Improving Efficiencies in Water-Based Separators Using

Mathematical Analysis Tools

by
Jaisen N. Kohmuench

Dissertation submitted to the Faculty of the

Virginia Polytechnic Institute and State University in partial
fulfillment of the requirements for the degree
of
DOCTOR OF PHILOSOPHY
in
Mining and Minerals Engineering

Approved
by:

G.H. Luttrell, Chairman

R.H. Yoon
G.T. Adel
E.C. Westman
M.J. Mankosa

December, 2000
Blacksburg, Virginia

Keywords: Coal Spirals, Circuit Analysis, CrossFlow Separator, HydroFloat Separator,

Classification, Density Separation.
Copyright 2000, Jaisen N. Kohmuench

Improving Efficiencies in Water-Based Separators Using

Mathematical Analysis Tools
by
Jaisen N. Kohmuench

Committee Chairman: Gerald H. Luttrell

Department of Mining and Minerals Engineering

(ABSTRACT)

A better understanding of several mineral processing devices and applications was gained
through studies conducted with mathematical analysis tools.

Linear circuit analysis and

population balance modeling were utilized to remedy inefficiencies found in a number of popular
mineral processing water-based unit operations. Improvements were made in areas, including
unit capacity and separation efficiency.
One process-engineering tool, known as linear circuit analysis, identified an alternative
coal spiral circuit configuration that offered improved performance while maintaining a
reasonable circulating load. In light of this finding, a full-scale test circuit was installed and
evaluated at an existing coal preparation facility. Data obtained from the plant tests indicate that
the new spiral circuit can simultaneously reduce cut-point and improve separation efficiency.
A mathematical population balance model has also been developed which accurately
simulates a novel hindered-bed separator. This device utilizes a tangential feed presentation
system to improve the performance of conventional teeter-bed separators.

Investigations

utilizing the mathematical model were carried out and have predicted solid feed rates of up to 71
tph/m2 (6 tph/ft2) can be achieved at acceptable efficiencies. The model also predicts that the

unfavorable impact of operating at low feed percent solids is severely reduced by the innovative
feed presentation design. Tracer studies have verified that this system allows excess feed water
to cross over the top of the separator without entering the separation chamber, thereby reducing
turbulence.
A hindered-bed separator population balance model was also developed whose results
were utilized to improve the efficiencies encountered when using a teeter-bed separator as a
mineral concentrator. It was found that by altering the apparent density of one of the feed
components, the efficiency of the gravity separation could be greatly improved. These results
led to the development of a new separator which segregates particles based on differences in
mass after the selective attachment of air bubbles to the hydrophobic component of the feed
stream.

Proof-of-concept and in-plant testing indicate that significant improvements in

separation efficiency can be achieved using this air-assisted teeter-bed system. The in-plant test
data suggest that in some cases, recoveries of the plus 35 mesh plant feed material can be
increased by more than 40% through the application of this new technology.

iii

ACKNOWLEDGEMENTS

The author wishes to express his deepest thanks and gratitude to Dr. Gerald H. Luttrell
for his guidance and advice during this investigation. The opportunity to work on a project of
such breadth was greatly appreciated as was the freedom allowed the author in completing this
research. The author also wishes to thank the rest of the Luttrell Clan, Kay, Sarah, and Greg, for
their fellowship and friendship.
The author is also grateful to Dr. Greg Adel for his friendship and advice, especially his
counsel in the area of population balance modeling. Thank you also is expressed to Dr. RoeHoan Yoon for his helpful suggestions and recommendations.
A sincere thank you is also expressed to Dr. Mike Mankosa for his friendship and
guidance. His in-field instruction and insight were invaluable. Thanks also to Cathy Mankosa,
his wife, for her continued support and encouragement.
Thanks are also expressed to several companies whose support, both monetary and
otherwise, made this work possible.

This gratitude is expressed to Eriez Magnetics, PCS

Phosphate, and the Pittston Coal Management Company.

Individual thanks must also be

expressed to Mr. Joe Shoniker, Mr. Fred Stanley and particularly, Mr. Richard Merwin.
The author wishes to acknowledge Wayne and Billy Slusser for their technical advice,
assistance, and instruction. Their effort and ability are greatly appreciated.
The author would like to thank his parents, William and Carolyn Kohmuench, for their
continued support and encouragement.

And finally, the author expresses his deepest

appreciation to his wife, and most loyal fan, Kathryn, for her support, encouragement and love.

iv

TABLE OF CONTENTS
ABSTRACT ...............................................................................................................................ii
ACKNOWLEDGEMENTS .......................................................................................................iv
TABLE OF CONTENTS ............................................................................................................v
LIST OF FIGURES ..................................................................................................................vii
LIST OF TABLES.....................................................................................................................ix
ORGANIZATION ......................................................................................................................x
CHAPTER 1: Improving Spiral Performance Using Circuit Analysis ........................................1
1.1 Introduction.....................................................................................................................1
1.2 Literature Review ............................................................................................................2
1.3 Theoretical Framework....................................................................................................7
1.3.1 Circuit Analysis .......................................................................................................7
1.3.2 Direct Procedure to Determine Optimum Circuitry ................................................12
1.3.2.1 Reid Equation ....................................................................................................17
1.3.2.2 Circuit Comparison and Optimization ................................................................18
1.4 Circuit Testing...............................................................................................................25
1.4.1 Site Description .....................................................................................................25
1.4.2 Test Program .........................................................................................................26
1.4.3 Experimental Results .............................................................................................28
1.5 Circuit Simulations........................................................................................................32
1.6 Conclusions...................................................................................................................38
1.7 References.....................................................................................................................40
CHAPTER 2: Improving Performance of Hindered-Bed Separators ........................................42
2.1 Introduction...................................................................................................................42
2.2 Literature Review ..........................................................................................................45
2.2.1 General ..................................................................................................................45
2.2.2 Hindered-Settling...................................................................................................49
2.3 Comparative Studies......................................................................................................55
2.3.1 In-Plant and Laboratory Testing.............................................................................55
2.3.2 Tracer Studies........................................................................................................61
2.4 Population Balance Model.............................................................................................67
2.4.1 Model Description .................................................................................................67
2.4.1.1 Feed Section ......................................................................................................69
2.4.1.2 Teeter-Bed and Underflow Sections...................................................................71
2.4.2 Calculations...........................................................................................................72
2.4.3 Model Validation...................................................................................................76
2.4.4 Model Accuracy ....................................................................................................79
2.4.5 Model Investigations..............................................................................................83

2.5 Conclusions...................................................................................................................88
2.6 References.....................................................................................................................89
CHAPTER 3: Improving Coarse Particle Recovery in Hindered-Bed Separators .....................94
3.1 Introduction...................................................................................................................94
3.2 Literature Review ..........................................................................................................97
3.2.1 General ..................................................................................................................97
3.2.1.1 Recovery by Flotation........................................................................................97
3.2.1.2 Recovery by Gravity Concentration .................................................................100
3.2.1.3 Phosphate Recovery.........................................................................................104
3.2.1.4 Carbon/Coal Recovery.....................................................................................109
3.3 Theoretical Framework................................................................................................115
3.3.1 Flotation Fundamentals........................................................................................115
3.3.2 Theoretical Advantages of the HydroFloat Cell....................................................122
3.4 Population Balance Model...........................................................................................127
3.4.1 Model Description ...............................................................................................127
3.4.1.1 Feed Section ....................................................................................................128
3.4.1.2 Teeter-Bed and Underflow Sections.................................................................129
3.4.2 Calculations.........................................................................................................129
3.4.3 Modeling Insight and Investigation......................................................................130
3.5 Proof-of-Concept Testing ............................................................................................133
3.5.1 Phosphate Recovery.............................................................................................133
3.5.1.1 Testing of a North Florida Phosphate Matrix....................................................133
3.5.1.2 Testing of a Central Florida Phosphate Matrix .................................................136
3.5.2 Coal/Carbon Recovery.........................................................................................138
3.5.2.1 Testing of Anthracite Slag................................................................................138
3.5.2.2 Testing of Central Appalachian Coal................................................................139
3.5.2.3 Testing of Australian Coal ...............................................................................142
3.5.2.4 Testing of Heavy Mineral Sands ......................................................................143
3.6 Pilot-Scale Testing ......................................................................................................145
3.6.1 Northern Florida Phosphate Matrix......................................................................145
3.6.2 Appalachian Coal ................................................................................................149
3.7 Conclusions.................................................................................................................154
3.8 References...................................................................................................................156
CHAPTER 4: General Summary ...........................................................................................163
CHAPTER 5: Recommendations for Future Work.................................................................169
APPENDIX A: Mass-Balance Equations ................................................................................172
APPENDIX B: Model Input/Output Examples........................................................................180
VITA ......................................................................................................................................188

vi

LIST OF FIGURES

Chapter 1: Improving Spiral Performance Using Circuit Analysis

Figure 1.1 - Cross-section of a spiral trough flow (Chedgy et al., 1990). .....................................3
Figure 1.2 - Feed rate versus cut-point of a spiral separation (Mikhail et al., 1988) .....................5
Figure 1.3 - Analysis of a single-stage separator..........................................................................7
Figure 1.4 - Analysis of a rougher- cleaner circuit. ......................................................................9
Figure 1.5 - Schematic of middlings reclean circuit. ..................................................................13
Figure 1.6 - Sigmoid partition data generated for the circuit depicted in Figure 1.5. ..................15
Figure 1.7 - Normalized partition data for determining best fit. .................................................18
Figure 1.8 - Traditional (a) and modified traditional (b) spiral circuits without recycle. .............20
Figure 1.9 - Traditional (a) and modified traditional (b) spiral circuits with recycle...................20
Figure 1.10 - Circuit efficiency (a) and SG50 (b) results for varying splitter positions.................22
Figure 1.11 - Winoc coal preparation plant rougher-cleaner spiral circuit. .................................26
Figure 1.12 - Test #1 experimental partition data.......................................................................29
Figure 1.13 - Test #2 experimental partition data.......................................................................29
Figure 1.14 - Test #3 experimental partition data.......................................................................30
Figure 1.15 - Normalized partition curve obtained using the data from the rougher spirals. .......34
Figure 1.16 - Normalized partition curve obtained using the data from the cleaner spirals. ........34
Figure 1.17 - Overall circuit partition curves for the rougher-cleaner spiral circuits with (Test #1
Circuit) and without (Test #2 Circuit) middlings recycle for simulations conducted
at constant clean coal ash. ..................................................................................... 37
Chapter 2: Improving Performance of Hindered-Bed Separators
Figure 2.1 - Schematic diagram of a conventional hindered-bed separator.................................43
Figure 2.2 - Schematic of CrossFlow hindered-bed separator. ...................................................44
Figure 2.3 - Schematic of Floatex hindered-bed classifier (Littler, 1986)...................................47
Figure 2.4 - Laboratory-scale CrossFlow separator....................................................................56
Figure 2.5 - Initial comparative test data. ..................................................................................57
Figure 2.6 - Efficiency comparison of CrossFlow and conventional separators..........................58
Figure 2.7 - Cut-point comparison of CrossFlow and conventional separators. ..........................58
Figure 2.8 - CrossFlow and Whirlsizer solids feed rate test comparison.....................................59
Figure 2.9 - CrossFlow and conventional teeter-bed separator solids feed rate test results. ........60
Figure 2.10 - Overflow residence time data for conventional feed system..................................66
Figure 2.11 - Overflow residence time data for the CrossFlow feed system. ..............................66
Figure 2.12 - Schematic depicting the primary divisions and flows for the CrossFlow separator.
...............................................................................................................................68
Figure 2.13 - Arrangement of zones for the feed section............................................................69
Figure 2.14 - Photo of dye experiment showing feed section flow regime. ................................71
Figure 2.15 - Arrangement of zones for the teeter-bed and underflow sections. .........................72
Figure 2.16 - Flowchart illustrating procedure needed to complete the population balance model.
...............................................................................................................................75

vii

Figure 2.17 - Correlation between cut-point and maximum concentration of solids. ..................78
Figure 2.18 - Comparison of predicted and actual cut-points. ....................................................80
Figure 2.19 - Actual versus predicted efficiency........................................................................81
Figure 2.20 - Model predicted curve and test data. ....................................................................82
Figure 2.21 - Effect of solids feed rate on Ep. ...........................................................................83
Figure 2.22 - Effect of solids feed rate on Imperfection. ............................................................84
Figure 2.23 - Effect of feed percent solids on cut-point. ............................................................85
Figure 2.24 - Effect of feed percent solids on efficiency. ...........................................................86
Chapter 3: Improving Coarse Particle Recovery in Hindered-Bed Separators
Figure 3.1 - Schematic illustration of the HydroFloat separator. ................................................96
Figure 3.2 - Relationship between particle size and recovery (Ahmed and Jameson, 1989)........98
Figure 3.3 - Recovery versus size data (Mankosa et al., 1999).................................................105
Figure 3.4 - Effect of feed rate on maximum floatable particle size (Luttrell, 2000).................112
Figure 3.5 - Size by size recovery data for multiple flotation feed samples (Phillips, 1998). ....113
Figure 3.6 - Effect of particle size on the probability of bubble-particle adhesion for induction
time values of 1, 2 and 5 msec..............................................................................117
Figure 3.7 - Influence of turbulence on the maximum particle size that may be recovered by
froth flotation (after Schulze, 1984)...................................................................... 118
Figure 3.8 - Relationship between flotation recovery and the dimensionless quantity kp for
different numbers of mixers-in-series. ..................................................................120
Figure 3.9 - Size by size recovery of components for varying density ratios (1/2).................131
Figure 3.10 - Circuit recovery data for varying component density ratios. ...............................132
Figure 3.11 - Comparison of separation data for batch test units for a phosphate matrix. .........134
Figure 3.12 - Comparison of separation data from continuous test units for a phosphate matrix.
.............................................................................................................................135
Figure 3.13 - Grade and recovery curve for central Florida phosphate sample. ........................137
Figure 3.14 - Optimum results of HydroFloat tests on central Florida phosphate sample. ........137
Figure 3.15 - Testing of anthracite slag using the HydroFloat separator (6.35 mm x 200 mesh).
.............................................................................................................................139
Figure 3.16 - Test results (fractionated) obtained using spiral feed from Central Appalachia. ..141
Figure 3.17 - Test results (cumulative) obtained using spiral feed from Central Appalachia. ...141
Figure 3.18 - HydroFloat results for the removal of impurities from mineral sands (Sample #1).
.............................................................................................................................144
Figure 3.19 - HydroFloat results for the removal of impurities from mineral sands (Sample #2).
.............................................................................................................................144
Figure 3.20 - Photograph of the pilot-scale HydroFloat separator. ...........................................146
Figure 3.21 - Pilot-scale flotation circuit for phosphate investigation.......................................146
Figure 3.22 - HydroFloat and plant comparison data. ..............................................................148
Figure 3.23 - Pilot-scale HydroFloat used for coal testing. ......................................................150
Figure 3.24 - Recovery data for the Appalachian coal tests......................................................151
Figure 3.25 - Coarse particles found in the HydroFloat product...............................................151
Figure 3.26 - Ash data for the Appalachian coal testing...........................................................153
Figure 3.27 - HydroFloat and coal spiral comparative data. .....................................................153

viii

LIST OF TABLES

Chapter 1: Improving Spiral Performance Using Circuit Analysis

Table 1.1 - Theoretical relative separation efficiencies for various circuit configurations...........11
Table 1.2 - Comparison of simulated and directly determined partition factors..........................15
Table 1.3 - Description of two-stage spiral circuits....................................................................21
Table 1.4 - Circuit comparisons for Ep and SG50 using the Reid partition model. ......................23
Table 1.5 - Circuit feed rates used during spiral testing. ............................................................27
Table 1.6 - Splitter settings used during spiral testing. ...............................................................27
Table 1.7 - SG50 values obtained from the spiral tests. ...............................................................31
Table 1.8 - Ep values obtained from the spiral tests. ..................................................................31
Table 1.9 - Summary of simulation results. ..............................................................................35
Chapter 2: Improving Performance of Hindered-Bed Separators
Table 2.1 - Mean underflow residence time for CrossFlow and conventional combinations.......63
Table 2.2 - Mean overflow residence time for CrossFlow and conventional combinations.........64
Table 2.3 - Particle size distribution of limestone used in laboratory validation test work..........76
Table 2.4 - Comparison of actual and predicted results..............................................................81
Chapter 3: Improving Coarse Particle Recovery in Hindered-Bed Separators
Table 3.1 - Size-by-size HydroFloat results obtained using an Australian coal........................143
Table 3.2 - Parameters for in-plant pilot test program. .............................................................145
Table 3.3 - Comparison of typical plant data and pilot-scale HydroFloat test results................147
Table 3.4 Operating conditions for in-plant pilot test program. .............................................149

ix

ORGANIZATION

This dissertation is separated into three major chapters, each addressing a different aspect
of improving efficiency in water-based separations. Chapter 1 discusses alternative circuitry
which can improve overall spiral performance in coal preparation plants. Chapter 2 addresses
the development of an efficient novel hindered-bed classifier. In Chapter 3, the improvements
gained by the addition of air to hindered-bed density separators are discussed.
A great majority of each chapter has been constructed from articles that the author has
published in several journals and proceedings. As a result, there is a literature review and a
reference section for each of the three major chapters in this dissertation. Listed below are the
reference data for the works from which these chapters were constructed.

Chapter 1: Improving Spiral Performance Using Circuit Analysis.

1) Luttrell, G.H., Kohmuench, J.N., Stanley, F.L. and Trump, G.D., 1998. "Improving
Spiral Performance Using Circuit Analysis," SME Annual Meeting and Exhibit, Orlando,
Florida, March 9-11, 1998, Preprint No. 98-161, 8 pp, (accepted on basis of abstract).
2) Luttrell, G.H., Kohmuench, J.N., Stanley, F.L. and Trump, G.D., 1998. "Improving
Spiral Performance Using Circuit Analysis," Minerals and Metallurgical Processing,
November 1998, Vol. 15, No. 4, pp. 16-21, (full peer review).
3) Luttrell, G.H., Kohmuench, J.N., Stanley, F.L. and Trump, G.D., 1999. "An Evaluation
of Multi-Stage Spiral Circuits," Proceedings, 16th International Coal Preparation
Conference and Exhibit, Lexington, Kentucky, April 27-29, 1999, pp.79-88, (accepted on
basis of abstract).

Chapter 2: Improving Performance of Hindered-Bed Classifiers

1) Dunn, P.L., Stewart, S.O., Kohmuench, J.N. and Cadena, C.A., 2000. "A Hydraulic
Classifier Evaluation: Upgrading Heavy Mineral Concentrates," Preprint No. 00-155,
SME Annual Meeting, February 28 - March 1, 2000, Salt Lake City, Utah, (accepted on
basis of abstract).
2) Kohmuench, J.N., Mankosa, M.J., Luttrell, G.H. and Adel, G.T., 2001. A Process
Engineering Evaluation of the CrossFlow Separator, SME Annual Meeting, February
26-28, 2001, Denver, Colorado, (accepted on basis of abstract-in preparation).

Chapter 3: Improving Coarse Particle Recovery in Hindered-Bed Separators

1) Kohmuench, J.N., Luttrell, G.H. and Mankosa, M.J., 1998. "Testing of the HydroFloat
Cell for Recovery of Coarse Phosphate," Proceedings, Engineering Foundation
Conference on Beneficiation of Phosphate, Palm Coast, Florida, December 6-11, 1998,
pp. 1-6, (accepted on basis of abstract).
2) Kohmuench, J.N., Luttrell, G.H. and Mankosa, M.J., 2000.

"Coarse Particle

Concentration Using the HydroFloat Separator," Preprint No. 00-100, SME Annual
Meeting, February 28 - March 1, 2000, Salt Lake City, Utah, (accepted on basis of
abstract).
3) Mankosa, M.J., Kohmuench, J.N. and Luttrell, G.H., 2000.

"Evaluation of the

HydroFloat Separator for Coarse Coal Recovery," Proceedings, 17th International Coal
Preparation Conference and Exhibit, May 2-4, 2000, Lexington, Kentucky, (accepted on
basis of abstract).
4) Mankosa, M.J., Merwin, R.A., Kohmuench, J.N. and Luttrell, G.H., 2000. "In-Plant
Testing of the HydroFloat Separator," 21st International Mineral Processing Congress,
Rome, Italy, July 23-28, 2000, 9 pp., (full peer review).

xi

CHAPTER 1
Improving Spiral Performance Using Circuit Analysis

1.1

Introduction
Spirals have become one of the most effective and low-cost methods for cleaning 1 mm x

100 mesh coals.

Unfortunately, the specific gravity cut-points obtained using spirals are

typically much higher than those employed by the coarse coal dense medium circuits. This
imbalance creates either a loss of clean coal or a decrease in product quality. Also, water-based
separators such as spirals tend to be much less efficient than dense medium devices due to
misplaced coal and refuse. As a result, spirals are often used in multi-stage circuits in which the
clean coal and/or middling streams from primary spirals are rewashed using secondary spirals.
Plant operators are then faced with the decision to either (i) discard the secondary middlings and
sacrifice yield or (ii) retain the middlings and accept a lower coal quality.
Studies carried out at Virginia Tech indicate that a third alternative exists for handling the
middlings problem. This option involves the use of a rougher-cleaner configuration in which the
middlings from the cleaner spirals are recycled back to the feed of the rougher spirals.
Preliminary analyses indicate that this approach can improve separation efficiency (i.e., lower
Ep) while simultaneously reducing cut-point.

1.2

Literature Review
Since its introduction by Humphreys in the 1940s (Thompson et al., 1990), spirals have

proved to be a cost effective and efficient means of concentrating a variety of ores. Their
success can be attributed to the fact that they are perceived as environmentally friendly, rugged,
compact, and cost effective (Kapur et al., 1998). During the 1980s, there had been an increased
interest in recovering coal fines. Since then, spirals have become a common method for the
concentration of 0.1 mm 3 mm coal. Spirals are able to maintain high combustible recoveries
while treating material too coarse for flotation and too fine for dense media separation.
Nonetheless, coal spiral efficiencies have not been able to match the separation results generally
found in metalliferous concentration processes (Holland-Batt, 1995).
A spiral is comprised of helical conduit of semicircular cross-sections (Wills, 1992).
Feed is introduced at the top of the spiral between 15-45% solids and is allowed to flow
downward.

Complex mechanisms, including the combined effects of centrifugal force,

differential particle settling rates, interstitial trickling, and possibly hindered-settling (Mills,
1978), effect the stratification of particles. Generally, high density material reports to the inner
edge of the spiral, while lower density material reports to the highwall of the spiral.
Classification can also occur, predominantly misplacing the coarse, high density particles to the
outer edge of the spiral. The center of the spiral trough contains any middling material present in
the feed. The schematic cross-section seen in Figure 1.1 illustrates this separation. The band of
high density material that forms near the inner edge can be removed through the use of
adjustable splitters. Ep (Ecart Probable) values generally range between 0.10 and 0.15, with cutpoints ranging between 1.70 and 2.00 SG.

Figure 1.1 Cross-section of a spiral trough flow (Chedgy et al., 1990).

Several improvements in coal spiral performance have been seen over the years. Recent
studies have concentrated on optimizing the number of turns required on a spiral. This effort is
an attempt to standardize the required number of turns needed on a spiral for different ores. As
recent as the 1960s, Australian coal spirals had as few as 2 full turns, while modern spirals can
employ as many as 7 turns to achieve the required separation (Holland-Batt, 1995).
Improvements in mineral spiral efficiencies have also been noted by Edward, et al.
(1993) after the removal of products and subsequent repulping of the remaining flow after
approximately four spiral turns. Generally, without repulping, the spiral flow can reach steadystate after only two turns. However, the recovery of the mineral can continue slowly for up to
four or more turns. Repulping the spiral flow after only a few turns can restore the initial high

rate of recovery (Holland-Batt, 1995). Several spiral manufacturers have introduced designs that
have successfully incorporated repulping.
Repulping in coal applications is less effective. When treating coal, the number of
necessary turns increases due to the relatively low specific gravity of the pulp. For instance, if a
mineral spiral, treating 4.0 SG material, requires 2 to 3 turns in order to effectively make a
separation, it can be expected that a coal spiral will need 5 to 6 turns. Repulping after only 3
turns can destroy a partial separation occurring in the finer material which would normally
require 6 turns to complete. In addition, repulpers in mineral spirals add solids and water to the
concentrate zone, while repulpers in coal spiral applications add slurry to the reject zone thereby
decreasing combustible recovery.

Holland-Batt (1995) confirmed this in his work, which

showed repulpers do not improve efficiency in coal spiral separations, and can actually decrease
efficiency.
Studies have also shown that the feed rate, especially the total volumetric flow,
introduced onto a spiral can greatly affect the performance of a spiral. Walsh and Kelly (1992)
have stated in their work that the total mass feed rate is among one of the most important factors
for determining coal spiral capacity. Their work goes on to show that for any feed pulp density,
there is an associated optimum feed rate. Further studies by Holland-Batt (1990) show that there
is indeed a performance envelope that is greatly affected by slurry density, and further indicates
that a more dominant control of spiral performance is seen when combining slurry density with
the solids flow rate (i.e., volumetric feed rate). As volumetric feed rate is increased, an increasing
amount of entrained material will report to the outer wall and effectively reduce efficiency.
These misplaced particles find it hard to escape the high velocity flow regimes and ultimately
report to the clean coal product.

Mikhail, et al. (1988) also studied the performance of spirals with relation to feed rate
and feed solids. This work found that feed rate may actually have a greater effect on separation
cut-point than even splitter position.

If solids are held constant as feed rate increases, the

volumetric velocity of slurry down the spiral also increases. Consequently, the effect of the
centrifugal force exerted on the slurry particles is greater, forcing more material to report to the
clean coal launder. This higher recovery reflects a higher cut-point. Figure 1.2 illustrates this
point.

2.3
2.2

SG(50 )

2.1
2.0
1.9
1.8
1.7
1.6
1.5
0

Solids F eed R ate (tph/start)

Figure 1.2 Feed rate versus cut-point of a spiral separation (Mikhail et al., 1988)

Unexpectedly, little or no literature was found in the area of advanced coal spiral
circuitry for the reason of improving separation efficiency. However, there were a few
exceptions. A new process, utilizing rotating spirals, was studied by Holland-Batt (1992). His
studies suggest that by rotating the downward volumetric flow (the spiral flow turns over itself
during its descent), rotating spirals can improve the separation potentials by applying one or
more additional force to the flowing film of particles. These studies were an extension of work
completed in the early 1980s. Ultimately, it was found that the finer feed particles benefited
from a flow that rotated over itself. Unfortunately, little or no improvement was found for the
coarser feed particles.
Another advance in spiral circuitry was the advent of the compound spiral. The
compound spiral is essentially a two-stage, middlings reclean circuit arranged on one column
(MacNamara et al., 1995, 1996). A short primary and short secondary spiral are positioned on
the same center tube, where a first stage clean and reject product can be removed, after which,
the first stage middlings are repulped and retreated on the secondary spiral. Advantages of this
design include lower cut-points, reduced floor space, elimination of interstage pumping, and
improved recovery (Weldon et al., 1997).

1.3

Theoretical Framework

1.3.1 Circuit Analysis

Circuit analysis can be used to evaluate the overall effectiveness of various
configurations of unit operations in mineral and coal processing circuits. This powerful tool,
which was first developed and advocated by Meloy (1983), has regretfully seen only limited
application in the analysis of coal processing circuits. Strictly speaking, this method can only be
applied if particle-particle interactions do not influence the probability that a particle will report
to a particular stream. In other words, the partition (or Tromp) curve should remain unchanged
during variations in the characteristics of the feed stream. This assumption is generally valid for
dense medium separations. This may also be a reasonable assumption for water-based processes
such as spirals provided that the changes in feed characteristics are not too large. In fact, circuit
analysis will always provide useful insight into how unit operations should be configured in a
multi-stage circuit, even if the exact numerical predictions are not completely accurate.
Consider the one-stage unit operation shown below. The concentrate-to-feed ratio is
given by:

C/F = P

[1.1]

As a result, the mass of particles of a given property reporting to either the concentrate (C) or
refuse (R) streams can be calculated as seen in Figure 1.3:

F
C

C = (P) F
R = (1-P) F

Figure 1.3 - Analysis of a single-stage separator.

where P is a dimensionless probability function that selects particles to report to a given stream
based on their physical properties. For density-based separations, the probability function can
often be estimated from an S-shaped transition function commonly referred to as the Lynch-Rao
equation (1975), i.e.:

P = (e-1)/(eX+e-2)

[1.2]

in which X is the SG/SG50 ratio and is a sharpness index. Note that the specific gravity cutpoint (SG50) is represented by a value of X=1 at which P=0.5.
The slope of the probability function evaluated at X=1 can be used to represent the
separation efficiency of the process. The slope is obtained by taking the derivative of the
concentrate-to-feed ratio at X=1. For the Lynch-Rao (1975) equation, this gives:

(C/F)/X = P/X = e / (4-4e))

[1.3]

However, efficiencies of dense medium separators are more commonly reported in terms of an
Ecart probable error (Ep). Ep values may be calculated directly from the probability function
using the expression:

Ep = SG50 (X25-X75) / 2

[1.4]

where X25 and X75 are defined at P=0.25 and P=0.75, respectively. Therefore, the following
approximation may be used in this case:

(C/F)/X = P/X P/X = -0.25 SG50/Ep

[1.5]

Now consider a similar analysis of the rougher-cleaner circuit shown in Figure 1.4.

M
F

F
R

P1

P2
C

F = F + M
C = (P2)(P1)F
R = (1-P1)F
M = (1-P2)(P1)F

Figure 1.4 - Analysis of a rougher-cleaner circuit.

By simple algebraic substitution, the concentrate-to-feed ratio (C/F) for this circuit can be
calculated as:

C/F = P2P1 / (1-P1+ P1P2)

[1.6]

If the probability function (P) is assumed to be the same for both separators, then the separation
efficiency for the rougher-cleaner circuit is:

(C/F)/X = (2P-P2) / (P2-P+1)2 P/X

[1.7]

Using the Lynch-Rao (1975) equation and noting that P=0.5 at X=1, the following relationship
may be obtained:

(C/F)/X = 1.33 P/X = 1.33 e / (4-4e)

[1.8]

Likewise, for the case involving Ep values, the following approximation will also hold:

(C/F)/X = 1.33 P/X 1.33 (-0.25 SG50/Ep)

9

[1.9]

According to this analysis, the separation efficiency (defined by the slope of the circuit partition
curve) of a rougher-cleaner circuit should be 1.33 times that of the single-stage circuit.
The relative efficiencies of other circuit configurations can be evaluated by circuit
analysis using the same approach. Several of these are summarized in Table 1.1. As shown, the
standard rougher-cleaner (Circuit 2) and rougher-scavenger (Circuit 3) configurations each have
efficiencies 1.33 times greater than the single-stage process. Note that the rougher-scavengercleaner (Circuit 4) configuration incorporating three stages has an efficiency that is twice that of
the single-stage process.
The most common multi-stage spiral circuit used in industry today is the rougher-cleaner
configuration (Circuit 5). However, unlike the circuits discussed above, the cleaner spirals are
used to treat only the middlings from the rougher spirals. The clean coal streams from both
spirals are combined to produce an overall clean product, while both reject streams are discarded.
The circuit is normally configured so that no cleaner middlings are produced and no products are
recycled. Surprisingly, circuit analysis indicates that this configuration is no more efficient than
a single-stage unit. In fact, no improvement in efficiency is obtained even when both the rougher
concentrate and middlings streams are passed to the cleaner spirals (Circuit 6). According to
circuit analysis, the only configurations inherently capable of improving separation efficiency are
those which have product streams that are recycled back to the feed of a previous stage. These
recycle streams are shown as the dotted lines in Table 1.1.

