EPA-600/2-77-230
November 1977
Environmental Protection Technology Series
              APPLICATION OF HIGH-GRADIENT
                      MAGNETIC SEPARATION TO
                         FINE PARTICLE CONTROL
                             Industrial Environmental Research Laboratory
                                  Office of Research and Development
                                 U.S. Environmental Protection Agency
                             Research Triangle Park, North Carolina 27711

-------
                      RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental Protection
Agency, have been grouped into five series. These five broad categories were established to
facilitate further development and application of environmental technology. Elimination of
traditional grouping was consciously planned to foster technology transfer and a maximum
interface in related fields. The five series are:

    1.   Environmental Health Effects Research
    2.   Environmental Protection Technology
    3.   Ecological Research
    4.   Environmental Monitoring
    5.   Socioeconomic EnvironmentalStudies

This report has been assigned to the ENVIRONMENTAL PROTECTION TECHNOLOGY
series. This series describes research performed to develop and demonstrate instrumenta-
tion, equipment, and methodology to repair or prevent environmental degradation from point
and  non-point sources of pollution. This  work provides the new or improved technology
required for the control and treatment of  pollution sources to meet environmental quality
standards.
                            EPA REVIEW NOTICE

This report has been reviewed by the U.S. Environmental Protection Agency, and approved
for publication. Approval does not signify that the contents necessarily reflect the views and
policy of the Agency, 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 Information Service,
Springfield, Virginia 22161.

-------
                                         EPA-600/2-77-230
                                            November 1977
APPLICATION  OF  HIGH-GRADIENT
     MAGNETIC SEPARATION TO
      FINE  PARTICLE CONTROL
                        by

             C.H. Gooding. T.W. Sigmon. and LK. Monteith

                  Research Triangle Institute
                    P.O. Box 12194
               Research Triangle Park. N.C. 27709
                  Contract No. 68-02-1879
                   ROAP No. 21AOL-029
                 Program Element No. 1 ABO 12
               EPA Project Officer: Dennis C. Drehmel

             Industrial Environmental Research Laboratory
              Office of Energy, Minerals, and Industry
               Research Triangle Park, N.C. 27711
                     Prepared for

             U.S. ENVIRONMENTAL PROTECTION AGENCY
               Office of Research and Development
                  Washington, D.C. 20460

-------
                               ABSTRACT
     Experiments were conducted and theoretical  models were analyzed to
assess the potential use of high gradient magnetic separation (HGMS) as
a means of collecting gas stream particulates.
     Viable alternatives to the conventional particulate collection
technologies are continually being sought to improve the efficiency and
reduce the cost of fine particle emission control.  HGMS has been
demonstrated in prior research and commercial applications to be an
effective and economical method of removing small paramagnetic particles
from selected liquid process streams.
     In Phase I of this study existing theoretical explanations of HGMS
and published reports of liquid-system applications were evaluated.  A
bench-scale apparatus was constructed, and HGMS experiments were con-
ducted using redispersed dust from a basic oxygen furnace.  High-efficiency
collection of fine particles was achieved with a high throughput and
with reasonable projected energy requirements relative to conventional
devices.
                                                          o
     In Phase II the experiments were scaled up to a 0.8 m /s (1700 CFM)
capacity unit.  Dusts from basic oxygen and electric arc furnaces were
redispersed and collected.  The results show that submicron particles
can be collected with greater than 90 percent efficiency using applied
magnetic flux densities of 0.2 to 0.4 T.  With superficial gas veloci-
ties up to 11 m/s, the pressure drop across the HGMS device was typi-
cally less than 1.5 kPa (6 in. HgO).  Even lower fields can be used
successfully at the expense of higher pressure drop or reduced through-
put.
     A theoretical model was utilized to correlate the experimental data
and to evaluate the economic and collection-efficiency trade-offs associated
with operating parameter variations.  The analysis  indicates  that  HGMS has
a good potential for successful application  to  selected  full-scale fine
particulate emission sources with reasonable capital  and operating costs.

-------
     This report is submitted in fulfillment of Contract  No.  68-02-1879
by Research Triangle Institute under the sponsorship of the  U.  S.
Environmental Protection Agency.  The report covers  the period  June  30,
1975 to August 29, 1977, and work was completed as of September 29,
1977.
                                   IV

-------
                               CONTENTS
                                                                 Page
Abstract	   ti
Figures	   vi
Tab! es	   1x
Acknowledgments 	    x
SECTION
   1.     Introduction and Summary	    1
   2.     Conclusions	    3
   3.     Recommendations 	    5
   4.     Background Development	    7
               Basic concept of the process	    7
               HGMS development and experience	    9
               Theoretical aspects of the process	   12
               Potential applications of HGMS to fine
                 particle emission control	   33
   5.     Experimental Work	   35
               Basis of experiment design	   35
               Summary of Phase I experiments and results	   37
               Description of Phase II apparatus
                 and experiments	   48
   6.     Pilot-Scale Experimental Results	   59
               Supportive matrix and particle data	   59
               Comparison of results with theoretical models...   65
               Discussion of individual parameter effects	   89
               Operating difficulties	   "
   7.     Economic Comparison of HGMS and Conventional
            Control Technology	  103
               Basis of economic comparison	  103
               Cost and energy estimates for conventional
                 techno! ogy	  104

-------



Q
.7 •
i n
1 U.
Aooendix.
uost ana energy estimates rur nui-u 	
Comparison of HGMS and conventional





IT T
11
	 115
	 123
	 125
vi

-------
                                FIGURES

Number                                                             Page
  1       Schematic representation of a high gradient magnetic
           separator	     8
  2       Critical  trajectory of particle impacting on a
           cylindrical collector	    14
  3       Inertia!  impaction efficiency as a function of Stokes
           number in potential flow	    15
  4       Interception efficiency in potential flow	,.    16
  5       Geometric basis of HGMS trajectory model	    18
  6       Magnetic force contours around a ferromagnetic cylinder
           oriented perpendicular to the magnetic field	    20
  7       Dimension!ess capture radius from Watson's trajectory
           mode!		    22
  8       Lines of constant capture radius with inertia! forces
           i ncl uded	    27
  9       Force balance on collected particle with no magnetic
           field.	    28
 10       Differential matrix segment	    32
 11       Experimental apparatus used in preliminary HGMS
           experiments	    38
 12       Results of bench-scale HGMS tests with packing density
           of 0.0088	    42
 13       Results of bench-scale HGMS tests with packing density
           of 0.0132	    43
 14       Results of bench-scale HGMS tests with packing density
           of 0.0174	    44
 15       Bench-scale variation in collection efficiency and
           pressure drop wi th matrix 1 oadi ng	    46
 16       Schematic representation of pilot-scale HGMS facility...    51
 17       HGMS matrix construction and cleaning apparatus	    53
 18       Photomicrographs of stainless steel wool matrix fibers..    60
 19       Photomicrographs of stainless steel wool matrix fibers..    61
 20       Photomicrographs of stainless steel wool matrix fibers..    62
 21       Magnetization curves of steel wool material and in-situ
           matrix	   64
                                    vii

-------
Figures (continued)

Number                                                             Page.
 22       Magnetization curve of basic oxygen furnace dust	    66
 23       Magnetization curve of electric arc furnace dust	    67
 24       Zero-field pilot-scale results with BOF dust (F=O.OQ5,
           L=l 5 cm)	    69
 25       Zero-field pilot-scale results with BOF dust (F=0.010,
           L=l 5 cm)	    70
 26       Zero-field pilot-scale results with BOF dust (F=0.005,
           L=30 cm)	    71
 27       Zero-field pilot-scale results with EAF dust (FO.005,
           L=l 5 cm)	    72
 28       Lines of constant capture radius with inertia!  forces
           included and A=2.0	    74
 29       Pilot-scale HGMS results with BOF dust (BQ=0.05 T)	    75
 30       Pilot-scale HGMS results with BOF dust (BQ=0.10 T)	    76
 31       Pilot-scale HGMS results with BOF dust at elevated
           temperature	    77
 32       Pilot-scale HGMS results with BOF dust (B =0.20 T)	    78
                                                   o       '
 33       Pilot-scale HGMS results with BOF dust (F=0.010)	    79
 34       Pilot-scale HGMS results with BOF dust (L=30 cm)	    80
 35       Pilot-scale HGMS results with BOF dust (duplicates of
           earl ier runs)	    81
 36       Pilot-scale HGMS results with BOF dust (replicate
           runs)	    82
 37       Pilot-scale HGMS results with EAF dust (B =0.10 T)	    83
 38       Pilot-scale HGMS results with EAF dust (B =0.20 T)	    84
 39       Pilot-scale HGMS results with both dusts at B =0.40 T..    85
 40       Pilot-scale variation in BOF collection efficiency and
           matrix pressure drop with matrix loading	    86
 41       Pilot-scale variation in EAF collection efficiency and
           matrix pressure drop with matrix loading	    87
 42       Comparison of collection efficiency results of
           replicate runs	    90
 43       Correlation of matrix pressure drop with gas velocity,
           matrix packing density and matrix length	'.    92
                                   vm

-------
Figures (continued)


Number                                                             Page

 44       Experimental effect of matrix length on collection
           efficiency with and without an applied magnetic
           field	   94

 45       Experimental effect of matrix packing density on
           collection efficiency with and without an applied
           f 1 el d	   95

 46       Experimental effect of applied magnetic field on the
           collection efficiency of both dusts	   97

 47       Illustration of the theoretical effect of increasing
           gas velocity	   98
 48       Experimental effects of gas velocity on collection
           efficiency with and without an applied magnetic
           field	  100

 49       Reported and estimated power requirements  for iron-
           bound solenoids of various configuration	   108
 50       Lines of constant capture radius with inertia!  forces
           included and A=1;5	   HO

 A-l       Coordinate system for HGMS single-fiber model	   128

-------
                                TABLES
Number                                                             Page
  1       Potential Applications of HGMS to Fine Particle
           Emission Control	  34
  2       Estimated Magnet Power Requirements for Several
           Field/Velocity Combinations	  36
  3       Ranges of Operating Parameters - Phase I Experiments	  41
  4       Preliminary Economic Analysis for HGMS Applied to EOF
           Dust Control	  47
  5       Compari son of Pi tot Traverse Resul ts	  54
  6       Experiment Design for Pilot-Scale HGMS	  58
  7       Values of the Near Field Correction Parameter	  63
  8       Pool ed Stati stics on Particl e Penetration	  91
  9       Design and Cost Summary of Conventional EOF
           Control Equipment	
 10       Cost Estimates for Large Iron-Bound Solenoids
           Applicable to Fine Particle Control	
 11       Comparison of HGMS and Conventional Technology*	

-------
                            ACKNOWLEDGMENTS

     The authors wish to acknowledge the assistance and support of the
following individuals in various segments of this work.
     Dr. Dennis C. Drehmel, EPA Project Officer, provided support and
direction throughout the project.
     Both sources of industrial dust were obtained through the cooper-
ation of Mr. John W. Leming, Jr., Manager of Pollution Abatement of
Bethlehem Steel Corporation, Bethlehem, Pa.
     Mr. William F. Lawson, Jr., of the USERDA Morgantown Energy
Research Center and West Virginia University provided a copy of his
unpublished master's thesis.  Mr. Lawson's prior theoretical research
on HGMS was of critical importance to this work.
     Dr. Herbert Hacker, Jr., of Duke University, Durham, N. C. provided
experimental determinations of particle and steel wool magnetization.
     Mr. R. Duncan Hay of Magnetic Engineering Associates, Cambridge,
Mass, made valuable contributions to the Phase I data interpretation and
preliminary economic analysis.
     Messers. Don Zanders, Billy Bowles, Wally Merritt, and Paul Johnson
of Monsanto Research Corporation were very cooperative in providing
coordination, operation, and troubleshooting of the EPA wind tunnel
utilized in the Phase II experiments.
     Other RTI personnel who made major contributions include Dr.
Forest 0. Mixon, Manager, Process Engineering Department, who provided
internal direction as Laboratory Supervisor; Messers. Doug VanOsdell and
Martin Lee, who assisted in the design, construction, and operation of
the bench-scale apparatus; Mr. Fred Schwarz, who designed and fabricated
many components of the sampling equipment for the Phase II experiments;
Messers. Bill King, John Sauerbier, and James Armstrong who assisted in
the design, construction, and operation of the pilot-scale apparatus.
                                   xi

-------
                                   SECTION 1

                            INTRODUCTION AND SUMMARY

     In the last several years, particulate control technology has been
developed to the extent that several methods are now available to con-
trol large particle emissions from most industrial sources with an
efficiency of greater than 99 percent.  Emphasis on particulate control
has now shifted toward the fine particle size range, particularly to
particles that have diameters between 0.3 and 3 micrometers.  These
particles are of primary interest because they tend to remain in the
atmosphere for long periods of time, thus contributing to atmospheric
haze.  They also are suspected to be major contributors to numerous air
pollution related health problems.
     The three major conventional technologies for the control of fine
particulate emissions are electrostatic precipitation, wet scrubbing,
and fabric filtration.  The cost of applying these control methods and
the fractional collection efficiency obtained vary according to the
specific application.  Recent EPA research in particulate emission
control has followed two complementary courses:  (1) study of the fine
particle collection characteristics of the conventional technologies and
(2) evaluation of viable alternatives to the conventional technologies.
Both of these courses have as their goals the enhancement of fine
particulate collection efficiency and the reduction of the capital and
operating costs  of industrial emission control.
     In the last decade research and commercial applications have
demonstrated that high gradient magnetic separation (HGMS) is an
effective and economical method of removing small paramagnetic particles
from selected liquid streams.  In Phase I of this study published
theoretical and experimental reports of HGMS applications were utilized
to evaluate the potential success of the process in removing fine,
paramagnetic particles from gas streams.  Industrial sources of fine,

-------
paramagnetic particles were identified, and the steelmaking basic oxygen
furnace (BOF) was selected for initial  HGMS evaluation.   A bench-scale
apparatus was constructed, and HGMS experiments were conducted using
redispersed BOF dust.  With existing theoretical equations as a guide-
line, the experimental data were analyzed,  and favorable operating
conditions were identified.  A preliminary  economic analysis was made of
the full-scale application of HGMS for particulate control on a basic
oxygen furnace.  The analysis indicated that HGMS could  be successfully
applied to BOF particulate emission control with reasonable capital cost
and energy requirements.  Phase I of the study concluded with the design
of a pilot-scale apparatus and experimental program to verify and extend
the results obtained in the preliminary work.
     In Phase II the 0.8 m3/s (1700 CFM) capacity pilot plant was
constructed and operated.  A systematic series of experiments was
conducted using both BOF dust and dust from an electric  arc furnace
(EAF).  Fractional collection efficiency of BOF dust was measured in
conjunction with variations in five important process operating pa-
rameters:  magnetic field, gas velocity, collection matrix density and
length, and gas temperature.  A more limited series of tests was run
with EAF dust.  A theoretical model was utilized to correlate the
experimental data and to evaluate the economic and collection efficiency
trade-offs associated with operating parameter variations.  The economic
projections of Phase I were refined, and a  comparison was made of the
capital cost and energy requirements of HGMS as a particulate control
method versus the conventional technologies.  Potential  design and
operating problems of HGMS were also considered.
     The following sections of this report present conclusions of the
study and recommendations for continued work along with details of the
development and theory of the process, a description of the pilot plant
used in this work, and the experimental results and analysis.

-------
                               SECTION 2
                              CONCLUSIONS

     The central purpose of this work was to demonstrate on a small
scale the feasibility of applying high gradient magnetic separation to
the control of fine particle emissions.  Theoretical analysis, bench-
scale testing, and pilot-scale testing were conducted, and the following
conclusions were drawn.
     (1)  High gradient magnetic separation has been demonstrated
          experimentally to be an effective method of collecting
          submicron particles of relatively high magnetic suscepti-
          bility from high-velocity gas streams.
     (2)  Several industrial processes in the iron and steel  industry
          and the ferroalloy industry produce particulate emissions of
          sufficiently high magnetic susceptibility to make them
          potential candidates for the application of HGMS fine particle
          control.
     (3)  Preliminary economic analysis based on the pilot-scale HGMS
          experiments indicates that HGMS could be economically com-
          petitive with conventional particulate control technologies.
          Projected energy requirements for a full-scale application
          to a basic oxygen furnace are significantly less than those
          of a comparable electrostatic precipitator and much less
          than those of a wet scrubber.  Projected capital costs
          based on existing HGMS equipment are somewhat higher than
          either precipitators or scrubbers.  Sufficient data are not
          currently available on baghouse installations on the candidate
          processes to allow a comparison of HGMS and this control
          method.
     (4)  Current methods of regenerating HGMS collection matrices based
          on cyclical operation may not be practical in high throughput
          gas stream systems.

-------
(5)   H6MS system designs directed specifically  to  gas  stream
     applications are needed to reduce  projected capital  costs  and
     to determine the optimum method  of matrix  regeneration.
(6)   HGMS potentially offers at least two  important  advantages  ver
     alternate methods of fine particle control.   The  high  gas
     velocities demonstrated in the pilot-scale experiments indi-
     cate much smaller space requirements  than  either  electrostatic
     precipitators or baghouses,  which  could  lead  to a significant
     reduction in total  installed equipment costs.   As presently
     conceived, the process  collects  the particulate in a dry form
     and avoids problems with sludge  disposal and  liquid waste
     treatment.
(7)   An existing theoretical  model of the  HGMS  process provides a
     valuable tool that can  be used to  screen potential appli-
     cations, evaluate alternative system  designs, plan experi-
     ments, analyze experimental  data,  and conduct economic
     analyses.
(8)   No inherent health  or safety hazards  or  any other unfavorable
     environmental impacts are foreseen that might prevent  the
     application of HGMS to  fine  particle  control.

-------
                                   SECTION  3
                                 RECOMMENDATIONS

     The successful pilot-scale experiments described in this report
indicate a strong potential for the efficient and economically at-
tractive application of HGMS to fine particle control on several in-
dustrial processes.  Experimental development of the process should be
continued to identify the most promising industrial applications and to
provide additional data on process performance and economics.  To gain
the -interest and confidence of the industrial community, it is recom-
mended that the next experimental program include pilot-scale field
testing of HGMS on a slipstream of one or more industrial processes.  To
provide flexibility in the choice of field test sites and minimize the
cost borne by industrial participants, the pilot plant should be a self-
contained relocatable unit.  A recommended four-part pilot demonstration
is summarized below.
     (1)  Experiment and System Design
          On-site evaluations of candidate sources should be conducted
          to gather data on the industrial process conditions.  Using
          experimental data and correlations from the work described in
          this report plus supplemental wind tunnel tests, a relocatable
          pilot-plant should be designed and constructed.  An important
          part of the equipment design should be the evaluation of
          alternate magnet systems to determine the configuration most
          appropriate for a gas stream application.  When the equipment
          design is completed, a compatible plan for the experimental
          field tests should be drawn up with the concurrence of the
          industrial participants and the approval of EPA.

-------
     (2)  Design Verification Tests
          Using the pilot plant described in this report, additional
          tests should be conducted with dusts from candidate sources to
          select the most promising sources for initial field testing
          and to assist in the design and material selection for the
          relocatable pilot-plant.  These tests would proceed con-
          currently with the work recommended in item (1) above.  Tests
          should also be conducted to evaluate alternate matrix materials,
          configurations, and cleaning methods.
     (3)  Pilot-Scale Field Demonstration
          The relocatable pilot unit should be tested on slipstreams of
          one or more industrial sources.  Tests should be conducted to
          characterize the performance of H6MS fine particle control
          with respect to particle source and size; gas humidity,
          temperature, and velocity; magnetic field strength; and matrix
          configuration.  The performance of the unit under continuous,
          long-term operation should also be evaluated.
     (4)  Summary and Conclusions
          Using the field data, an analysis of fractional efficiency
          versus residence time and energy requirements should be
          performed. Limitations and advantages of the process, problem
          areas, and safety aspects should also be evaluated.  Cost
          estimates for application of HGMS to the tested industrial
          sources should be prepared and compared to available data on
          conventional control technology.
     Throughout the recommended project, theoretical and/or semi-empirical
modeling of the HGMS process should be continued to aid in experiment
design, data interpretation, equipment design and optimization, and
process economic evaluation.  As a first step in the continued theoretical
work, capture radii should be actually calculated (rather than extrapo-
lated from Lawson's work) for conditions that will be incurred in gas
stream HGMS applications.

-------
                                   SECTION  4
                             BACKGROUND DEVELOPMENT
BASIC CONCEPT OF THE PROCESS
     Magnetic separation  has  long  been  recognized as a method of re-
moving magnetic solids from nonmagnetic solids and fluids.  Generally
the process has been applied  to  the removal of relatively large,
strongly magnetic solid bodies.  The largest commercial application has
been the separation of magnetic  minerals from ores.  Within the last
decade a new technique, high  gradient magnetic separation (HGMS), has
been developed and proven capable  of removing very small, weakly
magnetic particles from nonmagnetic materials.
     The fundamental concept  of  HGMS is the interaction of the small
paramagnetic particles with a ferromagnetic wire in a uniform background
magnetic field.  The ferromagnetic wire induces regions of highly non-
uniform field intensity which results in a net force being exerted on
the particles.  The magnetic  force, in competition with viscous and in
some cases inertia! and gravitational forces, causes the particles to
migrate to the surface of the wire where they are retained until the
magnetic force is removed.
     In its most simple,  practical form, the high gradient magnetic
separator consists of a canister packed with fibers of a ferromagnetic
material (such as ferritic stainless steel wool) and subjected to a
strong external magnetic  field (Figure  1).  The strong magnetic forces
near the edges of the fibers  result in  very efficient collection of
fine paramagnetic particles.  Thus the  particle-laden fluid is cleansed
as it passes through the  canister.  When the fiber matrix becomes fully
loaded, the magnetic field is removed and the particles may be flushed
from the matrix.

-------
00
             PARTICLE-LADEN
                 FLUID IN
                            STAINLESS
                          STEEL WOOL
                             MATRIX
                                               MAGNET
                                                COIL
CLEAN FLUID
   OUT
                   Figure 1.   Schematic representation  of a high gradient magnetic separator.

-------
     In essence HGMS is a type of enhanced filtration.  However, the
void volume of the filter matrix is typically much larger than in a
conventional filter, allowing very high fluid velocities at relatively
low pressure drop.
HGMS DEVELOPMENT AND EXPERIENCE
     The chronological development of the high gradient magnetic
separation process has been recorded in some detail in recent articles
in the open literature [1-5] and will only be summarized here, drawing
particularly from the synopses of Oder [4] and lannicelli [5].
     The expansion of magnetic separation into new process areas can be
linked to the recognition of two concepts:  (1) high field strengths can
be utilized to polarize even weakly magnetic materials, and (2) non-
uniform magnetic fields can be employed to develop magnetic forces in
dipolar materials.  The forerunner of modern high gradient magnetic
separators is the device developed by Franz in 1937 [6].  Franz's
separator utilized stainless steel screens to induce field gradients
and yielded good ceramic mineral separations although it employed a
field strength of only 0.15 tesla.  Around 1955 Jones developed the
first high intensity wet magnetic separator [7], which allowed the
processing of mineral slurries with fields up to 1.0 tesla.  Jones's
separator employed grooved steel plates rather than screens, however;
and the resulting field gradients were actually lower than those in the
Franz device.
     In the mid 1960's experimentation was begun at the J. M. Huber
Corporation in Georgia to test the use of magnetic separation in
removing small paramagnetic color-bodies from kaolin clay [8-10].  Both
the Franz and Jones-type machines were tested and proven effective in
this application, but the economics of commercialization were not promising.
In 1967 lannicelli constructed a hybrid high gradient-high field unit
using a Jones magnet and screens from a Franz device.  The hybrid separation
significantly upgraded separation efficiency, and for the first time
indicated commercial practicality.

