostatic problem that has been solved in several
theoretical developments (e.g. Lawson, et a 1., 1977). For the geometry
shown in Figure B-2 the result is
w
e + H cos e
a
r +
s2M
w
2r
sin e - H3 sin
a
Substituting Equation (B-8) into (B-7) and performing the indicated
mathematical operations leads to
r +
20
Equations (B-2), (B-3), (B-4), and (B-9) can now be substituted into
Equation (B-l). To simplify the result it is helpful to introduce one
physical identity and several dimensionless quantities.
w = 4Trb3Pp/3,
R = r/s,
T = tVQ/s,
(B-8)
(B-9)
(B-10)
(B-ll)
(B-12)
W =
V*Ha
(B-13)
.118
-------
(B-14)
K =
2bVo
9sn '
K)
(B-15)
(B-16)
When these substitutions are made and Equation (B-l) is simplified, the
r-component becomes
d2R
de
- R 5^ = G cos 9 - -^
. ^ i cos e + 3-
. 2WA/A +cos 2eV (B_17)
The 9-component is
2
D
R
d e , ~ dR de r . . , 1
-j—r + 2 -T— -j— = - G sin 9 + TT
2WA
R3
sin 2e. (B-18)
Two new dependent variables, r = dR/dt, and n = de/dt are defined
and substituted into Equations (B-17) and (B-18), with the results
TT- = RJT + G cos
IT I COS 9 + T
2WA
-2- + cos 2e
(B-19)
?«sin2e.
(B-20)
Equations (B-19) and (B-20) and the definitions of r and n form a
set of four first-order differential equations that can be solved numerically
with the initial conditions
(B-21)
119
-------
= tan'1 () , (B-22)
r. = / (KG - 1) , (B-23)
1 Ki
Y.
n. = or (1 - KG) . (B-24)
1 Ki
The initial dimensionless velocities, r.. and ^, may be obtained from a
simplified form of Equation (B-l), which assumes that the steady-state
particle motion far upstream from the wire is parallel to the x-axis and
is determined solely by drag and gravitational forces. Equations (B-21)
and (B-22) simply state the initial position of the particle at T = 0.
Specification of the initial coordinates (X., Y.) and the four
dimensionless parameters, W, K, G, and A provides information sufficient
to determine a unique particle trajectory. Calculations for the experi-
mental conditions used in this work showed that three of the parameters
fell within the following ranges:
W : 0 - 1
K : 0.05 - 250
G : < 2 x 10"5.
The fourth dimensionless parameter A = M /(2H ), is a measure of
the relative magnitudes of the magnetic field induced in the ferro-
magnetic wire and the applied field. A is referred to as the near-field
parameter because its effect on particle trajectory is significant only
near the wire. The wire magnetization, M , increases with applied field
until the filter is magnetically saturated. Ferritic stainless steels,
120
-------
the most common materials used as HGMF filters, have saturation magneti-
c /-
zations ranging from 1.17 x 10 to 1.27 x 10 ampere turns per meter
(Lyman, 1961). Kolm (1975) and Clarkson, et al . , (1976) report that
magnetic saturation of a steel wool filter occurs at an applied field of
5 5
5.6 x 10 to 6.4 x 10 ampere turns per meter, which yields a value of A
(= M/2H ) wl at the saturation condition. With higher applied fields,
w a
M is constant at the saturation value and the value of A decreases.
Below saturation the magnetization of individual wires in a filter
is dependent on the shape of the wires, on their orientation with respect
to the applied field, and on the proximity of neighboring wires. To
determine an average or characteristic value of A for a filter under
these conditions, one can revert back to the analysis of Lawson, et a! . ,
(1977). The magnetization of a linear, homogeneous wire is given by
where y is the magnetic permeability of the wire. The interior field of
a cylindrical wire magnetized perpendicular to its axis, H. is given
in terms of the applied field by
Hin =
Combining Equations (B-25) and (B-26) leads to
(B-25)
Mw = 2Ha I TTT-^ ) (B-27)
In a ferromagnetic material u is actually not constant but is dependent
on the applied field and the magnetization history of the material. How-
ever, below saturation u » u0» which simplifies Equation (B-27) to M =
2Ha or A = 1 for the idealized cylindrical wire. This result conveniently
matches the experimental saturation condition with no discontinuity.