10

Table 1.1 - Theoretical relative separation efficiencies for various circuit configurations.
Circuit
(1)
Rougher
(2)
Rougher
Cleaner
With
Recycle
(3)
Rougher
Scavenger
With
Recycle

Flow Diagram
F

(6)
Rougher
Cleaner
Without
Recycle
(7)
Rougher
Cleaner
With
Middlings
Recycle
Note:

1.00

C
R
RC

CC

RR
F

1.33

CR
1.33
CS

(4)
Rougher
Scavenger
Cleaner
With
Recycle
(5)
Rougher
Cleaner
With
Middlings
Reclean

Relative Efficiency

RS
RC

CC

CS

2.00

RS

CC

CR

RC

1.00

RR

CC
RR

RC

1.00

M
F

C
RR

RC

1.22

F=Feed, R=Refuse, C=Clean, M=Middlings (Subscripts: R=Rougher, C=Cleaner)

11

The results of the linear circuit analyses should not be taken to imply that traditional
multi-stage spiral circuits have no value. The primary advantage of these traditional circuits is
that they provide an effective means for reducing the specific gravity cut-point (SG50) below that
which may be achieved using a single-stage spiral. Furthermore, the preferred configurations
identified by circuit analysis are not practical for spiral circuits due to large circulating loads and
the excessive number of spirals required.
Despite the practical shortcomings of recycle streams, the final configuration (Circuit 7)
included in Table 1.1 does appear to merit further study. In this circuit, both the concentrate and
middlings products from the rougher unit are passed to the cleaner unit. The clean stream from
the cleaner unit is taken as final product, while the cleaner refuse is combined with the rougher
refuse and discarded. The middlings stream from the cleaner spiral is recycled back to the head
of the rougher unit. As shown in Table 1.1, this configuration is capable of an efficiency that is
approximately 1.22 times that of the single-stage circuit. While not as efficient as a true
rougher-cleaner circuit, this configuration substantially reduces the amount of material that must
be recycled. In fact, this configuration was found to be the only practical circuit capable of
simultaneously improving separation efficiency while reducing cut-point.

1.3.2 Direct Procedure to Determine Optimum Circuitry

An alternative method exists for investigating these trends not only with respect to
efficiency, but also separation cut-point. Consider the popular coal spiral circuit shown in Figure
1.5. In this circuit, only the middling material is rewashed in the secondary spirals.

12

MP
F

S
P

CP

CS
RS

C = CP + CS
CP = F(PPC)
CS = MP(PSC)
MP = F(PPR-PPC)

RP
Figure 1.5 - Schematic of middlings reclean circuit.

Similar to the previous section, PPC, PPR, and PSC are the dimensionless probability functions that
select particles to report to a given stream. Namely, these are the partition values for the primary
spiral clean product, primary spiral refuse product and secondary spiral clean product,
respectively. By simple algebraic substitution described above, the overall concentrate-to-feed
ratio (C/F=PT) at a given specific gravity for this particular circuit can be represented as:

PT = PPC + PSC (PPR PPC)

[1.10]

Once a partition expression is established for a bank of spirals, Equation [1.10] can be
easily expanded by utilizing a transition function to depict the separations that occur within a
bank or circuit of spirals. A sigmoid equation was used for all of the preliminary calculations,
due to its symmetrical representation of an S-shaped partition (Tromp) curve that will not
"flatten out" at higher specific gravities. According to the sigmoid model, the partition curve for
a density separation may be represented by the following exponential transition function:

P = 1/(1+exp ((SG-SG50)/s))

[1.11]

where P is the partition factor, s is an empirical fitting constant, and SG-SG50 is the specific
gravity cut-point of the separation subtracted from the specific gravity of interest. A value of

13

0.0911 for s was found to provide a reasonable fit with experimental data available in the
technical literature. By substituting the sigmoid partition function for each of the separations
represented in Equation [1.10], the overall partition expression for this circuit now becomes:

PT = 1/(1+exp ((SG-SG50PC)/s))+1/(1+exp ((SG-SG50SC)/s))*

[1/(1+exp ((SG-SG50PR)/s))-1/(1+exp ((SG-SG50PC)/s))]

[1.12]

where SG50PC, SG50SC, and SG50PR are the specific gravity cut-points for the primary spiral clean
coal, secondary spiral clean coal, and the primary spiral refuse products.
An example of partition data for a two-stage, middlings reclean spiral circuit is shown in
Figure 1.6. This simulated data depict separations where the clean and refuse splitter positions
on the primary spirals are set for specific gravity cuts of a 1.6 and 2.0 SG, respectively. The
inner and outer splitter settings on the secondary spirals are set for an SG50 of 1.67. The fitting
constant (s) is 0.0911.
Suppose the specific gravity in question was at a 1.75 SG. By simple substitution into
Equation [1.12], the partition factor for this circuit can be calculated as 0.390. Simply stated,
only 39% of the 1.75 SG material is reporting to the clean coal launder. If the overall cut-point
of the circuit is needed, then it is only required to sweep through specific gravities until P is
equal to 0.5. The cut-point for this circuit was found to be 1.715 SG. More importantly, these
findings resulted independently of feed washability.
To validate this procedure for directly calculating circuit concentrate-to-feed ratios,
circuit partition factors were calculated both directly and through an iterative simulation
technique, which utilized feed coal washability. This was completed for the middling rewash
circuit described in Figure 1.5. The results can be seen in Table 1.2. For each technique, the

14

primary clean coal and refuse splitters were set to make separations at 1.65 and 1.95 SG,
respectively. The secondary spiral was set to make a separation at a 1.65 SG.

Primary
Clean
Primary
Refuse
Primary
Midds
Secondary
Clean
Overall
Circuit

1.00

Parition Factor (PF)

0.75

0.50

0.25

0.00
1.15

1.35

1.55

1.75

1.95

2.15

2.35

Specific Gravity (SG)

Figure 1.6 Sigmoid partition data generated for the circuit depicted in Figure 1.5.

Table 1.2 - Comparison of simulated and directly determined partition factors.

Specific Gravity

1.25
1.35
1.45
1.55
1.65
1.75
1.85
1.95
2.05
2.20

Circuit 5
Direct

Simulation

0.999
0.997
0.986
0.928
0.732
0.413
0.165
0.052
0.015
0.003

0.999
0.997
0.986
0.928
0.732
0.413
0.165
0.052
0.015
0.003

15

Clearly, as seen in Table 1.2, this alternative method of determining circuit partition
factors is mathematically equivalent to the simulation method which utilized feed coal
washability. The consistency of the directly calculated and simulated partition values verifies
that for any specific gravity cut-point, a circuit partition value can be calculated. This also
indicates that circuit results such as SG25, SG50, and SG75 can be ascertained by simply varying
the specific gravity of interest (SG in Equation [1.12]) until the indicated partition value equals
0.25, 0.50, and 0.75, respectively. More importantly, the Ecart Probable Error (Ep) and cut-point
(SG50) of the entire circuit can be determined completely independent of feed coal washability.
Naturally, the results will be more accurate provided that a proper transition function is used.
It becomes obvious that Equation [1.12] would be more useful in the form:

SG =

(P, , SG

50PC,

SG50PR, SG50SC)

[1.13]

where the specific gravity of interest is a function of the circuit partition factor (P), the fitting
constant (), and the specific gravity cut-points for the primary and secondary spirals, as
indicated by splitter position.

Unfortunately, the complexity of the ensuing mathematical

expressions prevented accomplishment of this task.

Mathematica, a powerful mathematical software package, was utilized in an effort to
achieve this goal. In order to derive an equation for the specific gravity of interest, the term SG
had to be separated from the other variables present in the partition expression (i.e., P and ).
Mathematica had great difficulty in completing this task, and was only able to successfully
calculate an equation for one of the circuits discussed above. Unfortunately, the form of the
exponential expressions constrained Mathematica to solve for SG using inverse functions. This

16

made solutions nearly impossible to obtain. On the occasion that Mathematica was successful in
deriving an expression for SG as a function of the remaining variables (i.e., splitter position), the
solution was not unique, and its sheer length made it impractical to use. Some calculated
solutions reached several pages in length. Discussions with several mathematical authorities
confirmed that a practical solution, unique or otherwise, was not possible.

1.3.2.1 Reid Equation

The sigmoid and Lynch-Rao (1975) partition functions were utilized throughout the
preliminary calculations and concept validation to represent density-based separations.
However, a more suitable partition model was needed to accurately depict spiral separations that,
when represented as Tromp curves, tend to be asymmetrical and "flatten out" at higher specific
gravity cut-points.

A partition model developed by Reid (1971) was found to provide a

reasonably good fit to experimental data available in the technical literature. This exponential
transition function is given by:

C/FReid = PReid = exp{ln(0.5)(SG/SG50)m}

[1.14]

in which m is an empirical fitting constant.

The Reid transition function is plotted adjacent to actual plant data in Figure 1.7. It
should be noted that the normalized data could not be well fit using either the sigmoid transition
function, or the Lynch-Rao (1975) expression. This is due to the inability of the symmetrical
Lynch-Rao or sigmoid models to fit asymmetrical partition data.
Using the Reid partition function, Equation [1.12] (expression for the circuit shown in
Figure 1.5) can now be rewritten as:

17

PT = (exp(ln(0.5)(SG/SG50PC)m))+ (exp(ln(0.5)(SG/SG50SC)m))*
[exp(ln(0.5)(SG/SG50PR)m)- exp(ln(0.5)(SG/SG50PC)m)]

[1.15]

1.0

Partition Factor

0.8
0.6
Plant Data
Reid
Lynch-Rao
Sigmoid

0.4
0.2
0.0
0.5

0.7

0.9

1.1

1.3

1.5

SG/SG(50)
Figure 1.7 - Normalized partition data for determining best fit.

1.3.2.2 Circuit Comparison and Optimization

Water-based separators such as spirals tend to be much less efficient than dense medium
devices. It is also common for single-stage spirals to suffer from high concentrate ash values due
to an inherently high specific gravity cut-point. To overcome these problems, spirals are often
used in two-stage circuits in which the clean coal and/or middling streams from primary spirals
are rewashed using secondary spirals. Typical examples of these circuits are shown in Figures

18

1.8(a) and 1.8(b). When operating these two-stage circuits, plant operators must decide whether
to discard the secondary middlings and sacrifice yield, or retain the middlings and accept a lower
clean coal quality. The theoretical studies conducted earlier utilizing linear circuit analysis
suggest that a third alternative exists for handling the middlings stream. This option involves the
use of a primary-secondary spiral configuration in which the middlings from the secondary
spirals are recycled back to the feed of the primary spirals. Figures 1.9(a) and 1.9(b) provide
illustrations of these particular configurations.
Table 1.3 highlights key differences between these four circuits. In the case of the
traditional circuit (Figure 1.8(a)), the secondary spirals are used to treat only the middlings
product from the primary spirals. The clean coal streams from both the primary and secondary
spirals are combined to produce a total clean product, while both the primary and secondary
reject streams are discarded. The traditional circuit is normally configured so that the secondary
middlings are sent to the reject stream, although it may also be diverted into the clean coal
product if the quality is acceptable. The modified traditional circuit (Figure 1.8(b)) is similar to
this configuration except that the primary clean coal is also rewashed with the primary middlings
using the secondary spirals. The modified traditional circuit does require more secondary spirals
than the traditional circuit, but may prove beneficial if significant amounts of high ash material
are misplaced into the total clean coal product by the primary spirals.

The circuits that

incorporate middling recycle streams (Figures 1.9(a) and 1.9(b)) are essentially identical to the
traditional and modified traditional circuits except that the secondary middlings are passed back
to the primary spiral feed.

19

Feed

Feed

Traditional
(Midds to Refuse)

Primary
Spirals
Primary
Refuse

Primary
Spirals
Primary
Refuse

Primary
Clean

Primary
Midds
Total
Refuse
Secondary
Spirals
Secondary
Refuse

Modified
(Midds to Refuse)

Primary
Clean

Primary
Midds
Total
Refuse

Total
Clean

Total
Clean

Secondary
Spirals
Secondary
Refuse

Secondary
Clean

Secondary Middlings

Secondary
Clean

Secondary Middlings

Figure 1.8 - Traditional (a) and modified traditional (b) spiral circuits without recycle.

Feed

Feed

Traditional

Modified

(Midds Recycle)

(Midds Recycle)

Primary
Spirals
Primary
Refuse

Primary
Spirals
Primary
Refuse

Primary
Clean

Primary
Midds
Total
Refuse
Secondary
Spirals
Secondary
Refuse

Primary
Clean

Primary
Midds
Total
Refuse

Total
Clean

Secondary
Spirals
Secondary
Refuse

Secondary
Clean

Secondary Middlings

Total
Clean

Secondary
Clean

Secondary Middlings

Figure 1.9 - Traditional (a) and modified traditional (b) spiral circuits with recycle.

20

Table 1.3 - Description of two-stage spiral circuits.

Figure

Circuit Configuration

Primary Clean

Secondary Middlings

1.8(a)
1.8(b)

Traditional
Modified Traditional

To Clean
To Secondary

To Refuse (or Clean)

To Refuse (or Clean)

1.9(a)
1.9(b)

Traditional (with Recycle)

Modified Traditional (with Recycle)

To Clean
To Secondary

To Feed
To Feed

In order to quantify the improvements gained by utilizing recycle streams, these four
popular coal spiral circuits described above were investigated using the direct method of
determining circuit partition values, as described in the previous sections. These circuits were
compared in terms of overall specific gravity cut-point and separation efficiency as defined by
Ep. Microsoft Excel, which can readily solve iterative problems, was used to carry out the
comparisons in a spreadsheet based format. For each test, the primary and secondary spiral clean
and refuse cut-points were varied for all SG50 combinations between 1.6 and 2.0 SG, inclusive.
Although specific gravity cut-points approaching 1.6 SG are considerably difficult if not
impossible to obtain for a given spiral in a typical operation, the theoretical results yield
important insights.
To simulate a circuit with no recycle streams, the clean and refuse specific gravity cutpoints for the secondary spiral (SG50SC and SG50SR, respectively) were held equal. Because both
cut-points were the same, no middlings product was created. In contrast, to simulate a recycle
stream, the cut-points of the secondary unit were allowed to vary, where SG50SC SG50SR. For
each variation and circuit, the SG50 and Ep were recorded and plotted. Figures 1.10(a) and
1.10(b) show a typical example of results that were obtained for the circuit shown in Figure

21

1.9(b). This circuit is equivalent to Circuit 7 in Table 1.1, which incorporates both rewashing the
clean and middlings stream in a secondary spiral and a recycle stream.
Figure 1.10(a) indicates that the separation efficiency for Circuit 7 is maximized when
the circulating load is highest, thus giving the recycle material a higher probability of reaching
the correct streams as determined by the feed washability. Allowing the primary spiral reject
splitters to cut at a high specific gravity allows more material to be recleaned in the secondary
spirals. Consequently, the circulating load is then maximized when the secondary spiral clean
coal splitters are set to cut at the lowest specific gravity possible. The lowest specific gravity
cut-point for the circuit also occurs when the secondary spiral clean coal splitters are set for a
low cut-point (Figure 1.10(b)). However, unlike Ep, the circuit SG50 rises as the circulating load
is increased. A higher circulating load increases the probability (through multiple passes) that
middling material will now reach the clean coal streams, thereby increasing the circuit SG50.

Circuit Ep

1.7
1.8
1.9

1.6

1.800-1.875
1.725-1.800
1.650-1.725
1.575-1.650
1.500-1.575

1.7
1.8
1.9
2

2
1.6

1.7

1.8

1.9

Primary Refuse (SG)

0.140-0.155
0.125-0.140
0.110-0.125
0.095-0.110

Circuit SG 50
Primary Refuse (SG)

1.6

1.6

1.7

1.8

1.9

Secondary Clean (SG)

Secondary Clean (SG)

Figure 1.10 - Circuit efficiency (a) and SG50 (b) results for varying splitter positions.

22

The resultant charts for the circuits seen in Figures 1.8(a), 1.8(b), and 1.9(a) were also
completed, but are not shown here.

Instead, for comparison purposes, the lowest possible

specific gravity cut-points and peak efficiencies for each circuit are shown in Table 1.4. It must
be noted that optimization of splitter positions is necessary since the splitter positions that yield
the lowest possible circuit cut-point do not necessarily maximize efficiency (i.e., minimize Ep).
In other words, the results in Table 1.4 are independent of one another. For example, the
modified traditional circuit with recycle is capable of achieving a minimum specific gravity cutpoint of 1.53. This circuit is also capable of maintaining a minimum Ep of 0.094. However,
these two results are generally not obtainable at the same primary refuse and secondary clean
coal splitter positions.

Table 1.4 - Circuit comparisons for Ep and SG50 using the Reid partition model.
Circuit
Figure

Circuit Label

Description

Min.
SG50

Min.
Ep

1.8(a)
1.8(b)

Traditional
Modified Traditional

Middling Rewash
Clean & Midds Rewash

1.60
1.48

0.105
0.128

1.9(a)
1.9(b)

Traditional w/Recycle
Modified Traditional
w/Recycle

Middling Rewash w/Recycle

Clean & Midds Rewash
w/Recycle

1.76
1.53

0.086
0.094

The results recorded in Table 1.4 indicate several findings.

Incorporating a recycle

stream raises the maximum possible efficiency of a circuit by increasing the probability that the
material being treated will report to the correct streams. By adding a recycle stream to the circuit
shown in Figure 1.8(a), the Ep dropped from 0.105 to 0.086. By adding a recycle stream to the
circuit shown in Figure 1.9(a), the Ep dropped from 0.128 to 0.094. These results are indicative
of efficiency increases of approximately 18% and 26%, respectively. However, by adding a

23

recycle stream, the lowest possible specific gravity cut-point of the circuit will rise slightly due
to the multiple passes of middling material in the circulating load that will now report to the
concentrate.
According to linear circuit analysis, the efficiencies of the traditional and modified
traditional circuits (Figures 1.8(a) and 1.8(b), respectively) should be relatively equal. However,
these results indicate that the modified traditional circuit has a slightly lower maximum
efficiency than the traditional circuit. This is most likely due to the increased loading of near
gravity material on the secondary spiral that is more difficult to treat. Nevertheless, allowing
more material to pass from the primary spiral to the secondary spiral for recleaning lowers the
minimum circuit SG50 dramatically. For example, without any recycle streams, recleaning the
concentrate and middlings (Figure 1.8(b)) from the primary unit lowered the SG50 of the circuit
from 1.60 to 1.48 SG when compared to exclusively recleaning the middlings material (Figure
1.8(a)). This same finding holds true when recycle streams are utilized, as seen in comparing
circuits shown in Figures 1.9(a) and 1.9(b). For these circuits, sending the concentrate and
middlings material to the secondary spiral units yields a potential SG50 reduction of 23 SG points
(i.e., a cut-point of 1.76 versus 1.53).
Though there are advantages to the recycle configurations that incorporate exclusive
rewashing of the middlings from the primary spirals, the greatest advantage comes from utilizing
recycle configurations that rewash both the concentrate and middlings from the primary spirals.
These configurations lower the Ep, but more importantly lower the specific gravity cut-point of
the entire spiral circuit. By bringing the normally high spiral circuit SG50 closer to the cut-points
found in the plant circuits that treat coarser material at greater tonnages, plant yields and

24

efficiencies become maximized. In addition, the gravities in the more efficient dense medium
circuits can now be incrementally raised resulting in an increase in total plant yield.

1.4

Circuit Testing

1.4.1 Site Description

The Winoc preparation plant located in southern West Virginia was identified as an ideal
site for the installation of a prototype test circuit for the proposed rougher-cleaner spirals with
middlings recycle. The feed coals treated at this plant contain a relatively high proportion of
middlings that tend to make small improvements in efficiency highly profitable.
The plant flowsheet for the spiral circuit is shown in Figure 1.11. The circuit is fed 1 mm
x 100 mesh material from a bank of 38 cm (15-inch) classifying cyclones. Cyclone underflow
travels to a distributor that overflows into six sets of triple-start MDL-4 spirals. The clean coal
and middlings streams from the rougher spirals flow by gravity into a cleaner feed sump. This
material is then pumped up to a second distributor that feeds six sets of triple-start MDL-4 spirals
located on the next floor. The clean coal from the cleaner spirals are taken as final product,
while the reject streams from both the rougher and cleaner spirals are discarded. The cleaner
middlings are allowed to flow by gravity back to the feed of the rougher spiral bank.

25

Cyclone Feed
(-1 mm)
Cyclones

Cleaner
Spirals

To Thickener
(-100 M)

Clean
Coal
Cleaner
Refuse

Rougher
Spirals
Sample
Point

Sump/
Pump

Rougher
Refuse

Figure 1.11 Winoc coal preparation plant rougher-cleaner spiral circuit.

1.4.2 Test Program

Three separate sets of detailed tests were performed to evaluate the circuit. The first run
(Test #1) involved the sampling of the complete rougher-cleaner circuit with partial recycle. In
the second run (Test #2), the splitters on the cleaner spiral were adjusted so as to produce no
middlings stream. The results obtained from this test run would be similar to those obtained
from the widely used traditional rougher-spiral circuit. Finally, a third test run (Test #3) was
performed under the same conditions as the second test run, but at a significantly reduced plant
feed rate. It is well known that reducing the feed tonnage can significantly reduce the SG cutpoint for spirals (Mikhail et al., 1988). Therefore, data from the third run was used to determine
whether the SG cut-point could be more effectively reduced using (i) rougher spirals in series

26

with cleaner spirals or (ii) the same total number of spirals in parallel. The operating conditions
for the three test runs are summarized in Tables 1.5 and 1.6.

Table 1.5 - Circuit feed rates used during spiral testing.

Run

Rougher
tonne/hr (ton/hr)

Cleaner
tonne/hr (ton/hr)

Middlings
tonne/hr (ton/hr)

Test #1
Test #2
Test #3

58.9 (64.9)
56.0 (61.7)
30.6 (33.7)

40.2 (44.3)
40.7 (44.8)
21.0 (23.1)

3.9 (4.3)
-----

Table 1.6 - Splitter settings used during spiral testing.

Run

Rougher
Outer,
cm (inch)

Cleaner
Outer,
cm (inch)

Cleaner
Inner,
cm (inch)

Test #1
Test #2
Test #3

27.9 (11)
27.9 (11)
27.9 (11)

34.3 (13.5)
22.2 (8.75)
22.2 (8.75)

26.7 (10.5)
-----

The test program generated 22 total samples (i.e., 8 samples for Test #1 and 7 samples
each for Tests #2 and #3). The diamond symbols shown in Figure 1.11 indicate the locations of
the sampling points. Each sample was sized at 16 mesh and 100 mesh. The 16 x 100 mesh
fraction was subjected to float-sink analysis to produce float 1.4, 1.4 x 1.6, 1.6 x 1.8, 1.8 x 2.0
and sink 2.0 gravity classes. Each class was analyzed for ash and sulfur content. The data were
then analyzed using a mass-balance program to ensure that the results were reliable and selfconsistent.
27

1.4.3 Experimental Results

Data from each test was collected and used to construct partition curves for each spiral
configuration. The partition curves for Tests #1, #2, and #3 can be seen plotted in Figures 1.12,
1.13, and 1.14, respectively. Partition curves were constructed for the rougher, cleaner and
combined circuit performance in each test case. The data presented in these figures indicate that
the rougher bank of spirals for each configuration consistently operated at a higher cut-point than
the corresponding cleaner bank of spirals. This dissimilarity was less pronounced in Test #3,
where a significantly lower feed rate was utilized.

This outcome is a direct result of the

corresponding decrease in volumetric feed flow rate.

A lower volumetric feed flow rate

decreases the effect of the centrifugal force exerted on the slurry particles, resulting in a lower
percentage of material reporting to the clean coal launder. This lower recovery reflects a lower
cut-point.
Figures 1.12, 1.13, and 1.14 also illustrate that the overall efficiency of test circuit #1 is
superior to that of test circuit #2. This is seen when comparing the relative steepness (Ep) of
each of the corresponding combined circuit partition curves. It also appears that the overall
efficiency of test circuit #3 was relatively high in comparison to both test circuit #1 and #2.
Unfortunately, twice as many spirals would be required to obtain this efficiency since Test #3
was conducted at a feed rate half of that utilized in Test #1 or #2.

28

1.0
0.9

Partition Factor

0.8
0.7
0.6
0.5
0.4
0.3
Rougher
Cleaner

0.2
0.1

Combined

0.0
1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2.0

2.1

2.2

Specific Gravity

Figure 1.12 Test #1 experimental partition data.

1.0
0.9
0.8

Partition Factor

0.7
0.6
0.5
0.4
Rougher
0.3

Cleaner
Com bined

0.2
0.1
0.0
1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2.0

2.1

Spe cific G ravity

Figure 1.13 Test #2 experimental partition data.

29

2.2

1.0
0.9

Partition Factor

0.8
0.7
0.6
0.5
0.4
0.3
Rougher
Cleaner
Combined

0.2
0.1
0.0
1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2.0

2.1

2.2

Specific Gravity

Figure 1.14 Test #3 experimental partition data.

Tables 1.7 and 1.8 summarize the performance data seen in Figures 1.12, 1.13, and 1.14.
The key comparisons are highlighted as bold numbers for each case. The data shown in the first
column of Table 1.7 indicate that a reduction in feed rate (Test #3) reduced the SG50 for the
rougher spirals from 1.95-1.97 down to 1.82. However, this cut-point was still considerably
higher than the 1.63-1.65 SG values obtained using the rougher-cleaner circuits (Tests #1 and
#2). It was possible to achieve SG50 of 1.55 for Test #3, but only at half of the feed rate (or with
twice the number of spirals) used in Tests #1 and #2. These results demonstrate that a greater
reduction in SG50 cut-point can be achieved with a rougher-cleaner circuit than with a singlestage circuit. Rougher-cleaner circuits are highly recommended for this reason.

30

Table 1.7 - SG50 values obtained from the spiral tests.

Run

Rougher

Cleaner

Overall

Test #1
Test #2
Test #3

1.97
1.95
1.82

1.66
1.70
1.65

1.65
1.63
1.55

Table 1.8 - Ep values obtained from the spiral tests.

Run

Rougher

Cleaner

Overall

Test #1
Test #2
Test #3

0.16
0.15
0.17

0.25
0.25
0.25

0.16
0.20
0.18

Table 1.8 shows that the Ep values for the rougher spirals remained relatively constant at
about 0.16 + 0.01. The Ep values were even more consistent for the cleaner spirals, although a
worse Ep (i.e., Ep=0.25) was obtained. This suggests that the greater loading of near-gravity
material adversely impacted the shape of the partition curve for the cleaner spirals. As a result,
the overall Ep values for the traditional rougher-cleaner spiral circuit (Test #2) are worse than
those obtained using single-stage spirals. Thus, a portion of the gain achieved by reducing the
cut-point is lost as a result of the lower overall circuit efficiency. On the other hand, data from
the modified rougher-cleaner circuit (Test #1) suggests that good efficiencies (i.e., Ep=0.16) can
be maintained through the use of a middlings recycle stream. It is also worth noting that the ratio
of the Ep values for Tests #1 and #2 is 1.25 (i.e., 0.20/0.16). This value is close to the theoretical
ratio of 1.22 predicted by circuit analysis.

31

1.5

Circuit Simulations
Natural variations in the washabilities of the feed coal made it difficult to calculate the

exact improvement offered by the new rougher-cleaner circuit. To overcome this limitation, a
series of partition model simulations were conducted using a fixed set of typical washability
data for the plant.

This was accomplished by developing regression equations for the

experimental partition curves obtained in each of the three test runs.

To properly simulate both the rougher and cleaner circuits, two different fitting
expressions were required.

The most adequate fitting expressions are shown as bold lines in

Figures 1.15 and 1.16 for both the rougher and cleaner circuits, respectively. For the rougher
spiral, the best fit to the rougher partition factor (PR) was obtained using the transition function
advocated by Reid (1971):

PR = exp{-0.693 (SG/SG50)m}

[1.16]

in which m is an empirical fitting constant.

Shown in Figure 1.15 is the plant data for the rougher spiral circuit. Also shown are the
best-fit curves for the Reid (1971) and popular Lynch-Rao (1975) equations. The rougher
spiral circuit data tended to be asymmetrical. In a spiral separation, material at a lower specific
gravity is efficiently partitioned to the clean coal launder; however, the slightly raised tail is an
indication of how spirals tend to misplace coarse, high density material to the clean coal product
stream. Since the Lynch-Rao equation is a symmetrical transition function, it did not fit the data
well.
Unlike the rougher spiral circuit, the Reid (1971) equation was not an adequate fitting
expression for the cleaner spiral circuit (See Figure 1.16). In this circuit, the increased amount of

32

near-gravity material caused an even higher lift in tail of the partition data. The near-gravity
material present in the cleaner spiral circuit makes an efficient separation more difficult to obtain
as evidenced by the flatter partition data of the cleaner spiral circuit in comparison to the
rougher spiral circuit.
For the cleaner spiral, a modified version of the Reid expression had to be developed in
order to obtain the best fit to the cleaner partition factor (PC), , i.e.:

PC = 1- exp{-0.693 / (SG/SG50)n}

[1.17]

in which n is an empirical fitting constant. Equation [1.17] is plotted along with actual plant data
in Figure 1.16. Both the Reid (1971) and Lynch-Rao (1975) equations are also plotted. It can
easily be seen that neither the Lynch-Rao nor the Reid equation adequately fit the data for the
cleaner spiral circuit. The modified Reid expression permits the low gravity portion of the
partition curve to remain relatively steep, while allowing the tail of the partition curve to lift.
Consequently, this equation accurately predicts how a coal spiral will misplace an increased
amount of high gravity and/or middling particles when treating a feed material of a tight specific
gravity range.

33

Partition Factor

1.0
0.8
0.6
0.4

Plant Data
Reid Eqn.
Lynch-Rao

0.2
0.0
0.5

0.7

0.9

1.1

1.3

1.5

SG/SG(50)
Figure 1.15 - Normalized partition curve obtained using the data from the rougher spirals.

Partition Factor

1.0

Plant Data
Mod. Reid
Reid Eqn.
Lynch-Rao

0.8
0.6
0.4
0.2
0.0
0.5

0.7

0.9

1.1

1.3

1.5

SG/SG(50)
Figure 1.16 - Normalized partition curve obtained using the data from the cleaner spirals.

34

Circuit simulations were performed for each of the three test runs using partition factors
obtained from Equations [1.16] and [1.17]. The experimental feed washability data obtained
during Test #2 was used in all of the simulations. Two sets of simulations were conducted. In
the first set, clean coal yield and ash was calculated using the actual SG cut-points from the
experimental runs. In the second set, the SG cut-points were adjusted slightly so that a consistent
product ash of 11.75% was obtained. The simulation results are summarized in Table 1.9.