-------
     In 1968 consultants from the Francis Bitter National  Magnet
Laboratory at Massachusetts Institute of Technology (MIT)  and from
Magnetic Engineering Associates (MEA) in Cambridge, Mass., were engaged
in the kaolin project.  Ferritic (430) stainless steel  wool was tested
as a collection matrix in the Huber hybrid separator and proved more
effective than the larger Franz screens.  In addition to the production
of large field gradients, the steel wool afforded large surface-to-
volume ratio (large collection surface with high porosity) and excellent
corrosion resistance.
     A 50-cm iron-bound solenoid was subsequently constructed for Huber
by MEA, and in 1969 the first commercial clays were produced using high
gradient magnetic separation as a wet beneficiation process.  Since that
time the HGMS process has achieved commercial acceptance in the kaolin
industry.  Oder reported in 1976 [4] that there was already in the
industry installed magnet capacity with a potential for handling 75
percent of the world production of coating quality kaolin  clays.  The
largest HGMS units now in service are iron-bound solenoids with magnetized
volumes measuring 2.1 m in diameter by 0.5 m  high (direction of flow).
The coils are surrounded by a solid iron box 3 to 4 m on a side.  Field
strengths up to 2 T are typically used with a power consumption of 400
to 500 kW per unit.
     Although the only significant commercialization of HGMS to date has
been in the kaolin clay industry, interest in the process  is growing
rapidly [11]. Pilot-scale and laboratory investigations have been conducted
in several areas.  Other mineral processing applications have received
considerable attention [12-16].  In addition to the work on HGMS minerals
beneficiation that has been reported in the open literature, there are
indications that other proprietary investigations are being conducted
[4].  Related to the minerals beneficiation work is the proposed appli-
cation of HGMS to the removal of iron pyrites from coal to achieve
partial desulfurization [17,18].  This concept has been extended to
include numerous proposals for ash removal from coal and cleaning of
solvent refined and liquefied coals [19-28].
                                     10

-------
     The possible use of HGMS in removing fine suspended solids from
steel-making wastewaters is also under active investigation [4,11,29-
32].  Other wastewater applications, including municipal sewage treat-
ment, may be amenable to HGMS provided the process stream can be seeded
with a magnetic material that will become physically or chemically
associated with suspended and/or dissolved particulates [11,33-38].
     Medical, pharmaceutical, and food-processing applications of high
gradient magnetic separation are also possible.  Experiments have been
conducted which demonstrate that red blood cells (containing iron-
bearing hemaglobin) can be separated without damage from whole blood
using HGMS [39].  Oder [4] cites reported interest in other biologically-
related areas.
     Despite rapid commercialization in the clay industry and numerous
experimental studies in other areas during the last several years, HGMS
is just beginning to be recognized by the technical community at large
as a viable technology of significant potential impact.  Kolm [40]
suggests that a lack of interdisciplinary communication is the primary
factor that has delayed other commercial applications.  HGMS extends the
applicability of magnetic separation from large particles of a few
ferromagnetic compounds to the much broader spectrum of chemical
engineering processes involving small paramagnetic particles [2,40].
     As new applications are conceived, continued research will un-
doubtedly be required for HGMS to reach its full potential.  A better
understanding of the fundamental mechanisms of the process is certainly
needed.  Magnet design, matrix size and configuration, and matrix
cleaning methods must be tailored to specific applications [4,40,41].
The large commercial clay units utilize steel wool matrices of roughly
95 percent porosity, and operate batch-wise with 70-80 percent duty
cycles [5].  For other applications, particularly in gas-processing
systems, more porous matrices and continuous processing may be required
to achieve economic practicality.
     Large-scale HGMS devices have been built for continuous operation
[[31].  In these units, the matrix is packed into an annular ring.  The ring
is continuously rotated through the magnetic field zone where paramagnetic
                                     11

-------
particles are removed from the fluid stream.   After passing through the
separation zone, each segment of the matrix rotates through a zero-field
flush zone for removal  of the collected particles.   While this design
shows promise for liquid processing, it might present sealing problems
in a gas stream so alternative continuous-processing schemes must be
considered.  Since the major developments of HGMS technology have
occurred over a relatively short time span, it is reasonable to believe
that many more important inventive contributions are forthcoming.
THEORETICAL ASPECTS OF THE PROCESS
     Various physical phenomena may be responsible for the transport of
particles from a fluid stream to the surface of a collector and for
holding then on the surface after they strike it.  The development of
theories for conventional filtration have been adequately described by
several sources [42-46].  The most commonly considered mechanisms are
inertia! impaction, interception, Brownian motion,  gravitational
settling, and electrostatic attraction.  Recent literature is also
replete with theoretical arguments specific to the high gradient
magnetic separation process [47-65], although little experimental data
has been available to adequately confirm or refute their validity.
     In the development of theoretical filtration models, several common
approaches and assumptions are generally used to obtain mathematical
equations which can be solved.  First 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 collector is
assumed to be a single clean fiber of cylindrical geometry with its axis
oriented perpendicular to fluid flow.  The collectional efficiency of
the fiber, n.p is then defined theoretically as the ratio of the cross-
sectional area of the fluid stream from which particles of a given size
are collected (by the mechanism under consideration) to the projected
area of the fiber in the direction of flow.  Mathematically, this
definition is equivalent to the equation
                                                                         (1)
                                    12

-------
 where NC = number of particles of a given size that are collected
             per unit length of fiber per unit time,
       a = radius of the fiber,
      V  = velocity of the undisturbed fluid stream upstream of
       0    the fiber,
       N = upstream concentration of particles of the given size
            (number of particles per unit volume).
Note that if all particles of a given size have the same mass density,
Equation (1) can be expressed equivalently in terms of particle mass—
i.e. nf is both the mass and number fractional efficiency.
     The following paragraphs discuss filtration mechanisms as they
relate to high gradient magnetic separation.  The analysis is directed
toward ranges of operating parameters that were encountered in the
experiments described in Sections 5 and 6.  Section 5 also states
reasons for operating within these particular parameter ranges.
Inertia! Impaction
     A fluid approaching a cylindrical body will be diverted from its
original path because of the presence of the body.  A particle flowing
with the fluid will tend to continue its steady, rectilinear motion,  and
if it has sufficient momentum, will cross the diverging fluid stream-
lines and impact on the cylinder.  To determine the theoretical impaction
collection efficiency, it is necessary to determine the initial position
of the particle whose centerline trajectory will just touch the collector
as illustrated in Figure 2.  Trajectory calculations are started at some
point sufficiently far upstream (x^, y) where the fluid is not yet
significantly affected by the presence of the collector.  All particles
of the same radius and mass density with initial y coordinates less than
y  will impact, and those with initial y coordinates greater than yQ
will be diverted around the cylinder.  Thus by the definition given in
Equation (1)

                                    n1 = y0/a.                     (2)
                                     13

-------
CYLINDRICAL
COLLECTOR
                                               PARTICLE
                                                    *0
                                                       -»• X
                Figure  2.  Critical trajectory of particle impacting
                          on a cylindrical collector.
        Several  authors  have determined particle trajectories for this case
   by integrating  the  equations of motion [42,44,45].  The results obtained
   are dependent on  the  flow field and initial conditions assumed.  For
   all solutions the efficiency of impaction can be written in the form
   where K =
             2b2p  V
   , Stokes number
               9na
         b = particle  radius
        p  = particle  density
         n = dynamic viscosity of fluid
             2aV_p
       Ref =
-, Reynolds number
                                                                            (3)
         p  =  fluid  density.
   (Note:   Using  a  different  normalization scheme some authors  [42,45,46]
   define the Stokes number as one-half the  value  given  above.   Care must
   be taken when  using  published correlations  involving  the  Stokes  number
   that the correct definition is used.)
        Davies  [43] has shown that  the  b/a term  in  Equation  (3)  is  important
   only for small Reynolds  numbers  and  very  small cylinders  --conditions
   which are  not  applicable to this work.  However,  the  relationship of
   n.  to K  is still  dependent on the  Reynolds  number.
                                       14

-------
     For Reynolds numbers on the order of 40 and above, potential flow
is often assumed, implying constant fluid density, zero viscosity (negli-
gible boundary layer at the cylinder), and irrotational flow.  Under
this assumption Langmuir and Blodgett obtained the curve shown in Figure
3 (obtained from Fuchs [44]), which can be well approximated by the
equation
                                           ,,3
                              ni =
K3 + 1.35 K2 + 0.21
                                 (4)
             1.0
           0.75
           0.50
           0.25
                                       J	I
                     J	I
                O.I  0.2   0.4 .6 1.0  2   4  6  10   20  4060 100
                                      K
             Figure 3.   Inertia!  impaction efficiency as a  function
                        of Stokes number in potential  flow.
 Interception
      In the development of the inertia!  impaction mechanism, it is
 assumed that no particle strikes the surface of the fiber unless its
 center!ine trajectory is tangent to or intersects the fiber.  Actually a
 particle will  contact the surface of the fiber when its trajectory
 passes at a distance less than b away.  This mechanism of capture is
 termed interception.
                                      15

-------
     If a particle is assumed to have finite size but no mass, it will
follow the fluid streamlines around the fiber.  The efficiency of
interception is therefore defined by those streamlines that pass within
y - TT-R
where R = b/a.  Figure 4 shows n0 as a function of R for small R values
in the range encountered in this work.
                                                           0.20
              Figure 4.  Interception efficiency in potential flow.

Brownian Motion.
     Very fine particles in a gas stream do not follow a direct path
because of collisions with gas molecules.  The resulting zig-zag
movement of the particles around their statistical mean path (termed
Brownian motion or diffusion) can result in collision with the fiber as
the gas stream is diverted around it.
                                     16

-------
      The characteristic parameter for this process is the Peclet number
 which is defined by
                                          2aVn
                                     Pe - -^                            (6)

 where D = diffusivity of the particle given by

                                     D
 where C = Cunningham correction factor
       k = Boltzmann's constant
       T = absolute temperature.
      For potential flow the diffusion collection  efficiency of a single
 fiber can be estimated by [46]
                                   nd = 3(Pe)-1/2

      It can be seen from Equation (8)  that nd  is  largest for small
 particles and low velocities.   Under the  most  favorable conditions
 experienced in this work (b =  0.1  ym,  VQ  = 5 m/s), nd is estimated by
 Equation (8)  to be 0.003.   It  can therefore be considered that the dif-
 fusion mechanism is negligible under the  conditions of interest.
 Gravitational  Settling
      In general,  gravitational  settling is  important only for large
 particles in  slow moving gas streams.  Ranz and Wong [42] showed that
 compared to inertia!  impaction,  gravitational  effects become important
 when  the quantity 2ga/VQ approaches  the  value of the Stokes number,
 where  g  is  the acceleration of gravity.   Under the most favorable conditions
                                  2                      5
 in this  work,  the quantity  2ga/VQ  never  exceeds 2 x 10" , which justifies
 the neglect of gravitational settling.
 Electrostatic  Effects
     Electrostatic  effects can occur in a filter when an external electric
field  is  imposed  or when either the particles or fibers or both carry
electrostatic charges.   Charge effects frequently occur in filters made
                                    17

-------
of nonconducting materials and can create spurious-results.  Since the-
filter in this case was conductive and grounded and no charges were
intentionally placed on the particles, electrostatic effects were
assumed negligible.
High Gradient Magnetic Attraction
     Several basically similar theoretical  models have been presented to
explain the capture of small  paramagnetic particles on the matrix wires
of a high gradient magnetic separator.  Like the theoretical approach to
conventional filtration, these models involve the calculation of particle
trajectories near a ferromagnetic wire in a magnetic field.  Differences
among the HGMS models arise in the assumptions made concerning the
system geometry, the flow field, and the significance of forces acting
on the particles.  The analysis presented in some detail here follows
most closely the work of Watson [48,51], who published the first such
HGMS trajectory model in the open literature.  (The original trajectory
model development for this case should probably be credited to Zebel
[66], who published a similar analysis for particles in an electrical
field several years earlier.)
     The Watson model considers a single cylindrical  wire oriented
perpendicular to the magnetic field and velocity vectors as indicated in
Figure 5.  The background magnetic field, H , is homogenous, and the
                                      y
             Figure 5.  Geometric  basis  of  HGMS  trajectory model.
                                        18

-------
wire is uniformly magnetized to saturation.   In the vicinity of the wire
the field is distorted, and the resulting field gradients  impose a
magnetic force on the spherical, paramagnetic particles entrained in the
fluid.  Watson neglected  the inertia! and gravitational forces acting on
the particles and developed an equation of motion for a single particle
in the form of Newton's second law as follows:
                                   ->•    ->•
                                   F . + F  =0
      *                             dm
where Fd = drag force,
      ->•
      F  = magnetic force.

     Since the relative velocity between the particle and fluid will in
general be small, Watson invoked Stokes law for the drag force with the
result
                                                                         (9)
                                        ->•    -»•
                               Fd  = -6trnb.(Vp - Vf)
                                                                         (10)
where V  =  particle  velocity
      Vf =  fluid  velocity.
Assuming potential  flow,  the  radial and angular components of the drag
force, Fdr  and  FdQ,  then  become
                     Fdr  =  "  6lTTlb
                     Fde  =  '  6wTlb
     The magnetic  force  terms  presented by Watson can be developed from
the fundamental  equation which describes  the  potential energy of a
particle placed  in a magnetostatic  field  in a  linear, homogenous, and
isotropic medium:
                                                                        (11)
                                                                        (12)
                         m
                                 m
= 1/27
 f   M  •
7v,
                                                    BQ  dv
                                                                         (13)
                                      19

-------
where Um = potential  energy,
      V. = particle volume,
       M = induced magnetization of particle,
      B  = magnetic flux density of applied  field.
The simplification of Equation (13) with  appropriate assumptions is
presented in Appendix A.  The resulting expressions for  the radial and
angular components  of  the magnetic force, Fmr and
                                                  FmQ, are
                         me
                             =  -  xB,
(H
                                                  sin2e
                                                                        (14)
(15)
where x = magnetic susceptibility  of  the particle
     y  = permeability of free space
     MS = saturation magnetization of wire.
Typical magnetic force contours are illustrated in Figure 6.  Note that
the force is repellent along the y axis; hence no particles would be
collected on the surface of the cylinder near these points.  In Equation
               2         2
(14) the term a n0Ms/2B0r  occurs  because of the magnetization of the
wire and is called the near field  correction.  It becomes significant as
                                   2  ?
the particle approaches the wire (a /r  -*• 1) and can be quite important
with relatively low background fields.
                                   ir/z
           Ho
                                   -TT/2
    Figure 6.
              Magnetic force contours around a ferromagnetic cylinder
              oriented perpendicular to the magnetic field.
                                    20

-------
     Watson substituted Equations (11), (12), (14) and (15) into
Equation (9) to solve for the trajectories of particles with initial
coordinates (xm, y. ) and determined the critical y coordinate
analogous to yQ in  Figure 2.  Normalizing the coordinates by the wire
radius, he defined  the dimensionless capture radius of the wire, R  =
                                                                  \f
yQ/a, which is identical to a magnetic collection efficiency, n »
defined in the form of Equation (1).
     Watson solved  the substituted form of Equation (9) numerically to
obtain values of R  in terms of two dimensionless groups:
                    o             and            os
          V  9wVo                           ^T
The former group is basically a ratio of the magnetic force acting on a
particle to the drag force.  V  is referred to as the magnetic velocity.
The latter group is the near field correction factor.  Solutions for R
as a function of V /V  are shown graphically in Figure 7 for several
values of the near field correction factor.  The curves corresponding to
PQM /2B  = 0, 0.2, 0.67, and 2.0 are from actual reported trajectory
calculations [3,48,51].  The curves for M0MS/2B0 = 0.88 and 1.5 were
obtained by graphical interpolation of the other four curves for
purposes discussed later in this report.
     It is important to note that for larger values of V /V , the value
of R  is greater than unity; that is, under favorable conditions the
    c
magnetic attraction not only keeps particles from diverging with the
fluid but also will actually draw them in from beyond the projected area
of the wire.
     Additional understanding of the practical HGMS process can be
gained from a review of other modeling efforts.  Oder and Price [4,50]
developed an intuitive model based on their HGMS experience at Huber.
Their model predicts the successive probability of particles striking the
wires and being held once they strike.  Parametrically the result is
identical to Watson's model solution for the case Vm/V0 <\2 and
y M /2B  = 0 where it can be shown that RC = Vm/2V0.  Oder and Price's
model  provides a simple expression that can be quite useful in correlating
high field, low velocity data, but it is not parametrically correct for
most of the conditions experienced in this work.

                                      21

-------
0.05 -
Q02 -
 QOI
  QOI   002
                                                                                     100   200     500   1000
                                                Vm
                                                Vo
                 Figure 7.   Dimensionless capture radius from Watson's trajectory model

-------
     Oberteuffer [1] developed a force balance for a paramagnetic
particle on the surface of the wire at the point x = a, y = 0
(Figure 5), and deduced the significant result that the maximum magnetic
force occurs when the wire radius is approximately 3 times the particle
radius.  This concept has two important implications:  (1) If possible,
the fiber size should be matched to the particles being collected; and
(2) extrapolation of the single-fiber/single-particle model to a real
HGMS system is complicated when both the wires and particles are actu-
ally characterized by a distribution of sizes.
     In terms of system geometry, Watson [51] also considered the case
where the background magnetic field vector is perpendicular to the fiber
and to the flow vector (i.e., in the y direction in Figure 5).  This
configuration corresponds to the use of an H-type magnet rather than a
solenoid and could offer some advantages in terms of access to the
matrix for cleaning.  Although the theory predicts higher values of R
for a given value of V /V  with this geometry, observed collection
efficiencies are reportedly lower than the parallel vector case [41].
Watson attributes this anomaly to the fact that particles are prefer-
entially collected in this geometry at the point x = 0, y = a
(Figure 5), where the shear forces of the fluid are greatest.
     Birss, et al. [60] as well as Uchiyama, et al. [61] have theo-
retically investigated the geometry in which the flow is parallel to the
fiber and the field is perpendicular to both.  This geometry does not
appear to offer any practical advantage in most applications, and the
collection efficiency is always less than it would be if the flow were
perpendicular to the fiber.  The remaining geometric permutations in
which the fiber and field were parallel are of no significance since no
magnetic gradients are produced.
     Cummings, Prieve, and Powers [58] demonstrated in an analysis
similar to Watson's that the assumption of creeping flow rather than
potential flow reduces the capture radius by as much as a factor of
three because of earlier diversion of fluid streamlines.  This factor
demands caution when using Watson's model with real liquid system data
in which the fiber Reynolds number is frequently as low as 1.  With the
conditions encountered in this work, the Reynolds number ranged from about
                                     23

-------
15 to 30.   Potential  flow is thus a better approximation than creeping
flow, but boundary layer effects may still be significant.
      In real  systems employing  steel wool fibers, the wires  are often
 highly irregular  in shape  in contrast to  the cylindrical shape assumed
 in Watson's model.  Stekly and  Minervini  [59] theoretically  investigated
 the shape  effect  on capture radius  using  elliptical shapes of varying
 aspect ratio.   For clean wires  the  capture radius varies little with
 shape (aspect ratio) so long as the projected' upstream  area  is constant.
 However, as particles  begin to  build up on the wires, the shape of the
 clean wire can be an important  factor in  determining the rate of deterior-
 ation in collection efficiency.
      Luborsky and Drummond [53,56]  have investigated this deterioration
 in collection efficiency with particle buildup in some  detail.  As
 weakly paramagnetic particles are  accumulated on the wire, the volume in
 which field gradients  are  large is  gradually filled.  Also the flow
 streamlines are diverted farther upstream as the wire "grows" in diameter.
 These two  factors combine  to give  a reduced capture radius in comparison
 to the prediction of Watson's clean fiber model.  Luborsky and Drummond
 developed  a modified version of Watson's  model in which particles are
 assumed  to accumulate  in successive layers.  Viscous and magnetic forces
 are compared  to determine  when  each layer is filled.  Using  two adjustable
 parameters -- the first representing the  fraction of fibers  that are
 active in  the capture  process and  the second to account for  mechanical
 entrapment of particles — they were able to obtain a reasonable fit  for
 Oberteuffer's experimental  data [49] on the collection  of cupric oxide
 particles  in  water.
      Cowen, Friedlaender,  and Jaluria [52,57,62] have succeeded in
 obtaining  photographs  of actual  particle  accumulation on a single wire
 using all  three flow,  field, and fiber geometries mentioned  earlier.   In
 general, the  photographs qualitatively confirm the validity  of Watson's
 model  and  Luborsky and Drummond's  approach to particle  buildup.  The
 most important discrepancy is the  absence of significant collection on
 the back side of  the wires  in the  experimental setup with  the geometry
 illustrated in Figure  5.   Theoretically some particle trajectories will
 result in  capture in this  region near x = -a, y =  0.
                                    24

-------
     The effect of the inertial terms, which were ignored in Watson's
model, on the equations of motion may be of paramount importance to gas-
stream applications of HGMS.  Watson argued [51] that these terms are
negligible in comparison to viscous and magnetic terms, but his develop-
ment was directed toward low velocity liquid systems.  In gas-stream
systems the achievement of high velocity flow may be critical to the
economic viability of the process, and the effects of inertia become
more significant.  Three groups have independently developed models
incorporating inertia in recent publications [54,63,65].  The work of
Lawson, et al. [65] is presented in the most detailed and usable form.
     Lawson [64,65] solved the equations of motion incorporating inertial,
viscous, magnetic, and gravitational terms and demonstrated that the
particle trajectories are uniquely determined by four dimensionless
groups:
                            2
                W =	—	5-  magnetic parameter                  (16)
                    (1 + l/3x)PpV02

                A =  	TJ—  near field correction                       (17)
                    2b2p V
                 K =  " 9ap °  Stokes number                               (18)
                6 = ^L. n _ £_\  gravitational parameter                (19)
                    V \   PP/
where x  is the dipole moment per unit length; the other quantities have
been previously defined.
     In  a high velocity gas-stream system the gravitational parameter
will be  on the order of 10"5, which is negligible.  Furthermore, it can
be shown that if the wire  is homogeneous and magnetized to saturation,
the term A reduces to the  near field correction factor introduced in
Watson's model; and the quantity 2AWK is equal to Watson's Vm/VQ.  The
                                     25

-------
appearance of x in the denominator of Equation (16) arises from a more
rigorous derivation of the particle magnetization, which is important
when the susceptibility is relatively large.
     For the purpose of this work, the most useful result of Lawson's
model is illustrated in Figure 8, which shows lines of constant capture
radius versus the magnetic parameter and Stokes number with fixed A and
negligible gravitational effects.  The curvature of the lines demon-
strates the intuitive effect of inertia on the capture radius.  With
log  K = -2, inertia! forces are negligible, and the R  values are
                                                     \f
identical to those which would be obtained from Watson's model.  Further-
more Watson's model predicts that with fixed A, RC is constant so long
as the sum (log K + log W) is constant.  Hence, ignoring inertia, all of
the  contours would be straight lines of slope = -1 and would pass
through the points as shown in Figure 8 where log K = -2.  The downward
curvature of the line for R_ = 0.5 thus illustrates that inclusion of
inertia! terms increases RC in this case as it should since inertia!
forces are beneficial in preventing the particles from being diverted
around the wire.  For R  > 1, however, inclusion of the inertial terms
                       \f
reduces R  since this case corresponds to the magnetized wire's ability
         c
to draw in particles from outside its projected area,..and inertial
forces will tend to keep the particles outside the projected area moving
in steady, rectilinear motion.  The line for R  = 1 is straight with
                                              \*
slope = -1.  The points at which the lines begin to curve indicate the
onset of significant inertial effects.  Note that for smaller values of
W (lower magnetic fields and higher velocities), inertial forces becomes
significant at lower values of K.
Reentrainment
     Theoretical models of conventional filtration systems normally
assume that every particle that touches the fiber surface adheres.  With
sufficiently high gas velocities the validity of this assumption becomes
questionable.  With no magnetic field the particle approaches  the fiber
under the influence of inertial and viscous forces.  When it  strikes  the
surface, the inertial forces are dissipated; and in the absence of
electrostatic effects, only Van der Waals forces are  left to  compete
                                    26

-------
       0>
       o
Figure 8.  Lines of constant capture radius with  inertia! forces included [64],
                                         27

-------
 with  viscous  drag  and  prevent  reentrainment.   The  importance of  re-
 entrainment can  be estimated by  constructing  a force  balance for
 particles  located  at the  surface of  the  fiber.
      Considering the geometry  illustrated  in  Figure 9,  it  is assumed
 that  particles have an equal probability of striking  the collector any-
 where in the  interval  -ir/2  < e < -n/2.  The illustrated  particle  of radius
 b will  then be held by the  radial  Van  Der  Waals force,  F  . while the
                                                        a
 tangential  drag  force, F, , acting at  r  =  a + b will  tend  to shear it
 off the fiber.   If Fde >  kFQ,  where  k  is the  coefficient of friction,
 the particle  will  be reentrained.  Otherwise  it will  adhere.   Since Fd
 increases  as  e goes from  0  to  ir/2 and  the  geometry is symmetrical, the
 fraction of fiber  surface area over  which  particles of  radius  b  will be
 held is given by 2e  /tr,  where  e   is  the  critical angle determined  by
                                                                       -
                                                     Vn
                       -IT/2
 Figure 9.   Force  balance on collected  particle with  no magnetic  field.