121
-------
A Fortran subroutine, HPCG (IBM, 1970), which is based on Hamming's
modified predictor-corrector method, was used to solve the model equa-
tions. Preliminary computer solutions showed that trajectories calculated
with G = 2 x 10 were indistinguishable from the corresponding 6=0
trajectories, so gravitational effects were subsequently ignored.
Figure B-3 shows a contour map of the dimensionless collision radius,
YC = yc/s (see Figure B-l), as a function of W and K with A = 1 and G =
0. It is interesting to note that the contours for Y < 1 asymptotically
\*
approach vertical lines at low values of W as the magnetic attraction
becomes insignificant. Solutions were obtained with W = 0, corresponding
to no applied field, and the results were identical to those obtained by
Langmuir and Blodgett for conventional inertia! impaction in potential
flow (Fuchs, 1964).
The single-wire results may be extended to a filter by considering
a filter element of differential length dx and cross-sectional area S.
A single wire of length h and radius s removes particles for an area
normal to the direction of flow equal to 2hsY The volume of the
differential element is Sdx, so the volume of wire is FSdx, where F is
the volumetric fraction of wire in the element. The total length of
2
wire in the element is FSdx/irs , and the number of segments of length h
2
is FSdx/Trs h. In a randomly packed filter, the fraction of total wire
length that projects on a plane perpendicular to flow, and is thus
active in particle collection, is 2/ir. A mass balance is then formu-
lated in a manner analogous to a plug-flow reactor analysis. For a
filter of length L, the collection efficiency, E, is calculated to be
E=l --
CQ "^ I TT^S (1-TF
where CQ and c, are the particle concentrations at the entrance and
exit of the filter, respectively. Equation (B-28) implicitly assumes
that there is no aerodynamic or magnetic interference between collecting
wires and that every particle that collides with a wire adheres.
(B-28)
122
-------
f-O
CO
CC
ill
LLI
10°
10
1-1
o.
O
ill
10
r3 -
10"
G = 0
A = 1
10"1 10° 101
STOKES NUMBER (K)
103
Figure B-3. Contour map of collision radius as a function of W and K (A=l, G=0)
-------
Data from the experiments were first compared to predictions of the
theoretical model as it is expressed in Equation (B-28). A typical
result is shown in Figure B-4. The data and uncorrected theory agree
well at small particle sizes, but the model overestimates the collection
efficiency of larger particles. A reasonable explanation for the dis-
crepancy is that larger particles may not always adhere when they strike
a collecting wire. Other researchers have observed this reentrainment
phenomenon in conventional fibrous filters and attributed it to a bounce
mechanism or to the effect of fluid drag on the collected particles
(Gillespie, 1955; Loeffler, 1974; Stenhouse and Freshwater, 1976). Both
reentrainment mechanisms were considered in the subsequent analysis.
To analyze the bounce phenomenon, the physical situation illustrated
in Figure B-5 was considered. To simplify the mathematics, the particle
is assumed to be moving radially toward the center of the wire at the
free stream velocity V at the instant of collision. Trajectory calcu-
lations showed this assumption to be reasonable for larger particles-at
moderate to high magnetic fields.
If the kinetic energy of the particle is not completely dissipated
at the instant of collision, the particle rebounds radially outward from
the wire with a velocity that depends on the coefficient of restitution.
The rebound kinetic energy, E^. , can be compared to the energy required
to overcome the van der Waals attraction, E , plus the energy required
to escape the magnetic field, E , to determine whether the particle will
be reentrained. The residual kinetic energy, E, , is given by
Ekr = Ekb + Ea + Em
The model analysis consists of writing an expression for E,, in terms of
the coefficient of restitution of the particle; writing a van der Waals
force expression for a sphere leaving the surface of a cylinder to obtain
E , and integrating the radial component of the magnetic force ? (from
Equation (B-9) from r = s to r -> « to calculate E . The resulting
expression for the resideual kinetic energy, Ekr, is set equal to zero
to determine the critical value of e (see Figure B-5) above which the
impacting particle is reentrained.