Table 1.9 - Summary of simulation results.

Simulated
Circuit

Yield
(%)

Ash
(%)

Ep

Organic
Efficiency

0.15
0.19
0.17

92.8
86.2
79.2

0.15
0.18
---

92.5
86.4
---

Using Actual Cut-Points:

Test #1
Test #2
Test #3

59.47
55.27
62.30

11.87
11.66
13.08

Using Adjusted Cut-Points:

Test #1
Test #2
Test #3

59.27
55.40
---

11.75
11.75
---

The simulation results conducted using actual plant cut-points indicate that the roughercleaner circuit with middlings recycle would produce a 59.47% yield at 11.87% ash. This result
compares favorably to the 55.27% yield and 11.66% ash that would be obtained using the
rougher-cleaner without recycle. In contrast, the simulation of Test #3 for the rougher spiral
circuit only (with no recleaning stages) operated under actual plant cut-points produced the
highest yield of 62.30%, but at a relatively high ash of 13.08%.

35

Although the organic

efficiencies for these simulation runs have been reported in Table 1.9, these values cannot be
directly compared because of the variations in clean coal ash content.
In order to improve the comparisons, a second set of simulations was conducted in which
the cut-points for the rougher spirals were adjusted so that a constant clean coal ash of 11.75%
was obtained in each case. Unfortunately, it was not possible to achieve an ash value this low for
the rougher spiral circuit only (Test #3) since it would require a substantial adjustment to the SG
cut-point to a value below that which is realistically achievable. On the other hand, only minor
adjustments to the SG values were necessary to achieve 11.75% ash for the rougher-cleaner
configurations. As shown in Table 1.9, the circuit with the middlings recycle (Test #1) produced
a yield of 59.27% at an organic efficiency of 92.8% compared to a yield of only 55.40% at an
organic efficiency of 86.2% for the circuit with no middlings recycle (Test #2). This represents a
yield increase of 3.87% at the same ash content. For a typical 3-shift operation with a circuit
feed rate of about 40 tonne/hr (44 ton/hr), this represents a revenue increase of approximately
$255,000 annually (i.e., 44 ton/hr x 3.87% x $25/ton x 6000 hr/yr = $255,000). Preliminary
economic analyses show that this additional revenue would offer an attractive payback on the
capital investment required to purchase additional spirals.
Finally, a few comments need to be made regarding the Ep values obtained from the
simulation runs. According to the circuit analyses conducted in the introductory section of this
chapter, the rougher-cleaner circuit with middlings recycle was expected to be 1.22 times more
efficient than the same circuit without a middlings recycle stream. A comparison of the values
reported in Table 1.9 for the circuit simulations shows that an Ep ratio of 1.20 was achieved (i.e.,
0.18 / 0.15 = 1.20). The close agreement between these Ep ratios further supports the use of
linear circuit analysis as an effective tool for evaluating spiral circuit performance.

36

For

comparison purposes, the resultant partition curves from the simulations of the overall roughercleaner circuits are shown in Figure 1.17. The superior performance of the modified circuit is
easily observed in this plot.

1.0
Test #1 Circuit
Test #2 Circuit

Partition Factor

0.8
0.6
0.4
0.2
0.0
1.2

1.4

1.6 1.8 2.0

Specific Gravity

2.2

Figure 1.17 - Overall circuit partition curves for the rougher-cleaner spiral circuits with (Test #1
Circuit) and without (Test #2 Circuit) middlings recycle for simulations conducted
at constant clean coal ash.

37

1.6

Conclusions
1. A theoretical study was conducted using linear circuit analysis to evaluate a variety of
different multi-stage spiral circuits. The study suggested that a modified rougher-cleaner
circuit incorporating a middlings recycle stream offered the best option for improving
spiral separation efficiency while maintaining a reasonable circulating load.

2. Linear circuit analysis allowed for the derivation of an alternative method for determining
the partition expression of a given spiral circuit without the requirement of a washability
based simulation. Moreover and more importantly, the efficiency (Ep) and cut-point
(SG50), of a given spiral circuit can be calculated independent of washability provided a
proper transition function (i.e., Reid, Lynch-Rao, and/or modified Reid expressions) is
used to simulate the mineral separation.

3. A two-stage spiral test circuit was installed at the Winoc preparation plant located in
southern West Virginia. The test circuit was designed so that a variety of different circuit
configurations could be compared under actual plant conditions.

4. For an equivalent number of spirals, the in-plant spiral test data indicate that roughercleaner circuits operated in series are superior to parallel circuits for reducing the SG50.
This capability is needed so that the spiral circuit cut-point can be brought into line with
the cut-points realized in the coarse coal dense medium circuits.

5. Test data was used to develop regression equations that were used to simulate the
experimental partition curve data produced during the on-site circuit testing. While a

38

rougher spiral circuit separation could be simulated using an equation developed by Reid
(1971), a new, modified version of this equation was developed to properly simulate
cleaner spiral circuits that generally treat large amounts of near-gravity material.

6. The in-plant test results also suggest that the SG50 for rougher-cleaner spiral circuits
operated with and without a middlings recycle are very similar (i.e., 1.65 SG in this
case). However, the separation efficiency (as measured by Ep) was approximately 1.25
times higher for the circuit incorporating a middlings recycle stream. This ratio compares
favorably with the theoretical ratio of 1.22 predicted by linear circuit analysis and a ratio
of 1.20 obtained from partition simulations.

7.

Preliminary calculations suggest that the rougher-cleaner spiral with middlings recycle is
capable of increasing circuit yield by 3.86% at the same ash. For a typical plant, this
would represent about $255,000 of additional revenues annually. Economic analyses
suggest that this additional revenue would offer an attractive payback on the capital
investment.

39

1.7

References

Chedgy, D.G., Placha, D.S. and Watters, L.A., Spiral Concentrators for Fine Coal Processing,
PCMIA/SME Joint Meeting, Washington, PA, November 1-2, 1990.
Edward, D., Li, M., Davis, J. and Kruitschnitt, J., Spiral Research: Technique Development and
Use, Pittston Coal Management Co., Paper B2, 1998, pp. 100-119.
Holland-Batt, A.B., Interpretation of Spiral and Sluice Tests, Trans. Instn. Mining and
Metallurgy, 1990, Vol. 99, pp. C1-C20.
Holland-Batt, A.B., A Study of the Potential for Improved Separation of Fine Particles by Use
of Rotating Spirals, Minerals Engineering, 1992, Vol. 5, Nos. 10-12, pp. 1099-1112.
Holland-Batt, A.B., The Effect of Feed Rate on the Performance of Coal Spirals, Coal
Preparation, 1994, Vol. 14, pp. 199-222.
Holland-Batt, A.B., The Dynamics of Sluice and Spiral Separations, Minerals Engineering,
1995, Vol. 8, No. 1/2, pp. 3-21.
Holland-Batt, A.B., Gravity Separation: A Revitalized Technology, Mining Engineering, 1998,
Vol. 50, No. 9, pp. 43-48.
Kapur, P.C. and Meloy, T.P, Spirals Observed, Inter. J. Mineral Processing, 1998, Vol. 53, pp.
15-28.
Lynch, A.J. and Rao, T.C., Modeling and Scale-Up of Hydrocyclone Classifiers, Paper No. 9,
11th Inter. Mineral Processing Congress, Cagliari, Italy, April 21-26, 1975.
MacNamara, L., Addison, F., Miles, N.J., Bethell, P. and Davies, P., The Application of New
Configurations of Coal Spirals, Proceedings, 12th International Coal Preparation Conference
and Exhibit, Lexington, Kentucky, May 2-4, 1995.

40

MacNamara, L., Toney T.A., Moorhead, R.G., Miles, N.J., Bethell, P. and Everitt, B., On Site
Testing of the Compound Spiral, Proceedings, 13th International Coal Preparation
Conference and Exhibit, Lexington, Kentucky, April 30-May 2, 1996.
Meloy, T.P., Analysis and Optimization of Mineral Processing and Coal-Cleaning Circuits
Circuit Analysis, Inter. J. Mineral Processing, 1983, Vol. 10, p. 61.
Mikhail, M.W., Salama, A.I.A., Parsons, I.S. and Humeniuk, O.E., Evaluation and Application
of Spirals and Water-Only Cyclones in Cleaning Fine Coal, Coal Preparation, 1988, Vol. 6,
pp. 53-78.
Mills, C., Process Design, Scale-Up and Plant Design for Gravity Concentration, Mineral
Processing Plant Design, 1978, AIMME, New York.
Reid, K.J., Derivation of an Equation for Classifier-Reduced Performance Curves, Technical
Note, Canadian Metallurgical Quarterly, 1971, Vol. 10, No. 3, pp. 253-254.
Thompson, J.V. and Welker, M., The Humphries Companies: Development and Application of
Humphreys Spiral Concentrator, Skillings Mining Review, Feb. 24, pp. 4-15.
Walsh, D.E. and Kelly, E.G., An Investigation of the Performance of a Spiral Using
Radioactive Gold Tracers, Minerals and Metallurgical Processing, 1992, Vol. 9, No. 3, pp.
105-109
Weldon, W.S. and MacHunter, R.M.G., Recent Advances in Coal Spiral Development, SME
Annual Meeting, Denver, Colorado, Preprint No. 97-82, Feb. 24-27, 1997.
West, T.W. and Apodaca, L.E., Cost Effective BTU Recovery by Fine Coal Washing Spirals,
117th Annual AIME Meeting, Proceedings, Phoenix, Arizona, Jan. 25-28, 1988.
Wills, B.A., Mineral Processing Technology, 5th Edition, 1992, pp.430-433.

41

CHAPTER 2
Improving Performance of Hindered-Bed Separators

2.1

Introduction
Hindered-bed hydraulic separators have been used in mineral processing applications for

years. Simply stated, a hindered-bed separator is a vessel in which feed settles against an evenly
distributed upward flow of water or other fluidizing medium. Typically, these devices are used
for size classification, however, if the feed size distribution is within acceptable limits, hinderedbed separators can be used for the concentration of particles based on differences in density.
A simplified schematic of a typical hindered-bed separator is shown in Figure 2.1. Most
hindered-bed separators utilize a downcomer to introduce feed material to the system. This
material enters the feed zone and may encounter either free or hindered settling conditions,
depending on the concentration of particles in the separator. The settling particles form a
fluidized bed (teeter-bed) above the fluidization water injection point.

Material is then

segregated based on terminal, hindered-settling velocities. Slower settling material reports to the
top of the teeter-bed while the faster settling particles descend to the bottom of the teeter-zone.
Specifically, low density and fine material reports to the overflow, while coarse and high density
material report to the underflow. Particles that settle through the teeter-bed enter a dewatering
cone and are discharged through an underflow control valve. The rate of underflow discharge is
regulated using a PID control loop.

42

Feed

Overflow

Fluidization Water
Underflow
PID
Control
Loop

Figure 2.1 - Schematic diagram of a conventional hindered-bed separator.

More recently, a new hindered-bed classifier separator has been developed that utilizes an
innovative feed presentation system. This device, which is known as the CrossFlow separator, is
shown in Figure 2.2. The CrossFlow utilizes a tangential, low-velocity feed entry system that
introduces slurry at the top of the classifier. This approach allows feed water to travel across the
top of the unit and report to the overflow launder with minimal disturbance of the fluidization
water within the separation chamber. To reduce the velocity of the feed flow, the feed stream
enters a side well before flowing into the separation chamber. The feed then overflows into the
top of the device. Solids settle into the separation chamber as they travel between the feed entry
point and overflow launder. The result of this feed presentation system is the elimination of
excess feed water in the separation chamber, which can adversely effect separation efficiency.

43

Feed
Overflow

Feed Well

Separation
Chamber

Fluidization Water

Dewatering
Cone

PID

Control
Loop

Underflow

Figure 2.2 - Schematic of CrossFlow hindered-bed separator.

44

2.2

Literature Review

2.2.1 General
Hydraulic classifiers are primarily categorized by the method in which the coarse
material is discharged from the separation zone of the unit (Heiskanen, 1993). The first category
is marked by a lack of underflow (or coarse fraction) control. This causes an underflow stream
of such high velocity to occur that no fluidized bed forms and no gradation of particles (by size
and density) manifests. The second group of hydraulic classifiers is marked by an attempt to
control the underflow discharge generally causing the appearance of a teeter-bed. Classifiers can
be further subdivided into mechanical or non-mechanical categories. In mechanical classifiers,
the underflow discharge is removed via mechanical means. In non-mechanical classifiers, the
underflow stream is removed through mass-action and gravity.
The CrossFlow separator is a non-mechanical, hindered-settling, counter-current
hydraulic classifier that utilizes a teeter-bed. There are several other classifying devices that fall
under this description, including the Floatex fluidized-bed classifier (or Floatex Density
Separator) and the allflux separator. In these classifiers, the underflow rate is restricted and a
teeter-column is formed by solids settling against elutriation water (teeter-water) that is fed
evenly across the entire cross-section of the unit. Generally, coarse particles are graded in order
of decreasing terminal velocity (Heiskanen, 1993), with the coarser particles settling through the
teeter-bed, and the finer particles reporting to the overflow. The high interstitial velocities of
water traveling between the particles in the teeter-bed ensure that there is little bypass of fines to
the underflow.

In fact, these types of classifiers often produce very clean underflows

(Schwalbach, 1965).

45

Typically, teeter-bed classifiers are capable of separations as coarse as 800 microns and
as fine as 75 microns (Littler, 1986). According to Heiskanen (1993), when the separation is
coarser than 800 microns, efficiencies drop dramatically as the separator begins to act as an
elutriator. When separations finer than 100 microns are conducted, low capacities become an
issue. Solid capacities typically range from 10 to 40 tph/m2 (0.85-3.40 tph/ft2) depending on the
cut-point of the separation.

Generally, as separations become coarser, the solids capacity

increases, and the opposite is true for finer separations.

Littler (1986) utilized a Floatex hindered-settling classifier and essentially summarized
the effects on classifier performance of operating at different separation cut-points. It is stated
that the Floatex separator is considered to be the most advanced commercial separator for
hydraulic particle classification and is able to treat material whose size is between what would be
considered optimal for either screens (coarse) or hydrocyclones (fine). A schematic of this
device can be seen below in Figure 2.3. As shown, mineral slurry is introduced to the teeter
chamber through a downcomer. A differential pressure cell and discharge valve controls the
bed-level in the unit. According to Littler (1986), there is a discernable drop in efficiency
(normally 80-90%) as the nominal mesh of the separation is increased from 140 to 16 mesh
(U.S). These results reflect the response of the Floatex separator, however, the same general
trend can be found in any hindered-bed classifier.

46

Figure 2.3 Schematic of Floatex hindered-bed classifier (Littler, 1986).

In Littlers work, efficiency (E) is defined as:

E=

O
of

F f (100 f )

[2.1]

where O is the separator overflow tonnage (t or %), F is the separator feed tonnage (t or %), o is
the material in the overflow finer than the separation mesh (%), and f is the material in the feed
finer than the separation mesh (%).

47

Improving the sharpness of classification has many benefits.

The greatest of these

benefits is the reduction of misplaced material to the product stream. With less misplacement,
more properly sized material (an amount proportional to the total reduction of misplace
material), can now report to the product launder. Littler (1986) goes on to state that improved
classification can be beneficial to closed-circuit grinding, by reducing circulating load and
improving the gradation of material that is treated in other downstream processes (i.e., flotation).
Since the advent of the original hydraulic classifier in 1927 by Fahrenwald (Taggart,
1950), hydraulic separators have been used most extensively in the classification of material
based on hindered-settling phenomena. However, it has been shown that these devices can also
be effectively applied to gravity separations provided that the size distribution of the feed is
within acceptable limits, depending on the application (Heiskanen, 1993).

An example of

successful density applications using hindered-bed separators can be seen in fine and coarse coal
processing, (Reed et al., 1995; Honaker, 1996) mineral sands beneficiation (Mankosa et al.,
1995), and the recycling of chopped wire (Mankosa and Carver, 1995). Wills (1992) considers
this gravity concentration component, commonly found in hindered-bed classifiers, an added
increment. According to Bethell (1988), the cleaning efficiency of a teeter-bed separator is
limited to a feed size range of 6 to 1 when used as a gravity separation device. This is due to the
fact that when treating wider size distributions, coarse, low density material will be misplaced to
the underflow due to its net greater sizing effect. In the same way, extremely fine, high density
material will report to the overflow irrespective of its overall density. This inherent disadvantage
is further discussed in Chapter 3 of this dissertation.

48

2.2.2 Hindered-Settling
In teeter-bed applications, the free settling rates of particles are greatly reduced. This is
in response to the presence of other particles that cause either particle-to-particle collisions
and/or near-misses (Littler, 1986). As the size of the particle decreases, the reduction in the
settling rate of that particle increases.

According to Littler (1986), the hindered-settling

phenomenon begins to take place at approximately 20% solids by mass. Classification utilizing
hindered-settling is an improvement over free-settling classification due to the fact that less fine
material can become entrapped by coarse particles that settle more slowly through a teeter-bed.
In free-settling applications, coarse material can settle quickly enough to entrain fine particles to
the underflow.
According to Zimmel (1983, 1990) five effects occur as the volume fraction of solids ()
in a slurry increases. This includes a decrease in the cross sectional area available for the
elutriation fluid (teeter water) which results in an increased net velocity as seen by the settling
particles. The apparent viscosity of the pulp is also increased. This increase of apparent specific
gravity toward the specific gravity of the particles causes a reduction of gravitational force
effects on the individual particles.

The last two effects include an increase of wall hindrance

and the occurrence of hydrodynamic diffusion.

The apparent slurry viscosity is very important in determining the hindered-settling
velocities of particles. When treating a slurry containing particles in the size ranges generally
used in mineral processing classification applications, many other factors and variables can also
be significant. The most important factors being the volume concentration of solids, particle size
and particle shape (Heiskanen, 1993). There are various expressions available in the literature

49

that describes slurry viscosity. Einstein derived the following equation for apparent viscosity
():
= 1 + 2.5

[2.2]

where is the fraction of solids by volume. Heiskanen and Laapas (1979) and Laapas (1983)
later went on to modify this formula with an empirical correction as seen below.
= 1 + 2.5 + 14.12 + 0.00273e16.6

[2.3]

Rutgers (1962) derived a simple equation for pulp viscosity in the form of the Arrhenius
equation as seen here:
= w exp( k)

[2.4]

where w is the viscosity of water or other fluidizing medium. The variable, k, is a fitting
parameter which has been given values of 5 (John and Goyal, 1975) to 14 (Plitt, 1976). This
equation provides values similar to the equations listed above when k is approximately 5.
In 1989, Swanson suggested this semi-empirical equation:

= w

2max +
2(max )

[2.5]

where max is the highest fraction of solids by volume obtainable for a specific material. An
incredible amount of work was found in the literature on determining this variable, max.
Disappointingly, most of the conclusions have been empirical in nature.

50

According to Sudduth (1993), many attempts have been made to predict the optimum size
distribution for packing material, but little work has been completed on determining the exact
value of the attainable maximum fraction of solids (Yu and Standish, 1993). As early as 1930, it
was concluded that size ratios of particle components was an extremely important factor in
determining the maximum packing of solids (Furnas, 1931; Westman and Hugill, 1930). The
most definitive work was completed by McGeary (1961), nevertheless it can be considered
empirical in nature, as it requires direct measurement and is only applicable for ideal spherical
particle systems. However, in his work, the packed density for monosized spherical particles
was approximately 62.5% that of the crystal density of the solid. Sudduth (1993) was able to
match the results summarized by McGeary by utilizing the size ratios of the first to nth size
fraction of a dry mineral sample in determining the maximum obtainable packing of solids.
Sudduth (1993) used an empirical process in choosing the proper value for n.
According to Low and Bhattacharya (1984), the determination of max has been calculated
from direct measurements and even graphical estimation. Work in estimating these values was
conducted by Lewis and Nielsen (1969) who concluded that the maximum concentration of
solids was far more accurately determined in air than in water. Another conclusion demonstrated
was that as particles increased in aggregation, the maximum packing of solids decreased. This
was a direct result of a lack of sphericity of the particles.
Other methods for determining the maximum concentration of particles include direct
measurement through sedimentation (Robinson, 1957) and a least square regression of the
experimental data. Essentially, most reliable means of determining max are empirical in nature.
According to Yu and Standish (1993), the packing density of the system is affected by
both the solids volumes as well as their particle size distribution. Yu and Standish (1988, 1991)

51

further demonstrate that linear models can satisfactorily predict the solids packing with the use of
a discrete or simple continuous size distribution. However, recent work by Swanson (1999) in
the area of hindered-settling phenomena advocates the determination of the maximum
concentration of solids through the direct measurement of teeter-bed expansion when
transitioning between a fully settled and fully elutriated state.
In modeling hindered-bed separations, several equations have been developed and
utilized for determining the hindered-velocity of a particle (vt). Masliyah (1979) utilizes the
expression:

vt =

gd 2 (s f )
f F( )
18f

[2.6]

where g is the force due to gravity, d is the diameter of the particle, s is the density of the solids,

f is the density of the fluidizing medium, f is the suspension voidage (1-), and f is the
viscosity of the fluid.

The term F() describes a function that accounts for particle

concentration. Usually, this function is in the form described by Richardson and Zaki (1954).
The above equation is for a laminar flow regime and can be corrected for non-stokes flow
as seen below:

vt =

gd 2 s susp

18f 1 + 0.15 Re0.687

)f F( )

[2.7]

where susp is the apparent density of the suspension and Re is the Reynolds number (Masliyah,
1979). Reynolds number can be calculated as:

52

Re =

df v t f
susp

[2.8]

where susp is the apparent viscosity of the suspension.

Richardson and Zaki (1954) define F() as (1-) where is an unknown function of
particle size and shape. For transitional flow regimes, normally found in hindered-bed separator
applications, can be determined as seen below:
= 4.36 Re-0.03

for 0.2 < Re < 1.0

[2.9]

= 4.4 / Re0.1

for 1.0 < Re < 500

[2.10]

Another accepted form of this function for transitional flow regimes comes from Barnea
and Mizrahi (1973) where:

5(1 f )
F(f ) = 1 + (1 f )1 / 3 exp

3f

[2.11]

Mondal (1997) lists several other expressions in his work for hindered-settling velocities,
including those derived by Steinour (1944) and Concha and Almendra (1979). However, these
settling equations are not valid for non-transitional flows. Work has also been conducted by
Brauer, et al. (1973) that shows the reduced (hindered) settling velocity depends on the free
settling velocity of a particle in a narrow size and density class i, the particle size, and the
distance to any other particle. According to Brauer, et al. the hindered settling velocity (vhi) can
be calculated as:

vhi = voi kif kic

53

[2.12]

where the parameter kif is a fluid counter flow factor which addresses the displacement of water
by settling particles. The parameter kic is another factor that addresses turbulence caused by
clusters of settling particles.
Recent population balance models by Mondal (1997) and Swanson (1999) each attempt
to accurately simulate hindered-settling. Mondal utilizes the equation set forth by Brauer, et al.
He was able to show significant upgrading when processing fine coal slurry using a Floatex
Density Separator. Swanson, on the other hand, utilized a semi-empirical process that was selfdeveloped and was able to show simulated consistency with 50 sets of test data. Both models
demonstrate impressive results.

54

2.3

Comparative Studies

2.3.1 In-Plant and Laboratory Testing

The CrossFlow concept was originally investigated in the laboratory, by way of a labscale separator. The laboratory setup is shown in Figure 2.4. This laboratory test unit was
constructed out of Plexiglas and has a cross-sectional area of 8 in2 (0.005 m2). The clear nature
of the Plexiglas allowed for the optical determination of varying flow regimes and the presence
of turbulence. A vibratory feeder was used to provide a constant feed flow rate to the separator.
A control loop, consisting of a PID controller, pressure sensor, and underflow pneumatic valve
was employed to maintain a constant bed level within the unit. A traditional downcomer could
be installed, which would allow this lab-scale unit to operate like a conventional teeter-bed
separator.
It was apparent, even at the earliest stages of experimentation, that efficiencies and
capacities for applications using teeter-bed separators could be improved with the CrossFlow
device. The earliest tests compared the conventional teeter-bed separator against the CrossFlow
separator at relatively benign test conditions. These initial investigations were completed using
either passing 14 mesh aggregate limestone or phosphate ore. Feed solids rates ranged from 1.0
to 2.3 tph/ft2 at a feed percent solids of approximately 50%. Figure 2.5 shows the results of these
initial comparative tests. For an array of separation cut-points at the solid feed rates described, it
can easily be seen that the CrossFlow has a tendency to perform slightly better in terms of
separation efficiency. The separation efficiency was defined as either Ecart Probable (Ep) or
Imperfection as calculated in Equations [2.13] and [2.14].

Ep =

d 75 d 25
2

55

[2.13]

I=

Ep
d 50

[2.14]

where d75, d50, and d25 represent the size at which 75%, 50%, and 25% of the feed mass,
respectively, reports to the underflow of the separator.

Feeder Tray
PI Loop Controller

CrossFlow

Pressure
Sensor

Control Valve

Figure 2.4 Laboratory-scale CrossFlow separator.

56

0.16
0.14
0.12

Ep

0.10
0.08
0.06
0.04

CrossFlow
Conventional

0.02
0.00
0.20

0.30

0.40

0.50

0.60

Separation Cut-Point (mm)

Figure 2.5 - Initial comparative test data.

The results shown in Figure 2.5 are representative of a multitude of random tests where
the variables, including feed rate, elutriation water rate, bed-level, and feed percent solids are not
necessarily equal. In an effort to fairly compare this data, points where these variables are
consistent for both separators are graphed below in Figures 2.6 and 2.7.
Figure 2.6 reveals that for tests where all variables are equal, the CrossFlow separator
repeatedly produced classification results higher in efficiency than that realized using a
conventional feed system. It is also interesting to note that the separation cut-point is generally
greater in the conventional separator tests, as seen in Figure 2.7. These results suggest that less
feed water is entering the separation chamber of the CrossFlow unit. In a conventional feed
system, the total volume of feed water is introduced directly into the separation chamber, adding
velocity to the rising current of elutriation water in the upper portion of the classifier. This
increase in velocity can increase the cut-point of the separation.

57

C o n v e n tio n al S ep arato r, E p

0.150

0.125

0.100

0.075

0.050

0.025
0.025 0.050 0.075 0.100 0.125 0.150

C ro s s F lo w , E p

C o n v en tio n a l S e p arato r d 50 (m m )

Figure 2.6 - Efficiency comparison of CrossFlow and conventional separators.

0.50

0.45

0.40

0.35

0.30

0.25
0.25

0.30

0.35

0.40

0.45

0.50

C ro ss F lo w , d50 (m m )

Figure 2.7 - Cut-point comparison of CrossFlow and conventional separators.

58

Only an incremental improvement in efficiency can be seen at the low feed rates utilized
in the initial laboratory tests. Further test work was completed in a north Florida phosphate
beneficiation plant where constant high rates of feed solids could be provided. For these tests, a
2 x 2 ft. (0.6 x 0.6 m) CrossFlow unit was constructed out of steel. Like the lab-scale unit, a PID
controller coupled with an air-actuated underflow valve and pressure sensor was used to control
bed level. In these investigations, the CrossFlow separator was compared to a Krebs Whirlsizer,
which had been previously installed at the plant. Both units were fed from a bank of dewatering
cyclones. Feed percent solids were highly variable, ranging from 20 to 60%.
Results for this test work can be seen in Figure 2.8.

In this figure, efficiency

(Imperfection) is shown as a function of solids feed rate. Much like the original laboratory tests,
the CrossFlow separator demonstrated the potential for increased efficiency when compared to
the Whirlsizer. More importantly, the data show that the CrossFlow is less affected by increases
in feed solid rates (especially in excess of 6.0 tph/ft2) than a more traditional water-based
classifier.

0.45
Imperfection

0.40
0.35
0.30
0.25
0.20
Whirlsizer
CrossFlow

0.15
0.10
0.0

2.0

4.0

6.0

8.0

10.0

Feed Rate (tph/ft )

Figure 2.8 - CrossFlow and Whirlsizer solids feed rate test comparison.

59

A second set of solid feed rate tests, comparing the CrossFlow separator to a
conventional hindered-bed separator, was completed at a Florida mineral sands plant. For this
work, a 4 x 16 inch (0.1 x 0.4 m) CrossFlow unit was fabricated. This unit was constructed with
a removable conventional feed pipe.

This modification allowed for operation as either a

conventional teeter-bed or CrossFlow separator. Typical heavy mineral concentrate, averaging

76% heavy mineral, was used for all comparative test work. It was the goal of this work to
maximize heavy mineral recovery when operating as a density separator. Operating conditions
were held constant for all tests with the feed rate and feed presentation being the only variables.
Figure 2.9 shows the results of this testing.

HM Recovery (%)

100
80
60
40

CrossFlow
Conventional

20
0
0

10
20
Feed Rate (tph/m2)

30

Figure 2.9 CrossFlow and conventional teeter-bed separator solids feed rate test results.

60

It can be seen (Figure 2.9) that the CrossFlow unit is able to maintain high levels of
heavy mineral recovery at significantly higher capacities than the identical unit with a more
conventional feed introduction system. Specifically, the CrossFlow unit was able to achieve a
heavy mineral recovery of 95% at a solids feed rate of 23 tph/m2 (1.94 tph/ft2) compared to 13
tph/m2 (1.09 tph/ft2) with the conventional system.

2.3.2 Tracer Studies

In an effort to explain the evident capacity and efficiency improvements of the
CrossFlow feed presentation system over traditional feed systems, tracer studies were conducted.
These studies investigated both the overflow and underflow streams for both traditional and
CrossFlow feed configurations. The laboratory-scale separator was utilized in this effort. For
comparative purposes, it was important to keep the operating conditions (i.e., solids feed rate,
feed percent solids, elutriation water rate and bed-level) consistent while conducting these tests.
The most important of these variables was volumetric feed rate. It was found that relatively
coarse, dry silica sand could be fed at a constant rate from a vibratory feeder.

This, in

conjunction with make-up water flow meters, facilitated a constant volumetric feed to the
separator.
Liquid residence times were calculated using the method advocated by Mankosa (1990).
In this method, a conductivity probe was used to measure the salinity (conductance) of the
overflow stream. Incremental samples of the stream were then taken with respect to time
(seconds) and assayed for concentration (% or ppt) of tracer. The data is normally corrected for
background or residual salinity, and then mathematically normalized with respect to the original
tracer concentration. Plotting time versus normalized concentration yields a response curve from

61

which the initial tracer concentration can be calculated by summing the area under this curve.
The mean residence time, m, is then calculated from Equation [2.15]:

m =

Ci t i t
Ci t

[2.15]

where Ci represents the concentration (salinity) at a time increment, ti.