      In general,  the magnitude of  the  free  stream drag force,  Fd ,
 acting on  an  sphere of radius b  is given  by
Fdo • Cd
                                                                          (20)
where Cd  = drag coefficient.  Various  expressions  have  been  offered for the
drag coefficient, each of which is valid over a finite  range of  Reynolds
numbers.  The best  known of these was  given  by Stokes as  C .  = 24/Re
where Re  is the particle Reynolds number defined  by 2bpVQ/n. This  ex-
pression  is equivalent to Stokes law, which was utilized  in  Equation (10)
and is valid with Reynolds numbers up to about 0.5.  For  particles  on
                                     28

-------
the surface of the wire relative velocity is increased, however; and
Reynolds numbers of interest will range from about 0.1 to 10.  For this
range a least-square curve was fit to experimental data obtained by
Moller and Davies and reported in Fuchs [44] with the result

                                   1 + 0.126 Re  - 0.009 Re 2)

In potential flow the tangential component of the drag force at any e
and r = a + b is then given by
Fde
= F
do
sine
1
+ a2
(a+b)

2

     According to Hamaker's microscopic theory, the Van der Waals force
between a sphere of radius b and a half-space is given by [67]
                                                                         (22)
                                                                         (23)
where A = Hamaker constant
      z = separation distance between the sphere and half -space.
The parameter z cannot be measured easily and is usually estimated to
be 4 x 10    m, which corresponds to the lattice constant for crystals
with Van der Waals bonds [46,67,68].  Hamaker constants vary by orders
of magnitude for different materials.  Visser [69] reports two values of
                                                             -19
interest to this work:  for Fe on Fe in vacuum, A = 2.12 x 10    J;
                                  **             19
and for Fe203 on Fe203 in vacuum, A=2.32xlO    J.
     The coefficient of friction required in this analysis likewise
cannot be predicted with confidence.  To initiate particle sliding a
static coefficient of friction on the order of 0.5 would be appropriate
[70].  If the particle were considered to roll off rather than slide,
the coefficient of friction would be closer to 0.01 [70].
     Summarizing the above equations and assumptions,  the fraction of
colliding particles that are retained on the fiber surface is given by
2e A, where
                                     29

-------
= Arcs in
c
kAb
6z2FdQ /I + a2 \
° \ (a+b)2/
                                                                         (24)
                                                           2
Using the estimates given above,  the lumped parameter kA/6z  should
have a value in the range of 0.002 to 0.1  N/m depending on the choice of
friction coefficients.   Uncertainty in the other parameters could broaden
this range.  More importantly, Equation (24) indicates the functional
relationship of the reentrainment phenomenon on system operating parameters.
     In the HGMS process, reentrainment will be greatly reduced or
eliminated by the presence of the magnetic force.   As a first approxi-
mation one might estimate no reentrainment since the capture radius is
calculated from a force balance involving  both the drag and magnetic
forces, and the magnetic forces are strongest at the surface of the
wire.  Luborsky and Drummond's arguments on particle buildup [53,56]
imply the use of an effective reentrainment factor with predicted
capture radii, however, if the capture radii are predicted from a clean
fiber model.
Combination of Mechanisms
     From the previous  arguments  it is apparent that in an HGMS filter
with no applied field,  three particle collection phenomena can still
function to a significant extent:  inertia! impaction, interception, and
reentrainment.  It is incorrect to add the single fiber collection
efficiencies of impaction and interception since this method could in
principle lead to some  particles  being collected more than once on the
same fiber.  If the mechanisms are considered independently and the
fraction of particles in the projected area of a fiber collected by
impaction is r\., then interception can collect no more than Ti(l-n-).
Summing these efficiencies and accounting  for reentrainment then gives
for the net single fiber efficiency
                                                 2er
                            nf =  (n, + n0  - n-^)-^ .                   (25)
                                     30

-------
     In a practical HGMS system in which the field is energized and
oriented parallel to bulk flow, the interception mechanism loses its
significance since it implies collection near the points 0 = -ir/2,
Tr/2, and these are points of magnetic repulsion (see Figure 6).
Lawson's solution of the type illustrated in Figure 8 includes both the
high gradient magnetic and inertia! effects.  Thus, if reentrainment is
assumed negligible in this case, the net single fiber collection efficiency
is simply given by
                                    nf = Rc .                             (26)

Collection Efficiency of a Total Filter
     Several slightly different approaches have been used to generalize
from the collection efficiency of a single fiber to a complete filter or
matrix [42,45,46,48].  For the limiting conditions of high porosity and
many fiber layers, these methods differ only by constants related to the
geometric assumptions.
     Consider the matrix of differential length and unit cross-sectional
area illustrated in Figure 10.  Fibers oriented parallel to the flow and
field will be ineffective for particle capture.  Hence the fraction of
total fiber length that is effective in the randomly packed matrix is
given by the fraction that projects on a plane perpendicular to the flow
and field, which is
                  ?  —
                  f  I  cos<|- - ) d
-------
                                              V = BULK  AVERAGE  VELOCITY

                                              N= PARTICLE  CONCENTRATION
                    Figure 10.  Differential  matrix segment.

     With collection efficiency defined in  the  form of  Equation  (1)
a mass or number balance for particles  of a single  size can be constructed
in the following form:
     Particle Flux In = Particle Flux Out + Particles Collected           ,  (28)
                  NV = (N + dN)V +
                                                                           (29)
where — , - F = particles collected per unit length  of  fiber per unit time

     — • —  = total length of fiber projected  perpendicular to flow in the
    ira
                unit cross  section  of  matrix
V = V (1 - F)= bulk average velocity upstream of matrix
           F = fractional  packing  density (1  - porosity).
Rearranging Equation (29)  and integrating over a matrix  of  length  L
gives
                                             4FLn,
                                NT = exp
                                                  -  F).
                                                                       (30)
where P = penetration
     N  = particle co
     N. = particle concentration at matrix inlet.
N  = particle concentration at matrix  outlet
                                         32

-------
Equation (30) expresses the theoretical fractional penetration on a
number or mass basis for a single particle size.  The applicable value
of nf is selected from Equation  (25) or (26).  Since the extrapolation
from a single fiber to the total matrix inherently assumes that there
are no particle-to-particle or fiber-to-fiber interactions and contains
                                                                    O
the idealized 2/ir fiber-orientation factor, the numerical factor 4/ir
should be considered as a first  approximation for an "effectiveness
factor" to be verified or determined empirically through experiment.
POTENTIAL APPLICATIONS OF HGMS TO FINE PARTICLE EMISSION CONTROL
     Probably the most important parameter to be considered in the
evaluation of potential applications of HGMS is the particle suscepti-
bility.  Particles with high magnetic susceptibilities will be most
easily and economically collected by the HGMS method.  Significant
emissions of such particulates are found in the iron and steel industry
and the ferroalloy industry.  Emissions from various processes in these
industries contain high percentages of FeQ, Fe203, MnO, Mn20,, MngO*,
Cr90o, and other highly paramagnetic compounds.  Table 1 lists several
  C* O
potential applications with typical reported dust compositions.  The
composition obviously varies widely among the different processes.
Within a single process category the composition will vary over a more
narrow range with changes in raw feed and operating conditions.  However,
the susceptibility of heterogeneous dust particles cannot be predicted
accurately even when a bulk composition is known.  The valence state of
ferromagnetic elements is extremely important because very small per-
centages of ferromagnetic species can produce order-of-magnitude increases
in susceptibility.  Fortunately, the susceptibility of small samples can
be determined experimentally using a Faraday balance or similar device,
providing a reasonably simple way to screen potential applications.
                                          33

-------
TABLE 1.  POTENTIAL APPLICATIONS OF HGMS TO FINE PARTICLE EMISSION CONTROL

Industrial Process           References       Typical Dust Composition,
                                                     mass basis
Sinter Plant                     -        Highly variable
Blast Furnace                [71]         35-50% Fe, 12% FeO, 0.5-0.9% Mn
Basic Oxygen Furnace         [71,72]      90% Fe203, 1.5% FeO, 4% Mn304
Open Hearth Furnace          [71,72]      85-90% Fe203, 1-4% FeO, 0.5% MnO
Electric Arc Furnace         [71,72]      20-55% Fe203, 4-10% FeO, 0.5% MnO
Silico-manganese Furnace     [71,72]      4-7% FeO, 30-35% MnO
Ferro-manganese Furnace      [71,73]      6% FeO, 34% MnO
Ferro-chrome Furnace         [71,73]      7-11% FeO, 3% MnO, 29% total Cr

     Particle size also varies widely in different types of dusts, and
the susceptibility can vary with size in a single process due to
compositional changes with size.  HGMS theory predicts in general a
decrease in collection efficiency with size, but experimentation with
particular dusts will-be required to determine the minimum feasible size
to collect.  Gas temperature may also be an important factor.  Trendwise,
particle susceptibilities are proportional to the reciprocal of absolute
temperature assuming Curie law behavior [74].  Gas viscosity increases
with increasing temperature.  Both of those trends theoretically reduce
HGMS efficiency at higher temperatures, but the effects could be minor.
Again experimentation will be required to determine the technical and
economic practicalities of the process.  Theoretical aspects, however,
can play a strong role in designing experiments and equipment and
understanding experimental results.
                                     34

-------
                                   SECTION 5
                                EXPERIMENTAL WORK
BASIS OF EXPERIMENT DESIGN
     The bench-scale experimental test program was planned to evaluate
the preliminary technical and economic feasibility of applying high
gradient magnetic separation to fine particle control.  Existing methods
of fine particle collection for air pollution control are generally
characterized by a limited range of capital costs and energy require-
ments.  Although the actual values vary greatly with the specific
technology applied, the type of source controlled, and the efficiency of
particulate collection, many common applications require a capital
investment of roughly $2000 to $8000/m3/s of gas flow ($1 to $4/CFM) for
uninstalled primary equipment.  Power consumption frequently ranges from
1.6 kW to 3.2 kW/m3/s of gas flow (1 hp to 2 hp/1000 CFM).  Total mass
collection efficiencies of greater than 99 percent are typical, and the
current trend is to design for at least 90 percent collection of sub-
micron particulate.  If HGMS is to become a viable alternative to
present control methods, satisfactory performance must be demonstrated
with competitive estimates of capital cost and power requirements or
some other significant advantage over present technology must be exhibited.
     To focus the HGMS evaluation on practical operating conditions, the
upper bounds of the above rule-of-thumb capital costs and power require-
ments were adopted as tentative performance goals.  With appropriate
design information and assumptions these goals were then transformed
into tentative limitations on major operating variables.  The trans-
formation is explained in the following paragraphs.
Capital Cost
     In late 1975 when this analysis was first conducted, the largest
commercially operating HGMS systems were iron-bound solenoids with a
2.13-m (7-«) diameter bore and a 50-cm (20-inch) matrix length.
                                     35

-------
According to one vendor of commercial-scale equipment, the cost of a
unit of this type with a field rating of 0.2 T would have been approxi-
mately $240,000 including the magnet, power supply, and cooling equip-
ment [75],  This estimate indicated that the superficial gas velocity
would have to be at least 8.4 m/s (1650 ft/min) to keep the capital
                  •5
cost below $8000/m /s ($4/CFM) with a duty cycle approaching 100 percent.
Power Requirements
     The major power requirements of HGMS in a particle/gas system would
be  generation of the magnetic field and operation of the gas moving
equipment.  To produce a background field,of 0.2 T, the large iron-bound
solenoid referenced above would require approximately 20 kW electrical
input [75].  A unit of the same size has been reported to consume 200 kW
of  power to produce a field of 1.0 T [76].  These two figures provided
baseline data that were used to calculate magnet power requirements for
the field/velocity combinations shown in Table 2.
                TABLE 2.  ESTIMATED MAGNET POWER REQUIREMENTS FOR
                       SEVERAL FIELD/VELOCITY COMBINATIONS
Background Field
T
0.2
0.2
1.0
1.0
Gas Velocity
m/s ft/min
5 984
10 1968
5 984
10 1968

kW
20
20
200
200
Magnet
kW/m3/s
1.12
0.56
11.23
5.61
Power
hp/1000 CFM
0.71
0.35
7.12
3.57
     Fan power requirements are directly proportional to the product of gas
flow rate and pressure drop.  With a 60-percent efficient fan a pressure
drop of 2.5 kPa (10 inches of water) is equivalent to a power requirement
of 4.17 kW/m3/s (2.63 hp/1000 CFM).
                                      36

-------
     The bench-scale experiment design and execution were strongly
influenced by the results of this analysis.  However, since the design
of large electromagnets has been primarily for research usage in the
past (with the exception of the several units now in use in the clay
industry), cost reductions and more energy-efficient designs might be
anticipated should HGMS emerge as a widely used technology.  Accordingly
the limiting conditions for the major operating variables were initially
specified at the following levels:

          Operating Parameter              .  Limiting Condition
        Superficial gas velocity             >5 m/s (984 ft/min)
        Pressure drop                        <2.5 kPa (10 inches  of water)
        Background magnetic field            90% for submicron particles
                                             >99% for particles greater than  1  ym

     Basic oxygen furnace dust was selected as the test dust from the
list of likely candidates because it satisfied three important criteria:
(1) it is associated with a major industrial  process, (2) it is difficult
to collect by conventional methods because of the high mass percentage
of submicron particles, and (3) it is one of  the strongest candidates for
successful HGMS application because of the typically large ferric oxide
content.
SUMMARY OF PHASE I EXPERIMENTS AND RESULTS
Apparatus and Procedures
     The experimental apparatus utilized in Phase I is shown schematically
in Figure 11.  Basically, the apparatus consisted of a fluidized bed
system to redisperse the dust followed by a section in which the dust
was diluted with clean air.  The dirty air stream was then passed
through the magnetic separator and vented.  Samples were extracted
upstream and downstream of the magnetic separator and analyzed to
determine the particle size distribution and concentration.
                                        37

-------
         ROOM
         AIR -
         INTAKE
co
00
                            HEATER
                           CONTROL
                               FILTER
                      FLUIDIZED
                       BED
                FILTER
                                   DUST FEEDER
                                           DRIVE
                                           O
      SCR
      MOTOR
      CONTROL
                                                               HGMS
I ENTRAIN WENT
 SEPARATOR
U	
                        t
                                                                           r
                                                              RARTICLE
                                                              ANALYZER

                             J*
                                                                          3-
                                                               BLEED—XI*
                                     DILUTER
Q.
      -CXJ-
                                                                                                          VENT
                 F1LTER
                  C.W.
-><3	I^BLEED
                                                                                                                 INTAKE
                                                                                                                 FILTER
                                                                                              COMPRESSOR
                      Figure  11.  Experimental apparatus used in  preliminary  HGMS experiments.

-------
     A dust redispersing system was chosen for source simulation over
artificial particle generators because it allowed the use of an actual
industrial dust at a comparatively high feed rate.  The fluidized bed
system was similar in principle, although somewhat different in appli-
cation, to equipment described by Guichard [77], Blann and Moreno [78],
and Willeke and co-workers [79].  The fluidized medium was nickel  metal
powder sieved to a size range of approximately 75 to 100 ym.  Fluidizing
air was compressed and cooled, passed through an entrainment separator
and submicron filter, and metered into the bed to provide a superficial
velocity of approximately 15 cm/s.
     The test dust was obtained from the hoppers of an electrostatic
precipitator which controls emissions from a basic oxygen furnace at a
Pennsylvania steel mill.  The dust was fed into the fluidized bed by a
small screw feeder.  The fluidized nickel spheres partially deagglomerated
the bulk dust, allowing it to be entrained in the superficial air flow.
The fluidized BOF dust was diluted by mixing with a stream of clean room
air.  The incoming air was preheated to 38° C (100° F) to reduce the
relative humidity and filtered to remove some of the background particulate.
To minimize the complexity of the experimental apparatus, all of the
Phase I tests were limited to 38° C.
     The dusty air stream was passed through the high gradient magnetic
separator, which consisted of a 35.6-cm long steel canister with an
inside diameter of 8.9 cm.  The canister was positioned in the bore of
an iron-bound solenoid where the background magnetic field could be
adjusted in six increments from 0.094 to 0.965 T.  The electromagnet and
its associated power supply constituted a commercially available laboratory
HGMS unit which was leased from Magnetic Engineering Associates of
Cambridge, Massachusetts.  The filter material was 430 stainless steel
wool similar to that used in previously reported experiments [49].
     Sampling and analysis were accomplished using a custom dilution
system and a commercial optical particle analyzer.  Sequential sampling
was conducted upstream and downstream of the HGMS apparatus to determine
the collection efficiency as a function of particle size.  Since rather
long sample lines were required, information on the larger particles
(3 vm and above) was unavoidably lost.
                                        39

-------
     Dilution of the samples was required to avoid coincidence loss in
the optical analyzer.  Specifications for the dilution system were
obtained through the courtesy of Southern Research Institute (SoRI) in
Birmingham, Alabama.  In the SoRI-designed diluter the sample stream is
fed into a conical vessel where it is mixed in highly turbulent flow
with a filtered stream of dilution air.  A dilution ratio of 50:1 was
used throughout these tests.  Calibrations by SoRI [80] have shown that
this type of diluter yields representative samples very close to the
predicted dilution ratio for submicron particles.  Some transmission
losses experienced by SoRI with 2.3 urn particles during calibration
caused greater uncertainty in the predicted dilution ratio.
     Single point, isokinetic sampling was initially attempted in this
preliminary work, but a problem developed because larger particles were
drawn through the sample lines and deposited in the diluter.  These
particles would then occasionally become broken up and entrained,
causing spurious counts.  To eliminate the problem, it was necessary to
turn the sample nozzles downstream to prevent large particles from
entering the sampling apparatus.  This extreme deviation from isokinetic
sampling probably led to some bias in the sampling of even the fine
particles.  However, since the upstream and downstream particle concen-
trations were ratioed to calculate a collection efficiency, it was assumed
that the sample bias was eliminated by the ratio (i.e., that the bias
for any given particle size was a constant factor independent of particle
concentration and equal at the upstream and downstream sample points).
     The particle analyzer was a Climet Model 208A which uses near-
forward scattering of visible light to count and size particles  in
discrete increments.  The optical particle analyzer was chosen primarily
because it allowed the measurements to be made in near-real time.  The
particular unit used in this study was a special-order model including
an extra increment in the submicron range.  Data were extracted  for four
optically determined ranges of particle diameter:  0.3 to  0.5 ym,  0.5
to 0.7 ym, 0.7 to 1.0 ym and 1.0 to 3.0 ym.
     During each HGMS run three or more particle counts were taken up-
stream and downstream in sequence.  A collection efficiency was  cal-
culated from each paired data set, and the results were averaged to
                                      40

-------
obtain the efficiency for the particular run.  Although the individual
data sets sometimes yielded  penetration results differing by as much as a
factor of two, they showed no trend with time; that is, all of the data
were collected before significant loading of the matrix occurred.
Between runs the filter matrix was forward-flushed (aided by manual
rapping) with the magnet off.  Occasional duplication of runs showed
that this method of cleaning returned the matrix to its original
condition within the limits  of experimental error.  One long-term run
was made to observe the effects of uninterrupted filter loading.
Salient Experimental Results
     The ranges of important operating parameters are summarized in
Table 3.  Figures 12, 13, and 14 show results of several runs that
encompass the ranges of these variables.  The dramatic improvement in
collection efficiency upon application of a low magnetic field is most
easily seen in those groups  in which zero-field runs were made.  In-
creases in the background field further reduced the particle penetration
as expected.  Fields higher  than 0.308 T were never needed because in
all tests the collection efficiency approached or exceeded detectable
limits with the 0.308 T field.
     The effect of gas velocity was less obvious.  Efficiencies ex-
ceeding 90 percent were achieved with both low and high velocities.
Increases in velocity seemed beneficial in some test series and detri-
mental in others.  At this point in the program the effect of gas
velocity on collection efficiency was not well understood.

         TABLE 3.  RANGES OF OPERATING PARAMETERS - PHASE I EXPERIMENTS
Magnetic Field, T
Matrix Length, cm
Packing Density
Gas Velocity, m/s
Pressure Drop, kPa
(in. H20)
0-0.308
20.3
0.0088
5.9-9.3
0.7-1.3
(2.8-5.3)
0-0.214
20.3
0.0132
6.1-10.6
1.2-2.9
(4.7-11.5)
0-0.214
20.3
0.0174
5.9-10.2
1.4-3.2
(5.4-12.8)
                                      41

-------
    1.0
    0.5
   0.2
    O.I
  0.05
  0.02
                                                       V=8-4 mfy
                                                       V = 9.3 m/s
IU
z
UJ
   QOI
 0-005
  O.OO2
  aooi
        B0=0
       0.094T
              I    	i
     0.2     0.4  0.6 0.8 I        2

         PARTICLE  DIAMETER, fj.m
                                               1.0
   0.5
   0.2
    O.I
  0.05
<0.02
LU
Z
   OOI
 0.005
                                             0.002
                                             0001
                                                  0.094T
     0.2      0.4   0.6  0.8 I       2

         PARTICLE  DIAMETER^m
                                              1.0
   0.5
   0.2
                                                                                         O.I
  OO5
                                                                                             0.094T
-0.214T
 0.308T
                                                                                      O
|0.02

UJ

UJ
a.
   OOI
 0005
                                           0002
                                           0.001
     O.2      0.4  0.6 0.8 I

         PARTICLE  DIAMETER,
             Figure  12.   Results  of bench-scale HGMS tests with packing density  of 0.0088.

-------
               V= 6.1 m/s
                                                      V=8.4m/s
                                                       V=I0.6 m/s
    1.0

DC
Ul
z
UJ
a.
    0.2
    O.I
   0.05
0.02
    0.01
  O.OO5
  O.O02
   0.00
         0.094T
       O.2     0.4  0.6 0.8  I       2
          PARTICLE  DIAMETER,/*m
                                            1.0
                                               05
                                            O2
                                            O.I
                                           005
£  002
a.
                                           aoi
                                             0005
                                             0.002
                                             O.OOI
                                                  B0=0
                                                    0.094
                                                    0.2I4T
                                              0.2      0.4  0.6  OS I        2
                                                  PARTICLE  DIAMETER,/im
                                              IX)
                                                                                      0.5
                                             0.2
                                              O.I
                                            0.05
                                                                                     O
0.02
                                                                                  LJ
                                                                                  a.
                                                                                        001
                                                                                    0005
                                                                                   0.002
                                                                                    0.001
                                                                                          OO94T
                                                                                             0.214T
                                               0.2      O4  0.6 0.8 I        2
                                                   PARTICLE DIAMETER,p.m
                 Figure  13.   Results  of bench-scale HGMS tests with packing density of 0.0132.

-------
     1.0
                V=5.9 m/s
    O.2
     0.1
   0.05
<  Q02
£C
Ul
Ul
a  o.oi
  0.005
  0.002
  O.OOI
0.094T
        0.2I4T
      O2     0.4  O.6 0.8  I       2
         PARTICLE DIAMETER,  fJ.m
                                                 1.0
                                                 0.5
                                                 0.2
                                                 O.I
                                                005
                                                            V=8.l m/s
                                       002
                                     UJ
                                        0.01
                                      0.005
                                      0.002
                                               O.OOI
                                                      0.094T
                                            0.2I4T
                                          0.2      0.4   0.6 0.8 I
                                             PARTICLE  DIAMETER,
                                                                                             1.0
                                                                                   0.5
                                                                                   0.2
                                                                                    0.1
                                                                                  0.05
                                                                                              V=I0.2  m/s
                                                                                        O
2  0.02
UJ
Ul
0.
   0.01
 0.005
  OXX>2
                                                                                  O.OOI
                                                                                                0.094T
                                                                                                      i    i   i
      0.2     0.4  O.6 0.8 I
         PARTICLE  DIAMETER,
              Figure 14.   Results  of bench-scale  HGMS  tests  with  packing density  of 0.0174.