124
-------
o
99.9
99.8
99.5
99
98
95
90
80
LLJ
O
EL 70
LJL
LJJ 60
o 50
P 40
UJ 30
I
20
O
O
10
UNCORRECTED
THPORV
TMfcURY
IIIIrr
- 0
0.2
c«yiom,r*Ai
EMPIRICAL
CORRECTION
BOUNCE CORRECTION
BOF DUST
Run No. 03021
F = 0.0080
L = 0.15 m
V = 8.7 m/s
Ba = 0.30 T
t 1 I
l I I I I I _L
0.5 1.0 2
PARTICLE DIAMETER,
10
Figure B-4. Comparison of experimental data to the uncorrected
theoretical prediction and to two corrected models.
125
-------
Figure B-5. Illustration of the particle bounce reentrainment model
126
-------
Details of this derivation are given by Gooding (1979). The final
expression for the critical approach angle is
= 1/2 cos
-1
12 2
3* P.V/
- 0.5
The parameters e (the particle coefficient of restitution) and U/& (a
quantity related to the magnitude of the van der Waals force between the
particle and the wire) cannot be predicted from first principles, but
may be treated as adjustable parameters within certain bounds.
A reentrainment correction to the collision radius, Y , may be
derived in terms of e . Trajectory calculations of the type that led to
the construction of Figure B-3 show that the collision points are more
or less equally distributed with respect to sin 0 (y of Figure B-2).
Thus, of the total number of collisions, only the fraction sin 0
\*
results in particle collection. The corrected form of Equation (B-28)
then becomes
E = 1 - exp
4FLY_ sin
\*
A (l-F)
Analysis of the experimental data showed that Equation (B-31) could
provide an acceptable correlation for isolated cases such as that illus-
trated by the "bounce correction" curve shown in Figure B-4; but over a
broader spectrum of operating conditions, the correction proved to be
too severe with respect to particle size and gas velocity. Values of
the adjustable parameters that gave suitable results for low-field,
high-velocity data provided almost no correction to the higher-field,
lower-velocity data. Conversely, parameters fitted to the higher field
data would indicate total reentrainment of larger particles at higher
velocities and lower fields.
(B-30)
(B-31)
127
-------
In an alternative approach that proved to be unproductive, the
reentrainment phenomenon was analyzed via a force balance written for a
particle that was assumed to have collided with and momentarily adhered
to the wire. One can postulate that the drag force acting on a particle
at the surface of the wire would tend to reentrain the particle in the
gas stream. The drag force is proportional to particle radius if Stokes1
Law applies, or to the radius squared at higher particle Reynolds numbers.
As Equation (B-9) indicates, however, the opposing magnetic force is
proportional to particle volume, or radius cubed. Hence, reentrainment
caused by fluid drag would tend to diminish at larger particle sizes,
rather than increasing in the manner indicated by the discrepancy between
uncorrected theory and data in Figure B-4.
Since neither the bounce model nor the force-balance model could
adequately predict the experimental results, an empirical reentrainment
correlation was developed. Equation (B-31) was inverted to calculate a
reentrainment correction from each data point; i.e.
2 2
p = sin 0 = - u ~P] An(l-E) . (B-32)
The reentrainment correction was re-labeled with the generalized symbol
p, signifying the probability of a particle adhering to a wire once it
collides. A multiple regression analysis was then conducted to correlate
p with the previously defined dimensionless groups W and K. The results
are given in Table B-l . The correlations were improved somewhat by sub-
dividing the data into the groups indicated. For the EAF dust the
exponent on W is not significantly different from zero, implying that
the applied magnetic field has little effect on reentrainment. The
probability of reentrainment is also higher for EAF dust than for BOF
dust under equivalent conditions. The EAF data may have been biased
toward lower collection efficiencies by a failure to maintain clean wire
conditions throughout the impactor sampling period as indicated earlier
in Figure 6.
128
-------
Table B-1. Correlation of reentrainment correction.
Data No. Data Points Correlation Regression Coefficient
EOF dust
field-on
EOF dust
field-off
EAF dust
field-on
..0.087
°fi7 n - 0 790
MI p u./ay 0<190
is.