In order to determine the non-liquid mean residence times for a particle from the feed
reporting to either the underflow or overflow stream of the separator, a modification had to be
made to this procedure. In determining the mean residence time of solids from the feed that
should report to the overflow, a particle had to be used, whose size and density assured its
appearance in the overflow launder. Likewise, a very dense particle tracer that would settle
through the teeter-bed was needed for determining the mean residence time of the feed solids
that should appear in the underflow launder.
To track solids reporting to the underflow of the separator, a dense, monosized, titanium
mineral sand was used as a tracer. This titanium mineral sand was dense enough to settle against
the upward current of elutriation water. A spike of titanium sand (approximately 100 grams) was
added to the feed stream of the separator. Samples were taken of the underflow stream with
respect to time and assayed for heavy mineral content. The calculated mean residence times for
particles reporting to the underflow stream are summarized in Table 2.1.
It can be seen in Table 2.1 that the mean residence time to the underflow for both the
CrossFlow separator and a conventionally fed hindered-bed separator are nearly identical. Two
tests were conducted for each configuration. There was some disparity between the first and
second tests, which can most likely be contributed to the size of the separation chamber in
conjunction with the rapidity with which the automatic control valve reacted. The PID loop

62

controller was extremely sensitive to variations in the feed rate, which was magnified by the
small size of the separator. Consequently, as the underflow control valve responded, the internal
flow regimes were altered, varying slightly differently for each test. Nevertheless, the average
values of the mean residence times were relatively consistent. This finding suggests that the
improved efficiency of the CrossFlow separator (i) cannot be attributed to an extended residence
time in the teeter-bed and (ii) may be due to differences in how particles overflow the unit.

Table 2.1 - Mean underflow residence time for CrossFlow and conventional combinations.
Test
No.

Mean Residence Time (sec)

CrossFlow
Conventional

1
2

33.05
29.66

31.23
34.48

Average

31.36

32.86

In order to investigate the behavior of the overflow stream, it was necessary to determine
the retention time of both the water and the solids associated with the feed. This was necessary
since it has been argued that the CrossFlow feed presentation system allows for the rapid
removal of excess feed water from the separator without the entrainment of solids. A salt tracer
(NaCl) was used to track the liquid accompanying the feed and to determine its residence time.
The feed solids reporting to the overflow were tracked using a limestone tracer. The appearance
of the limestone tracer in the overflow was assured due to its small size and lower density. In
each test, the limestone and salt tracers were added at the same time. Similar to the process in
the underflow tracer studies, incremental samples of the overflow stream were then taken with
respect to time (seconds) and assayed for concentration (%).

63

The mean residence times of the liquid and solids from the feed that report to the
overflow launder are reported in Table 2.2. The solid samples were screened at 50 mesh to
provide a fine and coarse fraction. The data suggest that the CrossFlow system reduces the mean
residence time of the feed water by nearly half. Upon closer examination, it can also be seen that
the solids reporting to the overflow of the CrossFlow separator arrive faster than those in a
conventional hindered-bed separator.

Table 2.2 - Mean overflow residence time for CrossFlow and conventional combinations.
Fraction
Water (Salt Solution)
Solids (-50 mesh)
Solids (+50 mesh)

Mean Residence Time (sec)

CrossFlow
Conventional
6.08
14.61
11.07

11.31
15.94
17.19

Plotting of the residence time curves for each of the separator configurations shows that
each of the units acts extremely different. These residence time curves for the conventional and
CrossFlow separators can be seen in Figures 2.10 and 2.11, respectively. According to Figure
2.10, feed water takes several seconds to appear in the overflow launder in a conventional
configuration. The water is followed after a short time by the finest material (-50 mesh), and
then the coarsest material (+50 mesh).

There appears to be a distinct delay between the

emergence of each tracer in the overflow stream. This suggests that in a conventional hinderedbed design, a separation is occurring to overflow material prior to its appearance in the launder.
According to the data given in Figure 2.11, there appears to be no separation between the
coarse and fine material reporting to the overflow in the CrossFlow separator.

As in a

conventional feed system, the CrossFlow separator allows for the quick removal of liquid

64

associated with the feed stream. In contrast to the conventional system, the fine and coarse
material reporting to the overflow exits the system at nearly the identical time and with like
mixing, as indicated by the similar curves (Levenspiel, 1962).

This lack of separation is

providing the CrossFlow separator with an increased rate of rejection of material that should
report to the overflow. This essentially increases the apparent size (volume) of the device
available for separating a greater number and tighter size range of particles.
In a conventional hindered-bed unit, separator volume is being inefficiently utilized for
the partitioning of material that should report to the overflow. Consequently, a wider size
distribution of particles is being treated, causing an increase in particle interference and
interstitial velocities. Conversely, in the CrossFlow, a greater amount of separator volume is
being utilized for treating a greater number of particles closer to the cut-point of the separation.
The increased amount of closely sized material in the separation chamber should decrease
particle interference and the range of interstitial velocities encountered by any particle within the
system. Essentially, the system becomes more homogeneous and less turbulent.

65

Normalized Tracer Concentration

(ppt)

12

Liquid
Solids (-50 mesh)
Solids (+50 mesh)

10
8
6
4
2
0
0

10

20

30

40

50

60

Residence Time (sec)

Figure 2.10 - Overflow residence time data for conventional feed system.

Normalized Tracer Concentration

(ppt)

12

Liquid
Solids (-50 mesh)
Solids (+50 mesh)

10
8
6
4
2
0
0

10

20

30

40

50

60

Residence Time (sec)

Figure 2.11 - Overflow residence time data for the CrossFlow feed system.

66

2.4

Population Balance Model

2.4.1 Model Description

A mathematical population balance model has been developed at Virginia Tech to help
understand this new tangentially fed hindered-bed separator.

This model utilizes general

equations for hindered settling in transitional flow regimes to accurately predict overflow and
underflow partitions, particle size distributions, and component recovery data. Input data include
feed rate, percent feed solids (by mass), feed size distribution (up to 9 size fractions), density of
components in the feed stream (up to 2 components), fluidization water rate, and underflow
discharge rate.
The CrossFlow model was principally constructed as a series of well-mixed zones. These
zones represent three distinct sections that have dissimilar mixing patterns and flow regimes.
Therefore, each section must be modeled accordingly. The three primary sections include the
feed inlet, teeter-bed, and underflow areas. Figure 2.12 depicts these primary sections and flows
for the CrossFlow separator.
The model was constructed using the Microsoft Excel spreadsheet, which is a powerful
engineering tool capable of performing iterative calculations (including compound iterations).
Advantages of using Excel include instant graphing of results, and more importantly, ease of
troubleshooting. Results of tens of thousands of calculations are readily seen in an array of
spreadsheet cells where mistakes and erroneous coding errors are easily seen and corrected.

67

Overflow, Qo

Underflow Section

Teeter-Bed Section

Feed Section

Feed, Qf

Teeter
Water, Qw

Underflow, Qu

Figure 2.12 - Schematic depicting the primary divisions and flows for the CrossFlow separator.

68

2.4.1.1 Feed Section

Figure 2.13 shows the zone arrangement for the feed inlet area. It was necessary to
divide the feed section into a series of 25 (5 x 5 configuration) zones to account for the flows
induced by the cross-flowing action of the feed water, the fluidization water, and to a larger
degree, the settling action of the feed solids. The settling solids are shown as dashed lines in
Figure 2.13.

Feed, Qf

Feed Section

Overflow, Qo

Transition Zone

Qx(n+1)
Qd

UpA
Ql(n+1)

(n+1)

Qln

Zone Xi,j
Qdn

UpA
Qx

Figure 2.13 - Arrangement of zones for the feed section.

69

An experiment with dye was conducted to view the flow patterns in the feed section of
the separator.

The Plexiglas lab-scale unit, previously described, was utilized in these

experiments. A photograph of one test run can be seen in Figure 2.14. From this experiment, it
was determined that the feed water predominantly travels in two directions, either across the top
of the classifier, or it is drawn directly down into the separator at the feed inlet point via drag.
The drag created by the settling of the feed solids is responsible for the downward flow. The
influx of solids and the associated liquid hinder the fluidization water from entering into the first
five vertical zones of the feed section. This downward flow (Qd) of liquid induced by the settling
solids was determined to be proportional to the total volumetric concentration of settling
particles within that zone as seen in Equation [2.16]. Test work to date indicates that this
proportionality constant (X) has an approximate value of 12-15.

Qd(zone) = X (UpA)zone
where Up
A

[2.16]

= hindered settling velocity

= area through which solids settle.

The upward flow of fluidization water that enters each zone is shown as Qxn. This flow
is counter-acted by both the flow induced by solids settling (Qd) and by the horizontal flows (Ql)
that can move to or from adjacent cells. Material suspended within the teeter-bed acts like a
distributor for the rising teeter water, evenly distributing Qxn over the entire cross-section of the
unit for each level of the feed inlet area. The horizontal flows can be calculated by conducting a
flow balance for each zone within the feed section, given the elutriation water rate (Qw), feed rate
(Qf), and the underflow discharge rate (Qu).

70

Overflow

Feed Well

Dye

Separation
Chamber

Figure 2.14 - Photo of dye experiment showing feed section flow regime.

2.4.1.2 Teeter-Bed and Underflow Sections

A schematic of the teeter-bed and underflow sections for the CrossFlow are shown in
Figure 2.15. The teeter-bed section was modeled as a plug-flow device containing five zones. It
is separated from the single underflow zone by a transition zone, where the fluidization water
flow (Qw) is added. It can be seen that the fluidization flow makes a split in this transition zone,
with the majority of the flow rising up through the teeter-bed (Qo). The remainder of the teeter
water enters the dewatering cone and ultimately exits the system as Qu. The underflow section
was modeled as a single perfectly mixed zone. This is appropriate given that material enters by
transport or settling from the above transition zone and can only exit due to mass-action through
the underflow valve.

71

UpA

Qz

Zone Xn

Teeter-Bed Section

UpA

Qz

Zone X(n+1)
UpA

Qz

Zone X(n+2)
UpA

Qz

Zone X(n+3)
UpA

Qz

Zone X(n+4)
Qz

UpA
Transition Zone X(n+5)

Underflow Section

UpA

Qw

Q(u+1)

Zone X (n+6)

Qu

Figure 2.15 - Arrangement of zones for the teeter-bed and underflow sections.

2.4.2 Calculations
An iterative dynamic technique (i.e., finite differencing) was used to solve for the change
in concentration of particles over time for each zone in the CrossFlow separator. The model was
configured to allow for an array of particle sizes and densities. The volumetric flow of settling
solids (UpA) from zone to zone is shown as dotted lines in Figures 2.13 and 2.15. Using the laws
of mass conservation (steady-state flow), the change in concentration for each of these cells was
mass-balanced with respect to one another. To accomplish this, all volumetric flows were
balanced and then multiplied by the total volumetric concentration of solids () in the cell from
which the flow originated. An iterative solution was necessary because the concentration in all

72

cells constantly changes until steady-state conditions are achieved. When steady-state conditions
are achieved, the change in concentration of solids for all zones approaches zero.
The hindered settling velocity of particles within the CrossFlow was determined from
Equation [2.17] advocated by Masliyah (1979):

Up =

gdi2 F(f )(i susp)

18f 1 + 0.15Res0.687

[2.17]

where di is the particle size, i is the particle density, susp is the density of the suspension, and f
is the apparent viscosity of the fluid. As stated previously, F(f) is a function defined by
Richardson and Zaki (1954) as (1-) where is the volumetric concentration of solids. Beta
() is dependent on Reynolds number (Re) in the zone as calculated below in Equations [2.18]
and [2.19].

For Res < 1

= 4.36 Re 0.03

[2.18]

For Res 1

= 4.4 / Re0.1

[2.19]

Substituting F(f) into Equation [2.17] yields the overall hindered settling equation seen in
Equation [2.20].

Up =

gdi2(1 ) i susp
18f 1 + 0.15 Res0.687

[2.20]

Reynolds number was calculated from Equation [2.21] and is dependent on the hindered
settling velocity, apparent viscosity of the suspension, and concentration of particles within each

73

zone. The apparent viscosity of the suspension is also dependent on the total concentration of
particles in each zone as seen in Equation [2.22] developed by Swanson (1989). An iterative
process is needed when calculating Equations [2.20], [2.21], and [2.22] due to their
interdependencies. This iterative process must be completed for each size and density class in
each zone for every time step that occurs during the overall governing mass balance steady-state
iteration. A flowchart illustrating the procedure necessary to complete the mathematical model
is shown in Figure 2.16, while the required mass-balance equations are available in Appendix A.

Re s =

di f U p (1 )

susp = water

susp
2 max +
2( max )

74

[2.21]

[2.22]

Start

Input
Operational
and Density
Data

Calculate
Apparent
Viscosity in all
Zones

Input
Separator
Geometry

Increase time
by t:
t = t+t

Calculate Up for
all Size/Density
Classes for all
zones.

Input
Feed Size and
Component
Distribution Data

Reset time &

Zone
Concentrations
to Zero

Calculate
Overall Volumetric
Flow of Overflow &
in Teeter Zone

Calculate Beta
Factor for All Zones
and Size/Density
Classes

Are UP & Re
Self-Consistent for
All Zones?

Calculate Reynolds
Number for all Zones
for each Size/Density
Class.

Yes

No - Iterate

Balance all Zones with

Respect to Concentration
of Solids in/out of each
Zone for all Size/Density
Classes

Determine Lateral
Flows from Flow
Balance of Feed
Zone

Determine QD
Values for Feed
Zones

Calculate the New

Concentration of Solids
for each Zone
Is the New
Concentration
of Solids for each Zone
and every Size &
Density Class Equal
to the Old
Concentration?

Calculate Partition
Data

No

Yes

Calculate Efficiency
Data

End

Calculate Product
Recovery & Particle
Size Distribution Data

Figure 2.16 Flowchart illustrating procedure needed to complete the population balance model.

75

2.4.3 Model Validation

Over sixty laboratory tests were conducted to validate the population model. A minus 1.5
mm aggregate limestone was used as the testing material. The feed size distribution is provided
in Table 2.3. The validation tests were performed over a wide range of operating conditions.
Feed rates generally ranged from 0.5 tph/ft2 (5.93 tph/m2) to over 6 tph/ft2 (71.2 tph/m2). Feed
percent solids (by mass) ranged from 35 to 65 percent. Various cut-points were obtained by
either raising or lowering bed pressure and/or fluidization water rate. Cut-points ranged from a
low of 0.257 mm to a high of 0.700 mm.

Table 2.3 - Particle size distribution of limestone used in laboratory validation test work.
Particle Size
(Tyler Mesh)

Mass
(%)

+14
14 x 20
20 x 28
28 x 35
35 x 48
48 x 65
65 x 100
-100

33.76
17.85
13.28
10.34
7.92
5.75
3.50
7.61

It is obvious that settled material would pack closest in the dewatering cone of the
CrossFlow separator where there is no elutriation. It is appropriate that as the cut-point (d50) of
the separation changes, so does the size distribution of the underflow stream, and hence the
maximum possible concentration of particles at the underflow (max).

Fine material will

generally fill voids that occur between coarser material, but as more fine material reports to the

76

overflow of a hydraulic classifier, these voids will remain proportionally empty.

Yu and

Standish (1993) discuss that both the fractional solid volumes and the particle size distribution
affect the maximum packing density. In their work, it is stated that mathematical models are
only recently relating particle size distribution to packing density; however, these linear models
have been used to accurately predict the packing density of solids provided a simple continuous
or discrete size distribution is available.
The maximum packing of solids was determined semi-empirically. Several laboratory
tests were conducted and subsequently simulated using the CrossFlow model. The max term was
varied until the simulated cut-point results were consistent with the laboratory cut-point results.
The d50/max relationship was then graphed as shown in Figure 2.17.
In general, a linear correlation was found to exist between the maximum volumetric
concentration of solids (max) and the target cut-point (d50). A linear fit to this data yielded an R2
value of 0.87. The three outlying data points that do not fit this data well, occurred at extremely
high feed rates of over 7 tph/ft2 (83.1 tph/m2). It is believed that, at this feed rate, the separator is
approaching its capacity limit and the necessary increase in fluidization water causes the entire
teeter-bed to act as a fluid, thereby causing deviation from the linear relationship.
Due to the apparent linearity, the max and d50 relationship can be determined by
conducting as few as two laboratory control tests. One test must provide a coarse cut-point,
while the other a fine cut-point. Once this relationship is known, it can be incorporated into the
CrossFlow model.

77

1.00
2

R = 0.8724

0.90

(max)

0.80
0.70
0.60
CrossFlow Data
Outliers

0.50
0.40
0.0

0.2

0.4

0.6

0.8

1.0

Cut Size (mm)

Figure 2.17 - Correlation between cut-point and maximum concentration of solids.

As stated previously, the term in Equations [2.20] and [2.21] represents the volumetric
concentration of solids. As approaches unity, the term (1-) approaches zero, rendering the
hindered-settling velocity (Up) zero. However, in the case of particle classification, material can
never achieve 100% solids due to particle packing constraints. The hindered-settling velocity
must become zero as the volumetric concentration of particles approaches the maximum
concentration of particles allowed. This maximum concentration cannot be exceeded in any
zone. In view of the fact that max is changing linearly with the separation cut-point, Equations
[2.20] and [2.21] were transformed to:

Up =

gdi2( max ) i susp

18f 1 + 0.15 Re0s.687

78

[2.23]

Re s =

d i f U p (max )

susp

[2.24]

where max is a function of cut-size (d50) and ultimately the resulting size distribution of the
underflow stream.

2.4.4 Model Accuracy

Utilizing the d50/max relationship, the simulated results provided by the mathematical
model were compared to those achieved using the laboratory-scale CrossFlow separator. The
mathematical model was able to accurately predict cut-point, efficiency and the partition curves
of the separations. The separation efficiency was defined as either Ecart Probable (Ep) or
Imperfection as discussed previously.
Figure 2.18 shows the relationship between the actual and predicted cut-points of several
tests conducted in the laboratory.

The closer that the test points are to the indicated line, the

better the correlation. As seen in this figure, there is evidence of a good correlation between the
actual separation cut-points and those calculated via the mathematical model.

79

Model Cut-Point (mm)

0.7

0.6

0.5

0.4

0.3
0.3

0.4

0.5

0.6

0.7

Actual Cut-Point (mm)

Figure 2.18 - Comparison of predicted and actual cut-points.

Figure 2.19 illustrates the accuracy of the model for predicting efficiency (i.e., Ep). The
mathematical model was able to accurately predict Imperfection and Ep, as well as the overall
partition curve of the simulated separation. Figure 2.20 shows a common result when comparing
the actual and predicted partition curves for a typical laboratory validation test. In addition to
cut-point and efficiency, good correlations were found between other key actual and predicted
results (i.e., overflow and underflow yields and percent solids), as seen in Table 2.4.

80

0.150

Model Ep

0.125

0.100

0.075

0.050
0.050

0.075 0.100

0.125 0.150

Actual Ep
Figure 2.19 - Actual versus predicted efficiency.

Table 2.4 - Comparison of actual and predicted results.

Result

Actual

Predicted

Cut-Point (mm)
Ep
Imperfection
Underflow Yield (%)
Overflow Yield (%)
Underflow Solids (%)
Overflow Solids (%)

0.279
0.035
0.128
80.31
19.69
55.71
3.88

0.279
0.050
0.178
77.87
22.13
50.65
4.52

81

Percent to Underflow (%)

100

75

50

25
Model
Test Data
0
0.10

1.00

10.00

Mean Size (mm)

Figure 2.20 - Model predicted curve and test data.

It must be noted that efficiency is much harder to predict (i.e., in comparison to d50) since
there is greater error associated with sampling and size analysis when investigating either the d25
or d75 point of a separation. This is due to the fact that Ep and Imperfection rely on cumulative
mass values for points in the overflow and underflow distributions where these values are small
(Heiskanen, 1993). Discretization errors associated with the number of cells used for simulating
the feed section of the separator may also cause the slight inconsistency and scatter found in
Figure 2.19, which shows the prediction versus actual efficiency values for the CrossFlow
separator. Nonetheless, the predicted efficiencies are generally slightly lower than the actual
efficiencies, producing a conservative result.

82

2.4.5 Model Investigations

The impetus of the CrossFlow design was to minimize the detrimental effects of
operating at low feed percent solids (by mass) and excessive overall feed rates. Therefore,
investigations were completed to determine the effects of these variables on the performance
(i.e., cut-point and efficiency) of the CrossFlow separator. An example of the input and output
for the population balance model are available in Appendix B.
A series of feed rate simulations was conducted while maintaining a constant cut-point.
The fluidization water rate and feed percent solids were held constant.

The underflow

volumetric flow rate was manipulated to produce the desired cut-points. Two cut-points were
investigated in these feed rate tests (0.5 mm and 0.35 mm). Figures 2.21 and 2.22 illustrate the
effect of feed rate on the efficiency of the CrossFlow separator in terms of Ep and Imperfection.

0.22
0.5 mm Cut-Point
0.35 mm Cut-Point

0.18

Ep

0.14
0.10
0.06
0.02
0

Feed Rate (tph/ft2)

Figure 2.21 - Effect of solids feed rate on Ep.

83

0.30

Imperfection

0.25

0.20

0.15

0.5 mm Cut-Point
0.35 mm Cut-Point

0.10
0

Feed Rate (tph/ft2)

Figure 2.22 - Effect of solids feed rate on Imperfection.

Efficiency (Ep) is better for the finer separation; however, the imperfection for the
coarser separation tends to be slightly superior. Nonetheless, the average imperfection of both
the coarse and fine separations increases from 0.200 to 0.280 as feed rate increases from 23.76
tph/m2 (2 tph/ft2) to 71.2 tph/m2 (6 tph/ft2). Unlike other hydraulic classifiers, the CrossFlow
separator is capable of high throughput capacities at acceptable efficiencies. Heiskanen (1993)
states that the solids capacities for hydraulic classifiers are only typically in the range of 10
tph/m2 to 40 tph/m2 for fine and coarse separations, respectively.
The effect of feed percent solids (by mass) on Ep and cut-point was simulated at two
different feed rates, 23.76 tph/m2 (2.0 tph/ft2) and 43.82 tph/m2 (3.7 tph/ft2). To complete these
tests, the solids feed rate and fluidization water rate were all held constant. It was also necessary
to hold the underflow volumetric flow rate constant as the feed percent solids were varied from a
low of 20% to a high of 80%.

84

Figures 2.23 and 2.24 illustrate the effect of feed percent solids on both cut-point and
efficiency, Ep. In these simulations, the feed solids content was varied at two different mass
feed rates. While operating at 3.7 tph/ft2 (43.82 tph/m2), the cut-size of the separation decreased
from 0.656 to 0.537 mm as the feed percent solids increased from 20 to 70% by mass (Figure
2.23). Although there was a change in cut-size, it must be noted that the greatest change in cutpoint occurred between 20 and 35% solids. Once the feed reached approximately 35% solids,
the separation cut-point was virtually constant and unaffected by changes in feed percent solids.
At 2.0 tph/ft2 (23.76 tph/m2), the separation cut-point remained virtually constant (0.460 mm)
even as feed percent solids approached 30% by mass.

0.70
3.7 tph/ft2
2.0 tph/ft2

0.65
0.60
0.55
0.50
0.45
0.40

0 10 20 30 40 50 60 70 80 90

Feed Solids (% by Mass)

Figure 2.23 - Effect of feed percent solids on cut-point.

85

The effect of feed percent solids on efficiency followed the same basic trends (Figure
2.24). At the higher feed rate, the efficiency of the separation remained constant (Ep0.120)
until the feed percent solids approached 35% by mass. At the lower feed rate, the efficiency of
the separation (Ep0.100) remained consistent even as the feed percent solids approached 25%
by mass.

0.30
3.7 tph/ft2
2.0 tph/ft2

0.26

Ep

0.22
0.18
0.14
0.10
0.06
0.02
0 10 20 30 40 50 60 70 80 90

Feed Solids (% by Mass)

Figure 2.24 - Effect of feed percent solids on efficiency.

Lower feed percent solids generally will cause an increase in the flow velocity of the
upper portion of a conventional hydraulic classifier. This increase causes some coarse material
to incorrectly report to the overflow and leads to a decrease in efficiency (Heiskanen, 1993).
However, the CrossFlow appears to lose very little coarse material to the overflow.

86

The

tangential feed system allows the excess feed water to flow across the top of the separation
chamber, causing only a minimal amount of vertical flow disturbance.

87

2.5

Conclusions
1. Comparative studies completed in the laboratory and in-plant suggested that the
CrossFlow feed presentation system offers several advantages over traditional hinderedbed separator feed systems. These advantages include increased capacity and separation
efficiency.

2. Solid and liquid tracer studies suggest that the unique feeding system used by the
CrossFlow is capable of rapidly discharging excess feed water and fines that should
report to the overflow. Comparative test work indicates that conventional teeter-bed
separators are less efficient in segregating this overflow material prior to discharge.

3. A mathematical population balance model was developed to simulate the CrossFlow

separator. Validation tests show a good correlation between laboratory results and model
simulations. Consistent results were found for separation cut-point (d50), Ecart Probable
(Ep), and Imperfection.

4. A correlation between the target cut-point (d50) and the maximum concentration by
volume of solids (max) was confirmed. This linear relationship appears to vary with
material, feed size distribution, and ultimately the cut-point of the separation.

5. The mathematical model has shown that the CrossFlow separator can maintain an
acceptable and less varied efficiency over a number of different operating conditions,
including low feed percent solids (approaching 25% by mass) and feed solids rates in
excess of 6 tph/ft2 (71.2 tph/m2).

88

2.6

References

Alexander, K.S., Dollimore, D., Tata, S.S. and Uppala, V., 1991, A Comparison of the
Coefficients in the Richardson and Zakis and Steinours Equations Relating to the Behavior
of Concentrated Suspensions, Vol. 26, No. 6, pp. 819-829.
Austin, L.G., Lee, C.H., Concha, F. and Luckie, P.T., 1992, Hindered Settling and
Classification Partition Curves, Minerals and Metallurgical Processing, Vol. 9, No. 4, p.
161-168.
Barnea, E. and Mizrahi, J., 1973, A Generalized Approach to the Fluid Dynamics of Particulate
Systems. Part 1, Chemical Engineering Journal, Vol. 5, pp. 171-189.
Bethell, P.J., 1988, Current and Future Processing Flowsheets, Industrial Practices of Fine
Coal Processes, Chapter 30, pp. 317-329.
Brauer, H. and Thiele, H., 1973, Bewegung von Partikel-schwarmen, Chem.-Ing.-Tech., Vol.
45, No. 13, pp. 909-912.
Cho, H. and Klima, M.S., 1994, Application of a Batch Hindered-Settling Model to DenseMedium Separations, Coal Preparation, Vol. 14, pp. 167-184.
Concha, F. and Almendra, E.R., 1979, Settling Velocities of Particulate Systems 2. Settling
of Suspensions of Spherical Particles, International Journal of Minerals Processing, Vol. 6,
pp. 31-41.
Furnas, C.C., 1931, Grading Aggregates: I, Mathematical Relations for Beds of Broken Solids
of Maximum Density, Ind. Eng. Chem., Vol. 23, No. 9, pp. 1052-1058.
Davis, R.H. and Gecol, H., 1994, Hindered-Settling Function with No Empirical Parameters for
Polydisperse Suspensions, AIChE Journal, Vol. 40, No. 3, pp. 570-575.
Galvin, K.P., Pratten S.J. and Nicol, S.K., 1999, Dense Medium Separation Using a TeeteredBed Separator, Minerals Engineering, Vol. 12, No., pp.1059-1087.

89

Heiskanen, K., 1993, Particle Classification, Powder Technology Series, Chapman and Hall,
London, England.
Heiskanen, K. and Laapas, H., 1979, On the Effects of Fluid Rheology and Flow Properties in
the Wet Gravitational Classification, Proceedings, 13th International Mineral Processing
Congress, Elsevier, Amsterdam, pp. 417-446.
Hyde, D.A., Wlliams, K.P., Morris, A.N. and Yexley, P.M., 1988, The Beneficiation of Fine
Coal Using the Hydrosizer, Mine and Quarry, Vol. 17, No. 3, pp. 50-54.
Honaker, R.Q., 1996, Hindered Bed Classifiers for Fine Coal Cleaning, Proceedings, 13th
International Coal Preparation Conference, Lexington, Kentucky, pp. 59-70.
John, P.T. and Goyal, V.K., 1975, An Empirically Modeled Stokes Equation for Mean Particle
Size from Hindered Settling Experiment at Any One Concentration, Indian Journal of
Technology, Vol. 13, pp. 69-71.
Laapas, H.R., 1983, Measurement Estimation of the Rheological Parameters of Some Mineral
Slurries, DSc Thesis, Helsinki University of Technology.
Levenspiel, O., 1962, Chemical Reaction Engineering, Second Edition, John Wiley, New York.
Lewis, R.M., 1983, Hydrosizing of Industrial Minerals, Preprint No. 83-18, SME Annual
Meeting, Mar. 6-10.
Lewis, R.M., 1982, Lewis Hydrosizer Simple Solution to a Common Problem, Mining
Congress Journal, Vol. 68, No. 4, pp. 29-32, 35-36.
Lewis, T.B. and Nielsen, L.E., 1969, Transactions of the Society of Rheology, Vol. 12., p. 421.
Littler, A., 1987, Sand Processing, Product Optimization and Waste Treatment, Part 2
Classification, Mine and Quarry, Vol. 14, No. 8, pp. 25-31.
Littler, A., 1986, Automatic Hindered-Settling Classifier for Hydraulic Sizing and Mineral
Beneficiation, Transactions, Institute of Mining and Metallurgy, Vol. 95, pp. 133-138.

90

Low, G.S. and Bhattacharya, S.N., 1984, On the Study of Maximum Solids Concentration in
Suspension Rheology, The Twelfth Australian Chemical Engineering Conference,
Melbourne, Australia, August 26-29, Paper 23b, pp. 805-812.
Mackie, R.I., Tucker, P. and Wells, A., 1987, Mathematical Model of the Stokes
Transactions of the Institute of Mining and Metallurgical Engineers, Vol. 96, pp. 130-136.
Mankosa, M.J. and Carver, R.M., 1995, Processing of Chopped Wire Waste Material Using the
Floatex Density Separator, Third International Symposium on Recycling of Metals and
Engineered Materials, ed. P.B Queneau and R.D. Peterson, The Minerals, Metals, and
Materials Society, pp. 111-120.
Mankosa, M.J., Stanley, F.L. and Honaker, R.Q., 1995, Combining Hydraulic Classification
and Spiral Concentration for Improved Efficiency in Fine Coal Recovery Circuits, High
Efficiency Coal Preparation, ed. S. K. Kawatra, SME, Littleton, Colorado, pp. 99-107.
Mankosa, M.J., 1990, Scale-Up of Column Flotation, Dissertation, Virginia Polytechnic Institute
and State University, Blacksburg, VA.
Masliyah, J.H., 1979, Hindered Settling in a Multi-Species Particle System, Chemical
Engineering Science, Vol. 34, pp. 1166-1168.
McGeary, R.K., 1961, Mechanical Packing of Spherical Particles, Journal of the American
Ceramic Society, Vol. 44, No. 10, pp. 513-522.
McKnight, K., Stouffer, N., Domenico, J. and Mankosa, M.J., 1996, Recovery of Zircon and
other Economic Minerals from Wet Gravity Tailings using the Floatex Density Separator,
SME Annual Meeting, Phoenix, Arizona, March 11-14.
Mondal, K., 1997, A Dynamic Population Balance Model for a Hindered Bed Classifier for Fine
Particle Separations, Thesis, Southern Illinois University, Carbondale, Illinois.
Plitt, L.R., 1976, A Mathematical Model of the Hydrocyclone Classifier, CIM Bulletin,
December, pp. 114-122.
91

Reed, S., Roger, R., Honaker, R.Q. and Mankosa, M.J., 1995, In-Plant Testing of the Floatex
Density Separator for Fine Coal Cleaning, Proceedings, 12th International Coal Preparation
Conference, Lexington, Kentucky, pp. 163-174.
Rutgers, R., 1962, Relative Viscosity and Concentration, Rheological Acts, Vol. 2, No. 4, p.
304.
Richardson, J. and Zaki, W., 1954, Sedimentation and Fluidization: Part I, Transactions of the
Institute of Chemical Engineering, Vol. 39, pp. 35-53.
Robinson, J.V., 1957, Transactions of the Society of Rheology, Vol. 1, p. 15.
Schwalbach,

W.W.,

1965,

Uber

die

Vorgange

beo

der

Sedimentations-

und

Fliessbettklassierung, Aufbereit.-Tech., Vol. 6, No. 4, pp. 209-212.