-------
     The important effect of matrix packing density can also be seen by
comparing runs with comparable velocities in Figures 12, 13, and 14.
Increasing the packing density yielded a definite improvement in col-
lection efficiency but also increased the matrix pressure drop.
     One long-term run was made to study the variation in sampling
results with time and the transient effects of matrix loading.  The
results are shown in Figure 15.  The penetration varied without any
pattern for the first eight hours of the test and then began to rise.
After 10 1/2 hours the superficial gas velocity could no longer be
maintained at its original value because of the increased pressure drop,
and the run was stopped after 11 1/2 hours.  At the end of the run the
matrix was carefully removed and weighed.  The clean weight was 120 g,
and a total of 81.8 g of dust had been collected.  Assuming the total
mass collection efficiency was constant over the 11 1/2 hour run (neg-
lecting the small decrease in the last four hours), these results
indicate that the matrix collected approximately 50 percent of its own
weight in dust before the collection efficiency was appreciably affected.
However, during this time the pressure drop across the matrix also
increased by about 50 percent.
Preliminary Economic Estimates
     A preliminary economic analysis was prepared based on the results
obtained with the median packing density matrix (F = 0.0132).  Satis-
factory collection efficiencies were obtained with the median density
matrix using superficial gas velocities of 6.1 to 10.6 m/s (1200 to 2090
ft/min).  The choice of operating velocity involves a trade-off between
operating and capital costs.  A high velocity reduces the total flow
area and thus the number of parallel collector modules, but results in a
higher pressure drop and thus greater energy requirements for fan
operation.  Lower velocities have the converse effect.  Cost estimates
were made for a hypothetical system using either low or high velocity
operation and taking into account the increased fan power requirements
with higher pressure drop.  The incremental costs of HGMS modules were
evaluated on the basis of 2.13-m (7-ft) diameter modules each of which
                                     45

-------
CT>
                 0.08
                 O.07 -
                                                                                                          2.00
                                                                                                        - 1.75
                                                                                                        -0.25
          Figure 15.  Bench-scale variation in collection efficiency and pressure drop  with matrix loading,

-------
would cost $240,000  (including  the magnet,  power supply, and heat ex-
changers) and would  require  20  kW to produce a 0.2 T background field
[75].  Since information on  matrix cleaning was not yet sufficient to
allow a detailed analysis of operating cycles, a 75-percent duty cycle
was assumed (25 percent downtime on each module for matrix cleaning),
which is comparable  to commercial clay unit operation [10].  The results
of the preliminary analysis  are summarized  in Table 4.
     Both the cost and power requirement projections shown in Table 4
are slightly higher  than the tentative goals stated previously, but
are well within the  range of potential feasibility.  As mentioned
previously, the cost of large-scale HGMS systems is not yet firmly
established, and the process has not been optimized for gas-stream
application so the cost projections could easily be ±50 percent.
Also calculations indicate that the total energy requirement of the
system could be reduced by using higher magnetic fields with lower
matrix packing density (hence,  lower pressure drop).   In the low
velocity case presented in Table 4, magnet energization accounts  for

                TABLE 4.  PRELIMINARY ECONOMIC ANALYSIS FOR
                          HGMS APPLIED TO BOF DUST CONTROL


Superficial


gas Velocity, m/s
(ft/mln)
Estimated Pressure Drop, kPa

Cost of HGMS

Total Power

(in H20)
Modules, $/m /s
($/CFM)
3
Requirements, kW/m /s
(hp/1000 CFM)
Low Velocity
6.1
(1200)
1.2
(4-7)
14,700
(6.94)
2.9
(1-8)
High Velocity
10.2
(2010)
2.5
(9.9)
8,800
(4-15)
4.7
(3.0)
approximately 32 percent of the total power requirement, and in the high
velocity case, approximately 12 percent.
                                    47

-------
     At this stage in the program it became apparent that more detailed
experimental results were needed on a larger scale system to expand the
understanding of the process and to provide more extensive and reliable
results on which to base economic projections.
DESCRIPTION OF PHASE II APPARATUS AND EXPERIMENTS
General Guidelines For Scale-Up
     The bench-scale experiments provided a baseline of information and
experience from which to design the larger system.  The following guide-
lines for the design of both the pilot-scale apparatus and the experi-
ment resulted.
     (1)  Contractual agreement stipulated a minimum system capacity of
          0.235 m /s (500 CFM).  To insure that this obligation could be
          met even if operating velocities had  to be reduced in the
          larger system, a 30-cm diameter matrix canister was specified.
     (2)  A more systematic experiment was designed to allow analysis of
          variance and direct comparison of runs that differed by only
          one operating parameter.
     (3)  Operation at four velocities in the nominal range of 5 to 10
          m/s was planned in an attempt to more accurately determine the
          velocity effect since it has an important impact on system
          optimization.
     (4)  The matrix packing density was reduced to yield lower pressure
          drops and a more equitable distribution of projected power
          requirements between the magnet and fans.  Two matrix densities
          (F = 0.005, 0.010) were chosen for evaluation.
     (5)  A magnetic field capability of 0.5 T was specified.  Zero-
          field operation was also planned to aid in the comparison of
          experimental data and theory.
                                     48

-------
     (6)  Two different types of dust were specified to broaden the
          information on potential applications.  BOF dust from the same
          source as the bench-scale test was selected as one dust.  A
          second dust was obtained from the hopper of a roof-system
          baghouse that controls emissions from electric arc furnaces at
          a Pennsylvania steel mill.
     (7)  Allowance was made for a matrix length of up to 30 cm.  Two
          matrix lengths (L = 15, 30 cm) were selected for testing.
     (8)  Plans were made for limited testing at an elevated temperature
          to allow preliminary evaluation of temperature effects.
     (9)  Selection of operating parameter levels was purposely directed
          toward lower efficiency operation than that observed in  the
          bench-scale tests to allow better observation of parameter
          effects.  When collection efficiencies exceed 90 percent
          (penetration = 0.1) good repeatability of experiments is more
          difficult and parametric effects are more likely to be masked
          by random error.
     (10) Cascade inertia! impactors were specified as the primary means
          of particle size determination since their range of particle
          size information is broader than the available optical system
          and since impactors are more widely used in field sizing
          studies.  Also the theoretical filtration and HGMS models are
          based on aerodynamic behavior of spherical particles, which is
          the same basis on which impactors are calibrated and operated.
          The optical system was also retained to provide preliminary
          results in near-real time and to allow observation of transients
          during extended runs.
Apparatus and Procedures
     To avoid the duplication of costly equipment that was already
available, the pilot plant was designed to operate as a slipstream off
the existing Particulate Aerodynamic Test Facility located at the EPA
Technical Center in Research Triangle Park, North Carolina.  This
facility basically consists of a low-speed, closed-loop wind tunnel.
Incorporated with the wind tunnel is a dust feed and redispersion

                                      49

-------
system that operates on the same fluidized-bed principle as  the system
used in the bench-scale experiments.   (In fact, the bench-scale system
was designed as a scaled-down model  of the EPA facility.) The wind
tunnel is also equipped with a baghouse that results in once-through
operation with respect to dust.   The entire system is described in
detail elsewhere [81,82], and the description will be repeated here only
to the extent that it pertains to the Phase II experiments.
     The experimental apparatus  is shown schematically in Figure 16.
The fluidized bed in this system had two separate stages. The lower-bed
was 20 cm in diameter and contained  2.5 cm of 6-mm glass beads.  It was
operated at incipient fluidization to break up the dust initially as it
was fed in by the screw feeder.   The dust then sifted up through a
screen to the 40-cm upper bed, which contained approximately 2.5 cm of
no. 9 lead shot and 2.5 to 5 cm of 700-ym glass beads.  The  larger
diameter of the upper bed lowered the gas velocity so that the glass
beads were fluidized but not elutriated, and the lead shot was near
incipient fluidization.  The upper bed further broke up the  dust ag-
glomerates.  Some early effort was devoted to optimizing the gas flow
rate and bed depths with an objective of maximizing the percentage of
fine particles in the output, but little success was achieved.  The bed
depths were finally set at the above values, and the fluidizing gas flow
was set at roughly 0.14 m /s (300 CFM).   These conditions provided the
most trouble-free operation and yielded a steady particle output with
                                                  •3
roughly 1 percent smaller than 1 pm (p  = 4.5 g/cm ).
     From the fluidized bed the dust was blown into the 61-cm diameter
wind tunnel where it was mixed with dilution air coming from the bag-
house and wind tunnel fan.  A slipstream of the dusty air was then drawn
off the wind tunnel into the pilot-scale HGMS apparatus.  The stream
passed through a cyclone separator,  which was designed to provide a 50-
percent cut point (D5Q) of approximately 5 ym (p  = 4.5 g/cm3).  The
combination of superisokinetic flow at the slipstream intake and scalping
by the cyclone reduced the percentage of large agglomerates in the  gas
stream.  The characteristics of the stream exiting the cyclone varied
                                      50

-------
    TO  .

BAGHOUSE'
WIND TUNNEL
             TEMPERATURE
              AND  FLOW o
              INDICATORS 0--
          BOOSTER
            FAN
                     BLAST
                      GATE
                 DOWNSTREAM
                 I TEST  PORTS
                                       HGMS
                                           UPSTREAM
                                           TEST PORTS
CYCLONE
SEPARATOR
                                                                               CLEAN
                                                      FROM
                                                   BAGHOUSE
                                      FLUIDIZED
                                        BED
                                                      DUST
                                                     FEEDER
                                               FLUIDIZING
                                                 AIR
        Figure 16.  Schematic representation of pilot-scale HGMS facility.

-------
from run to run with the slipstream flow rate, but typically the mass
                                                            o
median diameter was on the order of 2 to 4 ym (p  =4.5 g/cm )  and the
                                                "          3
total mass concentration was in the range of 50 to 100 mg/m .
     From the cyclone this dusty gas stream flowed up through a baffle
which was designed to break up the vortex.  It then flowed past the
upstream sample ports and into the HGMS canister, which was 30 cm in
diameter and 35 cm long in the direction of flow.  From the HGMS
canister the clean gas passed by the downstream sample ports; through a
return bend; through a blast gate, which was used for flow control;
through a booster fan; and back into the wind tunnel where it rejoined
the remainder of the dusty gas and continued around the wind tunnel loop
to the baghouse.
     The matrix inside the HGMS canister was subdivided into four 7.5-cm
long sections by expanded metal screens* which were connected by three
threaded rods running the length of the canister in the direction of
flow (see Figure 17).  A small air-operated vibrator was attached
through the duct wall via a rubber grommet to one of the threaded rods.
To construct a 15-cm long matrix, the first and third sections (count-
ing from the upstream or inlet end) were packed with the appropriate
mass of randomly oriented 430 stainless steel wool of the same type used
in the bench-scale experiments.  Imbedded in the first section were two
20-cm diameter loops of 0.6-cm (1/4-inch O.D.) copper tubing.  The loops
each had numerous 1.5-mm perforations on their underside and were con-
nected to a source of dry, compressed air via a vertical section of
tubing running up the side of the canister and through the duct wall.
The third section contained one of these loops of perforated tubing.  To
construct a 30-cm matrix, all four sections in the canister were employed,
and perforated tubes were added to the second and fourth sections.
After each run the vibrator and compressed air were used in  tandem  to
shake and blow the collected particulate off the matrix.
     To conduct an experimental run the wind tunnel was started  up  and
the slipstream fan was started.  The desired nominal velocity was  set
by adjustment of the blast gate from a reading on a stationary pi tot
tube located between the canister and the blast gate.  A 16-point  pi tot
                                      52

-------
TO EXTERNALLY
MOUNTED
VIBRATOR
                                               CANISTER
                TO  COMPRESSED
                  AIR  SOURCE
                                     EXPANDED
                                     METAL SCREENS
                                          CONNECTING
                                            ROD
                               \  PERFORATED
                               /—  TUBES
                              /
  Figure 17.  HGMS matrix construction and cleaning apparatus
                                  53

-------
traverse (eight points on each of two perpendicular diameters) was then
run at the downstream test ports.  A special  test was conducted early in
the program to compare the upstream and downstream average velocities as
well as 16- and 32-point traverses.  The results are shown in Table 5.
The apparent larger flow rate on the downstream side suggested the
possibility of in-leakage, but no leakage could be detected.  It was
more probably due to a bias on the upstream pi tot due to the slight
swirling flow component remaining from the cyclone.  In any event the
results were considered close enough to justify the use of the 16-point
downstream traverse as the routine method of flow determination.
     When the pi tot traverse was completed, the dust feed was started,
and the magnet was energized if appropriate.   A 10- to 15-minute time
interval was allowed to insure steady-state output by the dust system,
and then sampling was begun.
     The cascade inertial impactors utilized were MRI Model 1502 units
manufactured by Meteorology Research, Inc., Altadena, California.  Prior
to use the MRI substrates were coated with a thin layer of Apiezon L
gease dissolved in toluene, baked at 120° C for 2 hours, desiccated
overnight or longer, and tared.  Because of the relatively small duct
diameter (30 cm) and upward direction of gas flow, extractive sampling
was judged to be preferable to in-stack sampling.  The sample nozzles
and probes were specially constructed using recommended techniques to
minimize particle losses [83].  The impactors were supported in a
horizontal position and were moved carefully to avoid dislodging
collected particulate.  Eight-point traverses were used on the upstream
side and 12-point traverses on the downstream side, both divided between
                TABLE 5.  COMPARISON OF PITOT TRAVERSE RESULTS

                               16-point traverse           32-point  traverse
Upstream                            7.57 m/s                    7.39 m/s
Downstream                          7.84 m/s                    7.82 m/s
                                    54

-------
two perpendicular diameters.  Downstream sampling was started first and
lasted 96 minutes.  Upstream sampling was then begun and lasted 32 or 48
minutes.  Each of the impactors was operated with a nozzle and sample
rate combination calculated to yield isokinetic sampling at the down-
stream average velocity.
     After the runs the impactors were carefully dismantled, and the
substrates were desiccated overnight before weighing.  The probes were
washed with acetone, and the probe washings were dried and also desic-
cated overnight.  All of the weights were determined to 0.01 mg.  A
Mettler Model H51AR balance was used for the runs conducted prior to May
17, 1977, and then a new Perkin-ETmer microbalance, Model AD-2Z was
placed in service for the remainder of the tests.  The probe washings
averaged 7.7 percent of the total collected mass on the upstream side
and 13.0 percent on the downstream side.  This particulate matter was
most probably made up of larger particles lost in the probe bend.  Data
from the first two impactor stages (particle diameters of about 5 ym
and above) were not included in the data analysis.
     The cut points (DCQ) of the impactor stages were calculated using
the latest recommended EPA procedures [84], including published calibration
                                                                          2
factors for the MRI impactor stages [85].  Particle densities of 4.47 g/cm
                              3
for the BOF dust and 4.61 g/cm  for the EAF dust were determined from
helium pyknometer measurements (courtesy of Mr. Ray Grote of EPA) and
                                                            o
were used in the D5Q calculations.  A value of p  = 1.0 g/cm  is frequently
used in D5Q calculations and yields what is known as the aerodynamic
particle diameter.  However, in this case since the volume susceptibility
was used in magnetic calculations, it was decided that the Stokes diameter
based on the measured particle density would be a more representative
measure of particle size.  Hence, measured particle densities and Stokes
diameters were used throughout the calculations.
     Differential particle size distributions were calculated to reduce
the upstream and downstream data to the same concentration basis.  During
the sampling, identical nozzles were used on both sample trains, and
sample rates were adjusted so that D50's of corresponding upstream and
downstream stages were equal (±7 percent).  This technique avoided the
                                      55

-------
use of curve fitting procedures and allowed the calculation of particle
penetrations from a direct ratio of upstream and downstream differential
distributions at corresponding impactor stages.
     Optical particle counts in the ranges 0.3 to 0.5 ym, 0.5 to 0.7 ym,
and 0.7 to 1.0 urn were also made using the same dilution system and
Climet instrument employed in the bench-scale tests.  In these experi-
ments, however, the sample nozzles were turned upstream, and samples
were drawn at a constant nozzle velocity of approximately 5.0 m/s for
all runs.  This resulted in sample rates from approximately isokinetic
in the low-velocity runs to 50 percent isokinetic in the high-velocity
runs.  Four-point, single diameter traverses were made on both the
upstream and downstream sides.  To prevent large agglomerates from
entering the diluter and causing spurious counts, a small cyclone with a
                                        o
DgQ of approximately 1 ym (p  = 4.5 g/cm ) was located at the inlet of
the diluter.  The subisokinetic sampling rates, the single diameter
traverse, and the presence of the cyclone could have created systematic
errors in the optical data; hence, the impactor data were utilized for
primary data analysis.  The optical system was utilized to ensure..that
no upsets occurred during impactor sampling, to provide approximate
efficiency results in near-real time, and to study transient effects
during two special long-term matrix-loading tests.
     When each run was completed, the dust feed and gas flow were shut
off.  The magnet was deenergized, and the vibration and blowing system
was used to dislodge the collected particulate from the matrix and to
cause it to fall down into the cyclone hopper.  The gas flow was  then
restarted and the matrix pressure drop was compared to the value  ob-
served before the run to ensure that the matrix was clean.  If necessary
the cleaning procedure was completed.  Difficulties were sometimes
encountered in returning the matrix pressure drop to its original value.
Frequently a net gain in pressure drop of 25 to 125 Pa (0.1 to 0.5  in
F^O) was observed.  The small vibrator and blowing system were evidently
not completely adequate to overcome the combination of residual magnetism
and mechanical entrapment of dust.  The matrix loading tests conducted
later indicated that this problem probably was not severe enough  to
seriously bias the efficiency results, but the cleaning system should
still  be upgraded before any future work is undertaken.

                                     56

-------
Experiment Design
     The overall layout of the pilot-scale experimental program is shown
in Table 6.  Practical constraints prevented a complete randomization of
the experiments.  Changes in matrix length or packing density necessi-
tated a time-consuming dismantling of the system so the different matrix
configurations were separated into blocks, which are separated by the
dashed lines in Table 6.  With each matrix, the magnetic field was
always incrementally increased, since returning to a lower field could
have introduced errors because of hysteresis.  In each set of three or
four runs with the same matrix and same field, the velocities were
randomly ordered.  A few preliminary runs were made to debug the equip-
ment and establish operating procedures.  The actual test plan was
conducted between April 19, 1977, and July 18, 1977.  In Table 6 a run
number of 4201, for example, indicates the first run on April 20, while
4202 indicates the second run on the same day.
                                       57

-------
                        TABLE 6.  EXPERIMENT DESIGN FOR PILOT-SCALE HGMS

Run No.

4191
4201
4202
4211
4271
4272
4281
4282
5031
5032
5041
5042
5091
5101
5102
5111
5121
5131
5161
5171
5181
5182
5191
5192
5201
5202
5242
5251
5252
5261
5262
5271
5272
6011
6012
6021
6022
6031
6032
*6281
7111
7112
7121
7122
7131
7132
7141
7142
7151
*7181

Parti cul ate

BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
BOF
EAF
EAF
EAF
EAF
EAF
EAF
EAF
EAF
EAF
EAF
Packing
Density

0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.010
0.010
0.010
0.010
0.010
0.010
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
Matrix
Length
cm
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
30
30
30
30
30
30
30
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
Magnetic
Field
T
0
0
0
0
0.05
0.05
0.05
0.05
0.10
0.10
0.10
0.10
0.10
0.10
0.10
0.10
0.20
0.20
0.20
0.20
0
0
0
0.05
0.05
0.05
0
0
0
0.05
0.05
0.05
0.05
0.10
0.10
0.10
0.10
0.10
0.10
0.40
0
0
0
0.10
0.10
0.10
0.20
0.20
0.20
0.40
fias
Velocity
m/s
5.59
8.21
6.71
9.79
5.49
8.06
6.74
9.93
8.47
5.63
6.89
10.20
10.78
8.52
7.02
5.46
8.00
6.71
5.70
10.16
9.74
5.94
7.38
5.71
7.84
7.17
5.16
7.65
9.24
7.53
10.06
10.70
6.56
5.68
8.52
10.30 '
6.84
6.76
7.06
8.24
8.45
7.85
10.96
7.88
11.12
9.16
7.53
9.31
10.76
8.21

Temperature
°C
31.1
30.6
34.4
30.8
25.3
32.5
29.1
33.9
30.9
35.3
31.9
36.7
108.9
109.4
109.4
109.7
32.8
33.0
32.8
33.9
30.0
39.4
32.8
37.8
32.8
38.3
33.3
29.3
36.9
30.9
37.0
31.1
35.9
30.9
37.2
32.5
37.2
30.3
37.0
35.4
34.7
41.4
36.4
40.6
35.8
41.1
35.6
40.8
33.1
36.7
Matrix removed from canister, manually cleaned, and replaced before this run.
                                                    58

-------
                                    SECTION  6
                        PILOT-SCALE EXPERIMENTAL RESULTS


     Since they predict the effects of  individual parameters on col-
lection efficiency, the theoretical models discussed in Section 4
provide a logical framework within which experimental results can be
analyzed.  In this section supportive data on matrix and particle
characteristics are first presented, and then the comparison of indi-
vidual runs with theory is presented graphically.  This procedure not
only provides a convenient means of presenting all of the experimental
data but also gives an overview of the  extent to which observed experi-
mental results can be explained.  The effects of individual operating
parameters are then discussed  in more detail.
SUPPORTIVE MATRIX AND PARTICLE DATA
     The 430 stainless steel wool used  as matrix material in these
experiments is commonly used in HGMS systems.  It is a very non uniform
material, as evidenced by the  photomicrographs shown in Figures 18, 19,
and 20.  Some fibers are reasonably well approximated by the idealized
cylindrical geometry while others are more ribbon-shaped and still
others are shaped nearly like  a coil.   The theoretical investigation of
Stekly and Minervini [59] implies that  to a  first approximation the
assumed cylindrical geometry should give reasonable predictions provided
the diameter of the cylinder approximates the mean projected width of
the actual fibers.  However, Oberteuffer's [1] conclusions on the
importance of matching fiber size to particle size to achieve maximum
magnetic force implies the possibility  of different effective mean fiber
diameters for different particle sizes.  An  impractically large number
of photomicrographs would be required to define a meaningful distri-
bution of fiber sizes and shapes for incorporation into the theoretical
expressions.   Therefore, a single radius of  25 urn was selected for use
                                      59

-------
CTl
O
                    68X
100 urn
68X
lOOym
                            Figure 18.  Photomicrographs of  stainless  steel wool matrix  fibers,

-------
60X
 100
200X
50 urn
Figure 19.  Photomicrographs of stainless steel  wool  matrix fibers,

-------
CT1
I •
                            I	1
                  150X       100 ym
150X
100
                            Figure 20.   Photomicrographs  of  stainless  steel wool matrix  fibers,

-------
in the models.   This  figure is  reasonably consistent with the photo-
micrographs  and  results  in  good agreement between  theoretical  and
experimental  results.
     Experimental magnetization measurements  were  also  made  on seven
small samples of the  steel  wool  material.   The  range and  average of
these measurements  are shown in Figure  21.  The variability  of the
magnetization results from  inhomogeneities  in the  material,  cold-worked
effects, and the fact that  the  different  samples had different shapes.
In a total matrix a much higher applied field is required  to saturate
the steel wool.   Also shown on  Figure 21  are  three points  on the cal-
culated in-matrix magnetization curve for a matrix packing density on
the order of 0.01 [86].   The matrix,eventually  reaches  saturation with
an applied field on the  order of 5.6 to 6.4 x 105  A/m (7 to 8  kOe) [41,
63].
     With the low fields used in this work, the matrix was never satu-
rated.  To utilize  the theoretical model, the near field correction
parameter was thus  calculated as
                                        y0M
                                    A = 2B(31)
                                          o
where M = estimated in-matrix magnetization from Figure 21.
The values of the parameter  are  given in Table 7 for the four fields
used in the  experiments.