61 p = 0.534 K~°'317
,,0.013
CO n _ o /\r\c w
b^ p U.4U6 Q.122
l\
0.727
0.669
0.634
The degree to which the empirically corrected model fits experi-
mental collection data is illustrated in Figure B-4 and in Figure 7,
which was presented earlier in the text. The confidence intervals in
Figure 7 were calculated by first transforming the efficiency data to
£n(l-E) to equalize the weight of high and low-efficiency points, and
then pooling the variances calculated at the six particle sizes to get a
single estimate of variance with 12 degrees of freedom. Several other
fitted-data plots are presented by Gooding (1979).
The inclusion of inertial terms in the theoretical model gives
gas velocity a complicated role. Referring to Equations (B-13) and
(B-15), velocity appears as a squared term in the denominator of W and
as a linear term in the numerator of K. Hence starting from any point
on Figure B-3, an increase in velocity involves moving two units down
and one unit to the right. Normally this leads to a reduction in the
collision radius, Yr, since the slope of the constant Y curves is
<- c
between - 1 and 0 over most of the contour plot. However, in the lower
left portion of the plot, an increase in Y can result from an increase
t*
in velocity, since the slope of the curves is less than -2. In physical
terms, increasing velocity, which increases particle inertia, tends to
129
-------
keep a particle moving in a straight line. If the particle is approaching
in the shadow of the wire, it is more likely to collide with the wire at
a higher velocity. Conversely, if it approaches from outside the
projected area of the wire, it is more likely to be swept past the wire
at a higher velocity. Therefore, increasing velocity increases YC only
if YC < 1 and if inertia has a more significant effect than magnetic
attraction. This combination of events is most likely to occur with
weakly magnetic, submicron particles.
The effect of velocity is confounded further by the fact that the
probability of reentrainment, p, is also dependent on velocity. Increas-
ing velocity always reduces p, so the conditions under which higher
velocity is beneficial are limited even more. The final result is that
collection efficiency is reduced by increasing the superficial gas
velocity in nearly all cases of practical interest. The magnitude of
the velocity effect is illustrated in Figure 12, presented earlier in
the text.
130
-------
TECHNICAL REPORT DATA .
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/7-80-037
3. RECIPIENT'S ACCESSK
4. TITLE AND SUBTITLE
Pilot-scale Field Tests of High-gradient Magnetic
Filtration
5. REPORT DATE
March 1980
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Charles H. Gooding
8. PERFORMING ORGANIZATION REF
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Research Triangle Institute
P.O. Box 12194
Research Triangle Park, North Carolina 27709
10. PROGRAM ELEMENT NO.
EHE624A
11. CONTRACT/GRANT NO.
68-02-2650
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final; 9/77 - 12/79
14. SPONSORING AGENCY CODE
EPA/600/13
15. SUPPLEMENTARY NOTES IERL-RTP project officer is Dennis
919/541-2925.
C. Drehmel, Mail Drop 61.
. ABSTRACT The report gives results 01 using a olUO cu m/hr mobile pilot plant to eval-
uate the effectiveness and economics of applying high-gradient magnetic filtration
(HGMF) to particulate emission control. A 4-1/2 month test program was conducted
at a Pennsylvania sintering plant to characterize the performance of the pilot plant
and to demonstrate its practicality under long-term operation. The pilot plant col-
lected approximately 90% of the iron-bearing particulate under practical operating
conditions but achieved lower overall collection because the windbox gas contained an
unexpectedly high concentration of fine alkali-chloride aerosol. To collect the non-
magnetic aerosol, a finer filter had to be used under conditions that were conducive
to plugging. Under the practical conditions, the pilot plant operated over 450 hours
without significant problems. Analysis of the results indicates that high-efficiency
collection can be achieved economically if HGMF is applied to steel industry dusts
that are more homogeneous and more strongly magnetic than the tested sinter dust.
The report describes laboratory pilot-plant work that demonstrated collection effic-
iencies greater than 99% with basic oxygen furnace and electric arc furnace dusts.
The development of a filter cleaning system and the design and construction of the
pilot plant are discussed. Experimental data are reported.
7.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATl Field/Group
Pollution
Filtration
Vlagnetic Properties
Magnetic Separators
Testing
Dust
Aerosols
Sintering Furnaces
Iron and Steel In-
dustry
Pollution Control
Stationary Sources
High-gradient Magnetic
Filtration
Particulate
13B
07D
20C
131
14B
11G
13A
11F
8. DISTRIBUTION STATEMENT
Release to Public
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
141
20. SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
131
------- |