Steinour, H.H., 1944, Rate of Sedimentation.

Nonflocculated suspensions of Uniform

Spheres, Ind. Eng. Chem., Vol. 36, pp. 618-624.

Sudduth, R.D., 1994, A Generalized Model to Predict the Viscosity of Solutions with
Suspended Particles. IV. Determination of Optimum Particle-by-Particle Volume Fractions,
Journal of Applied Polymer Science, Vol. 52, pp. 985-996.
Sudduth, R.D., 1993, A New Method to Predict the Maximum Packing Fraction and the
Viscosity of Solutions with a Size Distribution of Suspended Particles, Journal of Applied
Polymer Science, Vol. 48, pp. 37-55.
Swanson, V.F., 1999, Settling Equation for Minerals Process Application, Minerals and
Metallurgical Processing, Vol. 16, No. 3, pp. 8-13.
Swanson, V.F., 1989, Free and Hindered Settling, Minerals and Metallurgical Processing,
Nov., pp. 190-196.
Taggart, A.F., 1950, Handbook of Mineral Dressing, John Wiley, New York.

92

Westman A.E.R. and Hugill, H.R., 1930, Packing of Particles, Journal of the American
Ceramic Society, Vol. 13, No. 10, pp. 767-779.
Wills, B.A., Mineral Processing Technology, 5th Edition, 1992, pp.407-450.
Yianatos, J.B., Finch, J.A. and Laplante, A.R., 1986, Apparent Hindered-Settling in a GasLiquid-Solid Countercurrent Column, International Journal of Mineral Processing, Vol. 18,
pp. 155-165.
Yu, A.B. and Standish, N., 1993, A Study of the Packing of Particles with a Mixture Size
Distribution, Powder Technology, Vol. 76, pp. 113-124.
Yu, A.B. and Standish, N., 1991, Industrial Engineering and Chemical Resources, Vol. 30, p.
1372.
Yu, A.B. and Standish, N., 1988, An Analytical-Parametric Theory of the Random Packing of
Particles, Powder Technology, Vol. 55, p. 171.
Zimmels, Y., 1990, Hydrodynamic Diffusion of Particulates in Sedimentation Systems,
Powder Technology, Vol. 61, pp. 131-141.
Zimmels, Y., 1983, Hindered Motion of Particulate System in Newtonian Fluids. Acceleration
and Steady Regimes, Powder Technology, Vol. 34, pp. 191-202.

93

CHAPTER 3
Improving Coarse Particle Recovery in Hindered-Bed Separators

3.1

Introduction
Hindered-bed separators are commonly used in the minerals industry as gravity

concentration devices. These units can be employed for mineral concentration provided that the
particle size range and density differences are within acceptable limits.

However, these

separators often suffer from the misplacement of low density coarse particles to the high density
underflow. This shortcoming is due to the accumulation of coarse, low density particles that
gather at the top of the teeter bed. These particles are too light to penetrate the teeter bed, but are
too heavy to be carried by the rising water into the overflow launder. These particles are
ultimately forced downward by mass action to the discharge as more particles accumulate at the
top of the teeter bed. This inherent inefficiency can be partially corrected by increasing the teeter
water velocity to convey the coarse, low density solids to the overflow. Unfortunately, the
higher water rates will cause fine, high density solids to be misplaced to the overflow launder,
thereby reducing the separation efficiency.
To overcome the shortcomings of traditional hindered-bed separators, a novel device
known as the HydroFloat separator was developed based on flotation fundamentals. As shown
in Figure 3.1, the HydroFloat unit consists of a rectangular tank subdivided into an upper
separation chamber and a lower dewatering cone. The device operates much like a traditional
hindered-bed separator with the feed settling against an upward current of fluidization water.
The fluidization (teeter) water is supplied through a network of pipes that extend across the
bottom of the entire cross-sectional area of the separation chamber. However, in the case of the

94

HydroFloat separator, the teeter bed is continuously aerated by injecting compressed air and a
small amount of frothing agent into the fluidization water. The gas is dispersed into small air
bubbles by circulating the water through a high-shear mixer in a closed-loop configuration with a
centrifugal pump. The air bubbles become attached to the hydrophobic particles within the teeter
bed, thereby reducing their effective density. The particles may be naturally hydrophobic or
made hydrophobic through the addition of flotation collectors.

The lighter bubble-particle

aggregates rise to the top of the denser teeter bed and overflow the top of the separation chamber.
Unlike flotation, the bubble-particle aggregates do not need to have sufficient buoyancy to rise to
the top of the cell. Instead, the teetering effect of the hindered bed forces the low density
agglomerates to overflow into the product launder. Hydrophilic particles that do not attach to the
air bubbles continue to move down through the teeter bed and eventually settle into the
dewatering cone. These particles are discharged as a high solids stream (e.g., 75% solids)
through a control valve at the bottom of the separator. The valve is actuated in response to a
control signal provided by a pressure transducer mounted on the side of the separation chamber.
This configuration allows a constant effective density to be maintained within the teeter bed.
The HydroFloat separator can be theoretically applied to any system where differences in
apparent density can be created by the selective attachment of air bubbles. Although not a
requirement, the preferred mode of operation would be to make the low density component
hydrophobic so that the greatest difference in specific gravity would be achieved. Compared to
traditional froth flotation processes, the HydroFloat separator offers several important
advantages for treating coarser material, including enhanced bubble-particle contacting,
increased residence time, lower axial mixing/cell turbulence, and reduced air consumption.

95

Feed

Interf ace
Teeter Bed

Flo at
Pro du ct

Sep aratio n
Ch amb er

H igh D ensity

Circu lation
Lo o p

Low D ensity
Bubble s

Elu triation
N etw o rk
D ewater in g
Co ne

W ater
A dd ition

Flo at
Reject

Figure 3.1 - Schematic illustration of the HydroFloat separator.

96

3.2

Literature Review

3.2.1 General
The improved recovery of coarse particles has long been a goal within the minerals
processing industry. An array of studies has been conducted in an effort to overcome the
inefficiencies found in modern processes and equipment. These studies range in scope from
simple force investigation to the introduction of novel equipment. Advancements in chemistry
and conditioning have also been developed and employed at a number of installations.

3.2.1.1 Recovery by Flotation

Research on the relationship between particle size and floatability began as early as 1931
with work conducted by Gaudin, et al. showing that coarse and extremely fine material is more
difficult to treat when compared to intermediate sizes. Twenty years after this original work,
Morris (1952) arrived at the same conclusion, that particle size is one of the most important
factors in the recovery of ores by flotation. An illustration of this trend is seen in Figure 3.2.
Generally, recovery is low for the finest particles (dp<20m), and is at a maximum for
intermediate sized particles. A definite decrease in recovery occurs as the particle diameters
continue to increase in size. This reduction in recovery on the fine and coarse ends is indicative
of a reduction in the flotation rate of the particles (Jameson, 1977). It can be seen that the
efficiency of the froth flotation process deteriorates rapidly when operating in the extremely fine
or coarse particle size ranges. This is especially so when operating below 10m and above
200m. These findings might suggest that current conventional flotation practices are only
fundamentally optimal for the recovery of particles smaller than 65 mesh.

97

Figure 3.2 Relationship between particle size and recovery (Ahmed and Jameson, 1989).

According to Soto and Barbery (1991), conventional flotation cells operate with two
contradictory goals.

A conventional cell has to provide enough agitation to maintain any

particles in suspension, shear and disperse air bubbles, and promote bubble-particle collision.
However, for optimal recovery, it is also required that a quiescent system be maintained to
reduce detachment and misplacement of fine gangue to the overflow.

To maintain these

conditions when conducting coarse particle flotation is much more difficult considering that
increased agitation is needed to maintain these particles in suspension. Furthermore, coarse
particles are more likely to detach under these turbulent conditions. In order to compensate for
this lack of recovery, some installations are using relatively small flotation devices operated at
very low feed rates (Lawver, 1984). The increased high velocity, random agitation often found
in large (i.e., 8.5m3) flotation cells reduces selectivity and increases misplacement via carry-over.

98

Research conducted by Schulze (1977) determined that the flotation of a mineral is a

resultant of forces acting on a bubble and particle in a flotation system. These forces include
gravity, buoyancy, hydrostatic pressure, capillary compression, tension, and shear forces induced
by the system. According to Schulze, particles in diameter of several millimeters should float (in
the absence of turbulence) provided the contact angles are in excess of 50. Later work by
Schulze (1984) shows that turbulent conditions, similar to those found in mechanical flotation
cells, drastically reduce the upper size limit of floatable material. Several other investigations
support these findings (Bensley and Nicol, 1985; Soto, 1988). This research has demonstrated
that turbulent conditions reduced the maximum floatable size to one tenth of that found in nonturbulent conditions (Ives, 1984; Ahmed and Jameson, 1989).
The recovery of fine and intermediate sized particles is a product of two phenomena,
flotation and entrainment. However, coarse material is recovered solely by genuine flotation
(Trahar, 1981).

The low recovery of coarse particles can be attributed to several factors.

Robinson (1959) observed that when coarse and fine particles are combined in one system, the
result is a low surface coverage of collector on all particles in response to the magnitude of
surface area generated by fine material. Generally, a lower floatability is realized for the coarser
particles. Unlike fine material, coarse particles are not as capable of floating at low collector
dosages. Data also suggest that when a soft mineral is attritted, overall particle surface area is
substantially increased by the presence of slimes. This causes a considerable increase in reagent
consumption and a reduction of floatability in some ores (Soto and Iwasaki, 1986).
Another theory is that small particles have a higher rate of flotation and, therefore, crowd
out coarse particles from the air bubbles. Soto and Barbery (1991) disagree with this assessment,
speculating that the poor recovery of coarse material is strictly a result of detachment. They

99

further advocate the use of separate circuits for fine and coarse processing in an effort to
optimize the conditions necessary for increased recovery.
Several new devices have been produced and tested for the sole purpose of improving the
recovery of coarse particles. Harris, et al. (1992) tested a hybrid mechanical flotation column,
which is essentially a cross between a conventional cell and a column flotation cell. In this
device, a column is mounted above an impeller type agitator. The column component offers the
advantage of an upper quiescent section optimal for flotation, while the mechanical impeller
offers the ability of reattachment and increased collection of any non-attached coarse material in
the lower zone. When compared to a release analysis curve, this hybrid mechanical column outperformed a conventional flotation cell, but was equivalent to a traditional flotation column.
Improvements in coarse particle recovery have also been seen with the advent of nonmechanical flotation cells. Success has been observed when using column flotation (i.e., Flotair,
Microcel, and CPT cells), Lang launders, Skin flotation, and the negative-bias flotation column.
Column flotation offers several advantages that can be useful in any application. Barbery (1984)
advocates that columns have no mechanical parts, easy automation and control, low turbulence,
easy bubble size control, simple flow patterns, well-defined hydrodynamic conditions and high
throughput. These advantages translate to ease of maintenance, scale-up, modeling, and a
reduction of short-circuiting usually witnessed in conventional flotation.

3.2.1.2 Recovery by Gravity Concentration

Water-based gravity concentration devices are used extensively throughout the minerals
industry to concentrate high density particles from a mixture of high and low density material.
Although many devices have been developed over the years, a technique gaining in popularity is

100

hindered/fluidized-bed separators. These separators, traditionally used for classification, work

reasonably well for mineral concentration if the particle size range and density difference are
within acceptable limits (Bethel, 1988; Mankosa et al., 1995; Reed et al., 1995).
Separators, such as coal spirals and water-only cyclones, have been widely used in the
coal preparation industry to upgrade coal feeds in the intermediate particle size range (e.g., 2 x
0.15 mm). Particles of this size are generally too small to be handled in conventional dense
medium circuits and too coarse to be efficiently recovered by froth flotation circuits.
Unfortunately, water-based separators often provide lower separation efficiency when compared
to other plant circuits. For example, while water-only cyclones tend to misplace significant
amounts of larger, low-ash coal particles to the reject stream, spirals tend to misplace coarse,
high ash particles to the clean coal stream. Spiral circuits also generally suffer from high
specific gravity cut-points, however they also tend to maintain high combustible recoveries. As a
result, water-based separators are often used in multi-stage circuits in an attempt to deal with
misplaced coal or rock (as described in Chapter 1).
A great deal of research has been devoted to the study of fluidized-beds and their use in
gas/solid contacting and in liquid/solid applications. Studies describing the latter have typically
focused on the classification aspects of fluidized-bed separators and less so on mineral
concentration. Recent work has shown that fluidized-bed separators can be used to effectively
separate mineral assemblages that have components with different densities. For instance, coal
can be separated from ash forming components (Honaker, 1996), silica from iron ore, and silica
from various heavy minerals such as zircon and ilmenite (McKnight et al., 1996). Results from
these studies indicate that efficient concentration can be achieved if the particle size ratio (top
size to bottom size) is less than 3 or 4 to 1 and in a range from 200 mesh to several millimeters.

101

Unfortunately, this is seldom the case and, as a result, separation efficiency is poor. To correct
this shortcoming, the valuable component (i.e., coal, iron ore, ilmenite and zircon) frequently
must be reprocessed to achieve the desired quality.
As stated previously, a hindered-bed separator is a vessel in which water is evenly
introduced across the base of the separator and rises upward. The separator typically has an
aspect ratio of two or more and is equipped with a means of discharging solids through the
bottom of the unit. Rising water and solids flow over the top of the separator and are collected in
a launder. Solids are typically introduced in the upper portion of the vessel and begin to settle at
a rate defined by the particle size and density. The coarse, higher density particles settle against
the rising flow of water and build a bed of teetering solids. This bed of high density solids has
an apparent density much higher than the teetering fluid (i.e., water). Since particle settling
velocity is driven by the density difference between the solid and liquid phase, the settling
velocity of the particles is reduced by the increase in apparent density of the teetering bed. As a
result, the low density component of the feed resists penetrating the bed and remains in the upper
portion of the separator where it is transported to the overflow launder by the rising teeter water.
Hindered-bed separators are also well recognized as low turbulence devices. For this
reason, they are used extensively for particulate processing as either gas/solid or liquid/solid
contact devices (Heiskanen, 1993). The high solids concentration in the separator limits particle
mobility. As a result, particles move through the separation chamber in a plug flow manner.
Previous work has shown that this type of motion results in an increase in process recovery due
to reduced back-mixing (Doby and Finch, 1990). Furthermore, particle detachment is also
minimized due to a reduction in localized turbulence.

102

The concept of improving coarse particle recovery through the use of bubble-particle
attachment in a rising current separator (flotation column) has been previously demonstrated
(Laskowski, 1995; Barbery, 1989).

Unfortunately, these approaches used an open-column

reactor operating in the free, not hindered, settling regime. As a result, these configurations do
not have the advantages associated with a teeter-bed approach. The distinctive advantage of
utilizing a teeter-bed is the greatly improved hydrodynamic environment within the separator.
To recognize this advantage, the fundamental difference between free and hindered-settling
conditions must be examined.
Particle settling is generally recognized as falling into one of two categories: free or
hindered-settling. Under free settling conditions, individual particles do not affect the settling
behavior of adjacent particles and, as such, the pulp has the rheological characteristic of the fluid.
Furthermore, the settling velocity is determined by particle size and particle density. Hinderedsettling is fundamentally different. At high solids concentrations, adjacent particles collide with
each other influencing the settling characteristics.

The settling path is greatly obstructed

reducing particle velocity. Additionally, the high solids concentration increases the apparent
viscosity and specific gravity of the pulp, thus further reducing particle settling. As a result, the
acceleration of particles becomes more important than the terminal velocity. This collision
phenomenon is the most important aspect of hindered-settling and provides favorable
hydrodynamic conditions that cannot be achieved in open-tank reactors, such as conventional
column cells. Specifically, particle collection rate, retention time and cell turbulence are all
improved.

103

3.2.1.3 Phosphate Recovery

The United States is the worlds largest producer of phosphate rock, accounting for
approximately 45 million tons of marketable product valued at more than $1.1 billion annually
(United States Geological Survey, Mineral Commodity Summaries, January 1999).
Approximately 83% of this production can be attributed to mines located in Florida and North
Carolina. The major U.S. producers are located in Florida and include Cargill Fertilizer, Inc., CF
Industries, Inc., IMC-Agrico, Inc., Agrifos, LCC, and PCS Phosphate, Inc. Of these, IMCAgrico is by far the largest single producer in the state.
Prior to marketing, the run-of-mine phosphate matrix must be upgraded to separate the
valuable phosphate grains from other impurities.

The first stage of processing involves

desliming at 150 mesh to remove fine clays. Although 20-30% of the phosphate contained in the
matrix is present in the fine fraction, technologies currently do not exist that permit this material
to be recovered in a cost-effective manner. The oversize material from the desliming stage is
typically screened to recover a coarse (plus 14 mesh) high-grade pebble product. The remaining
14 x 150 mesh fraction is typically classified into coarse (e.g., 14 x 35 mesh) and fine (e.g., 35 x
150 mesh) fractions that are upgraded using conventional flotation machines, column flotation
cells, or other novel techniques such as belt flotation (Moudgil and Gupta, 1989). The fine
fraction (35 x 150 mesh) generally responds very well to upgrading and, in most cases,
conventional flotation technologies can be used to produce acceptable concentrate grades with
recoveries in excess of 90%. On the other hand, high recoveries are often difficult to maintain
for the coarser (14 x 35 mesh) fraction. In fact, prior work has shown that the recovery of coarse
particles (e.g., >30 mesh) can be less than 50% in many industrial operations (Davis and Hood,
1992). For example, Figure 3.3 illustrates the sharp reduction in recovery as particle size

104

increases from 0.1 mm (150 mesh) to 1 mm (16 mesh) for one northern Florida phosphate
operation. In many cases, attempts by plant operators to improve coarse particle recovery often
produce an undesirable side effect of diminishing flotation selectivity.

Plus 16 M
16x24 M
24x32 M
32x42 M
42x60 M
60x80 M
80x150 M
0

20

40

60

80

100

Recovery (%)
Figure 3.3 - Recovery versus size data (Mankosa et al., 1999).

These findings are consistent with historical data (Gaudin et al., 1931) from other
flotation applications, which show coarse particles are more difficult to recover using traditional
flotation machines. Current research indicates that coarser material is lost due to the unfavorable
hydrodynamic conditions within the turbulent pulp of agitated flotation machines and/or the
competition with the fines for the available bubble surface area. For this reason, a split-feed
circuit arrangement is recommended when treating a wide feed particle size distribution.
Furthermore, new and/or improved technologies that are more efficient in treating coarser feeds
need to be developed.

105

Existing column cells used in the phosphate industry also have performance limitations
due to mechanical design. In most cases, air is introduced using venturi-type aspirators that
require a great deal of water. The majority of this aeration water reports to the column overflow
product.

This aeration water carries undesired gangue material into the froth product.

Additionally, the column aeration rate is intrinsically dependent upon the aspirator water flow
rate. As a result, an increase in aeration rate requires an increase in water flow rate which, in
turn, can have a detrimental effect on performance. Based on these limitations, it is apparent that
a flotation system is required that incorporates quiescent hydrodynamic conditions and provides
for a de-coupling of the aeration system from external water supplies.
One well-known method of improving flotation performance is to classify the feed into
narrow size fractions and to float each size fraction separately.

This technique, which is

commonly referred to as split-feed flotation, has several potential advantages such as higher
throughput capacity, lower reagent requirements and improved separation efficiency. Split-feed
flotation has been successfully applied to a wide variety of flotation systems including coal,
phosphate, potash and industrial minerals (Soto and Barbery, 1991).
The United States Bureau of Mines (USBM) conducted one of the most comprehensive
studies of the coarse particle recovery problem in the phosphate industry (Davis and Hood,
1993). This investigation involved the sampling of seven Florida phosphate operations to
identify sources of phosphate losses that occur during beneficiation. According to this field
survey, approximately 50 million tons of flotation tailings are discarded each year in the
phosphate industry. Although the tailings contain only 4% of the matrix phosphate, more than
half of the phosphate in the tailings is concentrated in the plus 28 mesh fraction. In all seven
plants, the coarse fraction was higher in grade than overall feed to the flotation circuits. In some

106

cases, the grade of the plus 28 mesh fraction in the tailings approached 20% P2O5. The USBM
study indicated that the flotation recovery of the plus 35 mesh fraction averaged only 60% for the
seven sites included in the survey. Furthermore, the study concluded that of the seven phosphate
operations, none have been successful in efficiently recovering the coarse phosphate particles.
There have been several attempts to improve the poor recovery of coarse (16 x 35 mesh)
phosphate grains through the addition of improved flotation reagents. One such study, which
was funded by the Florida Institute of Phosphate Research (FIPR), was completed by the
University of Florida in early 1992 (FIPR Project 86-02-067). These investigators also noted that
the flotation of coarse phosphate is difficult and normally yields recoveries of only 60% or less
when using flotation. The goal of the FIPR study was to determine whether the recovery of
coarse phosphate could be enhanced via collector emulsification and froth modification achieved
by frothers and fines addition. Plant tests conducted as part of this project showed that the
appropriate selection of reagents could improve the recovery of coarse phosphate (16 x 35 mesh)
by up to 6 percentage points. Furthermore, plant tests conducted with emulsified collector
provided recovery gains as large as 10 percent in select cases. Unfortunately, reports of followup work by industry which support these findings have not yet been published.
In 1988, FIPR also provided financial support (FIPR Project 02-070-098) to the Canadian
Laval University to determine the mechanisms involved in coarse particle flotation and to
explain the low recoveries of such particles when treated by conventional froth flotation. In light
of this study, these investigators proposed the development of a modified low turbulence device
for the flotation of coarse phosphate particles. Laboratory tests indicated that this approach was
capable of achieving recoveries greater than 99% for coarse phosphate feeds. In addition, the
investigators noted that this approach did not suffer from high reagent costs associated with other

107

strategies designed to overcome the coarse particle recovery problem. Although the preliminary
data was extremely promising, this work was unfortunately never carried through to industrial
plant trials due to problems with the sparging system and tailing discharge system.
Building on these early findings, Soto and Barbery (1991) have recently developed a
negative bias flotation column that improves coarse particle recovery (Barbery, 1989). It was
surmised that the only factors preventing conventional columns from being ideally suited for
coarse particle recovery were wash water flow and a thick froth layer. Wash water is used in
column flotation to wash fine gangue (i.e., clays) from the product froth. However, wash
water can also propel coarse particles back into the pulp resulting in a loss of recovery. Soto and
Barbery (1991) removed this wash water resulting in a negative bias flow (i.e., net flow rising
upwards). An added flow of elutriation water aids in propelling coarse particles to the overflow
by inducing drag on any bubble-particle in the pulp. In fact, Barbery (1989) has been able to
demonstrate a four-fold improvement in coarse particle recovery when utilizing negative bias.
Essentially, this device is operated in a flooded manner and in the absence of a froth zone.
Several other similar devices have also been developed (i.e., Laskowski, 1995).
A number of alternative processes have been used by industry in an attempt to improve
the recovery of the coarser particles. These techniques include gravitational devices such as
tables, launders, spirals and belt conveyors that have been modified to perform skin-flotation
(Moudgil and Barnett, 1979). Although some of these units have been successfully used in
industry, they normally must be supplemented with scavenging flotation cells to maintain
acceptable levels of performance (Moudgil and Barnett, 1979; Lawver et al., 1984).
Furthermore, these units typically require excessive maintenance, have low throughput
capacities, and suffer from high operating costs. Reagent consumption can also be a major

108

drawback, as up to 10 lbs/ton of chemical can sometimes be needed to facilitate skin-flotation

(Keating, 1999). Despite these shortcomings, the increased recovery of coarse phosphate matrix
can offer several benefits.
One of the most obvious advantages of improved coarse particle recovery would be the
increased production of phosphate rock from reserves currently being mined. For example, a
survey of one Florida plant indicated that 7-15% of the plant feed was present in the plus 35
mesh fraction (Mankosa et al., 1999). At a 2,000 tph feed rate, this fraction represents 140-300
tph of flotation feed. An improvement in coarse particle recovery from 60% to 90% would
represent an additional 50-100 tph of phosphate concentrate. This tonnage corresponds to an
additional $7.5-15 million of corporate revenues. This incremental tonnage and income could be
produced without additional mining or reserve depletion.

3.2.1.4 Carbon/Coal Recovery

In 1993, the total world coal reserves were estimated at over 1,039,182 tons, of which
23% were estimated to be in the United States alone (Kawatra, 1995). Prior to sale, run-of-mine
coal is generally upgraded in order to remove ash bearing minerals and to increase the BTU
value of the clean coal product. The removal of sulfur bearing minerals has also become of
greater importance since the advent of the United States Clean Air Act, which restricts the
emission of sulfur dioxide from coal fired power plants.
In a general flowsheet, a coal feed is crushed to a top-size of a few inches and then
further classified into several size fractions. The largest of these size fractions, approximately 2
inch x 6 mesh, is predominantly treated in dense media processes. Dense medium bathes and
cyclones are the most popular and most efficient; they are capable of Ep (efficiency) values

109

approaching 0.02. The 6 x 65 mesh size fraction is generally processed in coal spirals or wateronly cyclones, while the passing 65 mesh size fraction is generally treated with flotation.
Coal spirals suffer from specific gravity cut-points that are typically much higher than
those employed by the coarse coal dense medium circuits. This imbalance creates either a loss
of clean coal or a decrease in product quality. Spirals are capable of minimizing the rejection of
these coarser, low-ash particles due to the buffering action of the flowing film on particle
classification. Water-only cyclones tend to misplace significant amounts of larger, low-ash coal
particles to the reject stream due to the size classification within the cyclone. Because of this
particle misplacement, these water-based separators tend to be much less efficient
(approximately 0.16 Ep) than dense medium devices. Further discussion on this topic is found in
Chapter 1, Sections 1.1 and 1.2.
Froth flotation is used almost exclusively for the upgrading of coal in the passing 65
mesh size range. However, the maximum floatable size of coal particles depends on several
variables, including coal rank, collector addition, pulp density, cell turbulence, and retention
time. In coarse particle flotation, a bubble will rise through the pulp and encounter a particle of
coal and/or gangue. If the particle is hydrophobic, and if it passes within a close enough range of
the bubble, the particle will adhere to the bubble. Once attached, the particle will be swept to the
rear of the bubble by its relative motion through the pulp. If the force of adhesion is strong
enough, the particle will remain attached to the bubble and reach the surface.
Collision efficiencies of bubble and particles should increase as the coal particle size
increases. These probabilities dictate that capture and attachment should be expected to increase
along with recovery. However, according to Jameson, et al. (1984), this does not hold true for
the coarsest material. As a bubble unites with a coarse particle, attachment occurs. The bubble

110

and particle become a bubble-particle aggregate that has a higher buoyancy effect than that of the
particle alone. Unfortunately, even after attachment, this bubble-particle aggregate may now
only have an effective buoyancy and/or density equal to that of the pulp, resulting in a loss of
combustible recovery. This effective buoyancy of the bubble-particle aggregate most likely sets
the upper limit on the maximum floatable size. Thus, the maximum floatable particle size for a
given material is anywhere between 10 and 100m (Jameson et al., 1984).
Studies conducted by Sun and Zimmerman (1950) found that bituminous coal was able to
float at slightly larger sizes than anthracite coal particles (6.7 mm vs. 1.17mm). However, the
specific gravity of the bituminous coal was less than that of the anthracite coal, which may have
contributed to this finding. Even though these coarse particles were buoyant enough to float,
they were incapable of passing over the overflow weir into the clean coal launder due to their
size.
In studies conducted by Crawford (1936), it was shown that fine particles are more likely
to float before coarser particles. In fact, subsequent studies conducted by Brown and Smith
(1954) and Rastogi and Aplan (1985) concluded that flotation rates increase with a decrease in
particle size. The slower flotation rate of coarse coal leads to a loss in recovery of these
generally high quality, low ash particles.
Subsequent investigations by Luttrell (2000) have demonstrated how feed rate can
influence the recovery of coarse coal particles. Plotted in Figure 3.4 is the maximum floatable
particle size as a function of feed rate. It can be concluded from this plot, that if effective bubble
surface area remains constant as feed rate increases, the competition for this surface area also
increases. As a result, bubble surface area is first covered by the finer particles, which have a
higher flotation rate in comparison to the coarser particles.

111

This phenomenon results in

unattached coarse particles. As a consequence, as feed rate increases, the maximum floatable
particle size decreases as seen in this figure.

Particle Size (m)

120

110

100

90

80
0

10

15

Feed Rate (ton/hr)

Figure 3.4 Effect of feed rate on maximum floatable particle size (Luttrell, 2000).

Other data presented by Phillips (1998) further illustrate how this phenomenon is realized
in industrial practice. Presented in Figure 3.5 are size by size recovery data for a multitude of
industrial flotation feed samples. It can be seen that at smaller sizes, combustible recovery is

112

high for most coals (approaching 90%). However, as particle size continues to increase, there is
a corresponding drop in combustible recovery. It can be concluded from this trend that coal
particle recovery declines after a particle reaches an average size of approximately 65 mesh as
indicated by the vertical line in this figure.

90
80
70
60
50
40
30
20

+1mm

0.5 x 1mm

.250 x .5

.150 x .250

.045 x .150

0 x .045

10
Cumulative

Combustible Recovery (%)

100

Particle Size (mm)

Figure 3.5 Size by size recovery data for multiple flotation feed samples (Phillips, 1998).

113

As previously stated, agitation of the pulp in the flotation cell increases the flotation rate
of the particles. It also has a serious effect on the particle size and rate constant relationship. A
process modeled by Woodburn, et al. (1971) shows that a particle could be physically wrenched
from a bubble to which it had adhered due to sudden acceleration. Morris (1952) also concluded
that the stresses induced by turbulence were a cause for reduced recovery in the coarsest particle
size fractions for flotation. With more agitation, attachment probability continues to increase
due to the increased momentum of the particles within the flotation cell, but as a consequence,
bubble size decreases, allowing detachment forces to play a larger role within the system. This is
especially detrimental for coarse coal flotation. This is not a problem for the finer particles
because their enhanced momentum will overshadow these worsening detachment forces (Ahmed
et al., 1989).