        TABLE 7.  VALUES OF  THE  NEAR FIELD CORRECTION PARAMETER
Background Magnetic Field, tesla                   A =

              0.05                                   2.39
              0.10                                   2.26
              0.20                                   1.98
              0.40                                   1.54
                                    63

-------
               MEASURED
               SMALL-FIBER
               MAGNETIZATION
                                ESTIMATED
                                IN-MATRIX
                                MAGNETIZATION
) O.I
1
0.2
i
0.3
A/m X IO'6
i i
0.4
i
0.5
i
0.6
i
            0.1
0.2       0.3      0.4
             TESLA
         APPLIED  FIELD
0.5
0.6
0.7
0.8
Figure 21.   Magnetization curves of steel wool material and in-situ matrix.
                                      64

-------
     Experimental determinations of particle magnetization were also
obtained for each of the two dusts.  The results are shown in Figures 22
and 23 for the BOF dust and EAF dust, respectively.  For convenience the
results are shown in the mass cgs units in which they were determined.
The particle susceptibility is determined from these data by the equation'

                                  x = 4*Ppa/H0                           (32)

where x = volumetric particle susceptibility, dimensionless SI units
     Pp = particle density = 4.47 g/cm3 (BOF) or 4.61 g/cm3 (EAF)
      a = specific magnetization, emu/g
     HQ = applied field, Oe.
Figures 22 and 23 show that the susceptibility of both dusts is rela-
tively large and is not constant with applied field.  Rather than being
weakly paramagnetic, both dusts exhibit relatively high permeability.
Note that the magnetization of the BOF dust is about twice that of the
EAF dust at any given field.
COMPARISON OF RESULTS WITH THEORETICAL MODELS
     The no-field experimental data were fit to an equation of the form
With the fiber radius a = 25 ym and all of the operating parameters
substituted into Equation (33), two adjustable parameters remain:  E,
                                      **
the matrix effectiveness factor, and A, the lumped adhesive force factor
that is required for the calculation of ec-  The matrix effectiveness
factor would be affected by variations in the tortuous path that the gas
stream must follow through the matrix.  Under the idealized assumption
of randomly oriented fibers which do not interfere with each other
aerodynamically, a theoretical value of 4/ir  is predicted.  In reality,
the value of E that is required to obtain a good fit to the data is also
influenced by the assumed mean fiber radius.  The adhesive force
                                    65

-------
   o>
   z
   o
   H-  30
   N

   h-
   Ul
   z
   o
   o
   UJ
   Q.
                     1.0         2.0          3.0


                        APPLIED FIELD.kOe
4.0
Figure 22.  Magnetization curve  of basic oxygen furnace dust.
                                   66

-------
               i.o         2.0         ao
                  APPLIED   FIELD, kOe

Figure 23.  Magnetization curve of electric  arc  furnace dust.
                         67

-------
factor was shown in Section 4 to have a predicted value in the range of
0.002 to 0.1 N/m.
     The no-field experimental data for the two dusts are plotted in
Figure 24 through 27.  Theoretical curves are shown for comparison.  For
both dusts an effectiveness factor of 0.22 was used in Equation (33).
An adhesive force factor of 0.003 provided a reasonable fit to the EOF
data, and the lower EAF collection efficiencies were fit with an ad-
hesive force factor of 0.0005.  The results suggest that the rolling
coefficient of friction is a more realistic approach and indicate that
the EAF dust does not adhere as strongly to the wire fiber surface as
the BOF dust does.  This is presumably a result of differences in the
,composition of the two dusts which give rise to different Van der Waals
force characteristics.
     Lawson's version of the HGMS trajectory model [64,65] was used to
correlate the results of the experiments conducted with a magnetic field
applied to the matrix.  Lawson's results presented in Figure 8 were
calculated with a value of the near field correction, A, equal to 0.88.
Table 7 indicates that a more realistic value of A for the lower field
data is 2, while the 0.4 T data correspond to A = 1.5.  To accomplish
the transformation of Lawson's results to the appropriate near field
correction, both Lawson's and Watson's [48,51] results were used.  First
it was verified that Lawson's and Watson's results are identical at low
values to the Stokes number by interpolating Watson's results to obtain
the curve for A = 0.88 and comparing the capture radius under identical
conditions, recalling that 2AWK = Vm/VQ.  Then using Watson's results at
A = 2.0, selected values of the capture radius, RC, were positioned on
Figure 28 along the line log K = -2.  The R  contours were then extended
                                           V*
with slopes of -1 out to values of log K where the negligible inertia
assumption is no longer valid.  For RC = 1 the contour remains straight
for all K.  The contours for RC = 4 and 8 were completed accurately using
additional Lawson results in the form log (AW) vs. log K [65] and  his
arguments that the separate parameter A has no effect on large R  .
                                                                \*
                                      68

-------
en
                           UJ
                           2
                           UJ
                           CL
                            o
                            o:
 10


 05



 0.2


 0.1


0.05



0.02


 0.01

 1.0


 0.5



 O2


  Ol


 Q05



 O02
RUN NO. 4191
   BOF

V« 5.59 m/s
LMScm
F - O.OO5
E = O22. A'=O.O03 H/m
                               O.OI
  RUN NO. 4202
    BOF


  V=67l m/s
  L = l5cm
- F=O.OO5
  E;O.22,A=O.OO3 N/ra
                                                        _L
                                  01
                                                                        _L
     02      05     102       5
          PARTICLE  DIAMETER,  fitn
                                                                                 ui
   1.0


  O5



  0.2


   0.1


"Q05



 0.02


  0.01

   1.0


  0.5



  0.2
5

   0.1


  005



  Q02
                                                     O.OI
                                                10     0
                                                     RUN NO. 42OI
                                                        BOF
                                                                                         V = 8.2I m/t
                                                                                         L= 15cm
                                                                                         F = O.OO5
                                                                                         E=0.22.A' =
                                                                  N/m
                                                                                 UJ
                                                                                 z
                                                                                 UJ
                                                                                 "-
                                                     RUN NO. 4211
                                                       BOF


                                                     V=9.79 m/s

                                                     F- O.O05
                                                     t-- O.O03, A1: 0 003 N/m

                                                    	I	1	
                                                   I     02       0.5     10     2        5
                                                               PARTICLE  DIAMETER, p. m
                                                                                                                                  10
                                Figure 24.   Zero-field pilot-scale results with BOF dust (F=0.005,  L=15  cm).

-------
 1.0
 05
 0.2
 01
O05
002
 0.01
                   N/i
   OJ    O2      O5    TO     2
            PARTICLE  DIAMETER,
                        1.0
                        O.5
                        0.2
                        O.I
                       0.05
                       0.02
                        OOI
                                                   1.0
                                                  0.5
                                                  0.2
                                                  O.I
                      005
                                                 0.02
                                                  0.01
                  10
                          - F= O.OK)
                            E=022, A'=OO03N/m
Ql
02      05     1.0    2
   PARTICLE DIAMETER,
  RUN  NO. 5191
     BOF

  8 =O
  V 7.38 ra/$
  L=I5 cm
_ F = 0.010
  E=O.H2, A'=O.OO3 N/m
                          01    02      0.5    1.0     2
                                   PARTICLE  DIAMETER,
                                                                    10
                                                                 10
    Figure  25.   Zero-field  pilot-scale  results with  BOF  dust  (F=0.010,  L=15 cm).

-------
0.0
  01
     RUN NO. 5242
        BOF
     B0=0
     V= 5.25mA
     LOO cm
      = O.OO5
         . A'=O.OO3 N/m
                                                     E=0.2Z, A'= 0.003 N/m
0.2      0.5     1.0
     PARTICLE  DIAMETER,
                1.0
                                                        0.2      0.5    10
                                                            PARTICLE DIAMETER,
                            B0=o
                            V= 9.24 ro/s
                            t=3Ocm
                                        05    1.0     2
                                     PARTICLE   DIAMETER,
     Figure  26.   Zero-field  pilot-scale results  with  BOF dust  (F=0.005,  L=30  cm)

-------
--4
ro
1 -\S
0.5
o 0.2
H
a:
ui 0.1
• •I
UJ
o.
0.05


0.02
0.01
•
-

-
RUN NO. 7111
EAF
B0=0
V= B.45m/>
L=l5cm
F = 0.005
E= O.22, A'=O.O005 N/m
1 1 i * i
                          01
0.2
0.5
I.O
    PARTICLE   DIAMETER,
                  I.Oi	
                                              2
                                              o
                                                 0.5
                                                 02
                                                 O.I
                                              ui
                                              z
                                                OO5
                                                002
                                                 001
                                                                         1.0
                                                                         Q5
                                                                         02
                                                                       o
                                                                       t-
                                                                          Ql
                                                                       ui
                                                                       2
                                                                       Ul
                                                                       o.
                                                                        0.05
                                                                        0.02
                                                                    10
                                                                         QOI
                                               RUN NO. 7112

                                                   EAF
                                                                               V = 7.65ra/»

                                                                               L-IScm
                                                                               E = 0.22 ,A=0 0005 N/ra
                                                                                  1
                                                                                          1
O.I    0.2      0.5    1.0     2

          PARTICLE  DIAMETER,
                                                                                     10
                       RUN NO. 7i2i
                         EAF

                       80=0

                       V= 10.96m/s

                       L=l5cm

                    -  F=OOO5

                       £ = 022, A'=O.OO05 N/m
                                                   0.1    0.2       0.5    1.0     2
                                                            PARTICLE  DIAMETER, /.
                                                                                              10
                                Figure  27.   Zero-field  pilot-scale  results  with  EAF  dust (F=0.005,  L=15  cm)

-------
m
The remaining contours were completed  by  extrapolating with  approxi-
 lately the same curvature as Lawson's  curves at  corresponding  values of
RC-  The extrapolated sections of  the  contours are  shown as  dashed  lines
in Figure 28.  They have not yet been  verified by trajectory calculations
and should be regarded as approximate.  A similar transformation was
accomplished to model the two runs which  were conducted at 0.4 T using a
value of A = 1.5.
     Theoretical curves for all of the HGMS runs were calculated using
the expression
                              P  ~  exp
The theoretical curves  and  experimental  da.ta are  presented in Figures 29
through 39.  The  runs are presented  basically  in  the order in which they
were conducted except that  the  two 0.4 T runs  are shown together in
Figure 39.  A fiber  radius  of 25  ym  was  used both in the determination
of R  from Figure 28 and where  it appears explicitly in Equation (34).
    \f
Equation  (34) has one adjustable  parameter, E,  the matrix effectiveness
factor.   A value  of  E = 0.09 provided a  good fit  to the BOF data, while
the EAF data indicated  a lower  value of  E = 0.07.
     It is reasonable to expect that the effectiveness factor might not
be the same for the  two dusts or  for the same  dust under with-field and
no-field  conditions.  As argued above, the value  of E is influenced by
the assumed value of mean fiber size, by the validity of the assumed
geometric orientation of the fibers, and by the assumption of no inter-
ference among neighboring fibers.  In general  all  of these factors could
be different with respect to aerodynamic and magnetic phenomena, resulting
in different values  of  E for the  with-field and no-field cases.
     The  differences in effectiveness factors  for the two dusts with  the
field energized could be related  to  the  matrix loading effects discussed
by Luborsky and Drummond [53,56]  and by  Cowen,  et al. [52,57,62].  The
results of the matrix loading tests  presented  in  Figures 40 and 41
provide additional information.   Both of the matrix loading tests were
                                                                            (34)
                                      73

-------
                -2
                                                       A = 2.0
                                                       G = 0
                                                                    i -
                        -2
-I
 0
logK
Figure 28.  Lines of constant capture radius with inertia! forces included and A=2.0.
                                        74

-------
o
t-

-------
Z
o
Z
Ul
Q-
  1.0


  05


 0.2

  O.I

0.05


0.02

001
         RUN NO. 5O3I
           BOF
         B0=OIOT
         V= 8.47m/4
         L= IS cm
         F=O.O05
         E = ttO9
O
H
<
CE
I-
Ul
LJ
a.
                                                          1.0
                                                         0.5
    0.2
                                                          0.1
   Q05
  Q02
                                                         0.01
RUN NO 503Z
   BOF
BifO.IOT
V^ 5.63 mA
L - IS cm
F= O.OO5
E-0.09
     I	
    1.0
Z
g
<
a:
   0.5
   0.2
 01
  005
  0.02
   001
      RUN NO. 5O4I
        BOF

      B^O.IOT
      V= 689mA
      L- 15cm
      F= O.OO5
      E= O.09
           I	
                                   _L
      01    O2      0.5     1.0     2
                PARTICLE   DIAMETER,
                                                          1.0
                                                         0.5
                                                         0.2
z
g
<  Ol
<£.
UJ
5 0.05
a.
                                                        OO2
                                                         001
                                                              RUN NO. 5O42
                                                                 BOF
        V^ 10.20mA
        LM5cm
        f- O.O05
        E= O.O9
             I
                                                   10
                                                         0.1
                                                                  0.2
                     0.5
                    1.0
10
                                                                    PARTICLE  DIAMETER,
              Figure  30.
                            Pilot-scale HGMS  results with BOF  dust  (BQ=0.10T).

-------
z
o
 1.0



0.5






02




 0.1
ui 0.05
z
  Q02




   0.01



    1.0
O
   0.5 -
    02
    0.1
<
oc
   0.05
   0.02
   O.OI
      RUN NO. SIO2

        BOF

      80=0.10?

      V- 7.O2 m/4

      L- 15cm

      F = O.005

      E = O.09
                                                        1.0
                                                       05
                                                       02
§   o,
                                                 oc

                                                 UI
                                                 20.05

                                                 £L
  002




   OOI



   1.0




   0.5






   Q2


Z


I   Ol
£C
I-
Ul
g 0.05
Q.




  0.02




   0.01
                                                            RUN NO Sill

                                                              BOF
        V-S.46nv%

        L: 15cm

        F=O.005

        E=O.O9
                                                                        _L
      0.1    0.2      0.5     1.0     2

               (WRTICLE DIAMETER,
                                               10      0.1    0.2      0.5     1.0     2

                                                                PARTICLE  DIAMETER,
                                                10
       Figure  31.   Pilot-scale HGMS  results with  BOF  dust at  elevated temperature.

-------
--J
00
                         001
                           0.1
02      0.5     1.0    2
   PARTICLE DIAMETER,
                                                                        O.I
                                             0.2
    0.5
PARTICLE
 1.0    2
DIAMETER
                                                                                                              10
                                 Figure 32.
              Pilot-scale HGMS results with  BOF dust (BQ=0.20T).

-------
vo
                                                                                                                        0.01
                                                                                                                      - 0005
                                                                                                                      - 0.002
                                                                                                                        O.OOI
                                                 z
                                                 o
                                                    05
                                                    0.2
£   o.i
                                                   O05
                                                   0.02
                                                    001
        RUN NO. 52O2
           BOF
        B0= OO5T
        V= 7.ITm/»
        L= 15cm
        F=OOIO
        E = O.O9

            I
                                                                         NOTE SCALE CHANGE
                                                      0.1    02      0.5     1.0     2
                                                               PARTICLE  DIAMETER,
                                                                                                0.01
                                                                                                0005
                                                                                                 0.002
                                                0.001
                                               10
                                       Figure,33.   Pilot-scale  HGMS  results  with  BOF  dust  (F=0.010)

-------
                               1.0
                              0.5
                              0.2
                            z
                            °  0.1
                            ^0.05
                            UJ
                            a.
                             0.02
                              0.01
                                                                                   1.0
                                                                                   05
                                                                                  0.2
                                                                                   0.1
                                                 005
                                                0.02
                                                 0.01
RUN NO. 5262
   BOF
B0=OO5T

V= 10.06mA
L = 3Ocm
F=O.OOS
E = O.O9
00
o
                               1.0
                              0.5
                              02
                            z
                            a:
                            I-
                            ^005
                            ui
                            a.
                              002
                              0.01
O.I
                                       0.2      0.5     1.0     2
                                           PARTICLE DIAMETER
                                                                               z
                                                                               o
                                                  1.0
                                                                                   Q5
                                                                                   0.2
                                              oc   O.I
                                              U
                                              Z
                                              Ul
                                              °-  0.05
                                                                                  002
                                                                                  0.01
                                                                             10
                                                    O.I
   0.2      0.5    1,0     2
      PARTICLE  DIAMETER
10
                                         Figure  34.   Pilot-scale  HGMS results  with BOF  dust (L=30 cm).

-------
                                1.0
                                0.5
                               0.2
                             O
                             tr
                             z
                             UJ
                             Q_
                                0.1
                               002
                               0.01
  RUN NO 6021
    OOF
  BjO.IOT
  V= l0.3Om/i
  L-IScm
  F = O.O05
  E=O.Q9

      J	
                                                                                 K
                                                                                     1.0
                                                                                    0.5
                                                                                    0.2
                                                  O.I
                                                                                 ui 005
                                                                                 a.
     0.02
     0.01
  RUN NO. 6011
    BOF
  B0=OIOI
  V= 5.68 m/s
  LMScm
- f- 0.005
  £ = 0.09

      I
                                                                                                                         0.0088
OO
O.I    02      05     1.0     2
          PARTICLE  DIAMETER,
10       O.I    0.2       0.5     1.0     2
                   PARTICLE  DIAMETER,
                                                                                                                                   10
                                                        1.0
                                                        0.5
                                                        0.2
                                                     2   0.1
                                                     a
                                                     §0.05
                                                       O02
                                                        001
                           RUN NO. 6OI2
                         - B0=O.IOT
                           V=8.53m/5
                                                              = O.005
                                                            - E=O.09
                                                                                      TO.OOM  T
                                                                                             o.oooo
                                                          0.1
                               02       0.5    1.0     2
                                   PARTICLE DIAMETER
                                                                                                       10
                                 Figure 35.   Pilot-scale  HGMS results  with BOF dust (duplicates  of  earlier runs).

-------
OO
ro
                        2
                        o
DC
t-
UJ
z
UJ
                            1.0
                           05
                           0.2
                            O.I
                          O.05
                          0.02
                           0.01
     O.I    02       OS    1.0     2
               PARTICLE DIAMETER,
                                               •z.
                                               o
                                                   1.0
                                                   0.5
                                                   02
                                                   0.1
                                               ui
                                               o.
                                                  0.05
                                                  0.02
                                                  001
                               RUN NO. 6O32

                                  BOF

                               B0=O.IOT

                               V= 7.O6 m/s

                               L = l5cm

                               F - O.OO5
                                                                             1.0
                                                                             0.5
                                                  2 O.Z
                                                  o
                                                                          UJ
                                                                             O.I
                                                                            Q05
                                                                            OO2
                                                    0.01
                                                                        10
O.I    02      0.5    1.0     2
          PARTICLE  DIAMETER,
                                                                                                                        10
                                                           J-
                                                     0.1
                                   0.2      0.5     1.0    2

                                      PARTICLE DIAMETER.
                                                                                               10
                              Figure  36.   Pilot-scale HGMS results with BOF dust  (replicate runs),

-------
a:
fc
tu
                          1.0
                         0.5 -
0.2
                          0.1
00
Go
                        O.O5
   OO2
                         0.01
 RUN NO. 7122
    EAF

 B0=O.IOT
 \1- 7.88 m/i
 L= 15cm
' F = O.OO5
 £ = 0.07

	I	
                                                                          Z
                                                                          o
                                                                             1,0
                                                                             0.5
                                                                             0.2
                                                                             Ol
                                                                         ui
                                                                         z
                                                                            Q05
                                                                            0.02
                                                                            001
                                                        RUN NO. 7131
                                                           EAF
                                                        BO=OJOT
                                                        V: 11.12 m/s
                                                        L= IS cm
                                                        F= O.OO5
  0.1     02      05    1.0     2
            PARTICLE  DIAMETER,
                        1.0
                                                 10
                                                 O.I
02
   05
PARTICLE
                                                                                                    J_
                                                                                                           _L_
1.0     2
 DIAMETER,
                                                                                                    10
                                                 0.5 -
                                                    O.I
                                 02      0.3     1.0     2
                                    PARTICLE  DIAMETER,
                                  Figure  37.   Pilot-scale HGMS  results  with  EAF  dust (B  =0.10T),

-------
                           10
                           05
                           Q2
                       2   o,
                       UJ
                       z  0.05
                       ui
                       QL
                          0.02
00
                          001
                                        en
                                        I-
                                        Ul
                                        z
                                                                               1.0
                                                                              0.5
                                                                              0.2
                                                                               0.1
                                          0.05
                                                                             0.02
                                           0.01
RUN NO. 7142

   EAF

Ho=0.20T


V*9.3I m/s

L = IS cm


f = O.OO5

£=0.07
                            O.I
0.2
     0.5

PARTICLE
  1.0     2
DIAMETER,
                                                                         10
0.1
03
   05
PARTICLE
                                                  UJ


                                                  UI
                                                  a.
                                                      0.5
                                                      02
                                                       0.1
                                                     0.05
                                                     002
                                                      001
                        RUN NO. 7ISI

                           EAF

                        B0= 0.20T

                        V= 10.76 m/s

                        LMScm

                        F = O.OO5
                   1.0    2
                   DIAMETER,
                              [ 0.0057


                              	U.
                                                                                                                            10
                                                        0.1     02      05     1.0     2

                                                                    PARTICLE   DIAMETER,
                                                                 10
                                    Figure  38.   Pilot-scale HGMS  results  with EAF dust (BQ=0.20T),

-------
z
o
t-
<
a:

UJ
z
Ul
                      1.0
                      05
                      02
                      01
                     Q05
                     002
00
01
                      001
 RUN NO. 7181

    EAF

 B0= 0.4OT


 V = 8.21mA

 LM5cm


- F-OOO5

 E=O.O7



     J	
             J_
                          J_
                        0.1    02      0.5     1.0     2

                                  PARTICLE DIAMETER,
                                        10
                                                                                                                      001
                                                                                                                    - 0.005
                                                                                                                    - 0.002
                                                                         0.01
                                                                                                                      0.001
                                                                      0.5    1.0    Z

                                                                     PARTICLE  DIAMETER,
                                                                                                                     10
                                     Figure 39.
                          Pilot-scale  HGMS  results with both dusts at BQ=0.40T.

-------
            0.10 r-
co
en
0
                              100
200            300


     TIME, MINUTES
400
600
                                                            -.2.25
                                                                                                        - 2.00
                                                                                                               S.

                                                                                                               JH
                                                                                                               a:
                                                                                                               a
                                                                                                        - 0.25
600
                   Figure 40.  Pilot-scale variation in BOF collection efficiency and  matrix  pressure

                               drop with matrix loading.