114

3.3

Theoretical Framework

3.3.1 Flotation Fundamentals

The reaction, or flotation, rate for a process is indicative of the speed at which the
separation will proceed.

In mineral flotation the reaction rate is controlled by several

probabilities, e.g., collision, adhesion and detachment. The attachment of particles to air bubbles
is the underlying principle upon which all flotation processes are based. This phenomenon takes
place via bubble-particle collision followed by the selective attachment of hydrophobic particles
to the bubble surface. Particles may also detach if the resultant bubble-particle aggregate is
thermodynamically unstable. According to Sutherland (1948), the attachment process may be
described by a series of mathematical probabilities given by:

P = Pc Pa (1 Pd )

[3.1]

in which Pc is the probability of collision, Pa the probability of adhesion, and Pd the probability
of detachment. The attachment and detachment probabilities are controlled by the process
surface chemistry and cell hydrodynamics, respectively. In an open (free settling) system, the
collision probability is quite low due to the low particle concentration. However, at higher
concentrations, the crowding effect within the hindered-bed increases the probability of collision.
This phenomenon is due to the compression of the fluid streamlines around the bubbles as they
rise through the teeter-bed. The increased probability of collision can result in reaction rates that
are several orders of magnitude higher than found in conventional flotation.
After a particle contacts a bubble, the particle is swept over the bubble surface for a finite
period of time known as the sliding time. During this period, the thin liquid film separating the
bubble and particle must rupture if particle adhesion is to occur. This sliding time is a

115

reflection of the hydrodynamics of the system and is primarily a function of the particle and
bubble sizes. On the contrary, the length of time required for the liquid film to thin sufficiently
so that rupture occurs is a measure of the chemistry of the flotation system and is commonly
referred to as the induction time. The induction time is small for hydrophobic particles (e.g., 1
msec) and may approach infinity for extremely hydrophilic particles.
Utilizing the induction time concept, Yoon and Luttrell (1989) derived an analytical
expression for the probability of bubble-particle adhesion (Pa) as:

B U b ti
Pa = sin 2 2 arctan exp

D b ( D p / D p + 1)

[3.2]

in which Dp is the particle diameter, Db is the bubble diameter, ti is the induction time, Ub is the
differential velocity between the bubble and particle, and B is a constant that varies depending on
the particular flow regime (as dictated by Reynolds number). In most cases, Ub is simply
assumed to be the terminal rise velocity of the bubble. Since Equation [3.2] is expressed as a
sine function, the calculated value of Pa will always fall between zero and unity, the correct
limits for probabilities.
To illustrate the effect that particle size has on the probability of adhesion, and hence
recovery, Pa was plotted as a function of particle size for different levels of induction time
(hydrophobicity) as seen in Figure 3.6. As expected, Pa increases sharply as the induction time
is reduced from 5 to 1 msec. It is illustrated that for a given value of ti, Pa decreases steadily as
the particle size increases. The reduced Pa value is due to the fact that larger particles tend to
slide more rapidly over the bubble surface since they project further out into the high velocity
region of the streamlines that pass over the bubble surface. However, it can be concluded that if
116

the differential velocity between the coarse particles and bubbles can be reduced, the probability
of attachment will increase resulting in a corresponding increase in recovery.

P ro b ab ility o f Ad h e sio n

1.0
Db = 1 m m
0.8

0.6
1 m s ec
0.4
2 m s ec

0.2
5 m s ec
0.0
0.0

0.5

1.0

1.5

2.0

P a rtic le D ia m eter (m m )

Figure 3.6 - Effect of particle size on the probability of bubble-particle adhesion for induction
time values of 1, 2 and 5 msec.

Studies conducted by Woodburn, et al. (1971) suggest:

Pd = D p / D p * x

[3.3]

in which Dp is the particle diameter to be floated, Dp* is the maximum floatable particle
diameter, and x is an experimental constant (typically 3/2). Factors that influence the magnitude
of Dp* include pulp chemistry (surface tension and contact angle), physical particle properties

117

(size, density, composition and shape), and cell agitation intensity. Theoretical Dp* values have
been calculated by Schulze (1984) from the tensile and shear stresses acting on bubble-particle
aggregates under homogenous turbulence. The degree of turbulence was quantified in terms of
the induced root mean square velocity (RMSV).
A study conducted by Schulze (1984) concluded that turbulence had a tremendous effect
on the recovery of coarse particles. A typical set of results obtained by Schulze is presented in
Figure 3.7. In this figure, the maximum floatable particle size is shown as a function of
turbulence (RMSV) and contact angle. According to this data, the maximum size of particles
that may be recovered by flotation increases by more than an order of magnitude when changing
from high to low turbulence. In fact, according to Barbery (1984), the optimum conditions for
coarse particle flotation occur when cell agitation intensity is reduced to a point just sufficient to
maintain the particles in suspension (i.e., teeter-bed conditions).

10

Particle Size (mm)

Low Turbulence
(Static Condition)

1
Medium Turbulence
(RMSV = 0.2 m/sec)

0.1

High Turbulence
(RMSV = 1.0 m/sec)
0.01
0

20

40

60

80

100

Contact Angle ()

Figure 3.7 - Influence of turbulence on the maximum particle size that may be recovered by froth
flotation (after Schulze, 1984).

118

The maximum floatable particle size is also effected by buoyancy. In froth flotation, the
bubble-particle aggregate must have sufficient buoyancy to be lifted to the surface of the pulp.
Mathematically, the maximum particle diameter (Dmax) that may be floated may be estimated
from:

f
D max = D b
p f

1/ 3

[3.4]

in which p and f are the densities of the particle and fluid, respectively. This expression
suggests that 1 mm diameter bubbles are capable of carrying particles up to approximately 0.85
mm before the critical buoyancy limit is exceeded (p = 2.5 gm/cm3).
Particle retention time can also greatly influence the recovery of coarse particles. The
mixers-in-series model provides a convenient framework for analyzing this phenomenon (Arbiter
and Harris, 1962; Bull, 1966). According to this model, the cumulative fractional recovery (R)
of a given particle species can be determined using the expression:

R = 1 1 + k p n

[3.5]

in which k is the flotation rate constant, p is the particle residence time and n is the number of
equivalent mixers. Figure 3.8 shows recovery determined from Equation [3.5] for different
values of n as a function of the dimensionless product kp. In most cases, n is assumed to be
equal to the number of cells in the flotation bank. This assumption is generally valid for a cellto-cell flotation bank.

However, the magnitude of n is typically smaller for flow-through

flotation banks that have a significant amount of intermixing. The appropriate value of n can be

119

readily estimated for any cell configuration using residence time distribution (RTD) data that
have been collected using solid or liquid tracers. Details related to this procedure have been
described elsewhere (Mankosa et al., 1992).

100
n=8
n=4

Recovery (% )

80

n=2
n=1

60

40

20

0
0

k p

Figure 3.8 - Relationship between flotation recovery and the dimensionless quantity kp for
different numbers of mixers-in-series.

In column flotation, particles settle vertically through the cell either with the fluid flow
(co-current) or opposite to it (counter-current). The particle residence time (p) can be estimated
from the liquid residence time (l) using the expression:

p = K l = K (V / Q s )

[3.6]

in which V is the volume of each cell (corrected for gas holdup), Qs is the volumetric flow rate of
slurry, and K is a correction factor to account for particle settling behavior. For well-mixed cells,

120

K is approximately equal to one and may be ignored. For vertical flow cells such as columns, K
may be estimated from:

K=

K=

Qs
U p A + Qs

Qs
U p A Qs

{Co Current Mode}

[3.7]

{Counter Current Mode}

[3.8]

in which Up is the particle settling velocity and A is the cross-sectional area of the flotation cell.
In most flotation processes, feed particles move with the fluid flow towards the discharge point
(co-current mode). A counter-current arrangement has obvious advantages since the settling
velocity is reduced by the upward flow of liquid resulting in a higher retention time. Hinderedsettling, as previously explained, provides an environment in which the particles never achieve
their terminal free-fall velocity. As a result, the effective particle velocity through the cell is
greatly reduced providing a significant increase in retention time as compared to a free-settling
system.
Finally, the rate constant (k) is the most important term in determining flotation
performance. Studies conducted by Yoon, et al. (1997) indicate that this parameter can be
mathematically described by:

k = 14 P S b

[3.9]

in which P is the probability of bubble-particle attachment and Sb is the superficial bubble

surface area rate. The latter can be calculated directly using:

121

Sb =

6 Qg
Db Ac

[3.10]

in which Qg is the volumetric gas flow rate, Db is the bubble diameter, and Ac is the cell crosssectional area (Yoon et al., 1997). Equations [3.9] and [3.10] suggest that the same flotation rate
constant (k) can be maintained at a lower overall gas rate (Qg) provided that the attachment
probability (P) increases accordingly.

3.3.2 Theoretical Advantages of the HydroFloat Cell

The HydroFloat cell is a flotation device that operates much like a traditional hinderedbed separator with feed settling against an upward current of fluidization water. However, unlike
a conventional teeter-bed separator, the HydroFloat cell is continuously aerated by injecting
compressed air and a small amount of frothing agent into the fluidization water. As previously
described, the small air bubbles are evenly dispersed into the cell and attach to the hydrophobic
particles within the teeter-bed. These bubble-particle aggregates have an effective density much
lower than that of the sole particle. These bubble-particle aggregates rise to the top of the denser
teeter-bed and overflow from the top of the separation chamber. Hydrophilic particles that do
not attach to the air bubbles continue to move down through the teeter-bed and eventually settle
into the dewatering cone and are discharged. Compared to traditional froth flotation processes,
the HydroFloat separator offers several important advantages for treating coarser material,
including enhanced bubble-particle contacting, increased residence time, lower axial mixing/cell
turbulence, and reduced air consumption.
According to Equation [3.2], the probability of attachment increases as the differential
velocity between bubbles and particles (Ub) is reduced.

Unlike conditions found in froth

flotation where particles are allowed to settle freely opposite the direction of rising bubbles, the

122

hindered-settling/rise conditions realized within the teeter-bed of the HydroFloat cell slows the
velocity at which bubbles and particles travel. As dictated by Equation [3.2], the reduced
velocity will increase the probability of adhesion (Pa), thereby enhancing flotation recovery. As
shown in Figure 3.6, this phenomenon is particularly important for coarse particles that tend to
suffer from low Pa values.
Greater recovery can also be realized utilizing the HydroFloat separator due to a decrease
in the probability of detachment (Pd). This decrease in detachment is a direct result of the
reduction of localized turbulence generally seen in hindered-bed separators.

As stated

previously, the optimum conditions for coarse particle flotation occur when cell agitation
intensity is reduced to a point just sufficient to maintain the particles in suspension. Thus, a
teeter-bed is an ideal environment for minimizing particle detachment (Barbery, 1984).
The HydroFloat cell is both a flotation device and a density separator. The use of a
teeter-bed makes it possible to achieve separations based on small differences between the
density of free suspended particles and the density of bubble-particle aggregates. As a result,
separations can be achieved even if the buoyancy of the bubble-particle aggregate is too small to
lift the particle load. In other words, the density of the bubble-particle aggregate need only be
smaller than the effective density of the teeter-bed to achieve a separation. This capability
eliminates the buoyancy limitation described by Equation [3.4]. This feature is important for
very large particles that are difficult to carry to the top of a conventional flotation pulp.
The HydroFloat cell also operates under nearly plug-flow conditions because of the low
degree of axial mixing afforded by the uniform distribution of particles across the teeter-bed. As
a result, the cell operates as if it were comprised of a large number of cells in series (i.e., high
value of n). As shown in Figure 3.8, this characteristic allows a single unit to achieve the same

123

recovery as a multi-cell bank of conventional cells (all other conditions equal). In other words,
the HydroFloat cell makes more effective use of the available cell volume than well-mixed
conventional cells or open columns.
The hindered-bed environment also influences particle retention time (p), and hence,
particle recovery. In most flotation processes, feed particles move with the fluid flow towards
the discharge point (co-current mode). In contrast, particles move in the opposite direction to the
fluid flow within the HydroFloat cell (counter-current mode). As dictated by Equations [3.6] and
[3.8], the fluidization water within a hindered-settling regime provides a significant increase in
the particle retention time. The longer retention time allows good recoveries to be maintained
without increasing cell volume.
The HydroFloat separator can be theoretically applied to any system where differences in
apparent density can be created by the selective attachment of air bubbles. In summary,
compared to traditional froth flotation processes, the HydroFloat separator offers several
important advantages for treating coarser feed streams. These include:

Improved Attachment: The differential velocity between bubbles and particles is greatly
reduced by the hindered settling/rise conditions within the teeter-bed of the HydroFloat
separator. Consequently, the reduced velocity will increase the contact time between bubbles
and particles, thereby promoting the probability of adhesion and enhancing flotation
recovery. This phenomenon is particularly important for coarse particles. The high solids
concentration within the teeter-bed will also improve recovery by increasing the collision
probability between bubbles and particles (Yoon and Luttrell, 1986).

124

Reduced Turbulence: According to Barbery (1984), the optimum conditions for coarse
particle flotation occur when cell agitation intensity is reduced to a point just sufficient to
maintain the particles in suspension. Woodburn (1971) and Schultz (1984) have also shown
that reduced cell turbulence significantly increases the maximum particle size limit for
effective flotation. The use of fluidization water in the HydroFloat separator makes it
possible to keep particles dispersed and in suspension without the intense random agitation
required by mechanical flotation machines.

No Buoyancy Limitation: Unlike traditional flotation processes, the HydroFloat cell is both a
flotation device and a density separator. The use of a teeter-bed makes it possible to achieve
separations based on small differences between the density of free suspended particles and
the density of bubble-particle aggregates. As a result, separations can be achieved even if the
buoyancy of the bubble-particle aggregate is too small to lift the aggregate from the surface
of the teeter-bed. This capability eliminates the buoyancy limitation and is particularly
important for very large particles that are difficult to carry to the top of a conventional
flotation pulp.

Plug-Flow Conditions: The HydroFloat cell operates under nearly plug-flow conditions
because of the low degree of axial mixing afforded by the uniform distribution of particles
across the teeter-bed. Consequently, the cell operates as if it were comprised of a large
number of cells in series. Provided that all other conditions are equal, this characteristic
allows a single unit to achieve the same recovery as a multi-cell bank of conventional cells
(Arbiter and Harris, 1962; Mankosa et al., 1992). In other words, the HydroFloat cell makes

125

more effective use of the available cell volume than well-mixed conventional cells or open
columns.

Increased Retention Time: In most flotation processes, feed particles move with the fluid
flow towards the discharge point (co-current mode). In contrast, particles move in the
opposite direction to the fluid flow in the HydroFloat cell (counter-current mode). The
counter-current mode has obvious advantages since the effective settling velocity of the
particles is reduced by the upward flow of liquid.

In addition, the hindered-settling

conditions within the teeter-bed never allow the particles to achieve their terminal free-fall
velocity. Therefore, the fluidization water provides a significant increase in the particle
retention time. The longer retention time allows good recoveries to be maintained without
increasing cell volume.

126

3.4

Population Balance Model

3.4.1 Model Description

A population balance model was developed and utilized in an effort to more fully
understand the HydroFloat separator.

Although fundamentals of flotation were used in

developing the HydroFloat separator, the actual separation of particles is accomplished by

gravity, based on density differences of components in the feed stream after the selective
attachment of air bubbles.

These bubbles change the apparent density of the hydrophobic

components so that the gravity separation can be enhanced.

The HydroFloat model was constructed much like the population balance model of the
CrossFlow separator developed in the previous chapter. The HydroFloat model utilizes general
equations for hindered-settling in transitional flow regimes to accurately predict overflow and
underflow partitions, particle size distributions, and component recovery data. Input data include
feed rate, percent feed solids (by mass), feed size distribution (up to 9 size fractions), density of
components in the feed stream (up to 2 components), fluidization water rate, and underflow
discharge rate. The general geometry and feed characteristics of these units are nearly identical.
This model was also constructed as a series of zones occurring in three distinct sections. These
primary sections include the feed inlet, the teeter-bed, and the underflow area. An illustration of
these primary sections has already been presented in Figure 2.12.
The Microsoft spreadsheet, Excel, was used for all calculations. This powerful software
package is capable of performing the iterative calculations required to solve the steady-state
equations necessary to model the HydroFloat separator.

127

3.4.1.1 Feed Section

The configuration of zones in the feed section of the HydroFloat separator was arranged
similarly to that of the CrossFlow separator model. Again, the cross-flowing action of the feed
water and solids necessitated the need for a 5 x 5 configuration as seen in Figure 2.13. If an
inadequate number of zones was utilized, particles could be incorrectly partitioned and
mathematically misplaced into overflow or underflow launders.
As shown in Figure 2.13, the upward flow of fluidization water that enters each zone is
shown as Qxn. This flow is counteracted by both the flow induced by solids settling (Qd) and by
the horizontal flows (Ql) that can move to or from adjacent cells. Material suspended within the
teeter-bed acts like a distributor for the rising teeter water, evenly distributing Qxn over the entire
cross-section of the unit for each level of the feed inlet area. The horizontal flows can be
calculated by conducting a flow balance for each zone within the feed section, given the
elutriation water rate (Qw), feed rate (Qf), and the underflow discharge rate (Qu).
Unlike the previous CrossFlow model, an assumption had to be made when modeling the
feed section of the HydroFloat separator. In this separator, bubbles attach to hydrophobic
particles creating bubble-particle aggregates. From visual inspection, it can be concluded that
these agglomerates are created in the feed section of the separator and rarely penetrate the teeterbed or underflow sections. It is not known how the attachment of air bubbles will affect the
rise/sink characteristics of these agglomerates. Consequently, it was assumed that after contact
and subsequent attachment of an air bubble or bubbles, the rise/sink characteristics of these
agglomerates would be equal to that of a particle of equivalent apparent size and density.

128

3.4.1.2 Teeter-Bed and Underflow Sections

The teeter-bed and underflow sections of the HydroFloat were also arranged similarly to
that of the previous CrossFlow model. This configuration can be seen in Figure 2.15. A
transition zone, to which fluidization water flow is added, separates these two sections. This
fluidization flow makes a split in this transition zone, with the majority of the water rising up
through the teeter-bed. This fluid flow assists the activated bubble-particle aggregates in rising
from the top of the teeter-bed into the overflow launder.
In the HydroFloat separator, small air bubbles are introduced into the unit along with the
elutriation water. The bubbles that rise up through the teeter-bed occupy a certain fraction of
volume within the separation chamber and consequently alter the apparent gravity of the teeterbed. It can be concluded that changes in air fraction within the separation chamber can be
affected by a large number of variables (i.e., average bubble size, average particle size in the
teeter-bed, frother addition, elutriation water flow rate, etc.). Incorporating these variables into
the general hindered-bed population balance model would add impractical complexities to the
already burdensome computer code. Consequently it was assumed that the rising bubbles had no
overall effect on the teeter-bed characteristics. This assumption may be inaccurate; however,
conclusions drawn from trends while using this model can nevertheless provide useful insight
into the advantages offered by the HydroFloat separator.

3.4.2 Calculations
Similar to the CrossFlow population balance model, an iterative dynamic technique (i.e.,
finite differencing) was used to solve for changes in concentration of particles over time for each
zone of the HydroFloat separator.

Using the general equations for hindered-settling in

129

transitional flow regimes presented in Chapter 2, component recovery/rejection data could be

calculated. The volumetric flows and solids from each zone were mass-balanced with respect to
one another using the laws of mass conservation (steady-state flow). This technique is also
discussed in length in Chapter 2.

3.4.3 Modeling Insight and Investigation

In an effort to illustrate the advantages offered by the HydroFloat for recovering coarse
particles, a separation of two components was simulated. Feed stream characteristics were
inputted into the model. This feed consisted of two density components and was divided into
nine size fractions. Both components had an equivalent and flat particle size distribution. The
feed rate was assumed to be 2.5 tph/ft2 at an elutriation water rate of 26.75 gal/ft2. Simulations
were conducted while varying the density ratio of the two components. The density of the first
component was reduced (from 3.0 to 1.25 SG) while maintaining the density of the second
component constant (3.0 SG).
The results of these simulations are presented in Figure 3.9. It can easily be seen that as
the density ratio (1/2) decreases, the recovery of coarse particles increases. At a density ratio
of 0.75, only 12% of the plus 0.71 mm material (+20 mesh) reported to the overflow. This
density ratio is typically found in applications where a 2.25 SG material is being separated from
a 3.00 SG material (i.e., mineral sands).

However, using air bubbles, the density ratio (1/2)

can be altered to 0.50. At this ratio, nearly 72% of the coarse, lower density material is now
recovered to the overflow launder. These results are analogous to recovering the coarse, low
density material that is typically lost in a conventional hindered-bed density separator. As
presented in Figure 3.10, this change in apparent density of one component can represent an
increase in total circuit recovery of nearly 22.5%. Naturally, additional improvements in overall

130

recovery can be realized if the apparent density of the low density component can be further
altered.

Recovery to Overflow(%)

100
80
60
40
20

1.00
0.92
0.83
0.75
0.67
0.58
0.50
0.42

0
0.11 0.13 0.18 0.25 0.36 0.50 0.71 1.00 1.18
Mean Size (mm)

Figure 3.9 Size by size recovery of components for varying density ratios (1/2).

131

Recovery to Overflow (%)

100
90

Overall Recovery (%)

80
70
60
50
40
1.00

0.80
0.60
Density Ratio ( 1/ 2)

0.40

Figure 3.10 - Circuit recovery data for varying component density ratios.

132

3.5

Proof-of-Concept Testing
Several laboratory-scale tests were conducted to evaluate the potential of the HydroFloat

separator for upgrading several types of mineral samples.

These samples included mineral

sands, phosphate matrix, carbonaceous slag and coal. Conventional flotation tests, whose results
were used as baseline data in some investigations, were conducted using a laboratory flotation
cell (Denver Model D-12). Aerated teeter-bed investigations were conducted with a laboratoryscale HydroFloat cell operated in two different modes, i.e., batch and continuous. The test unit
was fabricated from Plexiglas with an open area of approximately 50 cm2. This test unit is the
same device described previously in Chapter 2 (Figure 2.4) except that a static in-line mixer was
added to the elutriation water line.

Using compressed air and a glycol frothing agent in

conjunction with the static mixer allowed for the creation of small bubbles which were dispersed
throughout the teeter-bed via the elutriation piping network.

3.5.1 Phosphate Recovery

3.5.1.1 Testing of a North Florida Phosphate Matrix
A sample of a run-of-mine north Florida phosphate matrix (5 mm x 65 mesh) was scalped
at 3 mm to remove the extreme oversize and debris material. The feed ore was classified to
remove the minus 35 mesh fines and conditioned with a fatty acid-diesel fuel mixture at a dosage
of approximately 0.50 kg/t (active fatty acid). In all tests, ammonium hydroxide was used for pH
control and a polyglycol frother was added to stabilize the bubble suspension. Batch HydroFloat
tests were conducted and compared against conventional flotation results.
Figure 3.11 compares the results of the conventional flotation tests with those obtained
using the HydroFloat cell operated in the batch mode. In this series of tests, the HydroFloat cell

133

achieved a BPL recovery of nearly 95% compared to less than 79% for the conventional cell.
This represents an increase in BPL recovery of more than 20%. Although the conventional cell
was floated to exhaustion, higher recoveries were not possible since many of the coarser particles
remained unfloatable. Furthermore, the recovery improvement was achieved while maintaining
a slighter higher concentrate grade (67.0% versus 65.7% BPL).

The insol content of the

HydroFloat concentrate was also lower (7.2% versus 7.7% insols). In fact, the test data show
that the batch HydroFloat cell produced concentrates with a lower insol content over the entire
range of BPL recoveries.

The unexpected improvement in the quality of the concentrate

produced by the HydroFloat cell has been attributed to the incremental recovery of very highgrade coarse particles that could not be floated by the conventional flotation technique.

BPL
Recovery
(%)

Grade
(% BPL)

Insol (%)

Conventional
HydroFloat
0

20

40

60

80

100

Figure 3.11 - Comparison of separation data for batch test units for a phosphate matrix.

134

In light of the promising results obtained in the batch tests, several follow-up experiments
were conducted by operating the HydroFloat cell in the continuous mode. Several different
combinations of reagent dosages, water rates, and teeter-bed levels were examined in this effort.
Due to slight variations in the BPL head assay, it was necessary to compare the test results based
on a concentration ratio. This parameter is mathematically defined as the ratio of the concentrate
grade (%BPL) divided by the feed grade (%BPL). As shown in Figure 3.12, the continuous unit
produced consistently higher BPL recoveries than the conventional cell. The average recovery
for the HydroFloat was 93.8%, compared to only 78.7% for the conventional test. The best
continuous HydroFloat test result provided a BPL recovery of 92.4% at a concentration ratio of
2.12.

B PL R e cove ry (% )

100
80
60
40
Conventional

20

HydroFloat
0
0.5

1.0
1.5
2.0
Conce ntration R atio

2.5

Figure 3.12 - Comparison of separation data from continuous test units for a phosphate matrix.

135

3.5.1.2 Testing of a Central Florida Phosphate Matrix

Considering the favorable results of the northern Florida phosphate sample, another
sample of run-of-mine central Florida phosphate matrix (5 mm x 65 mesh) was investigated. It
was deemed necessary in order to prove that these were not ore-specific results. Again, the feed
was scalped at 3 mm to remove the extreme oversize and debris material.

In this second

phosphate investigation, two samples were tested. One sample was classified to remove the
minus 35 mesh fines while the other was classified to remove the minus 28 mesh fines. Each
sample was conditioned with a fatty acid-diesel fuel mixture at a dosage of approximately 0.50
kg/t (active fatty acid). For all tests, ammonium hydroxide was used for pH control and a
polyglycol frother was added to stabilize the bubble suspension. For each feed sample, several
tests were conducted in an effort to produce a grade and recovery curve.
Figure 3.13 compares the results of how each of the feeds responded to HydroFloat
testing. Both the plus 35 and plus 28 mesh phosphate matrix responded extremely well to the
aerated hindered-bed separator. It can be seen that the high BPL recoveries (95%) could be
maintained at extremely low insol grades (5-10%). It can also be seen that the plus 28 mesh feed
responded slightly better than the plus 35 mesh material as indicated by the higher separation
curve. This is most likely due to the elutriation water misplacing some of the fine silica particles
that were present in the finer feed sample. A lower water rate may have improved this result.
The optimum results for these tests are presented in Figure 3.14. The BPL content of the
plus 28 and plus 35 mesh product were 69.5 and 68.5%, respectively. These results were
produced at a BPL recovery of 93%. At these high recoveries, the HydroFloat was able to
maintain low insol grades of 5.0% and 6.3% for the coarse and fine feed material, respectively.

136

BPL Recovery (%)

100

90

80
+35 Mesh
+28 Mesh

70
0

10

20

30

40

Insols (%)
Figure 3.13 - Grade and recovery curve for central Florida phosphate sample.

100.00
80.00
60.00
40.00
20.00
0.00

BPL (%) Grade

BPL (%) Recovery

Insol (%) Grade

+28 Mesh

69.54

93.29

5.08

+35 Mesh

68.59

93.03

6.36

Figure 3.14 - Optimum results of HydroFloat tests on central Florida phosphate sample.

137

3.5.2 Coal/Carbon Recovery

3.5.2.1 Testing of Anthracite Slag
An industrial slag sample (nominally 6 mm x 200 mesh) was screened at 6.35 mm to
remove oversize tramp material prior to testing. The feed sample contained 12% fixed carbon
in the form of anthracite coal and 30% Fe2O3 and 26% TiO2. This sample was ideally suited
for HydroFloat treatment due to the inherent hydrophobicity of the low density component
(anthracite). The objective of these tests was to recover the remaining fixed carbon at a product
quality greater than 80%. Only seven tests were conducted on this sample due to the small
amount of available material. Four tests were conducted utilizing the lab-scale HydroFloat with
full teeter-bed aeration. Three were conducted with the HydroFloat operating as a traditional
hindered-bed separator (i.e., no teeter-bed air was employed).
Figure 3.15 shows the product grade and recovery plot for the plus 28 mesh fraction of
the feed material. Without aeration, the hindered-bed separator was able to make a separation,
although a product quality of over 71% fixed carbon could not be achieved. In contrast, a quality
of over 80% could be achieved when the teeter-bed was aerated. Aeration allowed the product
recovery to be increased by an overage of 10-15%, while simultaneously improving the fixed
carbon content of the product.

138

Fixed Carbon Recovery (%)

100
80
60
40

With Air

20

Without Air
0
0

20

40

60

80

100

Fixed Carbon (%)

Figure 3.15 - Testing of anthracite slag using the HydroFloat separator (6.35 mm x 200 mesh).

3.5.2.2 Testing of Central Appalachian Coal

A sample of run-of-mine coal from central Appalachia was used to evaluate the
effectiveness of the HydroFloat separator in treating 2 mm x 0.15 mm coal from an existing
spiral circuit. The feed coal was classified to remove the minus 100 mesh fines and conditioned
with approximately 0.25 kg/t of diesel fuel to enhance particle hydrophobicity. In the first series
of comparison tests, the HydroFloat separator was operated without the addition of air. The
separation performance achieved in this mode of operation would be identical to that obtained
using a traditional hindered-bed separator. In the second series of tests, the HydroFloat was
operated with air bubbles added to the teeter-bed. In this case, approximately 0.1 kg/t of
polyglycol frother was injected into the teeter water to improve air dispersion and minimize
bubble coalescence.

139

Figure 3.16 shows the recovery-ash curves comparing the performance of the HydroFloat
and hindered-bed separators. For convenience, the data have been reported for both the coarse
(plus 50 mesh) and fine (minus 50 mesh) size fractions. As expected, both devices achieved
good recoveries (>90%) of the minus 50 mesh material. The HydroFloat separator also produced
good recoveries of the plus 50 mesh material. Combustible recoveries in the range of 87-97%
were readily attainable over a wide range of operating conditions. In contrast, the hindered-bed
separator was not able to achieve recoveries greater than about 75% for the plus 50 mesh
material.

Attempts were made to improve the recovery of the plus 50 mesh particles by

increasing the flow rate of the fluidization water or by raising the level of the teeter-bed.
However, these attempts generally produced unacceptably high ash products due to (i) shortcircuiting of mineral matter into the product launder and (ii) excessive turbulence within the
teeter-bed. Since more of the feed mass resided in the plus 50 mesh fraction (approximately
60%), the overall performance of the HydroFloat was far superior to that of the hindered-bed
separator in treating the overall 2 x 0.15 mm sample. As shown in Figure 3.17, the recoveries
obtained for the overall feed with the addition of air were approximately 20 percentage points
higher than those obtained without air injection.

140

100

Recovery (%)

80

60

40

HydroFloat (+50 mesh)

Hindered Bed (+50 mesh)
HydroFloat (-50 mesh)
Hindered Bed (-50 mesh)

20

10

15

20

Ash (%)
Figure 3.16 - Test results (fractionated) obtained using spiral feed from Central Appalachia.

100

Recovery (%)

80

60

40

20
HydroFloat
Hindered Bed
-

10

15

20

Ash (%)
Figure 3.17 - Test results (cumulative) obtained using spiral feed from Central Appalachia.

141

3.5.2.3 Testing of Australian Coal

A sample of froth flotation feed from an Australian coal plant was obtained for testing in
the HydroFloat separator. This sample was considered very coarse for traditional flotation
processes, with over 25% of the coal greater than 0.6 mm (28 mesh). Traditionally, coal
flotation suffers from low recoveries above 28 mesh. It was expected that the HydroFloat could
recover the coarse feed material that was currently being lost in the froth flotation circuit. The
feed to the unit was conditioned with diesel (0.25 kg/ton) and a glycol frother was used for
bubble generation.
Table 3.1 provides a summary of the size-by-size recoveries and qualities obtained using
the HydroFloat separator. Special attention was paid to the effect of particle size since the teeter
water could easily propel fine ash particles (i.e., clay and mineral matter) into the concentrate.
As shown, the preliminary data indicate the plus 28 mesh material could be cleaned to an
acceptable ash content with combustible recoveries well above 90%. However, the ash content of
the finer size fractions (below 65 mesh) increased sharply due to the carry-over of fine mineral
matter. To overcome this problem, a second series of tests were conducted with the lab-scale
HydroFloat in which the minus 65 mesh material was discarded. In this case, the HydroFloat
was capable of combustible recoveries greater than 90% when air was added to the teeter-bed.
In comparison, an average recovery of only 83% was achieved without the addition of air.
However, the separator was capable of producing a higher quality (lower ash) product when no
air was added. The higher ash values obtained using the HydroFloat can be attributed to the
increased recovery of coarse high density middlings that report to the product stream when the
teeter-bed is aerated.

142

Table 3.1 - Size-by-size HydroFloat results obtained using an Australian coal.

Sample
Mesh
+16
16 x 20
20 x 28
28 x 65
-65

Yield (%)
Per Size
Cumulative
Class
85.0
80.4
75.8
61.4
63.3

0.5
8.0
28.2
61.6
70.3

Recovery (%)
Per Size
Cumulative
Class
94.4
93.1
91.7
86.3
81.4

0.8
12.0
41.6
83.9
90.2

Product
Ash (%)
8.4
13.8
16.5
21.2
25.8

3.5.2.4 Testing of Heavy Mineral Sands

Two samples of heavy mineral sands were tested using the HydroFloat separator. The
first sample contained unwanted carbonaceous matter as well as an undesirably high pyrite
content (average 0.78% sulfur). To process this sample, the feed was first treated with sodium
isopropyl Xanthate to make the pyrite hydrophobic and then passed through the HydroFloat unit.
As shown in Figure 3.18, the HydroFloat separator achieved sulfur and carbon rejections of up to
80% and 55%, respectively.

These rejections were maintained at a high product yield of

approximately 97%. The second sample also contained an unacceptably high carbon content
(average 0.92% fixed carbon). However, no Xanthate was added for this sample since the sulfur
content was already within product specifications.

Figure 3.19 shows that the HydroFloat

separator was also effective in treating this sample. More than 81% of the carbon was removed
from the feed material at a product yield of nearly 95%. The lower sulfur rejections reflect the
low feed sulfur content (0.05%) of this particular sample.

143

100

Rejection (%)

80
60
40
20

Sulfur
Carbon

0
0.00

0.10

0.20

0.30

0.40

0.50

Product Assay (%)

Figure 3.18 - HydroFloat results for the removal of impurities from mineral sands (Sample #1).

100

Rejection (%)

80

60

40

20

Sulfur
Carbon

0
0.00

0.05

0.10

0.15

0.20

0.25

Product Assay (%)

Figure 3.19 - HydroFloat results for the removal of impurities from mineral sands (Sample #2).

144

3.6

Pilot-Scale Testing
After the successful completion of the laboratory-scale proof-of-concept test work,

sufficient data was obtained to justify pilot-scale testing of the HydroFloat separator. Pilot-scale
work was carried out in two specific areas, namely, coal and phosphate.

Two pilot-scale

HydroFloat separators were fabricated to complete these investigations.

3.6.1 Northern Florida Phosphate Matrix

Phosphate testing was conducted at a northern Florida phosphate operation. To this end,
a 0.60 m square by 2.0 m tall test unit was fabricated and installed at the processing facility. A
photograph of the HydroFloat unit is provided in Figure 3.20. A flowsheet for the 10-15 tph test
circuit is provided in Figure 3.21. Circuit feed was supplied from a port located in the feed line
to a bank of plant dewatering cyclones used to prepare feed for existing conventional flotation
cells. The feed slurry was passed through a hydroclassification circuit to produce a plus 0.6 mm
underflow product. This material flowed by gravity into a bank of conditioning tanks where
flotation and pH reagents were added. The operating conditions are summarized in Table 3.2.
The product from the conditioners was then directed to the HydroFloat cell.

Table 3.2 - Parameters for in-plant pilot test program.

Parameter

CrossFlow

HydroFloat

Feed Rate (tph/sqft)

Feed Solids Density (%)
Water Addition Rate (gpm)
Aeration Rate (scfm)
Fatty Acid Dosage (lbs/ton)

1-7
15-50
40-90
n/a
n/a

2-4
50-70
40-80
2-5
1-3

145

Figure 3.20 - Photograph of the pilot-scale HydroFloat separator.

Classification

Conditioning

Separation

FEED

FINES
(-0.6 mm)
Water
Addition

4-Cell Conditioner
Water
Addition

COARSE
(+0.6 mm)

HydroFloat

Cross-Flow

Ammonia & Fatty Acid/Diesel

Mixture Addition Points

FLOAT

REJECT

Figure 3.21 - Pilot-scale flotation circuit for phosphate investigation.

146

Table 3.3 provides a comparison of test data from the pilot-scale HydroFloat unit with
that typically achieved by the existing full-scale conventional flotation circuit currently in
operation at the phosphate plant. At present, the plant typically operates with a BPL recovery of
approximately 35% for the plus 16 mesh feed and approximately 60% for the 16 x 35 mesh feed.
In comparison, the HydroFloat unit achieved a BPL recovery of more than 60% for the plus 16
mesh feed and nearly 85% for the 16 x 35 mesh feed. This represents an increase in recovery of
more than 40%. It is also interesting to note that the plus 16 mesh fraction had a very high BPL
content (72.8% BPL) and very low insol content (4.0%). The combined (i.e., plus 35 mesh)
concentrate from the HydroFloat cell represented at a total recovery of more than 80% with a
BPL grade of 56.8%. This result compares very favorably with the existing plant recoveries of
80-85% normally achieved for the finer 35 x 150 mesh feed.

Table 3.3 - Comparison of typical plant data and pilot-scale HydroFloat test results.
Particle
Size
(mesh)

Plant Cells
Recovery
(%)

HydroFloat
Recovery
(%)

HydroFloat
Grade
(% BPL)

+ 16
16 x 35
Total

35%
60%
50%

61.4%
84.7%
80.5%

72.8%
54.6%
56.8%

147

In fact, BPL recoveries were generally higher using the HydroFloat when compared to
actual plant data as seen in Figure 3.22. The data in this illustration show that the HydroFloat is
capable of recovering the coarse phosphate particles that are generally lost in the conventional
flotation cells. Naturally, as the particle size decreases, the performance of the HydroFloat
matches that of conventional flotation. However, in the coarsest particle size range (10x16
mesh), the HydroFloat was capable of recovering nearly 70% of the available BPL, while
conventional flotation only achieved a 17% BPL recovery.

BPL Recovery (%)

100

HydroFloat
Plant

80
60
40
20
0
10x16

16x20 20x28 28x35

Particle Size (Mesh)

Figure 3.22 - HydroFloat and plant comparison data.

148

-35

3.6.2 Appalachian Coal

The coal testing was completed at an Appalachian coal processing facility located in
southwestern West Virginia. Tests were conducted using a round, 20 inch diameter HydroFloat
as seen in Figure 3.23. This test unit is equipped similarly to the laboratory unit, with a PID
controller, pressure sensor, and pneumatic valve. One outlet of a spiral distributor was redirected
to provide constant feed (nominally 1mm x 65 mesh) to the HydroFloat. A ball valve was used
to throttle the feed flow rate to the device. Feed rates generally ranged from 1-4 tph/ft2, with
diesel additions of approximately 0.25 lb/ton to enhance the recovery of the coarse middling
coal. These operating conditions are summarized in Table 3.4.

Table 3.4 Operating conditions for in-plant pilot test program.

Parameter

HydroFloat

Feed Rate (tph/sqft)

Feed Solids Density (%)
Water Addition Rate (gpm)
Aeration Rate (scfm)
Diesel Dosage (lbs/ton)

1-4
25-35
6-12
2-5
0.20-0.50

149

Figure 3.23 - Pilot-scale HydroFloat used for coal testing.

As in the laboratory coal investigation, the HydroFloat was tested with and without the
addition of air bubbles in order to quantify the improvement, if any, over traditional hinderedbed separators. As seen in Figure 3.24, substantial increases in coal recovery were observed.
Specifically, the recovery of the plus 28 mesh fraction of this sample was increased by nearly
40%. The 28 x 35 mesh fractional recovery increased by almost 20%. The data suggesting that
the HydroFloat improved the recovery of coarse coal was further vindicated several times during
this test work, as misplaced and extremely coarse coal particles in the spiral circuit arrived in the
overflow launder of the HydroFloat (See Figure 3.25).

150

Recovery (%)

100
80
60
40

Without Air
With Air

20
0
+28

28x35

35x65

Particle Size (Mesh)

Figure 3.24 - Recovery data for the Appalachian coal tests.

Figure 3.25 - Coarse particles found in the HydroFloat product.

151

The HydroFloat was able to process and recover more coal without significantly
increasing the ash content when compared to traditional hindered-bed separations (without air).
Data presented in Figure 3.26 shows that for the plus 28 mesh fraction of coal, ash content
increased by only a few percent. The ash contents of the 28x35 mesh size fractions were nearly
identical. It must also be noted that the HydroFloat produced a significantly lower ash product
when comparing the 35x65 mesh size fraction.
The HydroFloat was tested using feed diverted from a distributor, which was supplying
feed to a bank of spirals. This bank of spirals was part of a rougher-cleaner spiral circuit with
partial middlings, comparable to the spiral circuit advocated in Chapter 1 of this dissertation.
Considering that the HydroFloat was treating the same material as the spiral circuit, a
comparison was warranted. Data from a concurrent coal company directed, full-plant sampling
endeavor were utilized in this effort. Figure 3.27 shows this data. It can be understood from this
figure that the HydroFloat, operating without air, achieved recoveries far below those obtained
by the existing spiral circuit. The coal spiral circuit also generally produced a higher quality
(lower ash) product. However, when the HydroFloat operated with air, combustible recovery
increased by nearly five percent. At optimum operating efficiency, the HydroFloat was able to
slightly improve upon the product quality (ash content) produced by the coal spiral circuit.

152

Without Air
With Air

20
15
10
5
0

+28

28x35

35x65

Particle Size (Mesh)

Figure 3.26 - Ash data for the Appalachian coal testing.

100
90

Recovery (%)

Product Ash (%)

25

80
70
60

With Air
Without Air
Plant

50
40
0

10

15

20

Product Ash (%)

Figure 3.27 - HydroFloat and coal spiral comparative data.

153

25

3.7

Conclusions
1. A new separator, known as the HydroFloat unit, has been developed to overcome some of
the shortcomings associated with traditional flotation machines in recovering coarse
particles. The novel characteristic of this separator is the formation of a hindered teeter
bed of fluidized solids into which small air bubbles are introduced. The bubbles attach to
hydrophobic particles and create light bubble-particle aggregates that can be separated
from hydrophilic particles based on the principle of differential density. Benefits of this
new separator include enhanced bubble-particle contacting, better control of particle
residence time, lower axial mixing/cell turbulence, and reduced air consumption.

2. Results from simulations conducted with a population balance model show that a
decrease in the apparent specific gravity of one feed component can greatly increase the
recovery of that component. If the density of a feed component can be altered through
the attachment of air bubbles, the density ratio (1/2) decreases, resulting in large
improvements in coarse particle recovery.

3. Laboratory tests were conducted with both batch and continuous HydroFloat cells in
order to evaluate the potential of this new technology for upgrading mineral samples
from various sources (e.g., phosphate matrix, coal, anthracite slag, mineral sands, etc.).
The test data indicate that the HydroFloat cell is capable of increasing coarse particle
recoveries by 20% over conventional flotation. Furthermore, the concentrate grades were
also improved in some cases due to a reduction in coarse particle misplacement.

154

4. In light of promising laboratory data, a pilot-scale HydroFloat was installed at an

industrial phosphate beneficiation plant. Test data obtained with this unit to date suggest
that the BPL recovery of the plus 16 mesh feed can be nearly doubled (i.e., increased
from approximately 35% to more than 65%) through the application of this new
technology.

In addition, the data suggest that the HydroFloat cell may be used to

increase the BPL recovery of the 16 x 35 mesh material from about 60% to nearly 85%.
As a result, the combined recovery of plus 35 mesh product from the HydroFloat cell
compares very favorably with existing plant recoveries of 80-85% normally achieved for
the 16 x 150 mesh feed.

5. A pilot-scale HydroFloat cell was also installed at an Appalachian coal preparation

facility. Data obtained from these in-plant tests indicate increases of coarse coal (+28
mesh) recoveries of nearly 40% over traditional hindered-bed separators.

Overall

combustible recovery was increased by nearly 20%. Product quality and combustible
recovery were consistent or better than that produced by an existing coal spiral circuit.

155

3.8

References

Ahmed, N. and Jameson, G.J., 1989, Flotation Kinetics, Mineral Processing and Extractive
Metallurgy Review, 1989, Vol. 5, pp. 77-99.
Anazia, I. and Hanna, J., 1988, Innovative Process for Beneficiation of Dolomitic Phosphate
Ores, International Journal of Mineral Processing, Vol. 23, pp. 311-314.
Anazia, I. and Hanna, J., 1987, New Flotation Approach for Carbonate Phosphate Separation,
Minerals and Metallurgical Processing, Vol. 4, No. 4, pp. 196-202.
Arbiter, N. and Steinenger, J., 1965, Hydrodynamics of Flotation Machines, Mineral
Processing, Pergamon Press, p. 595.
Arbiter, N. and Harris, C.C., 1962, Flotation Kinetics, Froth Flotation 50th Anniversary
Volume, (D.W. Fuerstenau, Ed.), Chpt. 8, American Institute of Mining, Metallurgy and
Petroleum Engineers, Inc., New York, New York, pp. 215-246.
Barbery, G., 1989, Method for Separation of Coarse Particles, U.S. Patent No. 4,822,493.
Barbery, G., 1984, Engineering Aspects of Flotation in the Minerals Industry: Flotation
Machines, Circuits and Their Simulation, The Scientific Basis of Flotation, (K. J. Ives,
Ed.), NATO Advanced Institute Services, Series E: Applied Sciences, No. 25, Martinus
Nijhoff Publishers, Boston, Massachusetts, pp. 289-348.
Bensley, C.N. and Nicol, S.K., 1985, The Effect of Mechanical Variables on the Flotation of
Coarse Coal, Coal Preparation, Vol. 1, pp. 189-205.
Brown, D.J. and Smith, H.G., 1954, The Flotation of Coal as a Rate Process, Transactions,
Institution of Mining Engineers, Vol. 113, pp. 1001-1020.
Bull, W.R., 1966, The Rates of Flotation of Mineral Particles in Sulphide Ores, Proceedings,
Australian Institute of Mining and Metallurgy, No. 220, December 1966, pp. 69-78.

156

Crawford, R. and Ralston, J., 1988, The Influence of Particle Size and Contact Angle in Mineral
Flotation, International Journal of Mineral Processing, Vol. 23, pp. 1-24.
Crawford, J.T., 1936, Importance of Pulp Density, Particle Size, and Feed Regulation in
Flotation of Coal, Transactions of AIME, Vol. 119, pp. 150-162.
Damodaran, R.M., Fahey, M. and Moudgil, B.M., Bench-Scale Optimization of Mixing
Variables During Anionic Conditioning of Florida Phosphates, Minerals and Metallurgical
Processing, February 1996.
Davis, B.E. and Hood, G.D., 1994, Conditioning Parameter Effects on the Recovery of Coarse
Phosphate, Minerals and Metallurgical Processing, Vol. 11, No. 1, pp. 50-54.
Davis, B.E. and Hood, G.D., 1993, Improved Recovery of Coarse Florida Phosphate, Mining
Engineering, Vol. 45, No. 6, pp. 596-599.
Davis, B.E. and Hood, G.D., 1992, Conditioning Parameter Effects on the Recovery of Coarse
Phosphate, Proceedings, Regional AIME/AIChE/AIPG/FIPR Phosphate Conference,
Lakeland, Florida, Sept. 24-25.
El-Shall, H., Zhang, P. and Snow, R., 1996, Comparative Analysis of Dolomite/Francolite
Flotation Techniques, Minerals and Metallurgical Processing, Vol. 13, No. 3, pp. 135-140.
El-Shall, H., Moudgil, B. and Wiegel, R., Beneficiation of Phosphate: Theory and Practice,
Engineering Foundation Conference, Palm Coast, Florida, editors; December 1993.
Finch J.A. and Dobby, G.S., Column Flotation, Pergamon Press, 1990.
Florida Institute of Phosphate Research, Publication No. 01-102-112, Defining the MgO
Problem and Its Economic Impact on Phosphoric Acid Production, Prepared by Jacobs
Engineering Group, Inc. and A.N. Baumann, February 1995.

157

Florida Institute of Phosphate Research, Publications No. 02-090-121, Understanding the

Basics of Anionic Conditioning in Phosphate Flotation, Prepared by Jacobs Engineering
Group, Inc., May 1995.
Florida Institute of Phosphate Research, Publication No. 04-045-103, An Optical Sensor for
On-Line Analysis of Phosphate Minerals, Prepared by Virginia Polytechnic Institute &
State University, April 1994.
Florida Institute of Phosphate Research, Publication No. 02-080-109, On-Line Analysis of
Phosphate Rock Slurry by Prompt Neutron Activation Technique, Prepared by University
of Florida, July 1994.
Florida Institute of Phosphate Research, Publication No. 02-094-108, Evaluation of Dolomite
Separation Techniques, Prepared by Hassan El-Shall, Global Marketing and Consulting,
October 1994.
Florida Institute of Phosphate Research, Publication No. 02-067-099, Enhanced Recovery of
Coarse Particles During Phosphate Flotation, Prepared by University of Florida, June 1992.
Florida Institute of Phosphate Research, Publication No. 02-070-098, Development of Novel
Flotation-Elutriation Method for Coarse Phosphate Beneficiation, Prepared by Laval
University, June 1992.
Florida Institute of Phosphate Research, Publication No. 04-032-091, An Investigation of
Potential for Improved Efficiencies in Phosphate Rougher Flotation Through On-Line BPL
Measurements, prepared by Harrison R. Cooper Systems, Inc., March 1991.
Florida Institute of Phosphate Research, Publication No. 02-063-071, Anionic Flotation of
Florida Phosphate, prepared by Zellar-Williams, Inc., February 1989.
Furey, J. and Mankosa, M.J., Column Flotation in the Phosphate Industry, Presented to Tenth
Annual Regional Phosphate Conference, Lakeland, Florida, October 1995.

158

Gaudin, A., Grob, J. and Henderson, H., 1931, Effect of Particle Size in Flotation, Technical
Publication No. 414, AIME.
Harris, M.C., Franzidis, J.P., OConner, C.T. and Stonestreet, P., 1992, An Evaluation of the
Role of Particle Size in the Flotation of Coal Using Different Cell Technologies, Minerals
Engineering, Vol. 5, No. 10/12, pp. 1225-1238.
Heiskanen, K., Particle Classification, Chapman & Hall, 1993.
Hutwelker, J.F., High Throughput Flotation Column Process, U.S. Patent No. 5,307,937.
Ives, K.J., Editor, The Scientific Basis of Flotation, NATO Advanced Science Institutes Series,
Martinus Nijhoff Publishers, 1984.
Jameson, G.J., Nam, S. and Young, M.M., 1984, Physical Aspects of Fine Particle Flotation,
Principles of Mineral Flotation, Vol. 215.
Jameson, G.J., Nam, S. and Young, M.M., 1977, Physical Factors Affecting Recovery Rates in
Flotation, Minerals Science and Engineering, Vol. 9, No. 3, pp. 103-118.
Kawatra, S.K., Editor, High Efficiency Coal Preparation: An International Symposium, SME,
Littleton, CO, 1995.
Keating, J., 1999, Personal Communication, IMC-Agrico.
King, R.P., Hatton, T.A. and Hulbert, D.G., 1974, Bubble Loading During Flotation, Technical
Note, Transactions of the Institute of Mining and Metallurgy, Section C, March, p. C9.
Laskowski, J., 1995, Flotation Process for the Flotation of Coarse Fractions Potash Ore, U.S.
Patent No. 5,456,362.
Lawver, J.E., Bernardi, J.P., McKeraghan, G.F., Raulerson, J.D., Lynch, D. and Hearon, R.S.,
1984, New Techniques in Beneficiation of the Florida Phosphates of the Future, Minerals
and Metallurgical Processing, Vol. 1, No. 2, pp. 89-106.

159

Lawver, J., McClintock, W.O. and Snow, R.E., 1978, Beneficiation of Phosphate Rock: A State
of the Art Review, Minerals Science and Engineering, Vol. 10, No. 4, pp. 278-294.
Llewllyn, T.O., Davis, B.E. and Sullivan, G.V., 1984, Beneficiation of Florida Dolomite
Phosphate Ores, Minerals and Metallurgical Processing, Vol. 1, No. 1, pp.43-48.
Luttrell, G.H., 2000, Column Flotation Workshop, conducted for CONSOL Energy, Osage,
WV, Aug. 22.
Mankosa, M.J., Luttrell, G.L, Gruber, G. and Shoniker, J., 1999, In-Plant Testing of the
HydroFloat Separator for Coarse Phosphate Recovery, Proposal to Florida Institute of
Phosphate Research, FIPR #99-02-137.
Mankosa, M.J., Forrest, M.L. and Furey, J.T., Application of the Cominco Column Cell for
Coarse Phosphate Flotation, CIM 35th Annual Conference of Metallurgists, Montreal,
Quebec, Canada, August 1996.
Mankosa, M.J., Luttrell, G.H., Adel, G.T. and Yoon, R.-H., 1992, "A Study of Axial Mixing in
Column Flotation," International Journal of Mineral Processing, Vol. 35, pp. 51-64.
McHardy, J.C., 1983, New Trends in Florida Phosphate Mining, Mining Engineering, Vol. 35,
No. 8, pp. 1196-1200.

McKnight, K., Stouffer, N., Domenico, J. and Mankosa, M.J., Recovery of Zircon and Other
Economic Minerals from Wet-Gravity Tailings Using the Floatex Density Separator, SME
Annual Meeting, Phoenix, Arizona, March 1996.
Morris, T.M., 1952, Measurement and Evaluation of the Rate of Flotation as a Function of
Particle Size, Mining Engineering, Vol. 4, No. 8, pp. 794-798.
Moudgil, B.M. and Gupta, D., 1989, Flotation of Coarse Phosphate Particles, Advances in
Coal and Mineral Processing Using Flotation, Proceedings of an Engineering Foundation
Conference, Dec. 3-8, pp.164-168.

160

Moudgil, B.M. and Barnett, D.H., 1979, Agglomeration Skin Flotation of Coarse Phosphate
Rock, Mining Engineering, Vol. 31, No. 3, pp. 283-289.
Olsen, S.I. and Miller, H.H., 1981, Improved Recovery of Florida Phosphate Rock by
Modifying Standard 14 m3 (500 ft3) Wemco Cells with 28 m3 (1000 ft3) Cell Mechanisms,
SME-AIME Fall Meeting and Exhibit, Denver, Colorado, Nov. 18-20.
Oteyaka, B. and Soto, H., 1995, Modeling of a Negative Bias Flotation Column For Coarse
Particles Flotation, Minerals Engineering, Vol. 8, No. 1/2, pp. 91-100.
Phillips, D.I., 1998, Optimum Processing of 1mm by Zero Coal, Ph.D. Dissertation, Dept. of
Mining and Minerals Engineering, Virginia Polytechnic Institute and State University, 196
pp.
Rastogi, R.C. and Aplan, F.F., 1985, Coal Flotation as a Rate Process, Minerals and
Metallurgical Processing, Vol. 2, pp. 137-147.
Robinson, A.J., 1959, Relationship Between Particle Size and Collector Concentration,
Transactions, Institution of Mining and Metallurgy, No. 69, pp. 45-62.
Schulze, H.J., 1984, Physio-Chemical Elementary Processes in Flotation, Developments in
Mineral Processing, Vol. 4, Chpt. 5, Elsevier, New York, pp. 238-253.
Schulze, H.J. and Gottschalk, 1981, Proceedings of the 13th International Mineral Processing
Congress, Warsaw, Poland, June 1979, Part A, Elsevier, New York, New York, p. 63.
Schulze, H.J., 1977, New Theoretical and Experimental Investigations on Stability of
Bubble/Particle Aggregates in Flotation: A Theory on the Upper Particle Size of
Floatability, International Journal of Mineral Processing, No. 4, pp. 241-259.
Schumann, R., 1942, Flotation Kinetics, Methods for Steady-State Study of Flotation
Problems, Journal of Physical Chemistry, Vol. 46, pp. 891-902.

161

Soto, H. and Barbery, G., 1991, Flotation of Coarse Particles in a Counter-Current Column
Cell, Minerals and Metallurgical Processing, Vol. 8, No. 1, pp. 16-21.
Soto, H., 1988, Private Report to the Florida Institute of Phosphate Research.
Soto, H. and Barbery, G., 1987, Method and Apparatus for Separation of Coarse Particles,
U.S. Patent Application No. 125-709.
Soto, H. and Iwasaki, I., 1986, Selective Flotation of Phosphate from Dolomite Using Cationic
Collectors. Part II, International Journal of Mineral Processing, No. 16, pp. 17-27.
Sun, S. and Zimmerman, R.E., 1950, The Mechanism of Coarse Coal and Mineral Froth
Flotation, Mining Engineering, Vol. 187, No. 5, pp. 616.
Sutherland, K.L., 1948, Kinetics of the Flotation Process, Journal of Physical Chemistry, Vol.
52, pp. 394-425.
Trahar, W.J., 1981, A Rational Interpretation of the Role of Particle Size in Flotation,
International Journal of Mineral Processing, Vol. 8, pp. 289-327.
Trahar, W.J. and Warren, L.J., 1976, The Floatability of Very Fine Particles A Review,
International Journal of Mineral Processing, Vol. 3, pp. 103-131.
United States Geological Survey, Mineral Commodity Summary, January, 1999.
Woodburn, E.T., King, R.P. and Colborn, R.P., 1971, The Effect of Particle Size Distribution
on the Performance of a Phosphate Flotation Process, Metallurgical Transactions, Vol. 2,
pp. 354-362.
Yoon, R.-H. and Mao, L., 1997, Predicting Flotation Rates Using a Rate Equation Derived from
First Principles, International Journal of Mineral Processing, Vol. 51, No. 1-4, pp. 171-181.
Yoon, R.-H. and Luttrell, G.H., 1989, The Effect of Bubble Size on Fine Particle Flotation,
Mineral Processing and Extractive Metallurgy Review, Vol. 5, Gordon and Breach Science
Publishers, Inc., Great Britain, pp. 101-122.

162

CHAPTER 4
General Summary

Mathematical analysis tools, including linear circuit analysis and population balance
modeling, have been utilized to analyze and evaluate some water-based processes. The data
collected from these investigations were used to make modifications and/or improvements in
coal spiral circuitry, hydraulic classification and hindered-bed separation. These improvements
resulted in increased separation efficiency and unit capacity. A better understanding of two
novel pieces of mineral processing equipment (i.e., the CrossFlow and HydroFloat separators)
were also a result of these analyses. In review, several points of summary can be identified from
this work.

Improving Spiral Performance Using Circuit Analysis

1.

Linear circuit analysis, a theoretical tool for comparing the relative effectiveness of various
configurations of unit operations, was successfully applied to coal spiral circuitry. Early
studies identified several coal spiral circuits that had the potential to improve separation
efficiency. One circuit in particular, a rougher-cleaner configuration with partial middlings
recycle, was capable of improving separation efficiency (Ep) approximately 1.22 times that
of more traditional coal spiral circuits. A reduced circuit SG50 and reasonable circulating
load were also a benefit of this modified circuit when compared to other preferred unit
configurations.

163

2.

Based on circuit analysis fundamentals, an alternative method was derived for determining
the partition expressions for any given spiral circuit. This method allows for the prediction
of efficiency (Ep), circuit cut-point or any other partition based result (i.e., SG25 or SG75)
independent of washability data, provided a proper partition function is used.

3.

In-plant testing of a full-scale, two-stage spiral circuit allowed for the comparison of several
alternative circuit configurations.

Data obtained during the in-plant circuit evaluations

indicated that for an equivalent number of spirals, rougher-cleaner circuits operated in series
are far superior than parallel circuits for reducing circuit cut-point.

4.

In-plant test data also indicated that although the circuit SG50 for rougher-cleaner coal spiral
circuits operated with and without a middlings recycle are very similar (i.e., 1.65 SG), the
separation efficiency increased when a middling recycle stream was utilized.

In fact,

separation efficiency was approximately 1.25 times higher for the circuit incorporating a
middlings recycle stream.

5.

Using the in-plant test data, regression equations were developed that were used to simulate
alternative spiral configurations. Although a rougher spiral separation could be adequately
simulated using an expression developed by Reid (1971), the increased loading of neargravity material often found on a cleaner bank of spirals necessitated the development of an
alternative expression.

This expression (i.e., the modified Reid equation) accurately

simulates an asymmetrical coal spiral partition in which a large amount of high gravity
material is misplaced to the clean coal launder.

164

6.

Simulations demonstrated that the rougher-cleaner coal spiral circuit incorporating a

middlings recycle was capable of a separation efficiency (Ep) 1.20 times that of a roughercleaner spiral circuit without middlings recycle. This ratio compares favorably with both
the theoretical ratio of 1.22 predicted by circuit analysis and the ratio of 1.25 demonstrated
by actual in-plant performance.

7.

The simulations further demonstrated that at a constant ash, the rougher-cleaner spiral
circuit with middlings recycle is capable of increasing circuit yield by as much as 3.86%.
This relatively small increase in yield translates to a growth in revenue of nearly $255,000
for the coal preparation facility at which the on-site tests were conducted. Furthermore, this
increase does not reflect the additional salable coal yielded by the coarse coal dense medium
circuits, which can now operate at higher effective gravities as a result of the lower overall
cut-point of the spiral circuit.

Improving Performance of Hindered-Bed Separators

1.

Data from comparative in-plant and laboratory studies show that the feed presentation
system of the CrossFlow separator offers several advantages. Results from tests conducted
with phosphate matrix, limestone aggregate and heavy mineral consistently showed an
increase in capacity and separation efficiency when compared to traditional hindered-bed
separators.

165

2.