-------
00 i
                       0.8




                       0.7





                       0.6





                       0.5

-------
begun with a clean matrix, and operation was continued for several
hours.  The optical particle counter was used to make periodic col-
lection efficiency determinations.   With the BOF dust there was no
evidence of deterioration in the collection efficiency as the matrix was
loaded.  It has been suggested that this behavior,  which was also
observed in the bench-scale tests,  might be due to  the high suscepti-
bility of the dust [87].  Since the particles have  a high suscepti-
bility, they may actually distort the magnetic field sufficiently to
maintain the large field gradients  in the interstitial spaces rather
than fill the gradient region.  Thus the aerodynamic wire radius in-
creases slightly but the magnetic collection mechanism is not grossly
deteriorated.  Eventually the accumulation of particles and the pressure
drop become so large that sloughing of collected particles will probably
occur.  In the bench-scale test, which was conducted with a field of
0.094 T, the deterioration was apparently beginning when the matrix had
collected about half its own mass (see Figure 15).   In the pilot-scale
test, which was conducted with a field of 0.4 T, the matrix had col-
lected nearly twice its own mass of particulate when the run was stopped,
and no deterioration in collection efficiency had been observed.  It is
possible that large particles which were not within the range of the
optical system were being sloughed off at this stage.
     The matrix loading test with EAF dust yielded  distinctly different
results.  Figure 41 indicates a gradual deterioration of collection
efficiency from the start of the test.  The susceptibility of the EAF
dust is not as large as that of the BOF dust so the particles apparently
do not contribute as significantly to the production of field gradients.
When the EAF loading test was stopped after 450 minutes, the matrix had
collected about three times its own mass in particulate.  (Although the
collection efficiency was lower in the EAF test, the inlet mass concen-
tration was higher; hence, more dust was collected.)  Since the col-
lection efficiencies determined by inertia! impactors represent integrated
values over 2-hour runs, the theoretical capture radius calculated for

-------
clean fiber conditions is larger than the actual EAF dust capture radius.
In the theoretical equation the lower matrix effectiveness factor provides
a correction.
     It should be noted that nearly all of the data correspond to
capture radii that are in the extrapolated portion of Figure 28.
Considering that the extrapolations are a possible source of error, the
overall fit of the experimental data to the theoretical expression is
quite good.  There is a fairly consistent trend to overpredict the
collection efficiency of the larger particles, particularly in the low-
field runs.  Assuming that the extrapolated capture radii are not
significantly in error, there are at least two possible explanations for
this problem.  If the collected particles do contribute new regions of
high gradient, the gradients are most effective over a dimension
approaching the particle size.  Considering Oberteuffer's arguments on
the relative size of collector and particle [1], this effect is not as
strong for larger particles, and they are thus more vulnerable to
reentrainment (or failure to be collected) as the matrix becomes loaded.
Secondly, the possibility of particles bouncing off the wires or knocking
others off has not been included in the model.  This phenomenon could
possibly be significant with the larger, high momentum particles.
DISCUSSION OF INDIVIDUAL PARAMETER EFFECTS
     Analysis of individual parameter effects requires comparison of two
or more experimental runs.  Many combinations are possible because of
the experiment design.  In this section representative comparisons are
made, and the results are discussed.  Before drawing conclusions from
these comparisons, it is important to consider the repeatability of
experiments conducted under identical conditions.  Because of time and
cost limitations it was not possible to duplicate all runs.  Instead a
set of replicates was run at one set of conditions in the central range
of the operating parameters.  Figure 42 shows the results of these
replicate runs.  The solid symbols represent the results of three
consecutive runs with the same matrix.  In each case the field strength
was set at 0.1 T.  The temperature ranged from 30.3 to 37.2 ° C and
                                     89

-------
        o:
        I-
        LU
        z
        UJ
        Q.
             1.0
            0.5
            0.2
            O.I
           0.05
0.02
           0.01
          0.005
          0.002
          0.001
        RUN  NOS. 5041,6022,6031,6032

                    BOF

        B0=O.IOT

        V= 6.76-7.06 m/s

        L= 15cm

        F= 0.005

        T= 30.3-37.2 °C
              O.I     0.2      0.5     1.0      2

                          PARTICLE  DIAMETER,
                                                 10
Figure 42.  Comparison  of collection efficiency results  of replicate runs,
                                       90

-------
the velocity from 6.76 to 7.06 m/s.  These ranges indicate the limits of
controllability of the two variables.  The open symbols are the results
of a test run 4 weeks earlier with a different matrix which was sup-
posedly identical.  The field strength was also 0.1 T, the temperature
was 31.9° C, and the velocity was 6.89 m/s.
     Differences in the three runs with the same matrix reflect random
error due to variations in dust characteristics, fluctuations in the
dust feed rate and size distribution during individual runs, errors
associated with the sampling and weighing process, and probably other
unknown factors.  The additional run with another matrix reflects all of
the above sources of error plus the possibility of matrix-to-matrix
variation.  It is apparent that the matrix variation was significant.
     Three other sets of duplicates were run across the same two
matrices at three different velocities.  Pooled standard deviations for
the replicates in Figure 42 plus the three sets of duplicates are reported
in Table 8.  Dividing the standard deviation by the mean penetration in
each case gives the coefficient of variation, a measure of the variability
as a fraction of the mean penetration at each diameter.  Statistics were
not calculated for particle sizes less than 0.4 ym since data in this range
could not be obtained during all runs.

          TABLE 8.  POOLED STATISTICS ON PARTICLE PENETRATION

Nominal Particle Diameter, ym
Standard Deviation of Penetration
Coefficient of Variation
Degrees of Freedom
0.4
0.100352
0.229
5
0.8
0.047474
0.421
6
1.8
0.07766,5
0.994
6
3.6
0.169624
1.055
5
     Figure 43 shows the important practical relationship among matrix
pressure drop, matrix length, packing density, and superficial gas
velocity.  The data were collected at the beginning of the first several
runs with each matrix and thus represent clean matrix conditions.  The
                                      91

-------
   3.0
-s

°M


E
u

X
cc

I


u.
o
   2.0
Ul
a

a.
o
UJ
cr

05
c/)
UJ
a:
a.
    1.0


   0.8



   0.6
z  0.5
01
_I

t_  0.4
   0.3
   0.2
    O.I
                                                A
                                                        F= 0.010
                                                            F= 0.005
                             3    456     8   10


                          SUPERFICIAL   GAS VELOCITY,  m/s
                                                                    20
30
    Figure 43.  Correlation of matrix  pressure drop with gas velocity,
                 matrix  packing density and matrix length.
                                             92

-------
straight lines running through the data were obtained from linear
regression of all of the data shown expressed .in logarithmic form.
The equation is


                             U~= 13.07 F1'32 V1'79                      (35)

with the pressure drop expressed in cm H20, the matrix length in cm,
and the velocity in m/s.  The pressure drop data also provide an ad-
ditional indication of matrix variation among runs.  Of particular note
are the filled circles on the lower line, which are the data for the 30-
cm matrix.  Assuming that the linear relationship between pressure drop
and matrix length is correct, these data suggest that the 30-cm matrix
was more porous than the others, possibly due to leakage around the
extra cleaning rings or other construction anomalies.
     The experimental collection efficiency results support the contention
that the construction of the 30-cm matrix might have been faulty.
Figure 44 compares the penetration results of the 15-cm and 30-cm
matrices at the same velocity for both the no-field and with-field
cases.  Theoretical arguments suggest that in either case doubling the
matrix length should result in the penetration being squared.  For
example, a penetration of 0.2 would be improved to 0.04.  The improve-
ment with the longer matrix is not that dramatic in either case shown in
Figure 44.  Since there is a question about the effective porosity of
the 30-cm matrix, the quantitative effect of matrix length remains
uncertain.  Increasing the matrix length definitely improves collection
efficiency but perhaps not as much as the theory suggests.
     Matrix density effects at one velocity are shown in Figure 45, with
and without a magnetic field.  Again theoretical considerations imply
that doubling the matrix density should square the penetration.  Analysis
of Figure 45 reveals approximately that quantitative result in the no-
field case.  With the magnetic field applied, the improvement in efficiency
achieved by doubling packing density is even greater than predicted.  In
other comparable runs the improvement in efficiency was at least equal
to that predicted.
                                    93

-------
IQ
o
h-
cc
     z
     UJ
     0.
   1.0
         0.5
   0.2
   O.I
 0.05
        0.02
        0.01
   0.05
0.002
                                                  = l5cm
                                                 L= 30cm
               SYMBOL   RUN NO.
                  o      4202
                  A      5251
 fe,T     V,m/s    T,«C
  0       6.71    34.4
  0       7.65    29.3
       0.001
            J_
                       _J_
_L
                                   _L
          O.I     0.2       0.5      1.0     2
                      PARTICLE DIAMETER,
                                                   10
                                                                   1.0
                                                                   05
                                                                   0.2
                                                                   O.I
                                                                z 0.05
                                                                o
                                                          tr
                                                          UJ0.02
                                                          UJ
                                                          a.
                                                            0.01
                                                            0.05
                                                                0.002
                                                                 0.001
                                                                                                          L= I5cm
                                                                                                           L=30cm
                                                                  SYMBOL
                                                                      O
RUN NO.
 4281
 5272
B0,T
 0.05
 0.05
V,m/s
 6.74     29.1
 6.56     35.9
                                                                               _L
                                                                                        I
                                                                       0.2      0.5     1.0     2
                                                                           PARTICLE  DIAMETER,
                                                                                 10
             Figure 44.   Experimental  effect of matrix length on collection efficiency with and without
                          an applied magnetic field.

-------
cn
o

h-
      LU

      2

      IU
           1.0
          0.5
          0.2
           O.I
         0.05
         Q02
          0.01
        Q005
        0.002
        0.001
          SYMBOL


             O


             A
RUN  NO.


 42O2


 5191





   i
BO,T

 o

 o
                               j_
                                                  F= 0.005
             0.1      0.2       0.5     1.0      2

                        PARTICLE DIAMETER, fj.m
                                                   F= 0.010
 V, m/s    T,°C


  6.71    34.4


  7.38   32.8





_J	I	
                                                     10
                                                                     1.0
                                                               0.5
                                                               0.2
                                                                0.1
                                                              0.05
                                       o

                                       I-
                                                           E °-°2
                                                           z
                                                           UJ
                                                           CL

                                                               0.01
                                                             OD05
0.002
                               0.001
                                                                         SYMBOL


                                                                            O
RUN NO


 4281


 5202
0.05


0.05
                                                                                                             F= 0.005
                                              0.1     0.2      0.5    1.0     2


                                                         PARTICLE  DIAMETER,
                                                                                  10
             Figure  45.   Experimental effect of matrix packing density on collection  efficiency with and without

                          an applied magnetic field.

-------
     The effect of increasing magnetic field is shown in Figure 46 for
both dusts.  The particular runs were selected because they spanned the
widest range of fields.  The effect of increasing field was not as
dramatic as in some of the bench-scale runs because the packing density
was much lower in the pilot tests.   The experimental  scatter also masks
the effects to some extent.  The reversed position of the 0.2 and 0-4 T
EAF data was contradicted by the optical  particle analyzer and was
probably due to sampling error since there is no apparent reason for
this behavior.
     Despite the scatter the beneficial effects of increasing field are
apparent.  The theoretical effect of increasing magnetic field is not
straightforward.  To a first approximation the capture radius is pro-
portional to the product of the particle and wire magnetizations, and
Figures 21, 22, and 23 show that all of the magnetizations are dependent
on applied field to some power less than 1.  As both the particles and
wires approach saturation, the beneficial effects of increasing field
become less significant.  Hence beyond some point dependent on the
particular dust, it would be more beneficial to adjust other operating
parameters in order to improve collection efficiency.
     Velocity effects are the most complex of all theoretically.  In
both the no-field and with-field cases, velocity has two compensating
effects.  Without a field, increasing velocity increases the Stokes
number and thus enhances the probability of a particle colliding with a
wire.  Unfortunately, it also increases the probability of the particle
being reentrained.  In the practical HGMS system the effects of velocity
are perhaps best analyzed with reference to the contour plot of Figure
47.  Velocity appears as a squared term in the denominator of W and a
linear term in the numerator of K.   Hence starting from any point on the
log W vs. log K plot, increasing velocity implies that one move down two
arbitrary units and to the right one unit.  Normally this will result  in
a reduction in the capture radius as illustrated by the movement from
point 1  to point 2 on Figure 47, corresponding to a doubling of velocity.
However, in the extrapolated portion of the plot when the negative  slope
                                     96

-------
IXD
          0.05  -
       O
       o:
       2
       Ui
       Q_
                                BOF DUST

                             RUN NO.  BQ.T

                                       0

                                      0.05

                                      0.10

                                      0.20

                                      0.40
         0.02  -
          0.01  -
0.005 -
0.002 -
          0.001
    01
                    0-2      0-5     1.0     2

                        PARTICLE  DIAMETER,
                                                        z
                                                        o

                                                        <
                                                        cr
                                                        H-
                                                        tu
                                                        z
                                                        UJ
                                                        Q.
 005  -
 0.02  -
                                                            0.01 -
0.005 -
0.002 -
                                                           0.001
           0.2       0.5    1.0    2

                PARTICLE  DIAMETER,
10
               Figure 46.  Experimental  effect of applied magnetic field on  the collection efficiency of both  dusts.

-------
          en
          o
             -I
             -2
                                                    A=2.0


                                                    G = 0
                     -2
 0

log K
Figure 47.  Illustration of the theoretical effect of increasing gas velocity.
                                     98

-------
of the low capture radius contours becomes larger in magnitude than 2,
increasing velocity actually results in an increase in capture radius.
Physically this is due to beneficial inertia! effects of particles
approaching in the projected area of the fiber.
     Figure 48 illustrates the experimental effects of velocity.  These
plots are not typical in the strict sense of the word because a few
results showed contradictory effects.  In general, however, increasing
velocity has no strong effect on the smaller particles, which correspond
to the extrapolated portion of Figure 47.  Higher velocities tend to be
detrimental to larger particle collection when the capture radius is
greater than 1.  The same behavior is true with no field apparently due
to the greater tendency of larger particles to be reentrained.
     Three of the four elevated temperature runs indicated the possibility
of a beneficial temperature effect especially with the larger particles.
The fourth comparison at the highest velocity showed no significant
effect.  The theoretical arguments predict a very small decrease in
collection efficiency when the temperature is raised from 30° C to 110°
C.  The collection matrix was particularly difficult to clean after the
four higher temperature runs, suggesting that for some reason the
particles adhered more strongly to the matrix.  Industrial gas cleaning
applications of HGMS would be more economically attractive if the
process could be operated at temperatures well above the 110° C used in
these experiments.  If the observed improvement in efficiency can be
shown to be real in more extensive testing, it will be an important
result.  However, the observed difficulty in matrix cleaning would have
negative implications.
OPERATING DIFFICULTIES
Operating problems were encountered in three separate areas of the total
pilot-plant system.  The first two were associated with the dust generation
and wind tunnel system.  While these problems have no direct  implication
toward application of the HGMS process, they do represent limitations on
future pilot-plant work that could be significant.  The third difficulty
was associated with the HGMS matrix cleaning system and suggests the
need for considerably more developmental work in this area.
                                     99

-------
o
o
       z
       o
       UJ
       z
       UJ
       a.
           1.0
           0.5
           0.2
           O.
          0.05
          0.02
          0.01
         0.005
         0.002
         0.001
SYMBOL
   O
   D
    _L
RUN NO.
 4191
 4202
 4201
 4211
   _L
0
0
0
0
J_
V,m/s
 5.59
 6.71
 8.21
 9.79
 _L
              31.1
              34.4
              30.6
              30.8
             O.I    O.2       0.5    1.0     2
                         PARTICLE  DIAMETER,
                                           10
                                                z
                                                o
                                                oc
                                                »-
                                                UJ
                                                z
                                                UJ
                                                a.
                                                                     1.0
                                                                     0.5
                                                    0.2
                                                    O.I
                                                   0.05
                                         0.02
                                                   0-01
0.005
                                                  0.002
                                                                   0.001
SYMBOL
   O
   a
RUN  NO.
 4271
 4281
 4272
 4282
B,T
0.05
0.05
0.05
0.05
                                                                                                     V, m s  T,°C
                                                                                                      549   25.3
                                                                                                      6.74
                                                                                                      8.06
                                                                                                      9.93
                                          29.1
                                          32.5
                                          33.9
                                            0.1     0.2       0.5     1.0    2
                                                        PARTICLE  DIAMETER,
                                                                                10
    Figure  48.   Experimental effects of gas velocity on collection  efficiency with and without an applied magnetic  field.

-------
     The capabilities of the fluidized bed dust generation system have
been studied in some detail using fly ash as the test dust [82].  With
the iron oxide dusts used in these tests, however, several new diffi-
culties were encountered.  First, dust feed rates were limited to about
2 kg/hr.  Higher feed rates were attempted so that higher dust concen-
trations could be studied, but the dust would accumulate on the lower
bed and eventually force shutdown.  The percentage of submicron par-
ticles in the system output was also much lower than desired.  Several
different combinations of fluidizing media and fluidizing gas flow rate
were tried with little improvement.  Workable conditions were finally
achieved by keeping the feed rate as high as possible without causing
plugging, reducing the wind tunnel dilution air flow to as low a level
as possible, and using the slipstream cyclone to remove large agglo-
merates ahead of the H6MS canister.  With the present facilities the
system cannot be used to study the collection of higher concentrations
of submicron particles that would be more typical of industrial conditions.
     The second problem area involved the higher temperature work.  As
the wind tunnel is presently constructed, steam can be added to the air
to raise the humidity, but there is no means of determining how much
steam is being added or what the absolute humidity is when the gas
temperature is greater than 50° C.  The humidity can be determined with
a sample train, but it cannot be set at a desired point before the run
is made.  In this experimental work the humidity was allowed to fluc-
tuate with the ambient conditions during the lower temperature work, and
small amounts of steam were added in the higher temperature runs so that
the humidity would not be extremely low.  Better control of the steam
addition and real-time monitoring of the air humidity will be needed
before the system can be used to simulate industrial gas streams at an
elevated temperature and a realistic percentage of water vapor.
   '  Since the bench-scale HGMS matrices were cleaned by manual rapping,
the quantitative requirements of matrix cleaning were not well defined
when the pilot unit was designed and constructed.  As a result, the
vibrator and air blowing system installed proved inadequate.  Transmission
of vibrational energy from the externally mounted vibrator to the matrix
                                       101

-------
was poor, and the air blowing system was not sufficient to completely
overcome the residual magnetism holding the particles to the matrix
wires.  These same concepts can probably be applied successfully however,
with the experience gained from the pilot-plant operation and the
quantitative background provided by the reentrainment modeling studies.
     The complex relationship involving matrix length, packing density,
particle susceptibility, particle size, and particle mass concentration
needs considerably more attention.  Matrix cleaning is obviously more
difficult with higher packing densities or longer matrices.  Higher
particle concentrations will require more frequent cleaning, and the
difficulty of cleaning is most probably related to particle size and
susceptibility.  If parallel, cyclical  systems are to be applied to fine
particle control, all of these factors  will have a strong effect on the
duty cycle of the individual magnets and the design of the auxiliary
cleaning apparatus.  The economic viability of the process could well
hinge on the successful demonstration of either a cyclical system with
reasonable duty cycles and simple mechanical construction, or preferably
a continuously operating system based on the carousel principle described
in Section 4 or some other innovative design.
                                   102

-------
                                    SECTION 7
                        ECONOMIC COMPARISON OF HGMS AND
                        CONVENTIONAL CONTROL TECHNOLOGY
BASIS OF ECONOMIC COMPARISON
     To compare the capital cost and power requirements of alternate
methods of fine particle control, one must insure that the estimates are
based on equipment that will achieve comparable results in a specific
application.  The design of particulate control equipment is usually
highly specific to the particular application.  For example, the design
parameters of an electrostatic precipitator (ESP) for a coal-fired
boiler might be quite different from those for a basic oxygen furnace
because of differences in particulate and gas stream characteristics.
HGMS performance data are currently available only for the collection of
particulate that originated in basic oxygen and electric arc steel-
making furnaces.  Estimates of capital cost and energy requirements
derived from these data should therefore be compared only to estimates
for conventional technology applied to BOF's or EAF's.  Even for a
singular type of application, differences will certainly exist in the
actual cost and energy requirements of equipment designed for specific
furnaces.  The estimates presented in this report are therefore intended
to serve as an approximate economic comparison of HGMS with conventional
control technology applied only to steel-making furnaces.
     The preliminary nature of this analysis combined with the uncer-
tainty that would have been introduced by comparing different existing
industrial applications (many of which are retrofits) ruled out an
inquiry to the iron and steel industry to determine actual cost and
energy requirements of existing BOF and EAF particulate control equip-
ment.  Instead the analysis was based on a study conducted by the
                                    103

-------
Industrial Gas Cleaning Institute, (IGCI) for EPA in 1971  [88],   The
published report of this study contains actual  vendor quotations for
high-efficiency electrostatic precipitators and venturi  scrubbers
designed to control particulate emissions from a typical  basic oxygen
furnace.  At the time the study was conducted there were no baghouses in
use on BOF's in the United States.  Electric arc furnaces  were not
specifically addressed in the IGCI report.  Although the particulate
matter emitted from the two types of furnaces is similar,  the cost and
energy requirements for comparable control efficiencies  could be some-
what different.
     The IGCI study reports flange-to-flange control device costs, total
installed capital costs, and operating costs in 1971 dollars.  Since
information on HGMS auxiliary equipment (ducts, dampers,  collected dust
handling equipment, etc.) and installation costs are not yet established
for particulate control applications, the comparison presented here is
limited to flange-to-flange device costs and operating power require-
ments.  Also since HGMS collection of airborne particulate has been
demonstrated only on bench-scale and pilot-scale equipment, the scale-up
factor is very large.  The full-scale HGMS estimates were based on
modules consisting of commercially available magnets (complete with
power supplies and cooling equipment) arranged in a multi-path parallel
flow configuration for cyclical operation.  A 70-percent duty cycle
was assumed.  A continuously operating system might offer better
utilization of magnetized volume and hence lower capital  costs, but
the application of continuous units to gas streams has not yet been
demonstrated.
     All equipment costs reported here have been escalated from their
original reported date  of applicability to mid-1977 using the Marshall
and Swift equipment cost index [89].
COST AND ENERGY ESTIMATES FOR CONVENTIONAL TECHNOLOGY
     The IGCI study considered several alternative designs.   Electro-
static precipitators are installed only in open-hood BOF systems  to
avoid explosion hazards.  Wet scrubbers and, presumably, high gradient
                                     104

-------
magnetic separators can function safely and effectively in either open-
hood or closed-hood systems.  To allow better comparison, the ESP and
venturi scrubber with the most nearly comparable design features (open-
hood system, largest gas flows, equal collection efficiencies) were
selected.  A summary of the ESP and, scrubber designs is presented in
Table 9.  The 99.9 percent collection efficiency specified by the IGCI
is a very stringent requirement which in most cases will exceed the EPA
new source performance standards for basic oxygen furnaces [90].  The
designs reported by the IGCI may be optimistic in that higher pressure
drop requirements have been reported for wet scrubbers operating at
equivalent or lower collection efficiencies [91,92].
COST AND ENERGY ESTIMATES FOR HGMS
     The mathematical model that was used to analyze the pilot-scale
HGMS experimental results also provides a quantitative approach to the
optimization of process operating conditions for a particular appli-
cation.  The process cannot be completely optimized, however, until
additional information is developed on alternative magnet designs
oriented directly to gas-stream application and until a wider experi-
mental data base is obtained to further verify the model.  The HGMS
units presently in service in the clay industry are basically of a
singular design:  2.13-m diameter iron-bound solenoids that require 400-
500 kW to produce applied fields on the order of 2 T over a pole gap
compatible with a 50-cm matrix.  For fine particle collection in gas
streams, much lower fields may be sufficient, and the approach to cost-
optimized design could be significantly different.  Also, because of the
large quantities of gas that must be cleaned, matrix pressure drop is a
more critical parameter that could also affect magnet design and matrix
configuration.
     To conduct this analysis cost estimates were obtained from a com-
mercial vendor [93] for large iron-bound solenoids of the conventional  de-
sign with modifications that would make them more compatible with  large gas
flows and with the operating conditions proven effective in the pilot-
scale tests.  The estimates are summarized in Table 10.  Power requirements
                                       105

-------
                    TABLE 9.  DESIGN AND COST SUMMARY OF CONVENTIONAL BOF CONTROL  EQUIPMENT
Control Device
System of Units
Process Gas Flow
Outlet Particulate Loading
Total Mass Collection Efficiency
Pressure Drop
Total Power Requirement
Flange-to-Flange Cost (1977 $)
Electrostatic Precipitator
SI
481 m3/s^
0.011 g/m
99.9%
- -
1525 kl/4)
English
1,020,000 ACFM^
0.0046 gr/ACF^
99.9%
- -
2045 hp^
$1,944,000
Wet Scrubber
SI
309 m3/s^2^
0.016 g/m3^2^
99.9%
10.5 kPa^3^
5605 kW^
English
655,000 ACFM^
0.0072 gr/ACF^
99.9%
42 in. H20^3^
7518 hp(5)
$640,000
(1) At precipitator outlet:  343° C (650° F), -1.5 kPa (-6 in.  H20) gauge,  11  percent water by volume.
(2) At scrubber outlet:  82° C (180° F), -11.2 kPa (-45 in.  H20)  gauge,  57.4 percent water by volume.
(3) Across scrubber and mist eliminator.
(4) Report does not state explicitly what this includes.   It definitely  includes  precipitator energization
      and pressure drop, but may also include some duct pressure  drop.
(5) Report does not state explicitly what this includes.   It definitely  includes  liquid  pumping and scrubber
      pressure drop, but may also include some duct pressure drop.

-------
  TABLE 10.  COST ESTIMATES FOR LARGE IRON-BOUND SOLENOIDS APPLICABLE
             TO FINE PARTICLE CONTROL
Matrix Diameter
Matrix Length
Field Capability
Cost, (1977)
 3.15 m
 15 cm
 0.4 T
$320,000
 3.15 m
 30 cm
 0.4T
$370,000
of these units were scaled from existing data on 2.13-m diameter by 50-
cm matrix length units.  Several factors can influence magnet power
requirements, but with similarly designed solenoids to a good approxi-
mation the power requirement for a given field is directly proportional
to the diameter and to the length of the cylindrical magnetized volume.
Reported data for the 2.13-m units and estimates for the two larger
units are shown in Figure 49.  References for the data are indicated on
the figure.
     The operating temperature and gas volume of the I6CI electrostatic
orecipitator were taken as operating conditions for the HGMS.  Theo-
retically higher temperatures will have an adverse effect on HGMS
performance by increasing gas viscosity and decreasing particle magnetic
susceptibility.  A correction was made for the viscosity difference but
the temperature dependence of particle susceptibility has not yet been
determined.  The design was also based on basic oxygen furnace dust
since that was the basis of the IGCI designs.  The electric arc furnace
dust used in the pilot-scale tests would require a more expensive design
in terms of capital investment and/or power requirements because of the
lower magnetic susceptibility of the dust.  This does not necessarily
mean that all electric arc furnaces would be more expensive to control
or that the design developed here would be appropriate for all BOFs.
Since the composition of the dust varies from furnace to furnace, the
magnetic susceptibility most probably does also.
     Since the IGCI designs stipulated 99.9 percent collection efficiency
without stating the particle size distribution, the HGMS design was
                                      107

-------
 1000 i

 900

 800

 700


 600


 500



 400




 300






 200
85 100
O
0_


O
LJ
or


o
LJ
or
   90

   80

   70


   60
   40
   30
   20
    10
     01
                                          x 15cm
0.2      0.3   0.4  Q5 0.6   0.8   1.0


          APPLIED  FIELD, T
                                                                2.0
3.0
          Figure 49.   Reported and estimated power requirements for

                      iron-bound solenoids of various configuration.
                                      108

-------
based on a reasonably comparable specification of 99.9 percent collection
of 1.0 ym BOF particles.  The procedure for estimating flange-to-flange
costs and power requirements is outlined below.
     (1)  Basis:  99.9 percent collection of 1.0 ym particles
     (2)  Design equation:
EFLR^
                              (34)
                              P = exp
     (3)  Determination of operating parameters:  Seven of the 3.15-m
                                              o
          diameter units will handle the 481 m /s gas flow with a
          superficial gas velocity of 8.8 m/s.  With BOF dust and the
          same matrix material used in the pilot-scale experiments,
          E = 0.09 and a = 25 ym.  For either F * 0.005, L = 30 cm or
          F = 0.010, L = 15 cm, the field requirement is determined
          as follows.
          particle diameter                  1.0 ym
          required collection efficiency     99.9%
          from Equation (34), R  =           1.27
                               \*
          from Equation (18), K  =           2.94
                           log K =           0.47
          from Figure 50,  log W =          -0.64
                               VI =           0.229
          from Equation (16), HQ =           3.5 x 105 A/m
                              B  =           0.44 T
                               o
          For additional reference, with these operating parameters the
          calculated collection efficiency of 0.5 ym particles is
          98.7 percent.
     (4)  Calculation of power requirements:
          (a)  With a shorter, more dense matrix (L = 15 cm, F = 0.010):
               From Figure 49 power per magnet is 27.5 kW for a total
               magnet power requirement of 192 kW.  From Equation (35)
                                      109

-------
           5 -
      log W
           -2h
                                                       A=l.5
                                                       G = 0
                    -2
-I
  0

log K
Figure 50.  Lines of constant capture radius with inertial  forces included
            and A=l.5.
                                   no

-------
               corrected for lower gas density at 343° C, matrix pressure
               drop is 1.1 kPa  (4.4 in. H20).  With 60 percent fan
               efficiency, fan  power is 885 kW.  Total power requirement
               is thus 1077 kW.
          (b)  With a longer, less dense matrix (L = 30 cm, F = 0.005):
               From Figure 49 power per magnet is 55 kW for a total
               magnet power requirement of 385 kW.  From  Equation  (35)
               corrected for lower gas density at 343° C matrix pressure
               drop is 0.9 kPa  (3.5 in. H20).  With 60 percent efficiency,
               fan power in 702 kW.  Total power requirement is thus
               1087 kW.
     (5)  Calculation of capital cost:  Since the power requirements of
          the two alternatives  considered are virtually identical, the
          shorter, more dense matrix is clearly superior because of the
          lower cost of the magnets.  With a 70 percent duty cycle, a
          total of ten magnets would be required for a total flange-to-
          flange cost of $3,200,000.
COMPARISON OF HGMS AND CONVENTIONAL TECHNOLOGY
     The estimated cost and energy requirements of the HGMS device are
compared with conventional technology in Table 11.  Since the assumed
gas temperature, pressure, and water content were not the same in the
wet scrubber as in the electrostatic precipitator and HGMS, normal-
ization of costs and energy requirements by the actual gas flow in each
device does not provide a direct comparison.  To refer the estimates to
an equivalent application the scrubber gas volume was corrected to the
operating gas conditions of the precipitator and HGMS.
     Table 11 indicates that on a flange-to-flange basis the scrubber
has the lowest capital cost.  This may be misleading, however, since the
scrubber would require an associated water treatment system not required
by either the precipitator or HGMS.  With the design utilized for these
estimates the capital cost of the HGMS is somewhat higher than the pre-
cipitator but still well within the economic ballpark considering the
accuracy of the estimates.  The estimates of power requirement show that
the scrubber requires several times more energy than either the pre-
cipitator or the HGMS to do a comparable job.  The HGMS is estimated to

                                      111

-------
           TABLE 11.  COMPARISON OF H6MS AND CONVENTIONAL TECHNOLOGY

Device
Collection Efficiency, %
3
Flange-to-Flange Cost, $/m /s
$/CFM
Power Requirement, kW/m /s
hp/1000 CFM
ESP
99.9
4042
1.91
3.2
2.0
Scrubber
99.9
2762
1.30
24.2
15.3
HGMS
99.9
6653
3.14
2.2
1.4
All estimates are referred to clean gas at 343° C (650° F),  -1.5 kPa
(-6 in. HpO) gauge, 11 percent water by volume.
 be more efficient than the precipitator in terms of energy utilization.
 Again, however, the accuracy of the estimates must be considered.   Also
 the precipitator estimate may include some unintended contribution for
 duct pressure drop.
     Several general conclusions about optimization of the HGMS process
 can be drawn from the above results.  First, it would appear that matrix
 length should be kept short at the expense of other variables to achieve
 the most cost effective designs.  The limiting factors of this situation
 might be good flow distribution and homogeneity of the applied magnetic
 field, both of which become more difficult to achieve as the aspect ratio
 of the matrix increases.
     The relative importance of capital and operating costs can vary
 with individual applications according to the company's amoritization
 methods, tax situation and so forth.  Compared to the precipitator
 system, however, the HGMS estimates of Table 11 imply a need to lower
 capital costs perhaps at the expense of power requirement if necessary.
One way to accomplish this reduction would be to increase the superficial
gas velocity and thus reduce the required number of magnets.  Since the
fan power requirement is already considerably higher than the magnet power
requirement, this increase in velocity might be accompanied by a reduction
                                     112

-------
in matrix packing density to lower pressure drop and an increase in the
applied field of each magnet to maintain the high collection efficiency.
Another approach would be to cool the gas further before cleaning it.
The volume of gas cleaned could be reduced allowing a smaller number of
magnets to handle the flow.  This savings would of course be at least
partially offset by the investment required to accomplish the additional
gas cooling.
     While these adjustments could in principle lead to a more optimum
design, they also point to a potential problem.  The loading time of the
collection matrix (time between cleanings) is apparently directly pro-
portional to the mass of the matrix and inversely proportional to the
gas velocity and to the particulate concentration.  Other factors such
as applied field and particle susceptibility may also be important.   With
a matrix density of 0.010, a matrix length of 15 cm, and a superficial
velocity of 8.8 m/s (the design used to obtain the economic estimates
                                                             3           3
above), a matrix cleaning a dirty gas concentration of 10 g/m  (4.4 gr/ft )
would collect its own mass of dust in 2.2 minutes.  By the time the matrix
collected its own mass, increased pressure drop or decreased collection
efficiency would probably dictate that the matrix be cleaned.  To
achieve a 70 percent duty cycle would then require that the matrix be
cleaned and returned to service in slightly under 1 minute.  In batch-
wise operation the gas flow and magnetic field would be temporarily shut
off to allow in-situ"matrix cleaning.  The large iron-bound solenoids in
commercial use are reported to have a field decay time on the order of
60-90 seconds although technology has now been developed that accelerates
this decay time to 5 to 10 seconds [5].  Damper systems to accomplish
the necessary flow switching in this relatively short time frame could be
relatively costly and subject to maintenance problems.  The indicated
optimization technique of increasing gas velocity and decreasing matrix
density would shorten the loading time and the cleaning time required to
maintain a 70 percent duty cycle.  The reduction of gas temperature
would also shorten the loading and required cleaning times by creating a
more concentrated dust stream.
                                      113

-------
     This analysis points to the need of additional  developmental  work
in the design of magnets, matrices, and cleaning systems specifically
geared to application in high velocity gas streams with moderate and
high dust concentrations.  The short estimated cycle times provide a
strong incentive to develop continuously operating rather than cyclic
systems.
     The comparison of alternate BOF emission control  methods suggests
that HGMS is a viable alternative for fine particle control  in systems
where the particulate has a relatively high magnetic susceptibility.  In
addition to reasonable projected capital cost and energy requirements
the process offers other potential advantages over conventional  technology.
The high throughput suggests smaller space requirements than electrostatic
precipitators or baghouses.  This could be important in any application
but especially perhaps in a retrofit situation.  Stainless steel wool is
a common, relatively inexpensive material which is resistant to  high
temperature and corrosive environments.  The lack of any sparking
mechanism should make the process amenable to dry, combustible environ-
ments with proper design precautions.  Unlike many wet scrubber  systems
the HGMS process would not turn an air pollution problem into a  water
pollution problem.  Even if water should be employed to assist in matrix
cleaning the quantity would be much less than in scrubber systems.
     These attractive features of the HGMS process strongly support an
argument for the continued development of the technology with early
emphasis on identifying the most appropriate potential applications
within the iron and steel industry as well as the ferroalloy industry.
A program should be undertaken to demonstrate the fine particle control
capability of HGMS with a pilot-scale device operating on a slipstream
of one or more actual industrial processes.  Continued research should
also be supported to develop and test HGMS system designs specifically
oriented toward gas stream applications so that capital costs and power
requirements can be minimized and efficient, practical matrix cleaning
systems can be demonstrated.
                                     114

-------
                               SECTION 9

                              REFERENCES
1.    Oberteuffer, J.A., Magnetic Separation:   A Review of  Principles,
     Devices and Applications, IEEE Transactions on  Magnetics, Vol. MAG-
     10, No. 2, June 1974, pp. 223-238.

2.    Kolm, H., J. Oberteuffer, and D.  Kelland, High  Gradient Magnetic
     Separation, Scientific American,  Vol.  233, No.  5,  November 1975,
     pp. 47-53.

3.    Luborsky, F.E., High Field-High Gradient Magnetic Separation:  A
     Review, Magnetism and Magnetic Materials—1975, 21st  Annual Confer-
     ence, Philadelphia, America'n Institute of Physics Conference Proceedings
     1976.

4.    Oder, R.R., High Gradient Magnetic  Separation Theory  and Applications,
     IEEE Transactions on Magnetics, Vol.  MAG-12, No.  5, September 1976,
     pp. 428-435.

5.    lannicelli, J., New Developments  in Magnetic Separation, IEEE
     Transactions on Magnetics, Vol. MAG-12,  No. 5,  September 1976, pp.
     436-443.

6.    Franz, S.G., Magnetic Separator,  U.S.  Patent 2,074,085, March 16,
     1937.

7.    Jones, G.H., Magnetic Separators, U.S. Patent 3,346,116, October
     10, 1967.

8.    Oder, R.R., and C.R. Price, Brightness Beneficiation  of Kaolin
     Clays by Magnetic Treatment, The Journal of the Technical Association
     of the Pulp and Paper Industry, Vol.  56, No. 10, October 1973, pp.
     75-78.

9.    Oder, R.R., The Magnetic Processing of Kaolin Clay,  Pulp and Paper
     Magazine of Canada, Vol. 75, No.  10,  October 1974, pp. T366-T369.

10.  lannicelli, J., Development of High Extraction  Magnetic Filtration
     by the Kaolin Industry of Georgia,  IEEE Transactions  on Magnetics,
     Vol. MAG-12, No.  5, September 1976, p. 489.

11.  lammartino, N.R., New Tasks for Magnetism, Chemical  Engineering,
     Vol. 81, No. 1, January 7, 1974,  pp.  50-52.

12.  Kelland, D.R., High Gradient Magnetic Separation Applied  to  Mineral
     Beneficiation, IEEE Transactions  on Magnetics,  Vol.  MAG-9,  No.  3,
     September 1973, pp. 307-310.


                                    115

-------
13.  Lawver, J.E., and D.M.  Hopstock,  Wet Magnetic  Separation  of Weakly
     Magnetic Minerals, Minerals Science Engineering,  Vol.  6,  No.  3,
     July 1974, pp. 154-172.

14.  Kelland, D.R., and E.  Maxwell,  Oxidized  Taconite  Beneficiation by
     Continuous High Gradient Magnetic Separation,  IEEE  Transactions  on
     Magnetics, Vol. MAG-11, No. 5,  September 1975,  pp.  1582-1584.

15.  Vives, D.L., L.J. Hirth, and W.H. Summer!in, Direct Reduction and
     Magnetic Beneficiation of Alabama Brown  Ore with  Lignite, IEEE
     Transactions on Magnetics, Vol. MAG-12,  No. 5,  September  1976,
     pp.,490-492.

16.  Murray, H.H., Beneficiation of Selected  Industrial  Minerals and
     Coal by High Intensity Magnetic Separation, IEEE  Transactions on
     Magnetics, Vol. MAG-12, September 1976,  pp. 498-502.

17.  Ergun, S., and E.H. Bean, Magnetic Separation  of  Pyrite from  Coals,
     Report of Investigation No. 7181, U.S. Bureau  of  Mines, Pittsburgh,
     PA., September 1968.

18.  Trindade, S.C., and H.H. Kolm,  Magnetic  Desulfurization of Coal, IEEE
     Transactions on Magnetics, Vol. MAG-9, No. 3,  September 1973,
     pp. 310-313.

19.  Oder, R.R., Pyritic Sulfur Removal from  Coals  by  High  Gradient
     Magnetic Separation, Proceedings of Second National  Conference on
     Complete WateReuse, 1975, pp. 315-322.

20.  Cohen, E. and J.A. Good, The Application of a  Superconducting
     Magnet System to the Cleaning and Desulfurization of Coal,
     IEEE Transactions on Magnetics, Vol. MAG-12, No.  5, September 1976,
     pp. 503-506.

21.  Maxwell, E., D.R. Kelland, and  I.Y. Akoto, High Gradient  Magnetic
     Separation of Mineral  Particulates from  Solvent Refined Coal, IEEE
     Transactions on Magnetics, Vol. MAG-12,  No. 5,  September  1976,
     pp. 507-510.

22.  Lin, C.J., Y.A. Liu, D.L. Vives, M.J. Oak, G.E. Crow,  and E.L. Huffman,
     Pilot-Scale Studies of Sulfur and Ash Removal  from  Coals  by  High
     Gradient Magnetic Separation, IEEE Transactions on  Magnetics, Vol
     MA6-12, No. 5, September 1976,  pp. 513-521.

23.  Monteiro, J.L.F., M. Saddy, and S.C. Trindade,  Studies on Sulfur
     Recovery from Coal Wastes and Prospective Magnetic  Separation
     Applications, IEEE Transactions on Magnetics,  Vol.  MAG-12, No.  5,
     September 1976, p. 522.
                                     116

-------
24.  Fine, H.A., M. Lowry, L.F. Power, and G.H. Geiger, A Proposed Process
     for the Desulfurization of Finely Divided Coal by Flash Roasting and
     Magnetic Separation, IEEE Transactions on Magnetics, Vol.  MAG-12,
     No. 5, September 1976, pp. 523-527.

25.  Oder, R.R., Magnetic Desulfurization of Liquefied Coals:  Conceptual
     Process Design and Cost Estimation, IEEE Transactions on Magnetics,
     Vol. MAG-12, No. 5, September 1976, pp. 532-537.

26.  Liu, Y.A., and C.J. Lin, Assessment of Sulfur and Ash Removal  from
     Coals by Magnetic Separation, IEEE Transactions on Magnetics,  Vol.
     MAG-12, No. 5, September 1976, pp. 538-550.

27.  Maxwell, E., D.R. Kelland, and I.Y. Akoto, Magnetic Separation of
     Solids from Liquified Coal,  IEEE Transactions on Magnetics, Vol.
     MAG-12, No. 6, November 1976, p. 907.

28.  Kazarian,  R., Researchers Developing Method to Remove Impurities
     From-Coal, National Science  Foundation News, NSF PR77-50, May 10,
     1977, pp.  1-3.

29.  Marston, P.G., and J.A. Oberteuffer, The Application of High Gradient
     Magnetic Separation  (HGMS) to the  Treatment of Steel Industry Waste
     Waters, Second  International Congress on  Industrial Wastes, Stockholm,
     Sweden, February 1975.

30.  Oberteuffer, J.A.,  I. Wechsler,  P.G. Marston and M.J. McNallan,
     High  Gradient Magnetic Filtration  of Steel Mill Process and Waste
     Waters, IEEE Transactions on Magnetics, Vol. MAG-11, No. 5,
     September  1975, pp.  1591-1593.

31.  Oberteuffer, J.A.,  Engineering Development of High  Gradient Magnetic
     Separators, IEEE Transactions on Magnetics, Vol. MAG-12, No. 5,
     September  1976, pp. 444-449.

32.  Harland, J.R.,  L. Nilsson and M. Wallin,  Pilot-Scale High Gradient
     Filtration of Steel Mill Wastewater,  IEEE Transactions, Vol. MAG-12,
     No.  6, November 1976, pp. 904-906.

33.  deLatour,  C., Magnetic Separation  in Water Pollution Control, IEEE
     Transactions on Magnetics, Vol.  MAG-9, No. 3, September 1973, pp. 314-
     316.

34.  Oder, R.R., and B.I. Horst,  Wastewater Processing with  High Gradient
     Magnetic Separators, Proceedings of Second National Conference  on
     Complete WateReuse,  1975, pp. 887-897.

35.  deLatour,  C. and H.  Kolm, Magnetic Separation  in Water  Pollution
     Control-II, IEEE Transactions on Magnetics, Vol. MAG-11, No. 5,
     September  1975, pp.  1570-1572.
                                     117

-------
36.  Ricci, L.J., Antipollution Processes Top Boston  AIChE  Meeting,
     Chemical Engineering, Vol. 82,  No.  21,  October 13,  1975,  pp.  73-75.

37.  Petrakis,  L. and P.P. Ahnen,  Use of High Gradient Magnetic  Separation
     Techniques for the Removal of Oil  and Solids  from Water  Effluents,
     IEEE Transactions on Magnetics, Vol. MAG-12,  No.  5,  September 1976,
     pp. 486-488.

38.  Yadidia, R., A. Abeliovich, and G.  Belfort, Algae Removal by  High
     Gradient Magnetic Filtration, Environmental Science and  Technology
     Vol. 11, No. 9, September 1977, pp. 913-916.

39.  Melville, D., F. Paul, and S. Roath, High Gradient  Magnetic Separation
     of Red Cells from Whole Blood,  IEEE Transactions on Magnetics,  Vol.
     MAG-11, No. 6, November 1975, pp.  1701-1704.

40.  Kolm, H.H., Research Needs In Magnetic  Separation,  IEEE  Transactions
     on Magnetics, Vol. MAG-12, No.  5,  September 1976, pp.  450-454.

41.  Kolm, H.H., The Large-Scale Manipulation of Small Particles,  IEEE
     Transactions on Magnetics, Vol. MAG-11, No. 5, September 1975,  pp.
     1567-1569.

42.  Ranz, W.E. and J.B. Wong, Impaction of  Dust and  Smoke  Particles on
     Surface and Body Collectors,  Industrial and Engineering  Chemistry,
     Vol. 44, No. 6, June 1952, pp.  1371-1381.

43.  Davies, C.N., The Separation of Airborne Dust and Particles,  Proceedings
     of the Institute of Mechanical  Engineering, Vol. 18, 1952,  pp.  185-
     198.

44.  Fuchs, N.A., The Mechanics of Aerosols. Pergamon Press,  New York,
     1964.

45.  Strauss, W., Industrial Gas Cleaning, Pergamon Press,  New York,
     1966.

46.  Loffler, F., Collection of Particles by Fiber Filters, in Air Pollution
     Control, Part 1, edited by W. Strauss,  Wiley-Interscience,  New
     York, 1971.

47.  Bean, C.P., Theory of Magnetic Filtration, Bulletin of American
     Physical Society, Vol. 16, 1971, p. 350.  (abstract only)

48.  Watson, J.H.P., Magnetic Filtration, Journal  of  Applied  Physics, Vol
     44, No.  9, September 1973, pp.  4209-4213.

49.  Oberteuffer, J.A., High Gradient Magnetic Separation,  IEEE  Transactions
     on Magnetics, Vol. MAG-9, No. 3, September 1973, pp. 303-306.
                                    118

-------
50.  Oder, R.R. and C.R. Price, HGMS:  Mathematical Modeling of Commercial
     Practice, Magnetism and Magnetic Materials-1975, 21st Annual  Confer-
     ence, Philadelphia, American Institute of Physics Conference  Proceedings,
     1976.	a-

51.  Watson, J.H.P., Theory of Capture of Particles in Magnetic High-
     Intensity Filters, IEEE Transactions on Magnetics, Vol. MAG-11,
     No. 5, September 1975, pp. 1597-1599.

52.  Cowen, C., F.J. Friedlaender, and R. Jaluria, High Gradient Magnetic
     Field Particle Capture on a Single Wire, IEEE Transactions on Magnetics,
     Vol. MAG-11, No. 5, September 1975, pp. 1600-1602.

53.  Luborsky, F.E. and B.J. Drummond, High Gradient Magnetic Separation:
     Theory Versus Experiment, IEEE Transactions on Magnetics, Vol. MAG-
     11, No. 6, November 1975, pp. 1696-1700.

54.  Cummings, D.L., D.A. Himmelblau, J.A. Oberteuffer, and G.J. Powers,
     Capture of Small Paramagnetic Particles by Magnetic Forces from  Low
     Speed Fluid Flows, AIChE Journal, Vol. 22, No. 3, May 1976, pp.  569-
     575.

55.  Aharoni, A., Traction Force on Paramagnetic Particles in Magnetic
     Separators, IEEE Transactions on Magnetics, Vol. MAG-12, No.  3,
     May 1976, pp. 234-235.

56.  Luborsky, F.E., and B.<3. Drummond, Buildup of Particles on Fibers in
     a High Field-High Gradient Separator, IEEE Transactions on Magnetics,
     Vol. MAG-12, No. 5, September 1976, pp. 463-465.

57.  Cowen, C., F.J. Friedlaender, and R. Jaluria, Single Wire Model  of
     High Gradient Magnetic Separation Processes I,  IEEE Transactions on
     Magnetics, Vol. MAG-12, No. 5, September 1976,  pp. 466-470.