A population balance model was developed to study and understand the operation of the
newly developed CrossFlow hindered-bed separator. This model was based on general
hindered-settling equations for transitional flow regimes. Model input data include feed
rate, feed percent solids (by mass), feed size distribution (up to 9 size fractions), density of
up to two feed components, fluidization water rate, and underflow discharge rate. Output
results included overflow and underflow partition data, size distributions, component
recovery, and classification efficiency in terms of Ep or Imperfection.

3.

Validation test work indicated a good correlation existed between the laboratory and model
simulation results.

Reliable consistency was found for separation cut-point (d50) and

efficiency (i.e., Ep or I). During the validation test work, a correlation between target cutpoint (d50) and the maximum concentration by volume of solids (max) was confirmed. This
linear relationship appears to vary with material, feed size distribution, and consequently,
separation cut-point (d50).

4.

Data produced from simulations using the population balance model indicate that the crossflow feed presentation system has several advantages when compared to those used in
conventional hindered-bed separators. These benefits include increased operational stability
and a unit capacity of up to 6 tph/ft2 (71.2 tph/m2). Simulation data also show that the
CrossFlow can maintain an acceptable and less varied efficiency over a wide range of
operating conditions, including low feed percent solids (i.e., approaching 24% by mass).

166

5.

Laboratory solid and liquid tracer studies of the CrossFlow separator suggest that excess
feed water and solids that should report to the overflow launder are quickly off-loaded by
the cross-flowing action of the feed presentation system. This occurs without disturbing the
volume of material within the separation chamber. In contrast, traditional hindered-bed
separators employing downcomer technology inefficiently use separation chamber volume
to manage excess feed water and segregate overflow material prior to discharge.

Improving Coarse Particle Recovery in Hindered-Bed Separators

1.

A hindered-settling population balance model was developed and utilized to identify an

approach to overcome the inherent disadvantages often found in conventional teeter-bed
separators when used as density separation devices. A feed consisting of up to two density
components and nine size fractions could be employed with this model. Output results
include component partition, recovery and rejection data.

2.

Results from the modeling investigations suggest that any alteration of apparent density of
any one feed component can greatly effect the recovery of that component. Data showed
that an increase in the recovery of coarse, low density material could be realized if the
apparent specific gravity of that component could be modified (i.e., lowered). In fact,
further simulation indicated that recovery could increase by up to 60%.

3.

To this end, the HydroFloat separator was developed based on flotation fundamentals. This
device uses an aerated teeter-bed through which bubbles can rise and attach to hydrophobic
particles. The attachment of air bubbles sufficiently reduces the apparent density of the

167

hydrophobic particles. These low density bubble-particle aggregates are then separated
from the hydrophilic particles based on the principle of differential density.

4.

Data from laboratory proof-of-concept testing indicate that the HydroFloat cell was
successful in upgrading various types of minerals, including phosphate matrix, coal,
anthracite slag, and mineral sands. Data further indicate the HydroFloat cell is capable of
increasing the recovery of coarse (2mm x 50 mesh) particles over that traditionally found in
either froth flotation or conventional hindered-bed separations.

Coarse particle

misplacement was also reduced resulting in improved product concentrate grades.

5.

In-plant testing was conducted at a north Florida phosphate beneficiation plant. Coarse
phosphate recovery increased substantially using the HydroFloat cell when compared to
existing conventional froth flotation cells. BPL recovery nearly doubled for the +16 mesh
size fraction, and an increase of 25% was also achieved for the 16 x 35 mesh size fraction.
Product quality also improved due to the increased recovery of substantially higher grade,
coarse phosphate.

6.

Further in-plant testing of the HydroFloat cell was conducted at an Appalachian coal
processing facility. Increases of up to 40% in coarse coal recovery were realized using the
HydroFloat cell when compared to a traditional hindered-bed separator. As a result, an
increase of 20% in combustible recovery was also achieved. A concurrent in-plant survey
showed that the HydroFloat cell was capable of achieving a product quality and combustible
recovery equivalent to that of an existing coal spiral circuit.

168

CHAPTER 5
Recommendations for Future Work

Several suggestions for follow-up efforts are offered here.

These recommendations

address all aspects of this investigation as seen below.

Improving Spiral Performance Using Circuit Analysis

1. It is recommended that a plant-wide study be conducted in order to quantify the
improvements seen throughout the coal preparation facility as a direct result of the reduced
cut-point in the spiral circuitry. As the specific gravity cut-point of the spiral circuit becomes
lower, the cut-points of other circuits in the plant should increase slightly to compensate for
the decrease in product ash content.

These higher cut-points should translate to increased

revenue since these circuits generally treat coarser material at higher tonnages and
efficiencies.

2. It is suggested that several other preferred circuit configurations be tested in plant. Several
configurations as indicated by linear circuit analysis had relative efficiencies better than the
rougher-cleaner configuration that incorporated middlings recycle (See Table 1.1). These
were discounted due to the impracticality and added cost associated with an increased
circulating load in the spiral circuit. Because the improved spiral circuit efficiency will
impact the entire plant, these alternative preferred circuits with high circulating loads may be
feasible when accounting for total improvement in plant performance.

169

Improving Performance of Hindered-Bed Separators

1. As stated in Chapter 2, the CrossFlow population balance model was capable of predicting
classification cut-point with a high degree of accuracy. However, the model had some
difficulty predicting efficiency, predominantly showing slightly lower efficiencies than what
is seen in actuality. This is most likely caused by a discretization error found in the feed
section of the population balance model.

Currently, the feed section of the model is

composed of 25 zones, in a 5 by 5 configuration. It is recommended that this feed section be

expanded to contain a higher number of zones to minimize the discretization error. In fact, a
continuous model would prove to be most useful.

2. It is also recommended that more in-plant test work be completed which directly compares
the CrossFlow to other conventional hindered-bed separators under identical conditions.
This test work would help to quantify the advantages offered by the CrossFlow separator.

Improving Coarse Particle Recovery in Hindered-Bed Separators

1. It is suggested that an advanced, full-scale, in-plant evaluation of the HydroFloat be
undertaken. A beneficiation plant will prove to be the only location where a high rate of
constant feed can be maintained for test purposes. The effect of treating coarse (1mm x 50
mesh) material with the HydroFloat can also be quantified with respect to any performance
changes in the conventional flotation cells.

2. The development of a hybrid, hindered-bed/flotation-based population balance model for the

HydroFloat separator is highly recommended. Although the separation of components in this

170

device is based on differences in apparent density, many operating characteristics are

flotation based. It is important to understand the effect, if any, that bubble size, air fraction,
aeration rate, etc., will have on the flotation and the recovery of coarse particles with respect
to the overall performance of this device.

171

APPENDIX A
Mass-Balance Equations

172

Flow Balances
Using the overall CrossFlow zone schematic shown in Figure A1, the following overall
volumetric flow balances can be written:
QF + Q W = QU + QL O

Q Z = Q W QU

where QF, QW, QU, and QLO5 are the feed, elutriation, underflow, and overflow volumetric flows,
respectively. The feed, elutriation and underflow flow rates are known at time t = 0.
Using the enlarged feed zone schematic seen in Figure A2, the following flow balance can be
written:
QL O = QF + Q X1
5

where QX1 is the vertical upwards flow rate exiting zone A and entering the feed zone sections.

The fluidized-bed evenly distributes all vertical flows evenly across the cross-sectional area of
the separator, except for the first five vertical zones (O1, B1, C1, D1, and E1). In these zones, it
can be shown that the falling action of feed solids and associated liquid prevents little or no
upward flow from entering. This assumption is valid as the cross-sectional area of these zones is
minimized. Using this information, the following flow balances can be equated:

Q
+ Q
+ Q
+ Q

Q X2 = Q X1 + QLB +

DC 2 4

Q X3 = Q X2 + Q L C

DD 2 4

Q X4 = Q X3 + Q L D

Q X5 = Q X4 + Q L E

DE 2 4
DO 2 4

Q
Q
Q
Q

DB 2 4

DC2 4
DD2 4
DE 2 4

where QD is the is the downward volumetric flow induced by the settling action of solids in to
and out of each respective zone.

173

Feed, Qf

Feed Section

Overflow, QLCO5
O5

O4

O3

O2

O1

E5

E4

E3

E2

E1

D5

D4

D3

D2

D1

C5

C4

C3

C2

C1

B5

B4

B3

B2

B1

Transition Zone, A

Teeter-Bed Section

UpA

Qz

G
UpA

Qz

H
UpA

Qz

I
UpA

Qz

J
UpA

Qz

K
Underflow Section

UpA

Qz

Transition Zone L

UpA

Qw

Qu

Qu
Figure A1 Overall zones and major flows in the CrossFlow population balance model.
174

Feed Section

Overflow, QLO5

Feed, QF

O5

O4

O3

O2

O1

E5

E4

E3

E2

E1

D5

D4

D3

D2

D1

C5

C4

C3

C2

C1

B5

B4

B3

B2

B1

Transition Zone, A

Qx(n+1)
UpA
Ql(n+1)

Qd(n+1)
Qln

Zone Xi,j
UpA

Qdn
0.25Qxn

Figure A2 Zone and flow schematic for feed section of CrossFlow model.

175

The horizontal flows that are found between each zone of the feed section can be determined
from simple flow balances around each zone. Completing these flow balances yield the
following equations:
Q L O = Q F Q DO
1

QLO = 0.25 * Q X5 + QL O QDO

2

QL O = 0.25 * Q X5 + QLO Q DO
3

QLO = 0.25 * Q X5 + QLO QDO

4

QL O = 0.25 * Q X5 + QLO QDO

5

Q L E = Q DO Q DE
1

QLE = 0.25 * Q X4 + QDO + QLE 0.25 * Q X5 QDE

2

QLE = 0.25 * Q X 4 + Q DO + QLE 0.25 * Q X5 QDE

3

QLE = 0.25 * Q X 4 + QDO + QLE 0.25 * Q X5 QDE

4

Q L D = Q DE Q DD
1

QLD = 0.25 * Q X3 + QDE + QLD 0.25 * Q X 4 QDD

2

QLD = 0.25 * Q X3 + QDE + QLD 0.25 * Q X 4 QDD

QLD = 0.25 * Q X3 + QDE + QLD 0.25 * Q X 4 QDD

Q L C = Q DD Q DC
1

QL C = 0.25 * Q X 2 + QDD + QLC 0.25 * Q X3 QDC

2

QLC = 0.25 * Q X 2 + QDD + QLC 0.25 * Q X 3 QDC

3

QLC = 0.25 * Q X2 + QDD + QLC 0.25 * Q X3 QDC

4

Q L B = Q DC Q DB
1

QLB = 0.25 * Q X1 + QDC + QLB 0.25 * Q X 2 QDB

2

QLB = 0.25 * Q X1 + QDC + QLB 0.25 * Q X2 QDB

3

QLB = 0.25 * Q X1 + QDC + QLB 0.25 * Q X2 QDB

4

176

Mass Balances
Utilizing the law of mass conservation, mass balance equations were written for each zone in the
dynamic population balance model (mass in = mass out). For each zone and for each size class,
all flows were balanced and multiplied by the concentration of solids (Czone) present in the zone
from which the flow emanated. This was completed for increments of time as small as 1/10000th
of one second. For each time increment (iteration), a new concentration for each zone could be
calculated and mathematically added to the concentration from the previous iteration. The term
UpA represents the volume of settling solids in a particular zone as defined by the hinderedsettling equations presented in the body of this work.

Using Figure A2, the following equations can be written:

C O1NEW = C O1 + (QF C F C O1 (QL CO + Q DCO + UP A ))t
1

C O 2NEW = C O2 + (QLCO CO1 + 0.25 * Q X5 CE2 C O2 (QL CO + QDCO + UP A ))t

1

C O3NEW = C O3 + (QLCO C O2 + 0.25 * Q X5 CE3 C O3 (QL CO + QDCO + UP A ))t

2

C O 4NEW = C O 4 + (QL CO C O3 + 0.25 * Q X5 CE 4 C O4 (QLCO + QDCO + UP A ))t

3

C O5NEW = C O5 + (QLCO C O 4 + 0.25 * Q X5 CE5 C O5 (QLCO + QDCO + UP A ))t

4

CE1NEW = CE1 + (QDO C O1 CE1 (QLE + QDE + UP A ))t

1

CE 2NEW = CE 2 + (QLE CE1 + 0.25 * Q X4 CD2 + QDO C O 2 C E2 (0.25 * Q X5 + QLE + QDE + UP A ))t
1

C E3NEW = CE3 + (QLE CE 2 + 0.25 * Q X 4 C D3 + QDO C O3 CE3 (0.25 * Q X5 + Q LE + QDE + UP A ))t

2

CE 4NEW = CE 4 + (Q LE CE 3 + 0.25 * Q X 4 CD4 + QDO C O 4 CE 4 (0.25 * Q X 5 + QLE + QDE + UP A ))t

3

C E5NEW = CE5 + (QLE CE 4 + 0.25 * Q X4 CD5 + QDO C O5 C E5 (0.25 * Q X5 + QLE + QDE + UP A ))t
4

CD1NEW = CD1 + (QDE CE1 CD1 (QLD + QDD + UP A ))t

1

CD2NEW = CD2 + (QLD CD1 + 0.25 * Q X 3 C C2 + QDE CE 2 CD2 (0.25 * Q X 4 + QLD + QDD + UP A ))t
1

C D3NEW = CD3 + (QLD C D2 + 0.25 * Q X3 C C3 + QDE CE3 CD3 (0.25 * Q X 4 + QLD + QDD + UP A ))t
2

CD 4NEW = CD4 + (QLD CD3 + 0.25 * Q X3 C C4 + QDE C E4 C D4 (0.25 * Q X 4 + QLD + QDD + UP A ))t
3

C D5NEW = CD5 + (QLD CD4 + 0.25 * Q X 3 C C5 + QDE CE5 CD5 (0.25 * Q X 4 + QLD + QDD + UP A ))t
4

C C1NEW = C C1 + (QDD CD1 C C1 (QL C + QDC + UP A ))t

1

C C2NEW = C C2 + (QLC C C1 + 0.25 * Q X 2 CB2 + QDD CD2 C C2 (0.25 * Q X3 + QLC + QDC + UP A ))t
1

C C3NEW = C C3 + (QLC C C2 + 0.25 * Q X 2 CB3 + QDD CD3 C C3 (0.25 * Q X3 + QL C + QDC + UP A ))t

2

C C 4NEW = C C4 + (QLC C C3 + 0.25 * Q X2 CB 4 + QDD CD 4 C C4 (0.25 * Q X3 + QLC + QDC + UP A ))t

3

C C5NEW = C C5 + (Q LC C C4 + 0.25 * Q X2 CB5 + QDD CD5 C C5 (0.25 * Q X 3 + QL C + QDC + UP A ))t

4

177

CB1NEW = CB1 + (QDB C C1 CB1 (QLB + QDB + UP A ))t

1

C B2NEW = CB 2 + (QLB CB1 + 0.25 * Q X1 C A + QDC C C2 CB2 (0.25 * Q X 2 + QLB + QDB + UP A ))t
1

CB 3NEW = CB3 + (QLB CB 2 + 0.25 * Q X1 C A + QDC C C3 CB3 (0.25 * Q X 2 + QLB + Q DB + UP A ))t

2

CB 4NEW = CB4 + (QLB CB3 + 0.25 * Q X1 C A + QDC C C4 CB4 (0.25 * Q X2 + QLB + QDB + UP A ))t
3

CB5NEW = CB5 + (QLB CB 4 + 0.25 * Q X1C A + QDC C C5 CB5 (0.25 * Q X2 + QLB + QDB + UP A ))t
4

C A New = C A + ( Q F C F + U P A * C B + Q Z C G C A ( Q X 1 + U P A )) t

Using Figure A3, the following equations can be written:

C GNew = C G + (Q Z C H + UP AC A C G (Q Z + UP A ))t
C HNew = CH + (Q Z C I + UP AC G C H (Q Z + UP A ))t
CINew = CI + (Q Z C J + UP AC H CI (Q Z + UP A ))t
C JNew = C J + (Q Z C K + UP ACI C J (Q Z + UP A ))t
C K New = CK + (Q Z C L + UP AC J C K (Q Z + UP A ))t
C LNew = C L + (UP ACK C L (Q Z + Q U + UP A ))t
C UNew = CU + (C P (QU + UP A ) C UQ U )t

These equations were solved continuously until the change in concentration for each and every
zone in the population balance model was equivalent to zero or CZONE(new) is equal to CZONE.
Steady-state conditions are realized when the change in concentration for the entire separator is
equal to zero.

178

UpA

Qz

Teeter-Bed Section

UpA

Qz

H
UpA

Qz

I
UpA

Qz

J
UpA

Qz

K
Qz

UpA
Transition Zone L

Underflow Section

UpA

Qw

Qu

Qu
Figure A3 Zone and flow schematic for teeter and underflow sections of the CrossFlow model.

179

APPENDIX B
Model Input/Output Examples

180

Operational Input (user input in bold font):

(tph/ft )
2.5

Solids (%)
50

Feed Tonnage Data

Solids Feed Rate
2
(tph/m )
tph
kg/hr
kg/s
26.91
0.139
126.0
0.0350

Slurry Feed Rate

3
m /s
kg/s
0.070
0.00005

Feed Input Data

Feed Concentration
Slurry
Density 1
Density 2
1.45
2.65
3.50

Water Rates
Elutriation Water Rate
3
3
m /s
ft /s
(gpm)
0.0022
0.00006
1.00

Starting Values (m /s)

Qf
0.00005
Qw
0.00006
Qo
0.00004
Qu
0.00002
Qz
0.00004

Lower Time Step (sec)

Higher Time Step (sec)
Current Time Step (sec)

Overall
2.65

Starting Values (kg/m )

Cf
726.03
Cw
0.0000
Co
0.0000
Cu
0.0000
Ca-Cp
0.0000

Starting Values (m /m )
0.2740
f
0.0000
w
0.0000
o
0.0000
u
0.0000
a-p

Time Limits
0.0001 Starting Delta Time (sec)
0.075
Total Time (sec)
0.0819
Total Time (min)

181

0.00001
10800
180

Qu (lpm)
1.22
Reset
1

Target D50
0.3
0.869

Unit Size Input (user input in bold font):

Number of Zones (#):
Separation Chamber Area
2
2
2
(ft )
(m )
(in )
0.056
0.0052
8.0

Cell
Length
Area

C5
0.75
0.0010

Zone
Zone Depth
(#)
(in)
O5
2.00
O4
2.00
O3
2.00
O2
2.00
O1
2.00
E5
1.00
E4
1.00
E3
1.00
E2
1.00
E1
1.00
D5
1.00
D4
1.00
D3
1.00
D2
1.00
D1
1.00
C5
1.00
C4
1.00
C3
1.00
C2
1.00
C1
1.00
B5
0.50
B4
0.50
B3
0.50
B2
0.50
B1
0.50
A
1.00
G
1.00
H
1.00
I
1.00
J
1.00
K
1.00
L
1.25
U
4.0
Calc. Depth
16.75
Act. Depth
16.75

Cell Geometry
3
13
Total Volume (m )
Vessel Outlet Area
2
2
2
Dia. (in)
(in )
(ft )
(m )
0.196
0.00136
0.00013
0.5
Horizontal Sectioning
C4
C3
C2
0.75
0.75
0.75
0.0010
0.0010
0.0010
Vertical Sectioning
Zone Depth Volume
Area
3
2
(m )
(m )
(m)
0.0508
4.92E-05
0.0010
0.0508
4.92E-05
0.0010
0.0508
4.92E-05
0.0010
0.0508
4.92E-05
0.0010
0.0508
6.55E-05
0.0013
0.0254
2.46E-05
0.0010
0.0254
2.46E-05
0.0010
0.0254
2.46E-05
0.0010
0.0254
2.46E-05
0.0010
0.0254
3.28E-05
0.0013
0.0254
2.46E-05
0.0010
0.0254
2.46E-05
0.0010
0.0254
2.46E-05
0.0010
0.0254
2.46E-05
0.0010
0.0254
3.28E-05
0.0013
0.0254
2.46E-05
0.0010
0.0254
2.46E-05
0.0010
0.0254
2.46E-05
0.0010
0.0254
2.46E-05
0.0010
0.0254
3.28E-05
0.0013
0.0127
1.23E-05
0.0010
0.0127
1.23E-05
0.0010
0.0127
1.23E-05
0.0010
0.0127
1.23E-05
0.0010
0.0127
1.64E-05
0.0013
0.0254
1.31E-04
0.0052
0.0254
1.31E-04
0.0052
0.0254
1.31E-04
0.0052
0.0254
1.31E-04
0.0052
0.0254
1.31E-04
0.0052
0.0254
1.31E-04
0.0052
0.0318
1.64E-04
0.0052
0.1016
0.0001
4.59E-04

182

C1

2.13E-03
Cell Depth
(in)
(m)
0.0508
2

Check

1
0.0013

4
0.0052

Comments
Overflow Zone

Feed Inlet

Transition Zone

Elutriation Point
Underflow Zone

Feed Size and Distribution Input (user input in bold font):

Screen
Passing
Size
Size
(mesh)
(mm)
Plus 18
***
18 x 20
1.180
20 x 30
0.850
30 x 40
0.600
40 x 50
0.425
50 x 70
0.300
70 x 100
0.212
100 x 150
0.150
Minus 150
0.106
Overall

Retained
Size
(mm)
1.180
0.850
0.600
0.425
0.300
0.212
0.150
0.106
***

Mean
Size
(mm)
1.180
1.001
0.714
0.505
0.357
0.252
0.178
0.126
0.106

Screen
Size
(mesh)
Plus 18
18 x 20
20 x 30
30 x 40
40 x 50
50 x 70
70 x 100
100 x 150
Minus 150
Overall

Comp 1
Rate
kg/s
0.002
0.002
0.002
0.002
0.002
0.002
0.002
0.002
0.002
0.017

Comp 2
Rate
kg/s
0.002
0.002
0.002
0.002
0.002
0.002
0.002
0.002
0.002
0.017

Feed
Rate
kg/s
0.004
0.004
0.004
0.004
0.004
0.004
0.004
0.004
0.004
0.035

Feed Size Distribution

Feed
Feed
Comp. 1
Mass
Mass
Assay
(g)
(%)
(%)
5.00
11.11
50.00
5.00
50.00
11.11
5.00
50.00
11.11
5.00
50.00
11.11
5.00
50.00
11.11
5.00
11.11
50.00
5.00
50.00
11.11
5.00
50.00
11.11
5.00
50.00
11.11
45.00
Feed
Conc.
3
kg/m
83.45
83.45
83.45
83.45
83.45
83.45
83.45
83.45
83.45
751.01

Comp 1
Conc.
3
kg/m
41.72
41.72
41.72
41.72
41.72
41.72
41.72
41.72
41.72
375.51

183

Comp 2
Conc.
3
kg/m
41.72
41.72
41.72
41.72
41.72
41.72
41.72
41.72
41.72
375.51

Comp. 2
Assay
(%)
50.00
50.00
50.00
50.00
50.00
50.00
50.00
50.00
50.00

Comp. 1
Mass
(%)
5.56
5.56
5.56
5.56
5.56
5.56
5.56
5.56
5.56
50.00

Comp. 2
Mass
(%)
5.56
5.56
5.56
5.56
5.56
5.56
5.56
5.56
5.56
50.00

Feed
Conc.
3
3
m /m
0.0277
0.0277
0.0277
0.0277
0.0277
0.0277
0.0277
0.0277
0.0277
0.2490

Comp 1
Conc.
3
3
m /m
0.0157
0.0157
0.0157
0.0157
0.0157
0.0157
0.0157
0.0157
0.0157
0.1417

Comp 2
Conc.
3
3
m /m
0.0119
0.0119
0.0119
0.0119
0.0119
0.0119
0.0119
0.0119
0.0119
0.1073
0.2490

Mixed
Feed
Density
3.02
3.02
3.02
3.02
3.02
3.02
3.02
3.02
3.02

Output Partition Data:

Mean Size
(m)
1.180
1.001
0.714
0.505
0.357
0.252
0.178
0.126
0.106

Feed Rate
(kg/min)
0.12
0.12
0.12
0.12
0.12
0.12
0.12
0.12
0.12

Mean Size
(m)
1.180
1.001
0.714
0.505
0.357
0.252
0.178
0.126
0.106

Feed Rate
(kg/min)
0.12
0.12
0.12
0.12
0.12
0.12
0.12
0.12
0.12

Component No. 1 Partitioning

O/F Rate
U/F Rate
O/F PF
(kg/min)
(kg/min)
(%)
0.00
0.12
0.03
0.00
0.12
0.07
0.00
0.12
0.49
0.00
0.11
3.57
0.02
0.09
19.78
0.07
0.05
58.27
0.11
0.01
91.26
0.12
0.00
99.53
0.12
0.00
99.93
0.000E+00
Component No. 2 Partitioning
O/F Rate
U/F Rate
O/F PF
(kg/min)
(kg/min)
(%)
0.00
0.12
0.01
0.00
0.12
0.01
0.00
0.12
0.10
0.00
0.12
0.78
0.01
0.11
5.95
0.04
0.08
30.14
0.08
0.03
72.30
0.11
0.00
96.28
0.12
0.00
99.23

184

U/F PF
(%)
99.97
99.93
99.51
96.43
80.22
41.73
8.74
0.47
0.07
0.081920

Balance Error
(%)
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00

U/F PF
(%)
99.99
99.99
99.90
99.22
94.05
69.86
27.70
3.72
0.77

Balance Error
(%)
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00

Output Partition Curves:

Yield to Underflow
100

PF

75

50
Component 1
Ef ficiency Points 1
Component 2

25

Ef ficiency Points 2
0
0.10

1.00
Mean Size (m m )

Output Separation Efficiency Data:

Comp. 1 Check
Feed
O/F
U/F
kg/min
kg/min
kg/min
1.0500
0.4351
0.6149
100.00%
41.44%
58.56%
Comp. 1 Sizing Efficiency
PF
(mm)
Ep
25.00
0.216
0.060
50.00
0.273
Imp.
75.50
0.337
0.220

Comp. 2 Check
Feed
O/F
U/F
kg/min
kg/min
kg/min
1.0500
0.3556
0.6944
100.00%
33.87%
66.13%
Comp. 2 Sizing Efficiency
PF
(mm)
Ep
0.050
25.00
0.171
50.00
0.214
I
0.230
75.00
0.270

185

Output Volumetric Concentration Data:

Volumetric Conc. by Zone

0.050

0.249

0.140
0.140
0.140
0.140
0.092
0.058
0.048
0.042
0.037
0.032
0.022
0.353

186

0.000

Output Product Recovery and Particle Size Distribution Data:

Combined Products
Feed Rate
Underflow Rate
Overflow Rate
Mean Size Combined Comp. 1 Comp. 2 Combined Comp. 1 Comp. 2 Combined Comp. 1 Comp. 2
(m)
kg/min
kg/min
kg/min
kg/min
kg/min
kg/min
kg/min
kg/min
kg/min
1.180
0.23
0.12
0.12
0.23
0.12
0.12
0.00
0.00
0.00
1.001
0.23
0.12
0.12
0.23
0.12
0.12
0.00
0.00
0.00
0.714
0.23
0.12
0.12
0.23
0.12
0.12
0.00
0.00
0.00
0.505
0.23
0.12
0.12
0.23
0.11
0.12
0.01
0.00
0.00
0.357
0.23
0.12
0.12
0.20
0.09
0.11
0.03
0.02
0.01
0.252
0.23
0.12
0.12
0.13
0.05
0.08
0.10
0.07
0.04
0.178
0.23
0.12
0.12
0.04
0.01
0.03
0.19
0.11
0.08
0.126
0.23
0.12
0.12
0.00
0.00
0.00
0.23
0.12
0.11
0.106
0.23
0.12
0.12
0.00
0.00
0.00
0.23
0.12
0.12
Totals
2.10
1.05
1.05
1.31
0.61
0.69
0.79
0.44
0.36

Mean Size
(m)
1.180
1.001
0.714
0.505
0.357
0.252
0.178
0.126
0.106
Totals

Mean Size
(m)
1.180
1.001
0.714
0.505
0.357
0.252
0.178
0.126
0.106
Totals

Feed Assay
Comp. 1 Comp. 2
%
%
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0

Underflow Assay
Comp. 1 Comp. 2
%
%
50.0
50.0
50.0
50.0
49.9
50.1
49.3
50.7
46.0
54.0
37.4
62.6
24.0
76.0
11.2
88.8
8.4
91.6
47.0
53.0

Underflow Overflow
Overflow Assay
Comp. 1 Comp. 2
PSD
PSD
%
%
%
%
83.7
16.3
17.82
0.01
83.7
16.3
17.81
0.01
83.5
16.5
17.77
0.09
82.0
18.0
17.43
0.64
76.9
23.1
15.53
3.80
65.9
34.1
9.94
13.05
55.8
44.2
3.25
24.13
50.8
49.2
0.37
28.89
50.2
49.8
0.08
29.39
55.0
45.0
***
***

Combined Product Recoveries

Total Mass Recovery
Comp 1 Recovery
Feed
U/F
O/F
U/F
O/F
%
%
%
%
%
100
100.0
0.0
100.0
0.0
100
100.0
0.0
99.9
0.1
100
99.7
0.3
99.5
0.5
100
97.8
2.2
96.4
3.6
100
87.1
12.9
80.2
19.8
100
55.8
44.2
41.7
58.3
100
18.2
81.8
8.7
91.3
100
2.1
97.9
0.5
99.5
100
0.4
99.6
0.1
99.9
100.0
62.3
37.7
58.6
41.4

187

Comp 2 Recovery
U/F
O/F
%
%
100.0
0.0
100.0
0.0
99.9
0.1
99.2
0.8
94.0
6.0
69.9
30.1
27.7
72.3
3.7
96.3
0.8
99.2
66.1
33.9

VITA

Jaisen Nathaniel Kohmuench, son of William C. and Carolyn A. Kohmuench was born in
Teaneck, NJ, on the 23rd day of May, 1973. He graduated from Hunterdon Central Regional
High School in the spring of 1991. The following fall, he was granted admission to Virginia
Polytechnic Institute and State University (Virginia Tech), where he went on to gain a Bachelor
of Science degree in Mining and Minerals Engineering. During his time as an undergraduate, he
was highly involved in the Burkhart Mining Society, the student chapter of the Society for
Mining, Metallurgy and Exploration (SME). It was during the 94-95 school year that he served
as the vice president of this organization. He also successfully passed the EIT exam.
After graduating in the spring of 1995, he remained at Virginia Tech to pursue a Master
of Science degree in Mining and Minerals Engineering with an emphasis in minerals processing.
He completed his degree in the fall semester of 1997 and immediately enrolled in the Ph.D.
program. He is currently a student member of SME. As a doctoral candidate, he conducted
several presentations at professional meetings in addition to authoring (or co-authoring) over 11
works, including two full peer-reviewed journal publications. While a student, he earned the
Graduate Student of the Year Award for both the 1996 and 1999 school years.
Upon completion of his doctoral dissertation, he and his wife, Kathryn, will relocate to
Erie, Pennsylvania. Here, having already accepted a position, Jaisen will begin his professional
career as a process engineer for Eriez Magnetics.

Jaisen N. Kohmuench

188