58.  Cummings, D.L., D.C. Prieve, and G.J. Powers, The Motion of Small
     Paramagnetic Particles in a High Gradient Magnetic Separator, IEEE
     Transactions on Magnetics, Vol. MAG-12, No. 5,  September 1976,
     pp. 471-473.

59.  Stekly, Z.J.J. and J.V. Minervini, Shape Effect of the Matrix on the
     Capture Cross Section of Particles in High Gradient Magnetic Separation,
     IEEE Transactions on Magnetics, Vol. MAG-12,  No.  5, September 1976,
     pp. 474-479.

60.  Birss, R.R., R. Gerber, and M.R. Parker, Theory and Design of Axially
     Ordered Filters for High Intensity Magnetic Separation,  IEEE Transactions
     on Magnetics, Vol. MAG-12, No. 6, November 1976,  pp. 892-894.
                                     119

-------
61.  Uchiyama, S., S. Kondo, M.  Takayasu, and I.  Eguchi,  Performance of
     Parallel Stream Type Magnetic Filter for HGMS,  IEEE  Transactions
     of Magnetics, Vol. MAG-12,  No.  6, November 1976,  pp.  895-897.

62.  Cowen C., F.J. Friedlaender, and R.  Jaluria, Single  Wire Model  of
     High Gradient Magnetic Separation Processes  II,  IEEE Transactions
     on Magnetics, Vol. MAG-12,  No.  6, November 1976,  pp.  898-900.

63.  Clarkson, C.J., D. Kelland, and T.B. King, Model  for Calculation of
     Capture Radii of High Gradient Magnetic Separator at Moderate
     Reynolds Numbers, IEEE Transactions  on Magnetics, Vol.  MAG-12,
     No. 6, November 1976, pp. 901-903.

64.  Lawson, W.F., The Dynamics  of Paramagnetic Particles Near a
     Magnetized Wire, M.S. Thesis, Department of Physics,  West Virginia
     University, Morgantown, WV, 1976.  (unpublished).

65.  Lawson, W.F., W.H. Simons,  and R.P.  Treat, The  Dynamics of a
     Particle Attracted by  a Magnetized  Wire, Journal of Applied
     Physics, Vol. 48, No. 8, August 1977, pp. 3213-3224.

66.  Zebel, G., Deposition of Aerosol Flowing Past a  Cylindrical Fiber
     in a Uniform Electric Field, Journal of Colloid  Science, 20,
     1965, pp. 522-543.

67.  Krupp, H., Particle Adhesion, Advances in Colloid and Interface
     Science, Vol. 1, 1967, pp.  111-239.

68.  Ketkar, A., and D.V. Keller, Adhesion of Micron-Sized Limestone
     Particles to a Massive Coal Substrate, Journal  of Adhesion,
     Vol. 7, 1975, pp. 235-251.

69.  Visser, J., On Hamaker Constants: A Comparison  between Hamaker
     Constants and Lifshitz - Van der Waals Constants, Advances in
     Colloid and Interface Science,  Vol.  3, 1972, pp.  331-363.

70.  Marks'  Standard Handbook for Mechanical Engineers,  T. Baumeister,
     Editor-in-Chief, 7th edition, McGraw-Hill Book  Company, New York, 1967.

71.  Katari, V., G. Issacs and T.W.  Devitt, Trace Pollutant Emissions
     from the Processing of Metallic Ores, EPA-650/2-74-115, October
     1974.

72.  Hedley, W.H., S.M. Mehta, C.M.  Moscowitz, A.D.  Snyder, H.H.S.  Yu,
     and D.L. Zanders, Sources and Characterization  of Fine Particular
     Test Dusts, EPA-650/2-74-117, November 1974.      "	~
                                    120

-------
73.  Dealy, J.O., and A.M. Killin, Engineering and Cost Study of the
     Ferroalloy Industry, EPA-450/2-74-008, May 1974.	

74.  Sears, F.W., and M.W. Zemansky, University Physics, Part 2, 3rd
     edition.  Addison-Wesley Publishing Company, Inc., Reading,
     MA, 1964.

75.  Oberteuffer, J.A. and M. Golay, A Review and Analytical  Evalu-
     ation of the Application of High Gradient Magnetic Separation
     to Fine Particle Scrubbing, report submitted by Magnetic Engi-
     neering Associates under subcontract to Research Triangle Insti-
     tute, December, 1975.

76.  Sala Magnetics, Inc.  Cambridge, MA, SMI Bulletin No.  D052401-
     7527GB/775.

77.  Guichard, J.C., Aerosol Generation Using Fluidized Beds, in Fine
     Particles-Aerosol Generation, Measurement, Sampling and Analysis,
     edited by B.Y.H. Liu, Academic Press, New York, 1976,  pp.173-194.

78.  Moreno, F. and D. Blann, Large Flow Rate Redispersion  Aerosol
     Generator, in Fine Particles-Aerosol Generation. Measurement
     Sampling and Analysis, edited by B.Y.H. Liu, Academic  Press, New
     York, 1976, pp. 195-218.

79.  Willeke, S., C. Lo, and K.T. Whitby, Dispersion Characteristics of
     a Fluidized Bed, Aerosol Science, Vol. 5, 1974, pp. 449-455.

80.  Ragland, J.W., D.H. Pontius, and W.B. Smith, Design, Construct,
     and Test a Field Usable Prototype System for Sizing Particles
     Smaller than 0.5 m Diameter  , Special summary report submitted
     by Southern Research Institute to EPA under Contract No. 69-
     02-2114, Task 8, EPA Task Officer Bruce Harris, February 1976,
     (unpublished).

81.  Blann, D.D., K.A. Green, and L.W. Anderson, Design, Fabrication,
     and Installation of a Particulate Aerodynamic Test Facility,
     EPA-650/2-74-103, October 1974.

82.  Gotterba, J., Characterization and Optimization of the  EPA
     High Mass Flowrate Aerosol Generator, Task  report  submitted
     by Aerotherm Corporation to  EPA under Contract No. 68-02-1318,
     Task 28, EPA Task Officer Dale Harmon, July 1976,  (unpublished).

83.  Roulliard, E.E.A., Experimental Errors in Sampling Dust Laden
     Gas Streams, South African Council for Scientific  and Industrial
     Research, Chemical Engineering Group, CSIR  Special Report  File
     No. 66/51/4510/2, Pretoria,  December  1971.
                                     121

-------
84-  Harris, D.B., Procedures for Cascade Impactor Calibration and
     Operation in Process Streams, EPA-600/2-77-004,  January 1977.

85.  Gushing, K.M., G.E.  Lacey,  J.D.  McCain,  and W.B.  Smith, Particulate
     Sizing Techniques for Control Device Evaluations  Cascade Impactor
     Calibrations. EPA-600/2-76-280,  October  1976.

86.  Hay, R.D., Magnetic  Engineering  Associates, Inc., Cambridge,
     MA, Personal communication, April  29, 1976.

87.  Oder, R.R., Bechtel  Corporation, San Francisco,  CA,  Personal
     communication, January 6, 1977.

88.  Hardison, L.C. and C.A. Greathouse,  Air  Pollution Control
     Technology and Costs in Nine Selected Areas, final  report of
     Industrial Gas Cleaning Institute to EPA under Contract No.
     68-02-0301, National Technical Information Service,  Springfield,
     Virginia, PB-222 746, September  1972.

89.  Economic Indicators, Chemical Engineering, Vol.  84,  No. 19,
     September 12, 1977,  p. 7.

90.  Federal Register, March 8,  1974, Vol. 39, p. 9308.

91.  Steiner, B.A., Air-Pollution Control in  the Iron  and Steel  Industry,
     International Metals Review, September 1976.

92.  Delaney, 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, June 1974.

93.  Wechsler, I.  Sala Magnetics, Inc.,  Cambridge, MA, Personal
     communication, October 18,  1977.
                                     122

-------
                              SECTION 10

                             BIBLIOGRAPHY
1.   Allen, J.W., Magnetic Separator, U.S. Patent 3,819,515,  June  25,
     1974.

2.   Aubrey, W.M., J.M. Karpinski, and D.S. Cahn, Magnetic  Separator
     Method and Apparatus, U.S. Patent 3,608,718, September 28,  1971.

3.   Davis, C.W.V., Magnetic Separator, U.S.  Patent 3,062,376,
     November 6, 1962.

4.   Fraas, F., Alternating Field Magnetic Separator,  U.S.  Patent
     3,988,240, October 26, 1976.

5.   Fraas, F., The Matrix-Type Magnetic Separator, Report  of Investi-
     gation No. 6722, U.S. Bureau of Mines, Pittsburgh,  Pennsylvania,
     1966.

6.   Kolm, H.H., Magnetic Device, U.S. Patent 3,567,026, March 2,
     1971.

7.   Kolm, H.H., Process for Magnetic Separation, U.S. Patent
     3,676,337, July 11, 1972.

8.   Marston, P.G., Magnet Design and Its Effect on the  Economics
     of High Gradient Magnetic Separation Processes, Proceedings
     of the High Gradient Magnetic Separation Symposium, Massachusetts
     Institute of Technology, Cambridge, Mass.,  May 1973.   pp. 25-37.

9.   Marston, P.G., I. Wechsler, and J.J. Nolan, Magnetic Separation
     Method, U.S. Patent 3,887,457, June 3, 1975.

10.  Marston, P.G., and J.J. Nolan, Moving Matrix Magnetic  Separator,
     U.S. Patent 3,920,543, November 18, 1975.

11.  Marston, P.G., J.J. Nolan, and L.M. Lontai, Magnetic Separator
     and Magnetic Separation Method, U.S. Patent 3,627,678,
     December 14, 1971.

12.  Nolan, J.J. and P.G. Marston, Flow Control  Unit for Magnetic
     Matrix, U.S. Patent 4,025,432, May 24, 1977.
                                  123

-------
13.  Nolan, J.J.,  P.G.  Marston,  and  L.M. Lontai, Multiple Matrix Magnetic
     Separation Device  and Method, U.S. Patent 3,770,629, November 6,
     1973.

14.  Oder,  R.R., Method for Magnetic Beneficiation of Particle Dispersions,
     U.S.  Patent 3,985,646, October  12, 1976.

15.  Sakata, S., A.  Yoshikawa, and A. Tasaki, Magnetic Separation of
     Aerosol Particles  from Air  Flow, Japan Journal of Applied Physics,
     Vol.  15, No.  10, 1976, pp.  2017-2018.

16.  Watson, J.H.P.  and D. Hocking, The Beneficiation of Clays Using
     a Superconducting  Magnetic  Separator, IEEE Transactions on
     Magnetics, Vol. MAG-11, No. 5., September 1975, pp. 1588-1590.
                                   124

-------
                                   APPENDIX
           DERIVATION OF THE EQUATIONS OF MOTION FOR A PARAMAGNETIC
                 PARTICLE IN A NONUNIFORM MA6NETOSTATIC FIELD

     The energy of a magnetostatic field established in a linear,
homogeneous, and isotropic medium where B  = y H is given by
where TQ  = energy, joule
      HQ  = magnetic field strength, ampere/meter
      Bn  = magnetic flux density, tesla
                                               -7
      UQ  = permeability of free space, 4Tr x 10   henry/meter.
      dv  = differential volume element.
The indicated integration is extended over all space.  If a paramagnetic,
spherical particle of volume V-| (meter  ) and radius R (meter) is placed
in the magnetic field,  the potential energy is given by
                         (J  =	i-  /  M • B  dv
                          m      2  \      "
where  Um = potential energy, joule
        M = induced magnetization, ampere/meter.
The integration is over the volume of the particle, and within this
volume
                         8 " PQ(H + M).                              (A-3)
                                  125

-------
     The magnetic  force F  (newton) acting upon the paramagnetic

particle is  given  by
               ->      ->    -.  -*• r •+   •+
               F  =  -  VU = j V /  M • B  dv                         (A-4)


                                vi


where  in cylindrical coordinates
                         Ir
                     Vl
Interchanging v and / gives


                    Ti
               ->•    ,   /*"*""*"   "*"
               Fm = ^ J  7(M  • BQ)dv                              (A-6)



According to equation (A-3)



                    B = MQ(H 4- M) =  yH         r
-------
In the linear, homogeneous, and  is^trbpic medium
                    ->     ->•
                    Bo  =  yoHo               r>>R                   (A-ll)

and equation  (A-6)  then becomes
          Fm =
                                            Vl
                                      = 1 XP0/V(H  •  HQ)dv.
 The dielectric  constant,  y,  and the susceptibility,  x»  are  properties of
 the particle, which  is  also  as1»"umed"to be  linear,  isotropic, and homogeneous.
 If the value of H  is assumed to be relatively constant  throughout the
 volume V-j  and approximately  equal  to H0i the  integration  can be performed
 to yield
                +         -». +
                   =  - «••  "'u  •  HQ)  /  dv
                                  .'/
and finally,

     The magnetic  field of interest is a combination of the  uniform
 background field H  and the nonuniform field  in the  vicinity of a
 ferromagnetic wire.   The geometric arrangement shown in Figure  A-l
 will be used to obtain  the appropriate expression for the  magnetic
 field strength.
                                       127

-------
        Figure A-l.   Coordinate  system for HGMS single-fiber model
     The radius of the cylindrical ferromagnetic wire is a (meter), and the
radius of the spherical particle is  R (meter).   For  r»a  the velocity
of the particle is V  and parallel  to but in  the negative x direction.
For r»a, the magnetic field strength is  H and  parallel  to and  in  the
positive x direction.
     Since the magnetic field strength is a vector,  we may  define a
magnetic potential $ such that
                          H = -V *.
(A-15)
                                       128

-------
As stated before for a linear,  isotropic,  and  homogeneous medium,



                          B =  uH.                                 (A-16)


The magnetic flux density B  is  solenoidal,  hence

                          -*•    -»•
                          V  •  B  =  0.                              (A-17)


 Introducing Equations  (A-15)  and  (A-16)  into  (A-17) yields
                     •-•-•••
                     V -  yH  =  -yV •  V « *  -uV  *  =  0,              (A-18)
 where  in  cylindrical  coordinates
                    3r2   r  3r   r
                   2
 Due to  symmetry,  -5-E = 0;  and we are left with
                  3r
    +        +  _
3r2    r  3r   r2 362
                            1  It + I_1« . o.                     (A-20)
                                        129

-------
A solution of the form
                           - rnCn(e)
(A-21)
can be shown to satisfy the partial differential equation where Cn(e)
are harmonic functions given by
Cn
Co
Cl
C2
C3
r"nC (0) rnCos (0)
n ii
log r
r'Xcos
r"2(A1Cos
r"3(A1Cos

0 + B,Sin 0)
20 + B-jSin 20)
0 + B^in 30)
1
r(A2Cos 0 + E
r2(A2Cos 20 H
•5
r (A2Cos 30 J

5pSin 0)
h B2Sin 20)
i- B2Sin 30)
where A-|, A2, B-j, and B2 are constants.

     Inside the wire the magnetic potential is -Mr, e).  If only first
order harmonic functions are considered and the origin is chosen as  zero
potential where $-](0,0) = 0, then the r   terms must disappear and the
potential is given by
                      $-|(r, 0) = r(A Cos 0 + B Sin 0)
(A-22)
and
     H-|(r, 0) =_(A Cos 0 + B Sin e)r -  (-A Sin  0 + B  Cos  0)0.     (A-23)
                                       130

-------
At the boundary of the wire the tangential component H   of  H  is zero for
6=0.  Evaluating     (a,0) from Equation (A-23)  yields
                         H-,j6(a,0) =-B = 0.                          (A-24)

Therefore Equation (A-22) becomes

                         ^(r.g) = rA Cos 9.                         (A-25)

     Outside the wire the first order approximation for the magnetic
potential is given by


          $2(r,9) = r-1(C Cos 9 + D Sin 9) + r (E Cos 9 + F Sin  9)    (A-26)

and

        H  (r,9)  = -F- ^2  (C  Cos  9 + D Sin 9) + (E Cos e +F Sin e)l r
                     r                                               (A-27)

                 _| r'^-C Sin 9 + D Cos  9)  + (-E Sin 9 + F Cos 9)1  9.

Outside the wire where  r»a

                              H2(r,0) - HQr.                          (A-28)

Therefore,
,0)  = HQ         r»a
                                                                      (A-29)

                         Ho ,(r,0)  = 0          r»a.                  (A"3°)
                                        131

-------
Evaluating H9 ^(r.O) and H~ Q(r,0)  from Equation  (A-27) yields
            £» I            £» 0
                 H2 r(r,0) =  H-^-- E = HQ         r»a.            (A-31)
                         .'.  E «-H0.                                (A-32)
Also
                    H9 .(r,0) = -- F = 0           r»a.             (A-33)
                     t. 5 0        I

By choosing F = D = 0, the boundary condition H0 Q(r,0)  =  0 is  satisfied
                                               £»°
and Equation (A-26) becomes

                    $2(r,e) = r  C Cos e - rHQ Cos  e.                 (A-34)

     Other information necessary to specify A in Equation  (A-25)  and C
in Equation (A-34) is obtained from analysis of the conditions  at the
boundary of the ferromagnetic wire.  Throughout the ferromagnetic wire
the magnetization is parallel and  in the positive x direction.  Thus the
tangential  component Mt and normal component Mn are related to  M  by

                         M = Mnr + Mt e                              (A-35)

                         Mn =  |M|   Cos Q                             (A-36)

                         Mt =  |M|  Sin e.                             (A-37)
                                       132

-------
It can be easily shown  that  the  normal  components  of  the magnetic  flux

density across the boundary  of the  wire must be equal  or
                               Bnl  =  Bn2
Inside the wire
                      nl
                    3r
                                                   r=a
                                        (A-38)
                                        (A-39)
Outside the wire
                        B   = yh
                         n2    on2  '
                                             r=a
                                        (A-40)
      In  addition, the tangential components of the magnetic field strength

 must  be  continuous  and equal across the boundary of the wire or
 Inside  the wire
 Outside the wire
                          Htl = Ht2
                                          i
                        Htl = Hl,e = " r  98
                                              r=a
t2 = H2,e = " r  39
                                                r=a
                                        (A-41)
                                        (A-42)
                                                                    (A-43)
                                         133

-------
     Equations  (A-38) and  (A-41) with the substitutions  indicated  by
Equations  (A-39),  (A-40),  (A-42), and (A-43) are solved  to yield values
for A and  C so  that the equations for the magnetic potential finally
become

                         ^ =  (_HO + |)r Cos e                       (A-44)

and
                             Cos 6 " V Cos
     To determine the force on the particle in the magnetic field, we  need
the magnetic field outside the wire or
                              -*•      ->
                              H2 = - V$2                             (A-46)

where
or
                     o       r +    —. e + —=-  z
                     2    9r      r  96       8z
        H2 =  (^-y Cos 6 4 HQ Cos e)r - 1 (-|^ Sin  6 + HQr  Sin  e)e.  (A-48)
Recall that
or
                    Fm = (Frr + FB9)-                                 (A-49)
                                       134

-------
Performing the appropriate operations with the results from Equations  (A-44)
and (A-45) yields
                                                                        (A-50)
and
                                        Sin 29
                                                                         (A-51)
     The equation of motion for the particle  in the air stream and in  the
magnetic field can be obtained for the  coordinate system established in
Figure A-l.
     The air stream velocity has components'
-*•
V  = V
 o    or
                                      Voe6
                                                (A-52)
 where assuming potential flow,
vor = * V1  - ¥  Cos  e
               r
                                                                         (A-53)
 and
                                                                        (A-54)
 The  Lagrangian form of the equation of motion in cylindrical  coordinates
 is
1"  [•
m d2r - m r/de\2 | : +  1 m r  d^e  + „  _dr  de
                                                      9 •
                                                (A-55)
                                       135

-------
where F  are the aerodynamic forces and F  are the magnetic forces.
The aerodynamic forces are assumed to be adequately expressed by Stokes
law
                             Fa = 6 TT n R V
                              a
(A-56)
where V is the relative velocity of the fluid and the particle.
Substituting for the aerodynamic and magnetic forces and rearranging
yields for the radial component (r)
,4 D3v
(I * R }Pp
6 ir n 1
! 1
V 9
I
Cos

a ,.<„.«._ ....._... _
2
dt
o Xy°(* 1
r/de ,
r^dt J
ir R3)M
6 y n Ra(f)3
a
-
+ a "dT~
M J- U P«^ OQ
o ' ' 'rt ^v»* *•• v
   = V,
wherep  is the particle density.

Defining r, = —  and
          a   a
                                                                       (A-57)
'm^ 2
a/ 9
X Vo H H0 R2
L ™2
                                                                       (A-58)
Equation (A-57) can be rearranged to


v
o n
a u
2PpR2
9n
d ra . de ^2
dt2 al dt } _
dr.

V /
1 \ r m 1 / M
^ a rj l*W
                                                 + Cos 2  e1
                                                                       (A-59)
                                      136

-------
Correspondingly, the angular component (e) is given by
204.
* Q i\
9 n
V
0 /
\r d2e
a dt2
i , 1
"a (l ' 2
dr
x o / S\ /d6 \
+ 2 (dt )(dt}
vm
} 9in fl m
. „ de
+ ra dt
Sin 29
; bin o a 3
                                                               (A-60)
     Equations  (A-59) and  (A-60) constitute the equations of motion for
 a  paramagnetic  particle under  the  influence of inertial, viscous, and
 magnetic forces.
                                      137

-------
                                TECHNICAL REPORT DATA
                          (Please read Inunctions on the reverse before completing)
 . REPORT NO.
 EPA-600/2-77-230
2.
                                                       3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
Application of High-Gradient Magnetic Separation
   to Fine Particle Control
                                                       5. REPORT DATE
                            November 1977
                           6. PERFORMING ORGANIZATION CODE
 . AUTHOR(S)
C. H. Gooding, T. W. Sigmon,  and L. K. Monteith
                                                       8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Research Triangle Institute
P.O. Box 12194
Research Triangle Park, North Carolina 27709
                            10. PROGRAM ELEMENT NO.
                            1AB012: ROAP 21ADL-029
                            11. CONTRACT/GRANT NO.
                            68-02-1879
 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: 6/75-8/77  	
                            14. SPONSORING AGENCY CODE
                             EPA/600/13
 is. SUPPLEMENTARY NOTES lERL-RTP project officer for this report is Dennis C.  Drehmel,
 Mail Drop 61,  919/541-2925.
 16. ABSTRACT
           The report gives results of an assessment of the potential use of high-
 gradient magnetic separation (HGMS) as a means of collecting gas stream particu-
 lates. The assessment included both experiments and analyses of theoretical models.
 Phase I included evaluations of theoretical explanations of HGMS and published
 reports of liquid system applications. A bench-scale apparatus was constructed, and
 HGMS experiments were conducted using redispersed dust from a basic oxygen fur-
 nace.5 High efficiency collection of fine particulates was  achieved  with both high
 throughput and reasonable projected energy rqeuirements relative to conventional
 devices. In Phase n, experiments were scaled up to 0. 8 cu m/s (1700 CFM). Dusts '
 from basic oxygen and electric arc furnaces were redispersed and collected. Results
 show that submicron particles can be collected with 90-plus % efficiency using applied
 magnetic flux densities  of 0.2-0.4 T; With superficial gas velocities up to 11 m/s, the
 pressure drop across the HGMS device was typically less than 1. 5 kPa (6 in. H2O).
 Even lower fields can be used successfully at the expense of higher pressure drop or
 reduced throughput.
17.
                              KEY WORDS AND DOCUMENT ANALYSIS
a.
                 DESCRIPTORS
               b.lDENTIFIERS/OPEN ENDED TERMS
c. COS AT I Field/Group
Air Pollution
Dust
Magnetic Separators
                Air Pollution Control
                Stationary Sources
                Particulate
                High-gradient Magnetic
                 Separation
 13B
 11G
 3. DISTRIBUTION STATEMENT
 Unlimited
                19. SECURITY CLASS (This Report)
                Unclassified
                                                                    21. NO. OF PAGES
                20. SECURITY CLASS (Thispage)
                Unclassified
                                                                          149
                                                                    22. PRICE
EPA Form 2220-1 (9-73)
                                          138

-------